consulting info with logo template etc in header and footer · web viewtungsten carbide with a...

Document for Submission to the University of Heriot-Watt Life Sciences’ Research Ethics Committee, v1.0

February 2012

Tungsten Carbide with Cobalt Binder: An Historical Cohort and Nested Case-Control Study of Lung Cancer – UK Component

In collaboration with:University of PittsburghUniversity of Illinois at Chicago

OUR IMPACT ON THE ENVIRONMENT

At IOM we seek to minimise our environmental impact. We produce thousands of reports every year and these consume a large quantity of paper. To minimise our impact on the environment, we prefer to only provide an electronic copy of reports, although we can provide a paper copy on request. If you have any additional requirements please let us know.

CONTENTS

EXECUTIVE SUMMARY 8

1 BACKGROUND AIMS AND OBJECTIVES 101.1 Objective 101.2 Specific Aims 101.3 Background 101.4 Significance 11

2 RESEARCH DESIGN AND METHODS 122.1 Study Interventions 122.2 Task 1 – Site Cohort Enumeration 122.3 Task 2 – Data Processing, Validation and Management 132.4 Task 3 – Vital Status Tracing 132.5 Task 4 – Integration of Exposure Data 132.6 Statistical Analysis 14

3 HUMAN SUBJECTS 153.1 General Characteristics 153.2 Inclusion/Exclusion Criteria 153.3 Identifying Subjects 153.4 Ethical clearance and Informed Consent 15

4 POTENTIAL RISKS 164.1 Confidentiality 164.2 Potential Benefits 164.3 Data and Safety Monitoring Plan 16

5 COSTS AND PAYMENTS 175.1 Research Study Payments 175.2 Qualifications of Investigators 175.3 Sources of Support 17

6 REFERENCES 18

7 APPENDIX 1 – RESEARCH PROPOSAL FOR THE INTERNATIONAL STUDY 19

Tungsten Carbide with a Cobalt Binder: an historical cohort and nested case-control study of lung cancer - UK component

EXECUTIVE SUMMARY

Information gathered in a feasibility study shows that a scientifically rigorous and comprehensive epidemiology study of workers employed in the manufacture or production of tungsten carbide with a cobalt binder (WCCo) is feasible and should include workers from 10 manufacturing sites in the United States and nine manufacturing sites in Europe, two of which are located in the West Midlands in England. This large epidemiology study, referred to as the Phase 3 study, will represent multiple companies, countries and manufacturing processes and will be larger, more robust and more definitive than any WCCo epidemiology study done to date.

The proposed historical cohort study is designed as an essential step in appraising the health implications of occupational exposure to WCCo. The primary research objectives of the Phase 3 study are:

1. To investigate the total and cause-specific mortality experience of current and former workers potentially exposed to WCCo at multiple US and EU industrial sites that produce(d) WCCo and/or manufacture(d) WCCo products, as compared with the experience of the corresponding national and local populations from which the workforces were drawn, with adjustment for potential confounding factors and with emphasis on malignant neoplasms of the lung. For the UK component, this will involve using mortality rates for the West Midland and for England and Wales combined. An analysis for all sites combined will be carried out by colleagues in Pittsburgh; a separate analysis of the UK sites will be carried out by IOM researchers.

2. To characterize as completely as possible the past and current working environment of the study members from the sites relative to work area, job title/function and potential for exposure to WCCo as well as potential co-exposures to several known or suspected human carcinogens including tungsten (W), tungsten carbide (WC), carbon (C) and cobalt (Co). Work to gather the data required at the UK sites will be undertaken by IOM, with the analysis being carried out by each unit using agreed upon methodology that will be proposed by colleagues at the University of Illinois at Chicago and will be modified or accepted by consensus at the forthcoming conference in Edinburgh in June.

3. To determine the relationship between level and duration of WCCo exposure and mortality from malignant lung neoplasms with analytic adjustment to the extent possible for potential co-exposures, including tobacco smoking habits. Whether a separate analysis will be carried out for the two UK sites will very much depend on a sufficient number of lung cancer cases being recruited to the study and therefore sufficient statistical power for a separate meaningful analysis.

4. To provide a framework for ongoing mortality surveillance of workers potentially exposed to WCCo with and without concomitant co-exposures.

Our statistical analysis of the study data will consist of two major parts, each of which is designed to address specific objectives of the study:

Part 1 Analysis of total and cause-specific mortality patterns in relation to basic demographic and work history factors (e.g., study site, race, gender, age, calendar time, year of hire, duration of employment and the time since first employment), with focus on cancer mortality and emphasis on the implicated site of interest (lung).

of 125


Part 2 Analysis of total and cause-specific mortality in relation to occupational exposure to WCCo with analytic adjustment for potential confounding and/or effect modification by smoking and co-exposures to known or suspected carcinogens including W, WC, C and Co also with focus on cancer mortality and emphasis on lung cancer.

The cohort will include approximately 15,000 U.S. workers employed since 1950 at one or more of included locations.

The racial, gender and ethnic characteristics of the study population reflect the demographics of the areas surrounding the plants. No exclusion criteria will be based on race, ethnicity, gender or HIV status.

of 125


1 BACKGROUND AIMS AND OBJECTIVES

1.1 OBJECTIVE

Information gathered in a feasibility study shows that a scientifically rigorous and comprehensive epidemiology study of workers employed in the manufacture or production tungsten carbide with a cobalt binder (WCCo) is feasible and should include workers from at 10 manufacturing sites in the United States and nine in Europe, including two from the United Kingdom. This large epidemiology study, referred to as the Phase 3 study, will represent multiple companies, countries and manufacturing processes and will be larger, more robust and more definitive than any WCCo epidemiology study done to date.

1.2 SPECIFIC AIMS


1. To investigate the total and cause-specific mortality experience of current and former workers potentially exposed to WCCo at multiple US and EU industrial sites that produce(d) WCCo and/or manufacture(d) WCCo products, as compared with the experience of the corresponding national and local populations from which the workforces were drawn, with adjustment for potential confounding factors and with emphasis on malignant neoplasms of the lung.

2. To characterize as completely as possible the past and current working environment of the study members from the sites relative to work area, job title/function and potential for exposure to WCCo as well as potential co-exposures to several known or suspected human carcinogens including W, WC, C and Co.

3. To determine the relationship between level and duration of WCCo exposure and mortality from malignant lung neoplasms with analytic adjustment to the extent possible for potential co-exposures, including tobacco smoking habits.


1.3 BACKGROUND

Several international health research agencies have recently acted to label tungsten carbide with a cobalt binder (WCCo), also referred to as “hardmetal,” as a probable human carcinogen. A review of the scientific basis for this decision reveals significant weaknesses in the primary occupational epidemiologic studies of French and Swedish workers on which it was based (Hogstedt and Alexandersson, 1990; Lasfargues et al., 1994; Moulin et al., 1998; Wild et al., 2000). To address these limitations, a 3-phase occupational epidemiology investigation of workers employed in the tungsten carbide (WC) industry was initiated in the early 2000’s by the International Tungsten Industry Association (ITIA).

Phase 1 of this investigation was a feasibility study conducted by BBL, Inc. in 2006 to determine the availability and accessibility of company records needed for the main epidemiology study that comprises Phase 3 of the investigation. The Phase 1 feasibility study was extended and enhanced in Phase 2 by the University of Pittsburgh (UPitt) and

of 125


the University of Illinois at Chicago (UIC) under a research contract from ITIA. Phase 2, conducted from October 2007 through October 2008, allowed researchers: (1) to communicate with plants directly to help clarify specific study needs; (2) to establish communication avenues with key plant personnel and enabled those personnel to prepare their facilities for inclusion in the Phase 3 study; (3) to apply different or additional feasibility criteria to the data obtained in Phase 1 and (4) to establish parameters that created the framework within which the Phase 3 study will proceed.

1.4 SIGNIFICANCE

Based on the conclusions of the feasibility study, sufficient demographic and WH data are available to conduct an historical cohort mortality study of former and current workers with potential exposure to WCCo at each of the study sites. Multiple sites were chosen to afford better opportunities for contrasting cohort attributes, processes, work practices and exposures; multiple sites also increases the likelihood of producing definitive and informative conclusions by increasing the statistical power and the precision of the risk estimates for detecting true excess risks overall and in relation to occupational factors.

of 125


2 RESEARCH DESIGN AND METHODS

Investigators will first completely ascertain the cohort of workers with potential exposure to WCCo (Research Objective 1). The University of Illinois at Chicago (1UIC) will perform a comprehensive exposure reconstruction of manufacturing processes and will develop task and time-specific estimates of exposure (Research Objective 2). The exposure matrix will enable UPitt investigators to construct summary measures of exposure to contaminants (Research Objective 3) and enable ongoing mortality surveillance of the cohort (Research Objective 4).

1The historical cohort study will provide the epidemiological platform for the proposed investigation, including a nested case-control study of lung cancer. The cohort study will focus on mortality from lung cancer and other cause of death categories (including total mortality).

2.1 STUDY INTERVENTIONS

UPitt under the direction of the Principal Investigator, Gary M. Marsh, Ph.D., and Co-Investigator, Jeanine M. Buchanich, Ph.D., will oversee all aspects of the study. For the UK sites, work will be overseen by Damien McElvenny, Principal Epidemiologist at the IOM.

2.2 TASK 1 – SITE COHORT ENUMERATION

The cohort will be enumerated by UPitt using all available corporate and plant-based employee records; IOM will enumerate the two UK cohorts and make data available to UPitt in a suitably anonymised form. The cohort will include active, terminated, retired and deceased employees. To the extent possible, relevant cohort data will be derived initially from records maintained by the companies in machine-readable format. To ensure that the cohort enumeration is complete and accurate, IOM staff will also review all hard copy employee records at the plants. Hard copy records will include work applications, detailed job service records and death certificates. Scanners will be used to capture any cohort data from the hard copy records.

The cohort enumeration will include four basic categories of employee data:

a. Data on personal identifiers and demographic factors (e.g., name, national insurance number, national health service number, birth date, gender, pay type and employment status)

b. Data on employment history and exposure (e.g., for each job held, beginning and end dates, department, work area and job title, types and levels of exposure (if known))

c. Data on health outcome (certification of death and cause of death)

d. Data on potential confounding factors (e.g., tobacco smoking history and previous and subsequent employment)

Due to the variations in available cohort data, each plant will follow a unique protocol during the cohort enumeration phase.

of 125


These sites were chosen because they have been in operation long enough for a meaningful epidemiological study (20+ years), have detailed work history available for all employees who ever worked at the site and have had more than 500 employees historically. Two UK sites will be included in the cohort enumeration for Part 1 of the Phase 3 study. The sites are the Sandvik plants at Coventry (hard metal production, powder production) and Featherstone (pressing, green shaping, sintering, hard grinding, coating).

At the end of Task 1, IOM staff will have visited both of the two UK sites included in Part 1 of the Phase 3 study and collected all relevant demographic and work history data.

2.3 TASK 2 – DATA PROCESSING, VALIDATION AND MANAGEMENT

2.3.1 Electronic records

Electronic records will be supplied directly from the sites. A process for electronically collecting data from each facility will be implemented to combine all data into single relational database system housed on IOM’s secure servers, a copy of which will be transferred, after appropriate anonymisation to UPitt colleagues. The electronic data will be exported into a common data format, for example, an ASCII text file, a Microsoft Excel file or a Microsoft Access database.

2.3.2 Hard copy records

Hard copy records will be produced from all scanned records. Hard copy records will be processed by IOM clerical staff or, if the volume is sufficiently large, the scanning will be subcontracted to an approved subcontractor. Abstracted data will be independently double-entered and discrepancies resolved against source documentation. Separate data files will be maintained according to the source of the data (e.g. payroll, pension, human resources). Data from different sources will be merged onto the database once deemed to have been accurately entered. If the data entry is outsourced, regular checks will be made with the subcontractor on progress and accuracy of their work. The subcontractor staff will sign appropriate confidentiality undertakings and when their work is complete, will provide all the data collected to IOM and not retain any data on their own systems. They will be asked to formally declare that they have complied with this. The method of transfer from the subcontractor to IOM will be agreed in advance and will be secure, making appropriate use of adequate encryption.

2.4 TASK 3 – VITAL STATUS TRACING

For the international analysis, the cohort’s vital status will be determined as of December 31, 2008 or a later date if deemed feasible. For the UK sites, it is likely that mortality data will be complete to the end of 2010 or possibly 2011, depending on the timing of death ascertainment for analysis. All identified employees will be sent for tracing at the National Health Service Central Register (NHSCR) in Southport. IOM will be notified of workers that have died, embarked or are untraceable on the NHSCR.

2.5 TASK 4 – INTEGRATION OF EXPOSURE DATA

UPitt investigators will work closely with UIC investigators to coordinate the integration and verification of data from the epidemiological and exposure assessment components of the study. This will involve linking the job/exposure matrix (JEM) created by UIC investigators with the individual WHs and epidemiological data compiled by the UPitt investigators and

of 125


exposure analysis provided by each unit. The JEM will be standardised based on the consensus between participating units so that the relevant parts can be shared with participating units for their analysis. UPitt will also serve as a clearing house for all exposure assessment-related study data, and will be responsible for maintaining an inventory and ultimately an archive of all project data. UIC/UPitt will make the relevant part of the job/exposure matrix for the UK sites available to IOM researchers so that a UK analysis can be undertaken in a way that is consistent with the international analysis.

2.6 STATISTICAL ANALYSIS

The following section applies to the statistical analysis plan for the entire Phase 3 study.

Because the recommended study sites are highly diverse relative to geographic location, cohort size and cohort entry period (facility start date in most cases), we propose to approach the statistical analysis in a site-specific manner, pooling data across sites only if warranted by evidence of sufficient homogeneity. The advantage of such diversity from an epidemiological standpoint is the ability to assess the consistency of our findings across the study populations. Efforts will be made to pool data when warranted, however, as this will improve the precision of the mortality risk estimates and increase the statistical power to detect epidemiologically important excess risks. A separate analysis will be undertaken for each of the two UK plants, and if appropriate, for the two plants combined.


Part 1 Analysis of total and cause-specific mortality patterns in relation to basic demographic and work history factors (e.g., study site, race, gender, age, calendar time, year of hire, duration of employment and the time since first employment), with focus on cancer mortality and emphasis on the implicated site of interest (lung).


Part 1 will include a descriptive analysis of externally standardized mortality ratios (SMRs), with SMRs based on both national and local standard population death rates (for the UK, this will be for England and Wales and the West Midlands, respectively). The descriptive phase of the analysis is useful for examining the basic structure and consistency of the overall and study factor-specific mortality risks and for determining appropriate cutpoints for categorizing continuous study variables.

Part 2 will include multi-variablee analyses of internal cohort rates to facilitate the simultaneous adjustment for multiple potential confounding factors and the assessment of effect modification among two or more study factors. In particular, Part 2 will include relative risk regression modeling (based on Cox proportional hazards model) of internal cohort rates.

The modeling of internal cohort rates provides mortality comparisons within the cohort that are unbiased by the “healthy worker effect” associated with external general population comparisons. The scope of the multi-variable analysis for lung cancer mortality will be determined largely by the corresponding number of observed deaths and their factor-

of 125


specific distributions. In addition, we may need to limit certain statistical analyses to subgroups with sufficiently large numbers of observed deaths (e.g., white males).

3 HUMAN SUBJECTS

3.1 GENERAL CHARACTERISTICS

The UK cohort will include between 1000 and 2000 workers employed since 1950 at one of the two UK factories.

The socioeconomic status of the study population is likely to reflect the demographics of the area surrounding the plants (the West Midlands). No exclusion criteria will be based on gender or length of employment. Contract staff may have to be excluded if details are not held at the factories concerned.

3.2 INCLUSION/EXCLUSION CRITERIA

Only those employees who were working in 1950 or hired later will be included in the study. Those employees who ended their employment prior to 1950 will be excluded. There is a possibility that some contract workers may be excluded.

3.3 IDENTIFYING SUBJECTS

All workers employed by Sandvik or its predecessor at one of the two factories beginning in 1950 or 1966 (depending which plant) will be included in the study.

3.4 ETHICAL CLEARANCE AND INFORMED CONSENT

Ethical clearance for the UK component of this study will be sought from an appropriate NHS research ethics committee in the West Midlands. In addition exemption from having to gain informed consent for the cohort study from current and former members of the workforces will be sought. There may be an expectation that presentations may have to be made to the current workforce about the proposed study, allowing current workers to opt-out of the study if they so wish (from the National Information Governance Board and/or the management or workforce representatives). Trades Unions or routes of publicising the study may have to be utilised to bring the study to the attention of former employees.

For the case-control study, informed consent will need to be obtained for all participants (or their proxies). These participants may have to be approached via their General Practitioners (GPs). It is anticipated that a small payment may be made to GPs to facilitate this, and that those who are interviewed are compensated for their expenses. The existence of the study may be advertised in advance to potentially relevant GP practices; advice on this will be sought from the Medical Research Information Service. Advice will also be sought as to whether ethical clearance for the case-control study should be sought separately from that for the case-control study, or it be sought at the same time.

of 125


4 POTENTIAL RISKS

Although every effort will be made to ensure the confidentiality of subject data, study members are subject to the risk of a breach of confidentiality.

4.1 CONFIDENTIALITY

All data relating to the study will be kept confidential and any statistical analyses carried out will be done so using a pseudonymised data set. Any data transferred to UPitt and/or UIC for the international analysis will also be via a pseudonymised data set. All electronic data held by IOM will be on a secure server and in full accordance with IOM’s system level security policy. This policy will be subject to the scrutiny of that National Information Governance Board’s Security Review Team. All paper records will be held in lockable cabinets only accessible by named researchers involved with the study. Any data that is published will be done in such a way that individuals cannot be identified.

4.2 POTENTIAL BENEFITS

As will all studies of this type, individual subjects will not directly benefit from participation but workers exposed to tungsten carbide and society as a whole will potentially benefit from the knowledge garnered in this study.

4.3 DATA AND SAFETY MONITORING PLAN

Data collection procedures and policies, vital status tracing progress and confidentiality assurances will be monitored. Study guidelines regarding patient confidentiality are strictly enforced and will also be reviewed quarterly to determine potential improvements.

of 125


5 COSTS AND PAYMENTS

5.1 RESEARCH STUDY PAYMENTS

No research subject will be charged nor paid to be a participant in this study, although GPs may be paid a small fee to pass on details of the study to potential participants and participants may be paid a small fee for expenses and inconvenience.

5.2 QUALIFICATIONS OF INVESTIGATORS

The Principal Investigator will be Gary M. Marsh, Ph.D. who has over 20 years of experience in biostatistical and epidemiological research. Dr. Marsh is a Professor of Biostatistics, Graduate School of Public Health and will supervise, coordinate and oversee all aspects of the core support program, and will play a major role in the annual mortality analyses and preparation of reports.

Jeanine M. Buchanich, Ph.D., Research Assistant Professor of Biostatistics and Deputy Director of Epidemiology for the Center for Occupational Biostatistics and Epidemiology (COBE) at the Graduate School of Public Health, University of Pittsburgh will serve as a Co-Investigator. Dr. Buchanich has more than 10 years experience in occupational health research and has served as project coordinator on numerous projects. She has authored more than 25 occupational health articles appearing in the peer-reviewed literature.

A second co-investigator of the proposed project is Ada O. Youk, Ph.D., Research Assistant Professor of Biostatistics. Dr. Youk will have primary responsibility for coordinating the analytical aspects of the core support program including the construction of standard rate files and the annual mortality analyses. She will also play a major role in the preparation of progress reports and scientific publications.

The lead investigator for the UK component of this study is Professor Damien McElvenny BSc(Hons), MSc, CStat, principal epidemiologist at IOM. Damien has over 25 years of experience in occupational health research and has been first-named author on several peer-reviewed scientific papers, and over 30 papers in total.

5.3 SOURCES OF SUPPORT

This study is being funded under subcontract from the International Tungsten Industry Association via the University of Pittsburgh.

of 125


6 REFERENCES

Breslow NE, Day NE. The design and analysis of cohort studies. In: Statistical Methods in Cancer Research, Vol. II. International Agency for Research on Cancer. Lyon, France: IARC Scientific Publications No. 82, 1987.

Buchanich JB, Dolan DG, Marsh GM, Madrigano J. Under ascertainment of deaths using Social Security records: A recommended solution to a little-known problem. Am J Epidemiol 162:193-194, 2005.

Cox DR. Regression models and life tables (with discussion). J R Stat Soc 34B:187-220, 1972.

Cox DR. Partial likelihood. Biometrika 62:269-276, 1975.

Doll R. Occupational cancer: A hazard for epidemiologists. Int J Epidemiol 14:22-31, 1985.

Esmen NA. Exposure estimation in four major epidemiologic studies in acrylonitrile industry. Scand J Work Env Hea 24:Sup2:63-70, 1998.

Esmen NA, Kennedy KJ, Hall TA, Phillips ML, Marsh GM. Classification of worker exposures. Chem-Biol Interact 166:245-253, 2007a.

Esmen NA, Hall TA, Phillips ML, Marsh GM. Chemical process based reconstruction of exposures for an epidemiological study: I. Theoretical and methodological issues. Chem-Biol Interact 166:254-263, 2007b.

Esmen NA, Hall TA, Phillips ML, Jones EP, Basara H, Marsh GM, Buchanich JM. Chemical process based reconstruction of exposures for an epidemiological study: II. Estimated exposures to chloroprene and vinyl chloride. Chem-Biol Interact 166:264-276, 2007c.

Hogstedt C, Alexandersson R. Mortality among hard metal workers. Arbete Hälsa, 21, 1-26, 1990.

IARC. Monographs on the evaluation of carcinogenic risks to humans. Volume 86, Cobalt in hard metals and cobalt sulfate, Lyon, France, 2006.

Lasfargues G, Wild P, Moulin JJ, Hammon B, Rosmorduc B, Rondeau du Noyer C, Lavandier M and Moline JJ. Lung cancer mortality in a French cohort of hard-metal workers. Am J Ind Med 26:585-595, 1995.

Marsh GM, Youk AO, Stone RA, Sefcik S, Alcorn CW. OCMAP-PLUS, A new program for the comprehensive analysis of occupational cohort data. J Occup Environ Med 40:351-362, 1998.

Marsh GM, Youk AO, Stone RA, Buchanich JM, Gula MJ, Smith TJ, Quinn MM. Historical cohort study of US man-made vitreous fiber production workers: I. 1992 fiberglass cohort follow-up: initial findings. J Occup Environ Med 43:741-756, 2001.

of 125


Marsh GM, Ehland J, Sefcik S, Alcorn C. Mortality and population data system (MPDS). Pittsburgh, PA: University of Pittsburgh (Department of Biostatistics Technical Report), 2005.

Marsh GM, Youk AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality patterns among industrial workers exposed to chloroprene and other substances. I. General mortality patterns. Chem-Biol Interact 166: 285-300, 2007a.

Marsh GM, Youk AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality patterns among industrial workers exposed to chloroprene and other substances. II. Mortality in relation to exposure. Chem-Biol Interact 166:301-316, 2007b. Moulin JJ, Wild P, Mur JM, Fournier-Betz M, Mercier-Gallay M. A mortality study of cobalt production workers: An extension of the follow-up. Am J of Ind Med 23:281-288, 1993.

Moulin JJ, Wild P, Romazini S, Lasfargues G, Peltier A, Bozec C, Deguerry P, Pellet F, Perdrix A. Lung cancer risk in hard metal workers. Am J Epidemiol 148:241-248, 1998.

Mur JM, Moulin JJ, Charruyer-Seinerra MP, Lafitte J. A cohort mortality study among cobalt and sodium workers in an electrochemical plant. Am J Ind Med 11:75-81, 1987.

Phillips ML, Esmen NA. Computational method for ranking task-specific exposures using multi-task time-weighted average samples. Ann Occup Hyg 43:201-213, 1999.

Santhanam AT. Cemented carbides. In: Kirk-Othmer Encyclopedia of Chemical Technology, Vol. 4, 4th Ed. Ed. Kroschwitz JI and Howe-Grant M. New York: John Wiley & Sons, 1992.

STATA Statistical Software: Release 11.0. College Station, TX, STATA Corp., 2009.

Wild P, Perdix A, Romazini S, Moulin JJ, Pellet F. Lung cancer mortality in a site producing hard metals. Occup Environ Med 57:568-573, 2000.

Youk AO, Marsh GM, Stone RA, Buchanich JM, Smith TJ. Historical cohort study of U.S. man-made vitreous fiber production workers III: Analysis of exposure-weighted measures of respirable fibers and formaldehyde in the nested case-control study of respiratory system cancer. J Occup Environ Med 43:767-778, 2001.

Youk AO, Buchanich JM, Marsh GM, Cunningham MC, Esmen NE. Pharmaceutical production workers and the risks of mortality from respiratory system cancer and lymphatic and hematopoetic tissue cancers. J Occup Environ Med 51:903-915, 2009.

of 125


7 APPENDIX 1 – RESEARCH PROPOSAL FOR THE INTERNATIONAL STUDY

7.1 BACKGROUND

Tungsten carbide (WC) is the most common of the hard metals, materials formed by binding or cementing metallic carbides with a soft and ductile metal binder (IARC, 2006), usually cobalt or nickel. Hard metals are manufactured by a powder metallurgy process consisting of a sequence of carefully-controlled steps designed to obtain a final product with specific properties, microstructure and performance (Santhanam, 1992). Several international health research agencies, including the International Agency for Research on Cancer (IARC, 2006), have recently acted to label WC with a cobalt binder (WCCo) as a probable human carcinogen (category 2A). IARC ruled there was limited evidence in humans and sufficient evidence in animals that WCCo acted as a lung carcinogen (IARC, 2006).

A review of the scientific basis for this decision revealed significant weaknesses in the primary occupational epidemiologic studies of French and Swedish workers on which it was based (Hogstedt and Alexandersson, 1990; Lasfargues et al., 1995; Moulin et al., 1998; Wild et al., 2000). Three of the four epidemiological studies were not mutually independent (Lasfargues et al., 1995; Moulin et al., 1998; Wild et al., 2000). The four epidemiological investigations also shared limitations which negatively impacted their interpretability, including: poorly defined exposure assessments; inconsistent and ambiguous smoking histories and classification; a lack of regional mortality comparisons; a lack of internal cohort rate comparisons; and incomplete job histories. To address these limitations, a 3-phase, international occupational epidemiology investigation of workers employed in the tungsten carbide (WC) industry was initiated in the early 2000’s by the International Tungsten Industry Association (ITIA).

7.1.1 Phase 1

Phase 1 of this investigation was a feasibility study conducted by BBL, Inc. in 2006 to determine the availability and accessibility of company records needed for the main epidemiology study that comprises Phase 3 of the investigation. The Phase 1 feasibility study was extended and enhanced in Phase 2 by the University of Pittsburgh (UPitt) and the University of Illinois at Chicago (UIC) under a research contract from ITIA.

7.1.2 Phase 2

In Phase 2, conducted from October 2007 through October 2008, UPitt and UIC developed and applied several criteria for a candidate site to be included in the Phase 3 study, including a minimum size of 100 or more employees historically for United States (US) sites and 500 or more employees historically for European (EU) sites. Sites also had to have been producing WCCo or WC products since at least 1980 to allow an adequate latency period for disease development. Additionally, detailed work history (WH) information had to be available for all employees who ever worked at the facility. The last consideration, for non-US sites, was whether vital status tracing was possible within that country. This information was gathered via the telephone survey and/or during a site visit.

The study questionnaire used in the telephone survey had separate epidemiology and industrial hygiene components and was administered by a professional interviewer on the

of 125


UPitt staff. Surveys were completed in April 2008. Following the surveys, UPitt and UIC investigators conducted four US and two EU site visit trips that included a total of 18 sites during the Phase 2 study. The sites visited were representative of the candidate sites with respect to country, company and process.

A final report on the results of the Phase 2 enhanced feasibility study was submitted to ITIA in November 2008 (Marsh et al., 2008). The Phase 2 report, which also included a draft protocol for the Phase 3 study, showed that a scientifically rigorous and comprehensive epidemiology study of hardmetal workers was feasible and should include workers from nine manufacturing sites in the US and nine manufacturing sites in EU including sites in Austria (AT), England (UK), Germany (DE) and Sweden (SE). Since the submission of the Phase 2 report, one company (ATI) chose to withdraw from the study, requiring us to replace two ATI plants with three plants owned by Kennametal. This brings the total number of plants in the US portion of the study to 10.

To ensure that the data collected from all US and EU can be combined ultimately for purposes of a pooled cohort data analysis, the Phase 3 protocol includes strict provisions for maintaining a common data collection and statistical analysis protocol. UPitt and UIC are responsible for coordinating these activities during the course of Phase 3.

7.1.3 Phase 3

7.1.3.1 US Component

In 2009, UPitt and UIC began limited work on the US portion of the Phase 3 epidemiology study under a grant from the Pennsylvania State Department of Health (PADOH). This work, termed Part 1 of the Phase 3 study, involves the collection and processing of relevant study data (demographic, work history and industrial hygiene data) from several of the US manufacturing sites and represents the first step to enumerating the US study population. Prior to data collection, the two participating companies in the US study, Sandvik and Kennametal, notified their employees of their possible inclusion in the historical cohort study. Table 1 shows the progress of the 10 sites participating in the US portion of the study.

of 125


Table A.1: Progress of US Sites Included in Tungsten Carbide Workers Study

Site (Company)Company Employee

NotificationSite Tour

Work History (WH) Data Collection

WH Data Processing

IH Data Collection

IH Data Processing

Asheboro, NC(Kennametal) Completed Not Done Not Done Not Done Not Done Not Done

Bedford, PA(Kennametal) Completed Completed Completed Ongoing Completed Not Done

Irwin, PA(Kennametal) Completed Completed Not Done Not Done Not Done Not Done

Fallon, NV(Kennametal) Completed Completed Mostly

completed Ongoing Completed Not Done

Henderson, NC(Kennametal) Completed Completed Completed Completed Completed Not Done

Johnson City, TN(Kennametal) Completed Completed Completed Not Done Completed Not Done

Orwell, OH(Kennametal) Completed Completed Completed Completed Completed Not Done

Roanoke Rapids, NC(Kennametal)

Completed Completed Completed Ongoing Completed Not Done

Traverse City, MI(Kennametal) Completed Not Done Not Done Not Done Not Done Not Done

West Branch, MI(Sandvik) Completed Completed Completed Completed Completed Not Done

of 125


7.1.3.2 EU Component

In 2010, data collection activities began at the Reutte, AT site owned by the Ceratizit Company; the AT study is being directed by Dr. Manfred Neuberger. Due to lack of funding, none of the other country-specific investigations has begun. The current proposal includes funding for these investigations. Table 2 shows the status of the nine plants participating in the EU component of the study.

Table A.2: Progress of EU Sites Included in Tungsten Carbide Workers Study

Site (Company)Company Employee

NotificationSite Tour

Work History (WH) Data Collection

WH Data Processing

IH Data Collection

IH Data Processing

Fagersta, Sweden(Sandvik) Not Done Completed Not Done Not Done Not Done Not Done

Gimo, Sweden(Sandvik ) Not Done Completed Not Done Not Done Not Done Not Done

Stockholm, Sweden (Sandvik) Not Done Completed Not Done Not Done Not Done Not Done

Coventry, England(Sandvik) Not Done Completed Not Done Not Done Not Done Not Done

Featherstone, England (Sandvik) Not Done Completed Not Done Not Done Not Done Not Done

Mistelgau, Germany (Kennametal) Not Done Completed Not Done Not Done Not Done Not Done

Ebermannstadt, Germany (Kennametal)

Not Done Completed Not Done Not Done Not Done Not Done

Essen, Germany (Kennametal) Not Done Completed Not Done Not Done Not Done Not Done

Reutte, Austria(Ceratizit) Unknown Completed Underway Underway Underway Underway

7.2 SIGNIFICANCE OF PHASE 3 STUDY

The Phase 3 historical cohort study was designed to overcome the methodological limitations of earlier studies by including a rigorous exposure assessment component, a nested case-control study of lung cancer cases and matched controls to obtain smoking information and the use of external and internal cohort rate comparisons. The Phase 3 study, which will represent multiple sites and manufacturing processes, will be larger, more robust and more definitive than any WCCo epidemiology study done to date.

Our preliminary data collection revealed that five main exposure agents should be evaluated and compared for potential adverse health effects: tungsten (W), WC, WCCo, carbon black (C) and cobalt (Co). Our large cohort size will involve more person-years than any previous study of tungsten carbide workers, giving us greater power when examining exposure contrasts and industry subsections. The large number of lung cancer deaths we expect to observe in the cohort study includes a surplus number of deaths

of 125


sufficient for maintaining good to excellent statistical power while accounting for the additional sample size requirements needed to assess confounding and effect modification.

This study will test the hypotheses that: 1) Co in the presence of WC is a carcinogen; 2) smoking explains the excess; 3) C, a suspected lung carcinogen, is associated with the excess; 4) Co alone is associated with the excess. If only hypothesis (2) is rejected and the lung cancer excess is corrected for smoking, then the project presents an opportunity to develop a carcinogenic index for the combination of three carcinogenic co-exposures. Potential candidates for such an index range from weighted linear combination of exposures to exposure matrices that consider different lag times for each toxicant. The validation and use of the exposure index requires outcome data, which in turn requires the completion of the initial analyses. Even though it may not move from the theoretical development phase, we plan to consider the necessary theoretical and mathematical aspects of this problem and propose solutions as part of the study.

7.3 PHASE 3 STUDY - PRIMARY RESEARCH OBJECTIVES




3. To determine the relationship between level and duration of WCCo exposure and mortality from malignant lung neoplasms with analytic adjustment to the extent possible for potential co-exposures, including tobacco smoking habits, via internal adjustment with a nested case-control study or external adjustment with a Monte Carlo sensitivity analysis.


The epidemiology component of the Phase 3 study will be complemented by a comprehensive and rigorous exposure assessment component conducted by UIC (Research Objective 2 above). The primary research objectives of the exposure reconstruction component are:

1. To generate scientifically sound estimates of exposure to WCCo and other potential carcinogens for all job and/or task categories on a site-specific and time-dependent basis. Average and cumulative exposure metrics will be developed and adjusted for country, company and site variability linearly. The interaction between company and country will also be tested.

2. To create exposure classes for subsequent statistical analysis in the epidemiology component of the study.

of 125


3. To assess the robustness of the exposure reconstruction models employed with respect to uncertainties arising from data gaps and inherent variability.

7.4 PHASE 3 STUDY - PROPOSED EPIDEMIOLOGIC STUDY DESIGN

This section includes a summary of the key aspects of the proposed study design. We provide more details about our study design, including the proposed exposure reconstruction, in our detailed Phase 2 report and Phase 3 study protocol included as Appendix A to this document (Marsh et al., 2008).

7.4.1 Historical Cohort Study

Based on the conclusions of the Phase 1 and 2 studies, sufficient demographic and WH data are available to conduct an historical cohort mortality study of former and current workers with potential exposure to WCCo at each of the study sites. Multiple sites were chosen to afford better opportunities for contrasting cohort attributes, processes, work practices and exposures; multiple sites also increases the likelihood of producing definitive and informative conclusions by increasing the statistical power and the precision of the risk estimates for detecting true excess risks overall and in relation to occupational factors.

As underway for the US sites in Part 1 of the Phase 3 study, UIC and UPitt investigators will first completely ascertain the cohort of workers with potential exposure to WCCo (Research Objective 1). UIC will perform a comprehensive exposure reconstruction of manufacturing processes and will develop task and time-specific estimates of exposure (Research Objective 2). The exposure matrix will enable UPitt investigators to construct summary measures of exposure to contaminants (Research Objective 3) and enable ongoing mortality surveillance of the cohort (Research Objective 4).

The historical cohort study will provide the epidemiological platform for the proposed investigation, including a nested case-control study of lung cancer. The cohort study will focus on mortality from lung cancer and other cause of death categories (including total mortality).

7.4.2 Nested Case-Control Study

Adjustment for potential confounding by smoking will occur primarily through nested case-control studies conducted in most countries (AT, SE, US or UK) or, in DE, via external adjustment. The nested case-control study of lung cancer can be considered as a special case of the corresponding relative risk regression analysis. That is, for each case (death) due to lung cancer identified in the cohort, we will randomly select a group of matched controls from the corresponding risk set formed for the relative risk regression analysis. The non-case members of each risk set are matched on the exact event (death) age of the case, gender and year of birth (caliper-matched as tightly as possible). We will select two controls for each case.

To obtain more detailed or otherwise unavailable information on potential risk factors for the cases and controls, an attempt will be made to locate and interview a knowledgeable informant, ideally the worker himself or a surviving member of the worker’s immediate family (proxy respondent). Potential informants will be instructed to return the consent form within two weeks of the date of the letter indicating their willingness to participate. Willing informants will be contacted by a professional interviewer for a brief (20-30 minute) telephone interview concerning such items as the subject’s smoking, non-occupational and

of 125


occupational history. The case-control study questionnaire proposed for this study will be based upon the one already used in a nested case-control study of lung cancer among workers with potential exposure to man-made vitreous fibers. We will modify the existing instrument as necessary to capture specific risk factors of interest for WCCo workers.

7.4.3 Exposure Reconstruction

The basic exposure model is job class (a), task (b) and year (y) normalized exposure levels adjusted for country (X1), company (X2) and location (X3). In specific terms the fundamental linear adjustments to exposure levels with country by company interaction may be expressed as:

E( a ,b , y )=βo (a ,b , y )+βcountry X1+βcompany X2+β site X3+βcountry∗Company X 1 X2+εwhereβ ,E , β0=f (a ,b , y0) or g (a ,b ,∞)for country, company and site calculations.

At this point, we do not know whether exposure values near the start-up (y0) can be found or whether an asymptotic value (y=∞) for exposure can be determined. Between start-up based or asymptotic functions which define the exposure and coefficients, we shall choose the function that is convenient and/or tractable. That formulation will be used consistently to adjust for the time decline of exposures first with the simplifying assumption that the exposure matrix is multiplied by a scalar function of time. If this approach does not provide reasonable results, then more complicated solutions will be sought. Once this function is determined, all deterministic or empirical models will be adjusted using this relationship.

The general exposure reconstruction process and the steps that may be involved in this process have been summarized in Table 3 below. Clearly, while some of the steps are sequential, many of the process steps are concurrent. It should be noted that while much of the methodology is written in general terms, the evaluation of each facility is independent. Some of the information shown in the table will be collected on an as-needed basis and some of the operations will be carried out on an as-required basis. The general approach is the least complicated path without sacrificing rigor. In order to aggregate the cohort across plants, the determinants of exposure will be adjusted to account for the differences in operations from one facility to another using the empirically modeled correction shown above.

Table A.3: Summary of Exposure Reconstruction Method

1. Selected process information collection a. Process operation documents b. Facility layouts and plans c. Technical memos and other sources d. General annual production records

2. Exposure information collectiona. All (occupational) IH monitoring data, including area and personal exposure

sampling data b. Documentation pertaining to

i. Sampling ii. Administrative orders

of 125


iii. Environmental exposure monitoringiv. Occupational hygiene inspection reportsv. Occupational health reports filed to or by regulatory agencies (as

applicable)

3. Selected job and task identificationa. Job descriptions as written in company recordsb. As described by experienced persons (as needed)

4. Projection of results backwards through time from contemporary data (as needed)

5. Operations analysis of the production processes

6. Mathematical modeling of the information gathered above

7. Verification of models from extant data

8. Generation of task-based exposures

9. Construction of a site-specific job/exposure matrix detailing exposures for each job title (or group of job titles) as a function of specific agents and appropriate time intervals

7.5 PHASE 3 STUDY- STATISTICAL ANALYSIS PLAN

This section includes a summary of the key aspects of the proposed statistical analysis. We provide more details about our study design, including the proposed exposure reconstruction, in our detailed Phase 2 report and Phase 3 study protocol included as Annex A to this document (Marsh et al., 2008).

Because the recommended US/EU study sites are highly diverse relative to geographic location, cohort size and cohort entry period (facility start date in most cases), we propose to approach the statistical analysis in a site-specific manner, pooling data across sites only if warranted by evidence of sufficient homogeneity. The advantage of such diversity from an epidemiological standpoint is the ability to assess the consistency of our findings across the study populations. Efforts will be made to pool data when warranted, however, as this will improve the precision of the mortality risk estimates and increase the statistical power to detect epidemiologically important excess risks.

Our statistical analysis of the study data will consist of two major parts, each of which we designed to address specific objectives of the study:

Part 1 Analysis of total and cause-specific mortality patterns in relation to basic demographic and work history factors (e.g., study site, race (US sites), gender, age, calendar time, year of hire, duration of employment and the time since first employment), with focus on cancer mortality and emphasis on the implicated site of interest (lung).


of 125


Part 1 will include a descriptive analysis of externally standardized mortality ratios (SMRs), with SMRs based on both national and local standard population death rates. The descriptive phase of the analysis is useful for examining the basic structure and consistency of the overall and study factor-specific mortality risks and for determining appropriate cutpoints for categorizing continuous study variables.

Part 2 will include multi-variable analyses of internal cohort rates to facilitate the simultaneous adjustment for multiple potential confounding factors and the assessment of effect modification among two or more study factors. In particular, Part 2 will include relative risk regression modeling (based on Cox proportional hazards model) of internal cohort rates.

The modeling of internal cohort rates provides mortality comparisons within the cohort that are unbiased by the “healthy worker effect” associated with external general population comparisons. The scope of the multi-variable analysis for lung cancer mortality will be determined largely by the corresponding number of observed deaths and their factor-specific distributions. In addition, we may need to limit certain statistical analyses to subgroups with sufficiently large numbers of observed deaths (e.g., white males).

To provide the most unbiased assessment of lung cancer risk possible from the available study data, we have included in Part 2 a nested case-control study of lung cancer in the AT, SE, UK and US sites. As described above, we will make an intensive effort to obtain the most complete and accurate individual worker-level data on tobacco smoking habits for all cases (deaths) of lung cancer and corresponding groups of non-cases (controls) selected at random from the remaining cohort members. In the DE sites, we will control for potential confounding by smoking via external adjustment.

While we plan to collect smoking information to the extent possible on all study members, these data may be incomplete. By performing this adjustment in the case-control setting we will be more likely to have complete data on smoking as most of the cases and corresponding controls will fall into the later time periods when these data are more complete. The statistical analysis of the case-control data will involve relative risk regression modeling of the matched sets with adjustment for potential confounding by smoking and co-exposures to several known or suspected carcinogens.

7.6 PHASE 3 STUDY - STRENGTHS AND LIMITATIONS

The strengths of the proposed study can be summarized as follows:

1. The study represents the joint efforts of a nationally and internationally recognized occupational health research team with more than 30 years of experience designing and conducting historical cohort and case-control studies of the types proposed herein, including the largest occupational cohort, incidence and case-control study of brain cancer ever conducted.

2. The UIC and UPitt investigators share a long history of highly successful and productive collaborative research, as evidenced by the many jointly authored peer-reviewed publications cited in their respective curricula vitae.

3. The proposed historical cohort study design and nested case-control study of lung cancer, which will enable analytic adjustments for smoking and co-exposures to known or suspected carcinogens, will provide the best available estimates of total

of 125


and cause-specific mortality risks among workers, overall, and in relation to occupational WCCo exposure.

4. A particular strength of this study is the ability to characterize the working history of study members relative to WCCo exposure and co-exposures to other known or suspected carcinogens. The UIC exposure assessment team plans to use a variety of statistical and other estimation methods to assign meaningful and scientifically defensible exposure estimates to workers over time.

5. Based on our estimates, the statistical power of the proposed cohort study to detect epidemiologically important (1.5-fold or greater) overall excesses in all cancers combined and in lung cancer is essentially 1.0. The study affords a sufficiently long time period for potential WCCo exposure and a sufficiently long observation period to observe cancer outcomes in relation to WCCo exposure. Therefore, our proposed study is able to detect a true increased risk or to conclude that there is no increased risk if one is not detected. Additionally, because the number of expected cancers is so large, the power necessary to detect 1.5 fold or greater risks in larger subgroups of the cohort will also be in the good to excellent range.

6. Also based on our independent estimates, the power of the nested case-control study of lung cancer to detect epidemiologically important (1.5-fold or greater) excesses in relation to key study factors is in the excellent range (95% or greater). This focused sub-study will also allow adjustment for potential confounding by smoking in addition to co-exposures to known or suspected carcinogens.

7. The addition of Co production companies to the study would be a large advantage of the Phase 3 study over previous studies of WCCo workers. However, the inability to include subjects with Co-only exposure does not represent a fatal flaw to the main study design. The Phase 3 WCCo worker epidemiology study will be comprehensive, scientifically sound and far superior overall to the existing epidemiology studies. The limited orthogonal contrasts for Co and WC simply suggests that we may be unable to sort out completely the contribution of each exposure alone if increased lung cancer mortality risks are observed among subjects with combined exposures.

8. The historical cohort study can form the basis for ongoing mortality surveillance of the workers with potential occupational exposure to WCCo.

The limitations of the proposed study can be summarized as follows:

1. Although it appears from the Phase 1 and Phase 2 studies that much of the data necessary to adequately address the question regarding WCCo exposure and lung cancer exists, we may find that some records are unavailable. We believe, however, that we have taken every reasonable approach to resolve or work around these limitations in designing the proposed study.

2. Although the cohort study has good to excellent statistical power to detect important excesses in mortality from all cancers combined and lung cancer in the total cohort and its larger subgroups, the power will be less to detect excesses in smaller subgroups of the total study population, such as factor-specific groups within study site. The same feature will also apply to the case-control study.

of 125


3. Because a large and diverse number of agents are used or produced at the plants, it may not be possible to identify the specific etiologic agent or combination of agents if an overall excess in cancer mortality is found. This is a limitation of all epidemiology studies of this type that attempt to relate mortality outcomes to specific chemical exposures. We have, to the extent possible, designed all aspects of our study to offset or mitigate this limitation.

4. The use of external adjustment for confounding by smoking in the DE sites has some limitations. Although we will attempt to estimate the patterns of smoking among WCCo exposed workers from information available in the peer-reviewed literature, there remains the chance that these estimates are not truly representative of the smoking prevalence in the worker cohorts. The smoking rates will most likely be based only on one time period which would not reflect temporal changes in smoking habits which can occur, particularly over the long duration of an historical cohort study. We may also be limited by the extent to which detailed smoking information is available in the literature.

5. On balance, we believe that the strengths of the proposed study outweigh its weaknesses. We are also confident that our investigation will produce scientifically sound, meaningful and defensible results, and will provide a significant contribution to the body of knowledge concerning the health implications of exposure to WCCo.

7.7 QUALIFICATIONS OF THE RESEARCH TEAMS

7.7.1 University of Pittsburgh

For more than 45 years, the UPitt, Department of Biostatistics (BIOS) has been one of the leading academic centers of occupational and environmental health research in the United States. BIOS faculty have been active in the development and application of biostatistical methods to study potential health effects of workplace exposures in a variety of industrial settings.

The Center for Occupational Biostatistics and Epidemiology (COBE) was established in February 2008 as a specialty research center within the Department of Biostatistics (BIOS) in UPitt’s Graduate School of Public Health (GSPH). The mission of the COBE is to build further upon existing departmental strengths in occupational biostatistics and epidemiology, to enhance collaborative research across departments and schools at UPitt, to promote both national and international recognition of these fields of strength and to increase opportunities for external collaboration and programmatic funding.

Gary M. Marsh, Ph.D., F.A.C.E., Professor of Biostatistics, Epidemiology, and Clinical and Translational Science is the Director of the COBE; Jeanine M. Buchanich, M.Ed., Ph.D., Research Assistant Professor of Biostatistics is Deputy Director of Epidemiology and Ada O. Youk, Ph.D., Assistant Professor of Biostatistics is Deputy Director of Biostatistics. COBE research and administrative staff include master’s level biostatisticians, master’s level information science specialist/computer programmer, research specialists and technical/clerical support staff, including a graduate student researcher and a professional telephone interviewer.

The UPitt group has conducted occupational studies to investigate the long-term health effects of exposure to such agents as man-made mineral fibers, formaldehyde, acrylamide, acrylonitrile, arsenic, petrochemicals, aromatic amines and pharmaceuticals. They have also applied their expertise in occupational epidemiological research to environmental

of 125


epidemiologic studies of communities exposed to industrial pollutants or to hazardous waste site materials.

Currently, the UPitt group is conducting an historical cohort study of nearly a quarter million jet engine manufacturing workers for the Pratt & Whitney Company, a mortality surveillance system for the Owens Corning Company and an historical cohort study of pharmaceutical workers for the Eli Lilly Company. The Pratt &Whitney study is a collaborative effort with the Department of Neuro-Oncology within UPMC and UIC.

7.7.2 University of Illinois at Chicago

UIC’s Environmental and Occupational Health Sciences (EOHS) Division is housed within the School of Public Health and its goal is to protect the environment and improve the health of workers and the general public. The Illinois Occupational and Environmental Health and Safety Education and Research Center (Illinois ERC) was established in 1977 as one of the first National Institute of Occupational Safety and Health (NIOSH) Educational Resource Centers in the US.

The Division has a nationally recognized program in IH that is accredited by the Accreditation Board for Engineering and Technology. Within the division, the research carried out by the occupational epidemiology team includes mathematical modeling, engineering, industrial and environmental field studies and general occupational hygiene. In addition to the mathematical, engineering, aerosol physics and industrial hygiene expertise, the team can draw upon the medical expertise within the division. Currently, the UIC team is engaged in several projects, three of which are in collaboration with the University of Pittsburgh team. Before the formation of the UIC team, the collaboration in large scale industrial epidemiology studies between the UPitt and UIC principal investigators started in 1975 and has continued more or less unbroken ever since.

Nurtan A. Esmen, Ph.D., FAIHA, FRSH, is Emeritus Professor of EOHS; Kathleen Kennedy, M.S. is project coordinator. The UIC team also includes a senior research scientist.

7.8 PROPOSED TERMS AND CONDITIONS

7.8.1 Research Agreement

We propose to conduct this study as a sponsored research project within the University of Pittsburgh (UPitt). The Sponsor (ITIA) will work with staff at the UPitt, Office of Research to develop a mutually acceptable research agreement.

7.8.2 Organization of Subcontractors and Collaborators

The UPitt component will be directed by Gary M. Marsh, Ph.D., F.A.C.E. Dr. Marsh is Professor of Biostatistics, Epidemiology, and Clinical and Translational Science, and Director of the Center of Occupational Biostatistics and Epidemiology at the University of Pittsburgh, Graduate School of Public Health. Jeanine Buchanich, Ph.D., Research Assistant Professor of Biostatistics and Ada Youk, Ph.D., Assistant Professor of Biostatistics will serve as co-investigators on the project.

The UIC component, which will provide the exposure reconstruction for the Phase 3 study, will be directed by Nurtan A. Esmen, Ph.D., Emeritus Professor of Environmental and

of 125


Occupational Health Sciences. Kathleen Kennedy, M.S. will serve as project coordinator. UIC will serve as a subcontractor to UPitt.

DataBanque (DB), a Pittsburgh-based company under the direction of Susan Allen, will serve as the data processing subcontractor to UPitt.

As noted above, the UPitt and UIC investigators will also be responsible for coordinating the efforts of the EU investigators to ensure that common US/EU data collection and analysis protocols are followed for purposes of a pooled data analysis.

The EU country-specific investigators will be responsible for enumerating their cohorts, collecting and processing the data and conducting the vital status tracing. They will also be responsible for the conduct of the country-specific case-control studies. The EU investigators are:

Manfred Neuberger, Ph.D., Professor of Environmental Health, Center of Public Health, and head of the Department of Preventive Medicine, Institute of Environmental Health at the Medical University of Vienna will direct the study of the Reutte Austria site;

Peter Morfeld, Ph.D., Head of the Institute for Occupational Epidemiology and Risk Assessment (IERA), Evonik Services GmbH will direct the study of the three German sites;

Magnus Svartengren, Ph.D., Professor of Environmental and Occupational Medicine, Karolinska Institute in Stockholm and Unit Head, Environmental Medicine, Department of Occupational and Environmental Health at Stockholm Center for Public Health will direct the study of the three Swedish sites;

Damien McElvenny, M.Sc., Emeritus Professor of Epidemiology, University of Central Lancashire, will direct the study of the two UK sites.

The organization and coordination between UPitt and the various subcontractors is displayed in Figure A.1.

of 125


Figure A.1: Organizational Structure of Phase 3 Study

7.9 PERIOD OF PERFORMANCE

The work proposed herein for the Phase 3 study is estimated to take three years to complete. Figure A.2 shows the proposed timeline and key activities of the Phase 3 study.

7.9.1 Reports and Manuscripts

Progress reports will be issued to the sponsor annually during the course of the Phase 3 study. Investigators will also prepare draft manuscripts for publication in peer-reviewed journals. The sponsor will be granted a customary 30-60 day sponsor review period on all draft manuscripts before they are submitted for publication in the peer-reviewed scientific literature.

7.9.2 Data Confidentiality

The US and EU investigators will maintain the confidentiality of all hard copy and electronic records, assuring that information is kept in locked files (and password protected electronic media files) and that persons working with these records are made aware of their confidentiality. A Confidentiality Statement Form will be executed and kept on file for each individual working with these records. All UPitt staff will also sign the required NDI-Plus Supplemental Assurance Form to ensure the confidentiality of NDI-Plus vital status and cause of death data.7.9.3 Quality Assurance and Quality Control Features of Proposed Study

7.9.3.1 Good Epidemiology Practices Guidelines

of 125


During the preparation of this proposal and the performance of the contract, UPitt and UIC have/will adhered/adhere strictly to the Good Epidemiology Practices Guidelines (GEP) described by Cook (1991). The GEP provides guidance regarding protocol development, reporting, quality assurance of data maintenance, and documentation of analytic procedures.

7.9.3.2 Metrics Used to Ensure Data Quality and to Determine Progress

Standard UPitt data checks will be used throughout the proposed study to ensure the integrity of all study data. This will include a formal check of the completeness and accuracy of the cohort enumeration and vital status tracing, as well as comprehensive audits of all primary study data coding and keying. For example, the coding and key-entering of all primary study data will be 100% verified by a second data coding clerk. UPitt staff will strive to achieve cohort completeness, follow-up and cause of death ascertainment rates of at least 95%. The UPitt guidelines will also be made part of the common data collection and analysis protocol followed by the EU investigators.

of 125


Figure A.2: Proposed Timeline for Phase 3 Study

of 125

Damien McElvenny, 01/09/26,

Need to Insert References?


ANNEX A

Cohort Mortality Study of Workers Exposed to Tungsten Carbide with a Cobalt Binder

Phase 2 Study Final Report &Phase 3 Study Proposal

submitted by

The University of PittsburghDepartment of Biostatistics

Center for Occupational Biostatistics & Epidemiology

and

The University of Illinois at ChicagoEnvironmental and Occupational Health Sciences

Division

November 7, 2008

of 125


Table of ContentsPage

List of Tables..........................................................................................................................xxxiii

List of Figures........................................................................................................................xxxiii

Acknowledgments..................................................................................................................xxxiv

Executive Summary................................................................................................................xxxv

1. Phase 2 Study - Background....................................................................................................1

2. Phase 2 Study – Results............................................................................................................2

2.1. Survey Development & Administration..............................................................................2

2.2. Site Visit Summaries...........................................................................................................2

2.3. Site Eligibility......................................................................................................................5

2.4. Site Selection for Phase 3 Study........................................................................................12

2.5. Country-Specific Issues Identified in Phase 2...................................................................12

2.6. Phase 2 Study Conclusions................................................................................................13

3. Phase 3 Study - Primary Research Objectives.....................................................................15

4. Phase 3 Study - Proposed Epidemiologic Study Design......................................................16

4.1. Overview............................................................................................................................16

4.2. Organization of Subcontractors and Collaborators............................................................16

4.3. Historical Cohort Study .....................................................................................................17

4.4. Basic Structure and Characteristics of the Cohort.............................................................18

4.5. Available Study Data.........................................................................................................19

5. Phase 3 Study - Proposed Exposure Reconstruction Study Design...................................21

5.1. Understanding of the Problem...........................................................................................21

5.2. Primary Research Objectives of the Exposure Assessment Component...........................21

5.3. Methodologic Approach....................................................................................................22

5.3.1. Basic Approach...........................................................................................................22

5.3.2. Collection of Information...........................................................................................25

5.3.2.1. Preliminary information......................................................................................25

5.3.2.2. Standardization of job and department titles.......................................................25

5.3.2.3. Collection of historical exposure data.................................................................26

5.3.2.4. Abstraction and analysis of historical exposure data..........................................26

5.3.2.5. Collection of historical job and task information................................................27

5.3.3. Preliminary Exposure Analysis..................................................................................28

5.3.4. Identification of Confounding Exposures...................................................................29

5.3.5. Identification of Missing Exposures and Information Gaps.......................................29

of cxxv


5.4. Modeling............................................................................................................................31

5.4.1. Exposure Metrics........................................................................................................31

5.4.2. Mathematical and Physical Models............................................................................31

5.4.3. Statistical Models........................................................................................................33

5.4.3.1. Monte Carlo modeling of interpolated, extrapolated and missing data..............33

5.4.3.2. Inversion with proportional extrapolation..........................................................33

5.5. Exposure Data Analysis.....................................................................................................34

5.5.1. Existing Usable Exposure Data..................................................................................34

5.5.2. Determination of Classification Limits.......................................................................34

5.5.3. Calculation of Job Specific Exposures.......................................................................35

5.5.4. Assigning Job Specific Exposures and Merging Exposures.......................................35

6. Phase 3 Epidemiology Study - Proposed Work Plan and Timetable.................................36

6.1. Task 1A - Cohort Enumeration..........................................................................................36

6.2. Task 1B – Plant Data Collection........................................................................................36

6.3. Task 2A - Data Processing, Validation and Management.................................................37

6.3.1. United States...............................................................................................................37

6.3.1.1. Electronic records................................................................................................37

6.3.1.2. Hard copy records...............................................................................................37

6.3.2. Austria.........................................................................................................................37

6.3.3. Germany......................................................................................................................38

6.3.4. Sweden........................................................................................................................38

6.3.5. United Kingdom.........................................................................................................38

6.3.6. All Study Sites............................................................................................................39

6.3.6.1. Data Integration..................................................................................................39

6.3.6.2. Data Storage........................................................................................................39

6.3.6.3. Data Backup........................................................................................................40

6.3.6.4. Documentation.....................................................................................................40

6.4. Task 2B – Exposure Modeling and Job Dictionary...........................................................40

6.5. Task 3 - Cohort Follow-Up................................................................................................41

6.5.1. United States...............................................................................................................41

6.5.2. Austria.........................................................................................................................42

6.5.3. Germany......................................................................................................................42

6.5.4. Sweden........................................................................................................................42

6.5.5. United Kingdom.........................................................................................................43

6.6. Task 4 - Integration of Exposure Assessment Data...........................................................43

of cxxv


6.7. Task 5 - Statistical Analyses..............................................................................................43

6.7.1. Overview of General Approach..................................................................................43

6.7.2. Descriptive Analysis of General Mortality Patterns...................................................45

6.7.2.1. Construction of Basic Person-Year Arrays.........................................................45

6.7.2.2. Analysis by Worker Type.....................................................................................45

6.7.3. Procedures to Account for Unknown Race.................................................................46

6.7.4. Calculation of Internal Cohort Rates, Expected Numbers of Deaths and SMRs.......47

6.7.5. Analysis of Mortality in Relation to Exposure to WCCo and Other Substances.......48

6.7.6. Descriptive Analysis...................................................................................................50

6.7.7. Relative Risk Regression Modeling...........................................................................50

6.8. Controlling for Confounding by Smoking.........................................................................51

6.8.1. External Adjustment for Confounding by Smoking...................................................52

6.8.2. Internal Adjustment for Confounding by Smoking....................................................52

6.8.2.1. Matching cases to controls..................................................................................52

6.8.2.2. Obtaining data on cases and controls.................................................................53

6.8.2.3. Statistical analysis of nested case-control study ..................................................53

6.9. Statistical Power Characteristics....................................................................................54

6.9.1. Cohort study............................................................................................................54

6.9.2. Case-control study..................................................................................................54

6.10. Strengths and Weaknesses of the Proposed Study...........................................................55

6.11. Qualifications of the Research Teams.............................................................................57

6.11.1. University of Pittsburgh............................................................................................57

6.11.2. University of Illinois at Chicago...............................................................................57

7. Phase 3 Study - Administrative Details.................................................................................59

7.1. Project Structure................................................................................................................59

7.2. Period of Performance.......................................................................................................59

7.3. Investigators.......................................................................................................................59

7.4. Progress Reports................................................................................................................59

7.5. Data Confidentiality...........................................................................................................59

7.6. Quality Assurance and Quality Control Features of Proposed Study................................60

7.6.1. Good Epidemiology Practices Guidelines..................................................................60

7.6.2. Metrics Used to Ensure Data Quality and to Determine Progress..............................60

8. Project Budget Justification...................................................................................................61

References Cited...........................................................................................................................1

Appendix A Phase 2 Survey Instrument ...................................................................................65

of cxxv


Appendix B ICDA Defined Cause of Death Categories ..........................................................68

Appendix C Biosketches of Investigators .................................................................................72

of cxxv


4. List of Tables

Table 1. Phase 2 Site Visit Examination of Work History and Industrial Hygiene Records.........3

Table 2. WCCo Sites Eligible for Inclusion in Phase 3 Study.......................................................9

Table 3. WCCo Sites Ineligible for Inclusion in Phase 3 Study...................................................10

Table 4. Summary of Exposure Reconstruction Method..............................................................18

5. List of Figures

Figure 1. Determination of WCCo Phase 3 Facilities: Hierarchical, Criteria-Based Selection.....7

Figure 2. Organizational Structure of Phase 3 Study....................................................................24

of cxxv


Acknowledgments

The International Tungsten Industry Association sponsored this research. The researchers would like to acknowledge the cooperation and assistance of the representatives and consultants of ITIA and the Alldyne ATI, ATI Stellram, Ceratizit, HB Carbide, Kennametal, Plansee and Sandvik personnel at the sites we visited. The research proposal and the conduct of Phase 2 were approved by the Institutional Review Boards (IRB) of the University of Pittsburgh and the University of Illinois at Chicago.

of cxxv


Executive Summary

Background


Phase 1 of this investigation was a feasibility study conducted by BBL, Inc. in 2006 to determine the availability and accessibility of company records needed for the main epidemiology study that comprises Phase 3 of the investigation. The Phase 1 feasibility study was extended and enhanced in Phase 2 by the University of Pittsburgh (UPitt) and the University of Illinois at Chicago (UIC) under a research contract from ITIA. Phase 2, conducted from October 2007 through October 2008, allowed researchers: (1) to communicate with plants directly to help clarify specific study needs; (2) to establish communication avenues with key plant personnel and enabled those personnel to prepare their facilities for inclusion in the Phase 3 study; (3) to apply different or additional feasibility criteria to the data obtained in Phase 1 and (4) to establish parameters that created the framework within which the Phase 3 study will proceed.

Three sites of the 60 from the Phase 1 study have been closed or operations have been consolidated (Sandvik Seco Carboloy Warren MI – closed; Kennametal Lorch Germany – closed; Sandvik Hard Materials Epinouze France – combined with Grenoble France site). We added one site not included in the EpiFS survey (Ceratizit Luxembourg) for a total of 58 sites evaluated.

Phase 2- Methods and Results

UPitt and UIC developed and applied several criteria for a candidate site to be included in the Phase 3 study, including a minimum size of 100 or more employees historically for United States (US) sites and 500 or more employees historically for European (EU) sites. Sites also had to have been producing WCCo or WC products since at least 1980 to allow an adequate latency period for disease development. Additionally, detailed work history (WH) information had to be available for all employees who ever worked at the facility. The last consideration, for non-US sites, was whether vital status tracing was possible within that country. This information was gathered via the telephone survey and/or during a site visit.

The study questionnaire used in the telephone survey had separate epidemiology and industrial hygiene components and was administered by a professional interviewer on the UPitt staff. Surveys were completed in April 2008. Following the surveys, UPitt and UIC investigators conducted four US and two EU site visit trips that included a total of 18 sites

of cxxv


during the Phase 2 study. The sites visited were representative of the candidate sites with respect to country, company and process.Overall, we determined in the Phase 2 study that 18 sites met the criteria for inclusion in the Phase 3 study and 40 sites did not meet the criteria. The eligible sites are: Kennametal, Orwell, OH; Kennametal, Johnson City, TN; Kennametal, Bedford, PA; Sandvik, West Branch, MI; Kennametal Roanoke Rapids, Weldon, NC; Kennametal, Fallon, NV; Kennametal, Henderson, NC; ATI Alldyne Powder Technologies, Huntsville, AL; ATI Metalworking, LaVergne, TN; Sandvik Coromant, Gimo, Sweden; Sandvik Vastberga, Stockholm, Sweden; Sandvik Seco Tools AB, Fagersta, Sweden; Sandvik Coromant, Feathestone, UK; Sandvik Powder Production, Coventry, UK; CERATIZIT Austria GmbH, Reutte, Austria; Kennametal, Mistelgau, Germany; Kennametal, Ebermannstadt, Germany; Kennametal, Essen, Germany.

Three additional US sites may be included after more information can be found about the whereabouts of archived records. The Kennametal facilities in Chestnut Ridge, PA, Irwin, PA and Rogers, AR sent some files off-site for storage and the exact location of these records has not yet been determined. We are also still working to finalize the eligibility of three French sites and have identified two French physicians who are interested in collaborating with us. We are still determining whether the requisite records are available at the sites and if these physicians will be able to gain access to the necessary vital status tracing services. For purposes of this report, these US and French sites will be considered ineligible.

Phase 2- Conclusions

The information gathered in Phase 2 showed that a scientifically rigorous and comprehensive epidemiology study of WCCo workers is feasible and should include workers from at least 18 manufacturing sites in the US and EU. The Phase 3 study, which will represent multiple companies, countries and manufacturing processes, will be larger, more robust and more definitive than any WCCo epidemiology study done to date. Additionally, Phase 2 revealed that five main exposure agents should be evaluated and compared for potential adverse health effects in the main epidemiology study: tungsten (W), WC, WCCo, carbon black (C) and cobalt (Co).

The efficiency and meaningfulness of these comparisons will be improved in the main study if the levels of exposures to the agents (termed “contrasts”) being compared are not highly collinear, that is, at least a portion of one exposure is uncorrelated with the other exposure. These uncorrelated components are termed “orthogonal contrasts.” While adequate orthogonal contrasts are possible between the first four agents, the uncorrelated portion of the WC and Co contrast is limited by small sample sizes. A straightforward and very efficient solution to this issue is to include subjects in the study from facilities at which there is Co exposure but no W or WC exposure. We believe that the supplier(s) of Co powder to the facilities we visited would be good candidates for inclusion in the main study. The limited orthogonal contrasts for Co and WC do not represent a fatal study flaw, but suggest that we may be unable to sort out completely the contribution of each exposure alone if increased lung cancer mortality risks are observed among subjects with combined exposures. The inclusion of subjects with Co only exposure will increase the sample size of the available orthogonal contrasts for Co and WC, which will increase the statistical efficiency or power of the main study.

of cxxv


Phase 3 Study - Primary Research Objectives






The epidemiology component of the Phase 3 study will be complemented by a comprehensive and rigorous exposure assessment component conducted by UIC (Research Objective 2 above). The primary research objectives of the exposure reconstruction component are:

1. To generate scientifically sound estimates of exposure to WCCo and other potential carcinogens for all job and/or task categories on a site-specific and time-dependent basis. Average and cumulative exposure metrics will be developed and adjusted for country, company and site variability linearly. The interaction between company and country will also be tested.

2. To create exposure classes for subsequent statistical analysis in the epidemiology component of the study.

3. To assess the robustness of the exposure reconstruction models employed with respect to uncertainties arising from data gaps and inherent variability.

of cxxv


Phase 3 Study - Proposed Epidemiologic Study Design


UIC and UPitt investigators will first completely ascertain the cohort of workers with potential exposure to WCCo (Research Objective 1). 1UIC will perform a comprehensive exposure reconstruction of manufacturing processes and will develop task and time-specific estimates of exposure (Research Objective 2). The exposure matrix will enable UPitt investigators to construct summary measures of exposure to contaminants (Research Objective 3) and enable ongoing mortality surveillance of the cohort (Research Objective 4).

1The historical cohort study will provide the epidemiological platform for the proposed investigation, including a nested case-control study of lung cancer. The cohort study will focus on mortality from lung cancer and other cause of death categories (including total mortality). Adjustment for potential confounding by smoking will occur primarily through nested case-control studies conducted in most countries (Austria (AT), Sweden (SE), US or the United Kingdom (UK)) or, in Germany (DE), via external adjustment.

Phase 3 Study- Statistical Analysis Plan

Because the recommended study sites are highly diverse relative to geographic location, cohort size and cohort entry period (facility start date in most cases), we propose to approach the statistical analysis in a site-specific manner, pooling data across sites only if warranted by evidence of sufficient homogeneity. The advantage of such diversity from an epidemiological standpoint is the ability to assess the consistency of our findings across the study populations. Efforts will be made to pool data when warranted, however, as this will improve the precision of the mortality risk estimates and increase the statistical power to detect epidemiologically important excess risks.




of cxxv



Part 2 will include multi-variate analyses of internal cohort rates to facilitate the simultaneous adjustment for multiple potential confounding factors and the assessment of effect modification among two or more study factors. In particular, Part 2 will include relative risk regression modeling (based on Cox proportional hazards model) of internal cohort rates.


To provide the most unbiased assessment of lung cancer risk possible from the available study data, we have included in Part 2 a nested case-control study of lung cancer in the AT, SE, UK and US sites. In this study, we will make an intensive effort to obtain the most complete and accurate individual worker-level data on tobacco smoking habits for all cases (deaths) of lung cancer and corresponding groups of non-cases (controls) selected from the remaining cohort members. In the DE sites, we will control for potential confounding by smoking via external adjustment.

While we plan to collect smoking information to the extent possible on all study members, these data may be incomplete. By performing this adjustment in the case-control setting we will be more likely to have complete data on smoking as most of the cases and corresponding controls will fall into the later time periods when these data are more complete. The statistical analysis of the case-control data will involve relative risk regression modeling of the matched sets with adjustment for potential confounding by smoking and co-exposures to several known or suspected carcinogens.

Phase 3 Study- Organization of Subcontractors and Collaborators

The UPitt component will be directed by Gary Marsh, Ph.D. and Jeanine Buchanich, Ph.D. UIC, under the direction of Nurtan Esmen, Ph.D. and Steven Lacey, Ph.D., will serve as a subcontractor to UPitt for the exposure reconstruction; DataBanque (DB), a Pittsburgh-based company under the direction of Susan Allen, will serve as the data processing subcontractor.

The EU site investigators will be responsible for enumerating the country-specific cohorts, collecting and processing the data and conducting the vital status tracing. Manfred Neuberger, Ph.D., Professor of Environmental Health, Center of Public Health, and head of the Department of Preventive Medicine, Institute of Environmental Health at the Medical University of Vienna will oversee the Reutte Austria site; Peter Morfeld, Ph.D., Head of the Institute for Occupational Epidemiology and Risk Assessment (IERA), Evonik Services GmbH will oversee the three German sites; Magnus Svartengren, Ph.D., Professor of

of cxxv


Environmental and Occupational Medicine, Karolinska Institute in Stockholm and Unit Head, Environmental Medicine, Department of Occupational and Environmental Health at Stockholm Center for Public Health will oversee the three Swedish sites and Thomas Sorahan, Ph.D., Professor of Occupational Epidemiology, The University of Birmingham, Institute of Occupational Health will coordinate the two UK sites. The investigators from the EU countries will report directly to the UPitt and UIC investigators. The identification of an Austrian collaborator occurred relatively recently and, as such, some details regarding the AT cohort enumeration and processing are still under development.

Phase 3- Projected Timeline and Budget

The proposed Phase 3 Study is anticipated to take approximately four to five years to complete. Part 1 of the Phase 3 study is anticipated to begin in November 2008 or following receipt of pending funds from the State of Pennsylvania Department of Health (PADOH). Subsequent parts of Phase 3 will be started as additional funding becomes available.

The PADOH has stated their intention to provide UPitt with $670,000 in direct costs for Part 1 of Phase 3 of the full epidemiology study and they have received a detailed application from UPitt for these funds. The initiation of Phase 3 Part 1 is dependent upon the receipt of the funds. Subsequent parts of Phase 3 are contingent upon procuring additional funding sources and, at that time, the specific tasks and detailed budgets will be provided to the sponsor.

We estimate that the total costs of the Phase 3 study will be in the $4-5 million range (direct costs), depending on the total number of manufacturing sites included.

of cxxv


1. Phase 2 Study - Background


Phase 1 of this investigation was a feasibility study conducted by BBL, Inc. in 2006 to determine the availability and accessibility of company records needed for the main epidemiology study that comprises Phase 3 of the investigation. The Phase 1 feasibility study was extended and enhanced in Phase 2 by the University of Pittsburgh (UPitt) and the University of Illinois at Chicago (UIC) under a research contract from ITIA.

Phase 2, conducted from October 2007 through October 2008, allowed researchers to communicate with plants directly to help clarify specific study needs; it also allowed the investigators to establish communication avenues with key plant personnel and enabled those personnel to prepare their facilities for inclusion in the Phase 3 study. Phase 2 also allowed researchers to apply different or additional feasibility criteria, based upon critical study components in an hierarchical structure, to the Epidemiological Investigation Feasibility Study (EpiFS) data (Schell et al., 2006) obtained in Phase 1 and to establish parameters that created the framework within which the Phase 3 study will proceed.

of 125


2. Phase 2 Study – Results

2.1. Survey Development & Administration

Due to some concerns that the earlier EpiFS may not have completely captured information on the extent to which powder production or use occurred in the potential study sites, UPitt independently contacted and surveyed the sites included in the original EpiFS survey to obtain information on powder production and related factors (dates of operation, type of powder, workforce size, co-exposures, etc.). We needed to be certain that we were not excluding any potentially informative sites when we developed the selection algorithm to identify the subset of plants for further evaluation in Phase 2.

The study questionnaire, shown in Appendix A, had separate epidemiology and industrial hygiene (IH) components and was administered by a professional interviewer on the UPitt staff. In April 2008, we concluded the survey phase after completing a survey for each of the sites. Three sites of the 60 from the Phase 1 study have been closed or operations have been consolidated (Sandvik Seco Carboloy, Warren, MI – closed; Kennametal, Lorch, Germany – closed; Sandvik Hard Materials, Epinouze, France – combined with Grenoble, France site). We added one site not included in the EpiFS survey (Ceratizit, Luxembourg) for a total of 58 sites evaluated.

2.2. Site Visit Summaries

The UPitt and UIC investigators conducted four United States (US) and two European (EU) site visit trips which included a total of 18 sites during the Phase 2 study. Table 1 below summarizes the information gathered during the visits regarding the availability and completeness of work history (WH) and IH data.

of 125


Table 4. Phase 2 Site Visit Examination of Work History and Industrial Hygiene Records

Trip Date Plant Location WH Records IH Records

US1 Feb. 27-29 2008

Kennametal USA Irwin (Carbidie)

Irwin, PA

On site; older hard copy, 2003+ electronic; WH details in active files, not in retiree files.

On site; older hard copy, 2000+ electronic.

US1 Feb. 27-29 2008

Kennametal USA Chestnut Ridge

Latrobe, PA

Hard copy; active on site with 9-10 yr WH gaps; inactive file availability unknown (at HQ?).

2000+ on site; earlier record availability unknown (at HQ?).

US1 Feb. 27-29 2008

Kennametal USA Bedford Bedford, PA

Last 2 yrs of files on site; inactive to Iron Mountain; active are electronic; WH detail included.

2005+ electronic; some earlier hard copies on site; older may be at HQs or Iron Mountain.

US1 Feb. 27-29 2008

Kennametal USA Orwell Orwell, OH

1999+ electronic; active and last 2 yr of terminations (terms) on site; WH detail included; files to HQ until ~ 1998 then to Iron Mountain.

Electronic ~ 2004+; current back to 1980s hard copy; fewer data for earlier years.

US2 April 1-3 2008

Kennametal USA Placentia Placentia, CA

1985+ electronic; 1986-2006 active, term hard copies on site; terms not available 1978-1986; WHs no job titles (JTs) only departments.

2004+ electronic; prior to 1991 no measurements.

US2 April 1-3 2008

Kennametal Fallon NV Fallon, NV

On site active and term hard copies; 1998+ electronic; WH details available.

Last 10 yrs electronic; 1980's/1990s records hard copy in Pittsburgh currently, to be returned in near future.

US3 April 28-30 2008

Alldyne ATI Powder Technologies

Huntsville, AL

All active and inactive hard copy files in Nashville; 1995+ records electronic.

2004+ electronic; 2001-04 records in Knoxville; prior to 2001 at TMC Huntsville.

of 125



US3 April 28-30 2008

Kennametal USA Henderson

Henderson, NC

Active and term hard copy records on site; some electronic, WH details available.

Sampling from 1980's; last 2-3 yrs electronic, older hard copy; on site.

US4 May 21-22 2008

Sandvik West Branch

West Branch, MI

Inactive files on site, contain WH details; active stored off site by Accenture, being moved to Fairlawn NJ, could not view WH details.

1998+ electronic; mid-1980's to 1998 hard copy on site.

EU1 June 6-18 2008

Sandvik Seco Tools Fagersta, SE

Records in Swedish; 1972+ active and term hard copies on site, WH details available; prior to 1972 may be at original company Fagersta AB; no electronic files.

2002+ electronic; older files from mid-1970's hard copy.

EU1 June 6-18 2008

Sandvik Coromant Gimo, SE

Records in Swedish; active and term hard copies on site; 2006+ electronic; WH details available.

Electronic and hard copy format; did not view on site, will folllow up in Phase 3 to verify format, extent and availability of records.

EU1 June 6-18 2008

Kennametal Germany Ebermannstadt

Ebermannstadt, DE

Records in German; 1991+ electronic; active and term hard copy on site; WH detail available.

Electronic database through workers' comp service; additional hard copy info on site; documentation back to 1981, earlier in archive.

EU1 June 6-18 2008

Kennametal Germany Mistelgau

Mistelgau, DE

Records in German; 1987+ electronic; active and term hard copy on site; WH detail available.

Electronic database through workers' comp service; on site hard copy records with additional details from 1983 to present.

of 125



EU1 June 6-18 2008

Sandvik Coromant UK

Featherstone, UK

Records may be on site or stored at Halesowen HQ; files not viewed; will need to verify format and availability in Phase 3.

2002+ electronic; earlier not thought to be available.

EU1 June 6-18 2008

Sandvik Powder Production UK

Coventry, UK

1970+ active and term on site, earlier terms may have been destroyed; some may be at Halesowen; ~1998+ electronic.

1999+ records, electronic and hard copy on site; some data early 1990's;

EU2 July 13-18 2008

CERATIZIT Austria GsmbH

Reutte, AT

Records in German; active and term on site; 1998+ electronic; prior to 1998 (back to 1950's or 1920's) scanned and entered into electronic system; WH detail available.

Hard copy 1992 to present on site; OSBS (federal service) can order measurements, at agency and also on site.

EU2 July 13-18 2008

Sandvik Hard Materials SAS Grenoble, FR

Records in French; active hard copy on site, detail WH available; term hard copy may be at Epinouze or destroyed; no electronic records.

Hard copy late 1960's to present on site; measurements in personal medical documents, some may exist outside of these.

EU2 July 13-18 2008

CERATIZIT Luxembourg Mamer, LU

Records in French; active hard copy on site, do not include WH details; records older than 12 yrs destroyed; 2000+ electronic.

Records on site from 1990's to present; hard copy and electronic.

2.3. Site Eligibility

UPitt and UIC developed specific criteria to evaluate the eligibility of the sites for inclusion in the Phase 3 study. We implemented a minimum size of 100 or more employees historically for US sites and 500 or more employees historically for EU sites. Including sites with fewer employees would dramatically reduce the efficiency of the teams to collect and process the site-specific information. Sites also had to have been producing WCCo or

of 125


WC products since at least 1980 to allow an adequate latency period for disease development. Additionally, detailed WH information had to be available for all employees who ever worked at the facility. The last consideration, for non-US sites, was whether vital status tracing was possible within that country. Figure 1 depicts the way in which using those three selection parameters reduced the number of eligible facilities to nine in the US and nine in the EU.

of 125


Figure 3. Determination of WCCo Phase 3 Facilities: Hierarchical, Criteria-Based Selection

Tables 2 and 3 provide specific information on each site regarding the selection criteria for inclusion in the Phase 3 study, including the total number of employees, the years of operation and the availability of detailed WHs. This information was gathered via the telephone survey and/or during a site visit. Table 2 shows the 18 sites that met the criteria

of 125


and Table 3 shows the 40 sites that did not meet the criteria; also shown in Tables 2 and 3 are whether the sites were deemed eligible in the Phase 1 EpiFS study (Selected (Sel) by BBL).

of 125


Table 5. WCCo Sites Eligible for Inclusion in Phase 3 Study

Eligible Facilities(Start - 1980 or earlier) Country Location Sel BBL

Total Employees

Year Plant Opened

WH Records Available

Kennametal United States Orwell United States Orwell, OH Yes More than 1,000 1960 YesKennametal United States Johnson City a. United States Johnson City, TN Yes 501 to 1,000 1966 YesKennametal USA Bedford United States Bedford, PA Yes 251 to 500 1952 YesSandvik West Branch MI United States West Branch, MI Yes 251 to 500 1975 YesKennametal Roanoke Rapids a. United States Weldon, NC Yes 501 to 1,000 1980 YesKennametal USA Fallon United States Fallon, NV No 101 to 250 1960 YesKennametal USA Henderson United States Henderson, NC No 101 to 250 1980 YesATI Alldyne Powder Technologies United States Huntsville, AL Yes 501 to 1,000 1960 YesATI Metalworking a. United States LaVergne, TN No 501 to 1,000 1976 YesSandvik Coromant Sweden Sweden Gimo Yes More than 1,000 1951 YesSandvik Sweden Vastberga a. Sweden Stockholm Yes More than 1,000 1953 Yes

Sandvik Seco Tools AB Sweden Fagersta Yes More than 1,000 1931Some (31-71

missing)Sandvik Coromant UK UK Featherstone Yes 501 to 1,000 1966 Not sureSandvik Powder Production UK UK Coventry Yes 501 to 1,000 1950 YesCERATIZIT Austria GmbH Austria Reutte Yes More than 1,000 1929 YesKennametal Germany Mistelgau Germany Mistelgau (Bavaria) Yes 501 to 1,000 1971 YesKennametal Germany Ebermannstadt Germany Ebermannstadt (Bavaria) Yes 501 to 1,000 1971 Yes

Kennametal Essen a. Germany Essen (North Rhine) Yes More than 1,000 1926Some (26-77

missing)

a. Record availability was not ascertained with a site visit

of 125


Table 6. WCCo Sites Ineligible for Inclusion in Phase 3 Study

Ineligible Facilities Reason Country Location Sel BBLTotal

EmployeesYear Plant

OpenedWH Records

Available

Sandvik Australia Size Australia Newcastle No 51 to 100 1960 s No

Sandvik Brazil Records Brazil Sao Paulo Yes 501 to 1,000 1963 No

Kennametal Canada Victoria Records Canada Victoria No 251 to 500 1941 No

Sandvik China Size/Age China Lang Fang No 251 to 500 1995 Yes

Seco - Pramet Tools, s.r.o. RecordsCzech

Republic Sumperk Yes More than 1,000 1951 No

Kennametal France Bordeaux Size/Records France Bordeaux No 101 to 250 1953 Don't knowKennametal France Andrezieux Records France Andrezieux Yes More than 1,000 1975 No

Sandvik SAS (France) Tracing France Fondettes Yes 501 to 1,000 1957 Yes

Sandvik Grenoble/Epinouze Records France Grenoble Yes 501 to 1,000 1950s Yes

CERATIZIT Horb GmbH Age/Records Germany Horb (Baden) Yes More than 1,000 1991 NoKennametal Germany Nabburg Size/Records Germany Nabburg No 251 to 500 1973 No

Kennametal Vohenstrauss Age GermanyVohenstrauss

(Bavaria) No 501 to 1,000 1985 YesSandvik Gunther & Co. GmbH Frankfurt Size/Age Germany Frankfurt No 51 to 100 1985 Yes

Sandvik WALTER AG Records Germany Tübingen (Baden) Yes More than 1,000 1924 No

Kennametal India Bangalore Size India Bangalore No 101 to 250 1967 Yes

Sandvik Asia Ltd Records India Pune Yes 501 to 1,000 1960 No

of 125



EmployeesYear Plant

OpenedWH Records

Available

Kennametal Israel Shlomi Size Israel Shlomi No 251 to 500 1950s/1988 No

Kennametal Italy Milano Size/Records Italy Milano No 101 to 250 1956 No

Sandvik K.K., Semine plant Size Japan Miyagi No 101 to 250 1976 Yes

CERATIZIT Luxembourg Records Luxembourg Mamer NA More than 1,000 1971 Some

Sandvik Mexico Size Mexico Tlalneplantla No 101 to 250 1962 YesKennametal Netherlands Arnheim Records Netherlands Arnheim Yes 501 to 1,000 1947 NoKennametal Netherlands Hardenberg Size/Records Netherlands Hardenberg No 251 to 500 1956 No

Sandvik Baildonit Poland Size/Age Poland Baildonit No 101 to 250 1994 Yes

Sandvik Russia Moscow Tracing Russia Moscow No More than 1,000 1936 Yes

Kennametal Spain Vitoria Size/Records Spain Vitoria-Alava No 251 to 500 1970 No

Sandvik Espanola Size Spain Caldes de Montbui No 101 to 250 1953 Yes

ATI Stellram SA Size Switzerland Gland No 101 to 250 1974 Yes

Sandvik Taiwan Size Taiwan Chung Li No 101 to 250 1970 YesATI Alldyne Powder Technologies Age United States Gurley, AL Yes 251 to 500 1981 Yes

HB Carbide Age United States Lewiston, MI No 101 to 250 1982 NoKennametal United States Asheboro Age United States Asheboro, NC No 101 to 250 1982 YesKennametal United States Chestnut Ridge Records United States Latrobe, PA Yes 501 to 1,000 1958 No

Kennametal United States Records United States Irwin, PA Yes 251 to 500 1976 No of 125



EmployeesYear Plant

OpenedWH Records

AvailableIrwinKennametal United States Placentia Records United States Placentia, CA Yes 251 to 500 1944 NoKennametal United States Rancho Cucamonga Records United States Monrovia, CA No 101 to 250 1947/1996 NoKennametal United States Rogers AR Records United States Rogers, AR Yes More than 1,000 1954 NoKennametal United States Traverse City Size United States Traverse City, MI No 51 to 100 1972 Yes

Sandvik Coromant Stafford Age United States Houston, TX No 101 to 250 2000 YesSandvik Hard Materials Valenite Records United States Westminster, SC Yes 501 to 1,000 1978 No

of 125


2.4. Site Selection for Phase 3 Study

As shown in Table 2, 9 US and 9 EU (Austria (AT)=1, Germany (DE)=3, Sweden (SE)=3 and UK=2) sites met all necessary criteria for inclusion in the Phase 3 study. With the inclusion of these 18 sites, the Phase 3 study will represent multiple companies, countries and manufacturing processes, and will be larger, more robust and more definitive than any WCCo epidemiology study done to date.

As noted in Table 3, many sites were deemed ineligible due to small facility sizes (n<100 for US sites or n<500 for non-US sites); record availability was also a limiting factor for many of the sites. The one site in Russia was deemed ineligible due to an inability to get detailed vital status and cause of death information.

Three additional US sites may be included after more information can be found about the whereabouts of archived records. The Kennametal facilities in Chestnut Ridge, PA, Irwin, PA and Rogers, AR sent some files off-site for storage and the exact location of these records has not yet been determined. For purposes of this report, these three sites will be considered ineligible.

We are still working to finalize the eligibility of three French sites currently listed as ineligible in Table 3 (Grenoble/Epinouze and Fondettes). We have identified two French physicians who are interested in collaborating with us on the study but we are still determining whether the requisite records are available at the sites and if these physicians will be able to gain access to the necessary vital status tracing services. We will be unable to include the French sites unless both aspects can be finalized in accordance with our study protocol although we remain hopeful about the possibility of including French sites. For purposes of this report, though, the French sites will be considered ineligible.

2.5 Country-Specific Issues Identified in Phase 2

One of the primary purposes of the Phase 2 study for sites located in non-US countries was to determine the availability and accessibility not only of the necessary WH and IH data but also the strengths and limitations of conducting vital status tracing in these countries.

As shown in Figure 1 and Table 3, 16 sites from 14 countries were eliminated because they did not meet the minimum size requirement. The UPitt and UIC investigators felt that it was necessary to have at least 500 employees ever employed in non-US sites because of the added difficulty of enlisting a country-specific co-investigator, developing a country-specific protocol for cohort enumeration and data collection and performing vital status. Two German sites did not start production until the mid-1980’s (Kennametal Vohenstrauss) and early 1990’s (CERATIZIT Horb) and were thus not eligible for inclusion. Nine other non-US sites destroyed WH and/or personnel records so did not have the necessary records available for study. The one site located in Russia (Sandvik Moscow) was excluded due to the difficulties in obtaining complete and accurate vital status information for the workers.

Facilities in four EU countries (AT, DE, SE and UK) met the entrance criteria for the Phase 3 study. UPitt and UIC have determined that the most time and cost-effective way in which

of 125


to collect the required data from these sites is to enlist the cooperation of country-specific co-investigators who will oversee the work for the sites in their respective countries. The information regarding these investigators and any remaining country-specific issues identified in the Phase 2 study are described in our discussion of the Phase 3 study.

2.6. Phase 2 Study Conclusions

The information gathered in Phase 2 showed that a scientifically rigorous and comprehensive epidemiology study of WCCo workers is feasible and should include workers from at least 18 manufacturing sites in the US and EU. The Phase 3 study, which will represent multiple companies, countries and manufacturing processes, will be larger, more robust and more definitive than any WCCo epidemiology study done to date.

The interviews, site visits and discussions with company officials to date indicate there are five main exposure agents that should be evaluated and compared for potential adverse health effects in the main epidemiology study: tungsten (W), WC, WCCo, carbon black (C) and cobalt (Co). The efficiency and meaningfulness of these comparisons will be improved in the main study if the levels of exposures to the agents (termed “contrasts”) being compared are not highly collinear, that is, at least a portion of one exposure is uncorrelated with the other exposure. These uncorrelated components are termed “orthogonal contrasts.” If exposures to agents are perfectly correlated then it is not possible to distinguish statistically the independent effects of each exposure.

Our investigations to date indicate that adequate orthogonal contrasts exist among the first three agents and adequate epidemiological studies exist on C to help overcome any remaining collinearity issues. However, we have found that exposures to WC and Co are generally highly correlated (collinear), and the uncorrelated portion of the relationship between these two exposures, or available orthogonal contrast, is limited by small sample sizes. Moreover, there is only very limited epidemiological data available on the independent health effects of Co powder exposure that would permit indirect analytical control for the independent effect of this agent (Moulin et al., 1993; Mur et al., 1987). These studies do not provide quantitative and/or interpretable exposure levels for the subjects. This lack of exposure information precludes their use as a source of exposure information for Phase 3 study purposes.

A straightforward and very efficient solution to this issue is to include subjects in the study from facilities at which there is Co exposure but no W or WC exposure. We believe that the supplier(s) of Co powder to the facilities we visited would be good candidates for inclusion in the main study.

It is important to note that the inability to include subjects with Co only exposure does not represent a fatal flaw to the main study design. As described in detail in subsequent sections, the Phase 3 WCCo worker epidemiology study will be comprehensive, scientifically sound and far superior overall to the existing epidemiology studies. The limited orthogonal contrasts for Co and WC simply suggests that we may be unable to sort out completely the contribution of each exposure alone if increased lung cancer mortality risks are observed among subjects with combined exposures. The inclusion of subjects with Co only exposure will increase the sample size of the available orthogonal contrasts for Co

of 125


and WC, which will increase the statistical efficiency or power of the main study.

Obviously, if the results of the study show no association between WCCo exposure and lung cancer mortality, then the inclusion of the Co only group would be irrelevant. However, if the study shows an association between exposure to WCCo and lung cancer, then it would be important to determine the contributions of both WC and Co alone and in combination. Unfortunately, if an association is found and if we are unable to sort out completely the independent Co effect, then retrofitting the study design to improve this contrast may not be a viable option.

The following sections of this proposal describe the objectives, design, work plan, timetable and budget associated with the epidemiology and biostatistical component of the Phase 3 study. When the exact information needed to estimate certain study design or cost parameters was unavailable from the Phase 1 or Phase 2 studies, we made reasonable assumptions as required to arrive at our best estimate.

3. Phase 3 Study - Primary Research Objectives






of 125


4. Phase 3 Study - Proposed Epidemiologic Study Design

4.1. Overview

This section discusses the study design proposed for this investigation and the associated work plan. The key study components were arranged to ensure that the stated study objectives will be achieved with the greatest precision and efficiency attainable from the available study data.


UIC and UPitt investigators will first completely ascertain the cohort of workers with potential exposure to WCCo (Research Objective 1). 1UIC will perform a comprehensive exposure reconstruction of manufacturing processes and will develop task and time-specific estimates of exposure (Research Objective 2). The exposure matrix will enable UPitt investigators to construct summary measures of exposure to contaminants (Research Objective 3) and enable ongoing mortality surveillance of the cohort (Research Objective 4).

4.2. Organization of Subcontractors and Collaborators

The organization and coordination between UPitt and the various subcontractors is displayed in Figure 2. The UPitt component will be directed by Gary Marsh, Ph.D. and Jeanine Buchanich, Ph.D. UIC, under the direction of Nurtan Esmen, Ph.D. and Steven Lacey, Ph.D., will serve as a subcontractor to UPitt for the exposure reconstruction; DataBanque (DB), a Pittsburgh-based company under the direction of Susan Allen, will serve as the data processing subcontractor.

The EU site investigators will be responsible for enumerating the country-specific cohorts, collecting and processing the data and conducting the vital status tracing. Manfred Neuberger, Ph.D., Professor of Environmental Health, Center of Public Health, and head of the Department of Preventive Medicine at the Institute of Environmental Health at the Medical University of Vienna will oversee the Reutte Austria site; Peter Morfeld, Ph.D., Head of the Institute for Occupational Epidemiology and Risk Assessment (IERA), Evonik Services GmbH will oversee the three German sites; Magnus Svartengren, Ph.D., Professor of Environmental and Occupational Medicine, Karolinska Institute in Stockholm and Unit Head, Environmental Medicine, Department of Occupational and Environmental Health at Stockholm Center for Public Health will oversee the three Swedish sites and Thomas Sorahan, Ph.D., Professor of Occupational Epidemiology, The University of Birmingham, Institute of Occupational Health will coordinate the two UK sites. The investigators from the EU countries will report jointly to the UPitt and UIC investigators. The identification of

of 125


an Austrian collaborator occurred relatively recently and, as such, some details regarding the AT cohort enumeration and processing are still under development.

Figure 4. Organizational Structure of Phase 3 Study

14.3. Historical Cohort Study

1As noted earlier, the historical cohort study will provide the epidemiological platform for the proposed investigation and will include a nested case-control study of lung cancer. The cohort study will focus on mortality from lung cancer and other cause of death categories (including total mortality). Adjustment for potential confounding by smoking will occur primarily through nested case-control studies conducted in most countries (US, AT, SE and UK) or, in Germany, via external adjustment.

4.4. Basic Structure and Characteristics of the Cohort

1Historical cohorts are generally defined as individuals who worked for a period of time (or any time) between two dates, the cohort entry and cohort end dates. The entry and end dates are determined by a number of factors including the relevant operating dates of the study plant, the availability of records, the availability of standard death rates and of information needed to trace individuals and the minimum observation time (latency period) associated with the disease(s) under study.

of 125


Table 2 summarizes the relevant epidemiological characteristics of each of the study sites, as gleaned from the Phase 2 study. Table 2 shows that the study sites are highly diverse relative to geographic location, cohort size and cohort entry period (typically plant start up). Because of this heterogeneity, we will approach all aspects of our investigation in a site-specific manner, combining activities or data across two or more sites only if warranted by evidence of sufficient homogeneity.

The cohort entry period for each study site, except Reutte, is defined as the earliest date (year) of hire until December 31, 2007. For the Reutte site, we will use a cohort entrance date of January 1, 1950 to obviate the difficulties associated with data collection, exposure reconstruction and vital status tracing in the 1929-1949 period. Workers hired at Reutte before 1950 who worked beyond January 1, 1950 will be included in the cohort, however. We propose to include in our study population all workers from each study site regardless of gender, race or duration of employment unless it presents an undue hardship for the investigator to enumerate the cohort in this way. Because the SE sites were enumerated for an earlier study using a six-month minimum duration of employment, we have agreed to uphold that entrance criteria for the SE sites in the Phase 3 study. Based upon the data gathered in the Phase 1 and 2 studies, we are estimating that the total Phase 3 cohort will include between 7,000 and 10,000 employees from the US facilities and between 10,000 and 15,000 employees from the EU facilities.

Many historical cohort studies of this type impose arbitrary minimum duration of employment restrictions (e.g., six months or one year) to screen out temporary, summer or short-term workers who are assumed to be unrepresentative of the longer-term workforce. We believe that a more scientifically defensible approach is to include all workers initially, and then handle any demonstrable differences among short-term workers analytically via stratification or modeling. In fact, short-term workers can provide unique and epidemiologically valuable information as they may have higher or otherwise different exposures to the agents of interest due to working in less desirable entry-level positions, for example. Collecting records on all employees also eliminates on-site work history screening by untrained data clerks who may miscalculate the duration of employment and exclude eligible employees or who may exclude a worker with multiple periods of employment and, therefore multiple records, whose total duration of employment cannot be accurately ascertained by examining any one WH record.

We propose to determine the mortality experience of the cohort from date of cohort entry until December 31, 2008. Our proposed study design affords ample WCCo exposure time and follow-up time to examine meaningful mortality trends and exposure-response relationships. The maximum follow-up time ranges from 29 years (Henderson, NC) to 59 years (Reutte, AT and Coventry, UK). The overall study also allows for the analysis of mortality patterns relative to WCCo and other exposures in earlier and later calendar time periods.

4.5. Available Study Data

Table 2 also indicates, with the variable entitled “WH Records Available,” the completeness and availability of the demographic and WH records necessary to conduct an epidemiologic

of 125


study. The Phase 2 survey and site visits revealed that all information necessary for tracing the cohort (i.e., name, birth date, Social Security number or government ID) is available for each plant meeting the selection criteria. Dates of hire and termination are available for every member of the cohort; therefore, mortality and exposure analyses can include time-dependent variables such as duration of employment and time since first employment (latency). Detailed WH records are also available at the job level for these sites, which will enable linking these data with the exposure reconstruction being conducted by UIC; this will also allow investigators to identify and calculate exposure levels properly for individual employees. However, analyses may be limited by incomplete or limited information available on subgroups of the cohort, such as females.

Complete adjustment for potential confounders of lung cancer (e.g., smoking history) may not be possible with the information available in the company-held records; however, to the extent possible, all relevant data will be microfilmed during our site visits. These data will then be processed accordingly and merged with available demographic information to form the master cohort file. Nested case-control studies will be conducted in most countries (AT, SE, UK and US) to examine more fully the extent to which our lung cancer risk estimates are confounded by smoking; for the German sites, the risk estimates will be externally adjusted for potential confounding by smoking via Monte Carlo simulation.

The availability of exposure data among all of the study sites is satisfactory. The available direct exposure measurements can be used for adjustment for time period, country, company and site variability considerations as well as for the final validation of the exposure model as structured in the job-exposure matrix. However, there are two perceived difficulties. The most important is the collinearity, or correlation, between Co and WC exposures which, if associated with an adverse outcome, might be difficult to differentiate from one another. As discussed earlier, our primary proposed solution would be to include in the Phase 3 study facility(ies) which use(s) Co but does not have WC exposure; we could also attempt to model the effects of Co using the very limited data found in the peer-reviewed literature (Moulin et al., 1993; Mur et al., 1987). The second perceived problem, of lesser importance, is that the co-exposure information is not complete so the exposure assessment portion of the study will more thoroughly attempt to adjust for potential co-exposures to known and suspected carcinogens, including C.

of 125


5. Phase 3 Study - Proposed Exposure Reconstruction Study Design

5.1. Understanding of the Problem

The proposed research is a hypothesis-testing study of the relationship between WCCo exposure and mortality experience. However, the ability of the study to identify the probable etiologic agent(s) for any observed excesses will depend upon a thorough and scientifically defensible reconstruction of exposure to WCCo, WC, W, Co and other agents such as C, coupled with a statistical analysis of the epidemiologic data that includes adjustment for potential confounders such as smoking and co-exposures to occupational carcinogens, as well as exposure-response analysis.

This study is a multi-site, multi-process study in which the mortality outcomes may be pooled to achieve the maximum statistical power. It is essential, therefore, that the exposure estimates for the members of the merged facilities have a sound empirical basis so that the delineation of exposure classes is meaningful and consistent across facilities. Moreover, the exposure reconstruction for the cohort must be sufficiently refined to provide an adequate number of exposure classes for the epidemiologic exposure-response analysis while keeping misclassification rates within acceptable limits. Uncritical use of limited or poorly documented exposure sampling data or quantification of exposures based on expert judgment may result in serious job exposure misclassification problems. Exposure reconstruction should draw upon on all available empirical data, including site-specific (occupational) IH measurements, site-specific historic process data, physico-chemical principles and accepted engineering models. Experimental empirical modeling may also be necessary to fill gaps in knowledge.

The proposed research therefore requires two components involving comparable levels of effort and sophistication: an exposure reconstruction component that synthesizes all available empirical data relating to historic and current exposures and a statistical component that analyzes the epidemiologic data in relation to this exposure reconstruction with adjustment for potential confounding factors.

5.2. Primary Research Objectives of the Exposure Assessment Component

To determine the relationship between occupational exposure to WCCo and mortality experience, the primary research objectives of the exposure reconstruction component are:

1) To generate scientifically sound estimates of exposure to WCCo and other potential carcinogens for all job and/or task categories on a site-specific and time-dependent basis. Average and cumulative exposure metrics will be developed and adjusted for country, company and site variability linearly. The interaction between company and country will also be tested.

2) To create exposure classes for subsequent statistical analysis in the epidemiology component of the study.

3) To assess the robustness of the exposure reconstruction models employed with respect to uncertainties arising from data gaps and inherent variability.

of 125


5.3. Methodologic Approach

5.3.1. Basic Approach

The basic exposure model is job class (a), task (b) and year (y) normalized exposure levels adjusted for country (X1), company (X2) and location (X3). In specific terms the fundamental linear adjustments to exposure levels with country X company interaction may be expressed as:

E( a ,b , y )=βo (a ,b , y )+βcountry X1+βcompany X2+β site X3+βcountry∗Company X 1 X2+εwhereβ ,E , β0=f (a ,b , y0) or g (a ,b ,∞)for country, company and site calculations.

At this point, we do not know whether exposure values near the start-up (y0) can be found or whether an asymptotic value (y=∞) for exposure can be determined. Between start-up based or asymptotic functions which define the exposure and coefficients we shall choose the function that is convenient and/or tractable. That formulation will be used consistently to adjust for the time decline of exposures first with the simplifying assumption that the exposure matrix is multiplied by a scalar function of time. If this approach does not provide reasonable results, then more complicated solutions will be sought. Once this function is determined, all deterministic or empirical models will be adjusted using this relationship.

The general exposure reconstruction process and the steps that may be involved in this process have been summarized in Table 4. Clearly, while some of the steps are sequential, many of the process steps are concurrent. It should be noted that while much of the methodology is written in general terms, the evaluation of each facility is independent. Some of the information shown in the table will be collected on an as-needed basis and some of the operations will be carried out on an as-required basis. The general approach is the least complicated path without sacrificing rigor. In order to aggregate the cohort across plants, the determinants of exposure will be adjusted to account for the differences in operations from one facility to another using the empirically modeled correction shown above.

of 125


Table 7. Summary of Exposure Reconstruction Method

1. Selected process information collection a. Process operation documents b. Facility layouts and plans c. Technical memos and other sources d. General annual production records

2. Exposure information collectiona. All (occupational) IH monitoring data, including area and personal exposure

sampling data b. Documentation pertaining to

i) Sampling ii) Administrative ordersiii) Environmental exposure monitoringiv) Occupational hygiene inspection reportsv) Occupational health reports filed to or by regulatory agencies (as applicable)

3. Selected job and task identificationa. Job descriptions as written in company recordsb. As described by experienced persons (as needed)

4. Projection of results backwards through time from contemporary data (as needed)

5. Operations analysis of the production processes

6. Mathematical modeling of the information gathered above

7. Verification of models from extant data

8. Generation of task-based exposures

9. Construction of a site-specific job/exposure matrix detailing exposures for each job title (or group of job titles) as a function of specific agents and appropriate time intervals

of 125


5.3.2. Collection of Information

5.3.2.1. Preliminary information

Any exposure reconstruction study must start with extensive familiarization of the investigators with the work tasks and operations history. Some of this work was performed and compiled during Phase 2 and will be informative. In addition, one of the team members is a mechanical engineer with extensive metals industry experience. Both of these considerations are time-saving.

For further specific information collection, additional visits will be scheduled during the Phase 3 study at the convenience of the personnel of the facility to be visited. In these visits an extensive documentation of the processes, operations and tasks will be made; again, this is likely simplified to some degree by the prior efforts. In addition, all extant information on sampling will be requested.

It is important to emphasize that we will need this information from all plants included in Phase 2 irrespective of whether the plant is included in Phase 3 or not. Although not all of the information gathered would be necessarily used in the initial exposure estimation, it is necessary for the empirical correction factor and for the basic exposure based stratification of the cohort. In addition, much of the information will be necessary for more extensive exposure reconstruction should that become necessary to understand an excess in mortality (if an excess is observed).

Obviously, the exposure reconstruction process may be more efficiently designed with the a priori knowledge of the quantity and quality of the data available. As it is planned, the exposure reconstruction is sequential and adjusted as the data evolves. In that respect it is efficient and allows considerable latitude for the development of improved mathematical methods. As the work tasks and operations information is collected and coded, exposure measurement information will be analyzed to determine the extent and quality of these measurements with respect to exposure reconstruction.

5.3.2.2. Standardization of job and department titles

To elucidate the exposure-specific information, a set of standardized job title codes will be developed. The highest order in the hierarchy of the standardization process is the spelling and word order of jobs. The second in this hierarchy is the query of the meaning of an uncertain job title to a person familiar with the specific process and job title information. The last resort is the professional judgment by the industrial hygienist to either decide the equivalency of one title to another or to deem the job title as undecipherable. In our experience in well over three-dozen facilities, only a fraction of a percent of jobs was ultimately deemed undecipherable. In these cases, together with untitled or unknown jobs, the contribution of the lost record to the aggregate person years in each facility was negligibly small.

5.3.2.3. Collection of historical exposure data

of 125


Historical monitoring data may be available only for more recent years, and the “quality” of the data with respect to the coverage of the jobs or tasks can only be determined through extensive examination of the available information. In general, the monitoring in the industry has been limited to several broad classes as defined by the purpose of the sampling campaigns carried out for some specific requirement. Typically IH monitoring has considered only the “known culprits,” such as Co, and rarely samples for compounds with undefined health effects. Environmental samples do not define exposure but may show the mixture of chemical challenges potentially present in each facility. Similar concerns exist with permits, inspections, etc. The historical exposure data in conjunction with the documentation of environmental concerns constitutes the basis of the investigation.

5.3.2.4. Abstraction and analysis of historical exposure data

Historical monitoring data from the facilities will be acquired where possible. We would like to obtain exposure data from all facilities included in Phase 2 (Tables 1 and 2) that are in the countries selected for the Phase 3 study. A substantial amount of monitoring data has been collected by the facilities and the results are available. After the determination of the data quality, we propose to analyze these data, in conjunction with the information collected as described in sections 5.3.2.1 to 5.3.2.3 with respect to the construction of mathematical interpolative and extrapolative models. The information available, extant and missing data, will determine the sophistication of the mathematical models that can be generated. In other words, in cases where the information is lacking, the mathematical model would of necessity be a crude one.

In addition, we propose to analyze these data with respect to the best classification scheme suggested by the distribution of the data itself. In this approach, arbitrary a priori definitions of exposure classes are avoided and the utilization of the information content of the data is maximized.

IH data are often happenstance measurements taken to satisfy other purposes. Therefore, even though the existing data set may have coverage over a long period of time, it might not be directly applicable to the exposure reconstruction. However, data spanning long periods of time have important uses in determining historical trends (Sections 5.4.3.1 and 5.4.3.2). These trends may be compared to calculated exposures as a secondary validation of the calculations. In addition, prediction from the mathematical models may also be compared to the existing data. In this respect, the data can serve as a direct validation. There are a few difficulties in this process, especially when the existing data cover only a small operational span of the process variables (Esmen, 1998). Unfortunately, the last problem is virtually impossible to predict without working with the actual data. Therefore, the resolution of any difficulties will be a part of this research project.

5.3.2.5. Collection of historical job and task information

According to the Phase 2 study, there are job, task and operations data available. Job or occupation descriptions are a key component to understanding the determinants of exposure for any worker. The main sources of information that describe jobs and their associated tasks can be found in company personnel files or in labor-management contracts. Typically, such job descriptions are general in nature, describing the overall tasks associated with the

of 125


occupation. Information relating to job/occupational categories can usually be found in the personnel and payroll files of the organization. Although the main source of general job/occupation information is at the corporate or facility level, specifics such as discrete tasks and how the tasks are performed, what equipment they are performed on, how often they are performed, etc. are not contained in the job/occupational descriptions. This information can often be collected during the site visits.

Company records at the facility and corporate level will be reviewed to determine the availability and completeness of the job/occupation data. Job titles and descriptions also undergo relatively routine rewriting and redefinition. This often occurs as a result of new labor-management contracts and with the changing or addition of new products. While in many organizations job/occupational descriptions are retained for some period after redefinition, there is no legal requirement for their long-term retention. Most companies undergo routine purging of unneeded records. Often records purges result in the loss of job/occupational information for those jobs/occupations that are no longer used. In order to fill these gaps, we will rely upon the job dictionary (as determined from WH or other records) to identify job or occupational titles that are not described in existing personnel records. For those jobs where no job descriptions are available, we will collect appropriate information from knowledgeable company personnel so that these jobs/occupations can be adequately described.

Job/occupational descriptions that are used for personnel or payroll purposes are generally reliable; however, they do not usually reflect routine operational variability, the personal preferences of workers in doing a task, daily changes in tasks performed, etc. Information about particular aspects of a job/occupation such as the equipment worked with, materials or processes where the job/occupation tasks were performed and the frequency of task performance may be available at the facility level; however, it is doubtful that information of this type is designated for long-term retention. Anecdotal evidence regarding how a task performance varied with respect to written job descriptions must be obtained through site visits, information from knowledgeable personnel and operations analyses.

5.3.2.5.1. Process information

Many of the tasks performed by workers at the process level are not described in the job descriptions. Additionally, this type of information is not contained formally within the documents available at the plant level. This type of information can be obtained by discussion with workers and supervisors who were actually involved in the processes or tasks. As an example, in one studied facility that produced a synthetic polymer, the job description related that a task of one category of workers was to clean the process strainers. Further definition of this task was obtained by process supervisors. Still further definition was obtained from workers who actually conducted the task. From the workers, it was noted that this specific task could be accomplished in a variety of ways, each having a different potential for exposure. This type of variability is observed in exposure studies of workers. Typically, variability of exposure measurements for a single worker is larger than that component of variability associated with sampling and analysis of samples. How workers conduct their assigned tasks explains why worker exposures for the same task in the same or a similar location can vary dramatically.

The tasks that a worker may be asked to perform can change periodically. These changes

of 125


may be related to new products, process changes or equipment changes – fundamental issues in an industrial facility. Typically, the determinants of process-related task changes can be identified by a review of process records that describe what products are being made.

5.3.2.5.2. Operations analyses

The work task analysis of production procedures, frequency of operations and changes in production figures provide a reasonably coherent picture of potential changes in the exposure levels over time. In addition, operations analyses are required to define the equivalence of operations with respect to a chosen contaminant (Esmen, 1979). The modeling methods used in this type of analysis are straightforward. Based on documented process changes, adjustments are made to exposure concentration levels. Usually, these adjustments are based on dimensionless scaling using either empirically determined coefficients (or powers) or production level scaling that suggests that all things being equal, the exposures increase linearly for tasks related to batch operations and as the 1/3 power for tasks related to continuous operations.

The ultimate goal is to develop an exposure profile for each job class. We will consider approaching the problem using generalized operation families and creating a task-based family tree of operations. The model developed on this basis will be tested against the existing exposure measurements; Monte Carlo analysis of potential outcomes is usually an efficient method to test the validity of a model as well as its sensitivity and specificity.

5.3.3. Preliminary Exposure Analysis

Exposure reconstruction studies usually have many levels of complexity. We propose to create and code an exposure measurement-based job/exposure dictionary in matrix form. Available measured exposures will be assigned to the corresponding dictionary entries strictly as placeholders. Analysis of the process and product information, confounder identifications and other potentially useful information will also be entered. Inspection of the job dictionary will provide the plant-specific jobs and time periods for which the exposures were not measured. In other words, the gaps in the data will be immediately apparent. The cohort mortality study team and the exposure reconstruction team will coordinate their efforts to create a job/exposure matrix that can be directly used in epidemiological software (OCMAP-PLUS (Marsh et al., 1998)).

5.3.4. Identification of Confounding Exposures

The potential exposures will differ from site to site depending on the processes present and material used. Since the Phase 2 study lacks details on the particular contaminants included in the exposure data, it is likely that some of the potential carcinogens may have to be modeled as a unit step (exposed/not exposed) as a function of time. The inclusion of a binary exposure characteristic in the worker exposure matrix does not create problems. The determination of the characteristics of the exposure metric function is a priori. The method of analysis used in gauging the number of exposure classes that can be generated is sufficiently general to include the possibility of dichotomous characterization of exposures. However, it might be possible that the rational analysis of exposure classification may suggest a different number of classes for the same compound in different facilities. The

of 125


epidemiologic analysis generated by such an outcome and the resolution of specific problems will be addressed as a part of the proposed research. At this point we speculate that the problem of different numbers of exposure classes for a given compound in different facilities will be confined to the exposures to confounding compounds.

5.3.5. Identification of Missing Exposures and Information Gaps

The Phase 2 study suggested that the availability of the historical monitoring data is limited to certain years and there is no indication of the quality of the data with respect to the coverage of the jobs or tasks. The exposure data that can be used directly will be the basis for all estimation processes. These data define the availability of direct exposure information and must be used as a basis for exposure reconstruction by other means. However, available monitoring data must be reviewed critically with respect to the sampling strategy employed (e.g., random, worst case, happenstance) and the completeness of documentation. The Phase 2 study also suggested that the availability of the historical process and environmental control information may be limited to certain years and there is no indication of the quality of the process data with respect to the coverage of both the production and control of the processes. Exposure reconstruction through modeling and by Bayesian methods require process, exposure control and environmental information that range from straightforward design and operational characteristics of the manufacturing processes to operations that may be unique to the plant. Therefore, the following date and facility specific exposure and process related information will be obtained to the extent possible.

i. The process equipment in use:Whenever necessary, plant design layouts and equipment drawings that may pertain to modeling will be collected and analyzed. Information on changes in process equipment, plant expansion and/or contraction, replacement of processes and dates corresponding to these events will also be sought. General and working area ventilation and installation dates will also be requested as a part of the information on process variables.

ii. The production variables:Change in production, plant repair and outages, production figures during constant operations without process malfunction. Start-up and shut-down procedures and dates will also be sought. Historical information on raw material input rates or product shipping rates will also be requested.

iii. Personal protective equipment usage and work practices:Information on the known personal protective equipment usage, work practice rules, and unusual practices will be requested.

The collected information will be analyzed for each facility. The information availability has three possible outcomes.

i. If the information is available for all plants that information is complete and no further action is necessary.

of 125


ii. If the information is available in some facilities and not others, then this is a true information gap. As a part of the research, ways and means to close this information gap will be sought.

iii. If the information is not available for any of the facilities, the importance of the specific piece of information will be evaluated. The lack of an essential or very important piece of information will be judged as a true and important information gap, and as a part of the research, ways and means to close this information gap will be sought. If the piece of information is judged to be not critical, it will be deemed an inconsequential information gap. Although there are risks associated with this classification, as the researchers may not be fully cognizant of the critical nature of missing information, this expediency may be heuristically justified in terms of efficiency and fiscal limits.

The information collected and analyzed as specified in this section will be used to derive process modeling based factors for extrapolation from known levels to unknown levels, for validation of process models between known levels and for general model building and adjustments, as well as serving as the basis for the design of experiments to obtain missing information.

In most cases, missing process information can be augmented from metallurgical and/or manufacturing sources. This approach, while not a primary information source, can provide invaluable information resources especially for operations that are no longer in use. In our experience, we have found that the literature-based augmentation of information is superior to expert opinion, which is the last resort for information augmentation.

5.4. Modeling

There are several distinct phases of modeling. During the first phase, the information collected to date (however complete or incomplete) is used in a binary exposed/not exposed scheme for job titles, years and pre-selected agent or agent/process combinations. The results of this phase are used to select the combinations of interest for further refinement. The hierarchically preferred methods of modeling may be listed as follows:

1. Extrapolation or interpolation of extant exposure measurements (for all combinations that span the entire time or entire life of the process)

2. Modeling of exposures

3. Verification of models

a. Comparison with existing measurementsb. Comparison of predictions with existing data from other operation specific

sources

5.4.1. Exposure Metrics

The primary exposure metric to be used is the time integrated cumulative exposure. The estimator for this exposure is the class average exposure multiplied by time. This estimate may be determined readily by the product of the exposure level (the average airborne

of 125


concentration) and the time spent at that job classification. However, the exposure level is a function of time; therefore the summation must be carried out at reasonable small intervals of time. The selection of the time interval for cumulating exposures depend on the availability of a number of parameters, some of which (such as ability to discern job titles as a function of time) are not related to the exposure reconstruction process. In that sense, it is impossible to assign an a priori target to the refinement of exposure levels as a function of time. The exposure reconstruction will compute average exposures for each job title in as fine of intervals as possible. Particle size dependent exposure levels (respirable, thoracic and inhalable) will be determined if possible.

5.4.2. Mathematical and Physical Models

Mathematical and physical models of exposure start from the premise that the exposure to a contaminant has two characteristic parts: one that is controlled by the properties of the source and the other controlled by the work practices of each worker. Super-imposed upon these two characteristic exposures, uncontrolled and/or uncharacterized variations generate the differences in exposure values. While the exposure of each worker is a continuous, highly auto-correlated function in time, the samples taken from this universe can be well represented by a right skewed distribution of measured exposure levels. The lognormal central tendency of exposure measurements has been both empirically and theoretically shown to be applicable to environmental sample analysis. Based on these observations, a known characteristic exposure may be scaled upwards or downwards to account for the process changes. It is not realistic to expect to model the characteristic exposure that is defined by the work practices. This however, does not present a problem.

It may be shown that the characteristic exposures based on individual work practice contribute to the between worker variance and the characteristic exposure that is due to the source parameter shifts the distribution. It may be stated that the changes due to process variables by and large change the median. The term “by and large” is a weak condition that is heuristically necessary. There are counter examples to the statement that suggest that process shifts the median. For example, controls that introduce high but intermittent releases will tend to reduce within worker variance. In the proposed research, the mathematical models to be developed and used will be wary of these types of important exceptions, and verify each step to the extent that it is verifiable.

Noting that mathematical modeling is a process of creating a symbolic representation of some phenomenon, its procedures include building a conceptual model. Therefore, the investigator would have to decide what factors are relevant to the problem and what parameters can be de-emphasized. Once a model has been developed, it must be challenged to the fullest extent to ensure that it is an accurate reflection of reality. It would not be wrong to claim that mathematical models evolve from simplifying assumptions to more refined representations of reality through the accumulation of insight gained at each step.

With the individual exposure reconstruction team members bringing different areas of expertise to the project, we will follow the procedural steps outlined below.

1. Construct a very simple model with a complete list of assumptions that go with the model. 2. Select several prominent parameters by sensitivity analysis.

of 125


3. Construct the relationship between the reconstructed exposure and these parameters.4. Does this make physical or chemical sense? If not change the assumptions. 5. Develop the controlling equations and make predictions.6. Is the result produced reasonable? Perform perturbation analysis (if necessary) and improve the model as need. 7. When the model is satisfactory, validate the model using existing measurements. If the validation is satisfactory then the model is ready to be used. If it is not satisfactory, a more complicated model can be sought.

It is important to note that the determining factor in the success of exposure estimation is not the precision of the estimates but the ability of the estimates to achieve the ordering of exposures and exposure bands assigned to exposure classes. The bounds of the exposure levels that can be assigned to a job class are very strongly influenced by the precision of the job dictionary rather than the precision of the assigned exposures (Esmen et al., 2007a).

A complete understanding of the underlying scientific and engineering principles of the exposure kinetics and the industrial processes being modeled is necessary. In some cases, the coupling between the two highly complicated systems may not be apparent or may be too complicated to be tractable. If a problem is not amenable to modeling, then other extrapolation techniques must be found. There are many predictive techniques and they are used often in making a wide range of predictions from simple industrial control predictions to complicated econometric forecasts. Some of these methods are directly adaptable to the extrapolative needs of exposure reconstruction. The order of preference in the method selection is based on the dependence of the method on conjecture. The remaining exposure reconstruction methods to be used in this study are listed in the order of preference.

5.4.3. Statistical Models

The statistical methods of interpolation or extrapolation usually require large amounts of data. If such data are available, or can be obtained through field measurements and/or experiments, then the two statistical methods we propose to use are listed in the order of preference.

5.4.3.1. Monte Carlo modeling of interpolated, extrapolated and missing data

Available IH sampling data records will be used to generate a distribution of historic plant exposure conditions. This distribution will be combined with a distribution of employee work patterns to generate a combined distribution of employee exposures over the years of operation of the plant.

5.4.3.2. Inversion with proportional extrapolation

The Principal Investigator of the exposure reconstruction (NE) previously developed a mathematical model to invert time averaged samples to rank the contributing tasks (Phillips & Esmen, 1999; Phillips et al., 2001). The method is designed to rank the tasks performed during sampling in the order of their influence. Extensive simulation results suggest that

of 125


the method is acceptably reliable in selecting one or two of the most influential tasks based on a rather modest number of time-weighted average samples (6 to 12 or so). While the method is not designed to estimate the absolute contribution level of each task to the composite, using the assumption that these influential tasks are the driving mechanism of exposure change, the exposure values may be extrapolated using equally weighted proportional change in these tasks.

The applications involving this method will assume that the change in the characteristic exposure potential of the contributing tasks are either known or can be calculated. It would be very unlikely to assume that data exists with respect to the exposure potential of each task. However, distinct tasks are usually easier to model than an integrated value for exposure. Therefore, a mathematical formula may be elucidated to represent the change in exposure level as a function of changes in the exposure potential of the dominant tasks of a job classification. Although arguable or partly defensible, both the concept and application of this procedure contains a number of assumptions that are either intuitive or can not be readily substantiated. Therefore, this is a less preferred extrapolation than the data based methods above.

5.5. Exposure Data Analysis

The necessary empirical data for the creation or validation of modeled exposures will be extracted from existing exposure measurements.

5.5.1. Existing Usable Exposure Data

Exposure data will be scrutinized to determine its usability. For example, area samples are not directly usable exposure measures. While they might be useful in modeling exercises, extreme caution is required in their inclusion in an exposure reconstruction study. Another common mistake is averaging the occasional short-term sample with a group of long-term samples. Combining samples with significantly different sampling times will introduce significant errors. For such combinations, the variances of short- and long-term sample must be known. Therefore, care will be taken in the calculation of long-term averages using different duration samples. While the procedure is not a difficult one, it must be observed since the errors propagate over time and across classes depending upon the type of extrapolation or interpolation methods used.

5.5.2. Determination of Classification Limits

The classification of jobs or workers by exposure is an important undertaking in any occupational epidemiology study. Hitherto, exposure classification designs have been strongly motivated by a desire to generate a sufficient number of exposure classes for the determination of a potential exposure-response relationship. Thus, the partitioning of exposures has been more or less arbitrary. The interpretative problems created by the selection of an arbitrary number of exposure assignment classes have been addressed only recently. This theory and its application to real data sets suggest that the choice of the number of exposure classes based on epidemiological convenience might not be

of 125


appropriate. However, by considering an acceptable level of exposure misclassification and by assuming the probable exposure distributions, it is possible to calculate the allowable number of exposure classes and the proper partitioning ratios between these classes. Thus, the trade-off between misclassification and number of exposure classes was shown to be a satisfactory solution. In this study, the number of exposure classes and the boundaries for each class will be determined by the methods published by the investigators of this grant (Esmen et al. 2007a, 2007b, 2007c).

5.5.3. Calculation of Job Specific Exposures

Upon final enumeration of the cohort, UIC researchers will prepare a job dictionary which they will utilize in the exposure assessment. The finalized job/exposure matrix containing job-specific exposures determined by UIC will be linked with the epidemiological data collected by UPitt. It has been our experience that many iterations between the epidemiologic and exposure assessment phases are necessary to finalize the job/exposure matrix, so UPitt and UIC investigators will work closely throughout its preparation.

Exposure profiles will be developed for each job title, or group of job titles, in order to estimate exposure levels for each worker. These exposure levels will be expressed as an exposure matrix and the exposure vector of the individual will be calculated based on the exposures indicated in the matrix for each job held by the individual.

5.5.4. Assigning Job Specific Exposures and Merging Exposures

As suggested above, job-specific time-weighted average exposure is the primary exposure metric. If it is at all possible, particle-size specific cumulative exposure levels based on three overlapping size groups (inhalable, thoracic and respirable) will be estimated for the three main component aerosols. C exposures are expected to be in the respirable size range, whereas Co and WCCo particulate exposures are expected to cover the entire spectrum.

After final validation, the exposure reconstruction results will be prepared in a job/exposure matrix usable in the cohort mortality study. Although this would theoretically be the official end of the exposure estimation procedure, the exposure estimation team and the epidemiology/statistical teams are expected to cooperate closely in the final analyses and interpretation of results. The two teams have cooperated in numerous studies and they are well aware of the fact that minor questions that require a coordinated effort tend to arise until the last version of the final report is prepared and ready to be submitted. Our experience has also shown that the cooperation and coordination between the two teams continues well past the final report through the publication of scientific papers.

of 125


6. Phase 3 Epidemiology Study - Proposed Work Plan and Timetable

Following is a description of the major project tasks associated with the proposed epidemiology study. The epidemiology and biostatistics component will run concurrently with the exposure reconstruction phase. Draft manuscripts for peer-reviewed publication and a final report for the sponsor will be produced in the last quarter of the project.

6.1. Task 1A - Cohort Enumeration

The WCCo cohort will be enumerated by the study investigators using all available corporate and plant-based employee records, including those currently stored in archives. The cohort will include active, terminated, retired and deceased employees who worked one or more days from cohort entry date through December 31, 2007.

To the extent possible, relevant cohort data will be derived initially from records in machine-readable format. To ensure that the cohort enumeration is complete and accurate, study staff will also review all hard copy employee records at the plants. Hard copy records will include work applications, detailed job service, insurance and death certificates.

The cohort enumeration will include four basic categories of employee data:

1) Data on personal identifiers and demographic factors (e.g., name, social security number, birth date, race, gender, pay type and employment status)

2) Data on employment history and exposure (e.g., beginning and end dates, department, work area, job title/function, types and levels of exposure (if known) for each job held)

3) Data on health outcome (certification of death and cause of death)

4) Data on potential confounding factors (e.g., tobacco smoking history and previous and subsequent employment)

Due to the variations in available cohort data, each plant will follow a unique protocol during the cohort enumeration phase.

6.2. Task 1B – Plant Data Collection

The exposure reconstruction component of the study will collect, as necessary and to the extent possible, data related to processes, exposures and jobs/tasks as detailed in Table 4. These data will be collected with the coordination and assistance of plant contacts established in Phase 2 of the study and the country specific collaborators identified for Phase 3. The UIC investigators will work with the plant contacts and collaborators to collect the relevant information in the most expeditious and cost-effective manner.

of 125


6.3. Task 2A - Data Processing, Validation and Management

6.3.1. United States

6.3.1.1. Electronic records

Electronic records will be supplied directly from each of the nine US sites. A process for electronically collecting data from each facility will be implemented to combine all data into single relational database system housed on UPitt’s servers. The electronic data will be exported into a common data format, for example, an ASCII text file, a Microsoft Excel file or a Microsoft Access database.

6.3.1.2. Hard copy records

UPitt scanners will be used to capture any cohort data from the hard copy records and hard copy records will be produced from all records scanned at the nine US sites. Hard copy records will be sent to DataBanque (DB), the UPitt data processing subcontractor, and prepared for abstraction. There, each file will be assigned to a box and will be given a file barcode. Abstractors will prepare the file for abstraction by segmenting the documents as critical or non-critical and in date order; they will then sort out relevant data from the employee files and transcribe it onto a standardized abstraction form. Abstraction will be double-blind verified to the minimum standard set by the UPitt (typically less than 5% error rate in any field). Abstracted information will be keyed exactly as it appears on the form and 100% double-blind verified. UPitt staff will visit the DB site weekly to confirm accordance with the agreed upon standards and to independently audit the abstraction and data entry. Master demographic and WH files will be created that will serve as the basis of files formatted for various statistical analyses. These files can be transferred to and from UPitt via CD, DVD and via secure web interfaces.

6.3.2. Austria

The Austrian investigator will work with plant personnel to obtain electronic copies of the data, to the extent possible, and to coordinate the hard copy data abstraction. We are continuing to work closely with the newly identified Austrian investigator to finalize the details regarding the Austrian cohort enumeration and data processing.

6.3.3. Germany

The study data from the three German sites will be collected using an electronic, web-based data collection system, the Health Study Application (HSA). It enables users at plant study sites to gather protected personal data, e.g. personnel or medical data, and contains user interfaces to input exposure, covariable and response data in German. Site users view and edit data (e.g. first and last name, date of birth, etc.) and then a unique identification pseudonym (PID) is created for every cohort member; this number can only be translated by a data protection commissioner. The German investigators will only have access to the PID and not the de-identified data. To ensure complete cohort enumeration and data linkage, site plants will send a list of eligible active and retired workers to the data protection

of 125


commissioner who imports eligible study members into the HSA-application, where the PID is automatically generated for each worker.

6.3.4. Sweden

The three Swedish sites in this study have already been evaluated in an historical cohort mortality study. Therefore, the Swedish investigator will begin the cohort enumeration with those study files. To prevent undue hardship and eliminate the possibility of missing or destroyed records, the same six-month minimum duration of employment criteria used in the original Swedish studies will be used in the Phase 3 study. The Swedish investigator will travel to the facilities and abstract the relevant study data for all employees who were hired after the end of the original study and will update work histories for those employees working as of the end of the original study.

6.3.5. United Kingdom

The UK investigator will visit the two sites (Sandvik Coromant and Sandvik Powder Production) with a team of survey assistants from the University of Birmingham. All available useful information from company personnel files will be transcribed onto a study form for all past and present employees who satisfy the study inclusion criteria (e.g., all workforce employees with one or more days worked from cohort entry date through December 31, 2007). Useful information includes identifying particulars (full names, dates of birth and last known address), work history details (dates of hire, first job/department, changes of job within company and dates of leaving employment) and smoking history.

Information abstracted by University of Birmingham personnel onto forms will be computerized at the university using a common database structure for all centers. All computerized information will be double-checked with the source forms. Duplicate entries will be identified and combined into a single entry.

6.3.6. All Study Sites

6.3.6.1. Data Integration

Initially, all sites will be provided with a standardized data layout. The layout will outline the data variables and data type fields (e.g. date fields, text fields or numeric fields) required by UPitt. Each site will abstract and enter the data, according to the layout, into a file or database with one of the following file formats:

Microsoft Excel spreadsheet Microsoft Access database Comma-delimited ASCII text file Tab-delimited ASCII text file Fixed-width ASCII text file

These files can then be transferred to UPitt's servers via CD, DVD or a secure web interface. A process for electronically collecting data from all sources and combining this

of 125


data into a single relational database system to be housed on UPitt’s servers will be implemented. When the transfer is complete, all data will be imported directly into MySQL tables. These various data sets are normalized through recoding and then they are combined to create the main research MySQL database(s). All data tables that are edited via human interaction (i.e. not programmatically) are done so via software that maintains an audit trail. This audit trail contains the original state of the record being updated, the user who is updating the record and the reason the record is being updated. The audit trail can be used to revert to an earlier version of a record should an input error occur. Potential problems of data management and analysis include: missing data, coding errors and inaccurately recorded data. UPitt has created software procedures that programmatically check for and correct these problems in the MySQL research datasets.

6.3.6.2. Data Storage

Data storage will include the coding, creating and updating of MySQL databases and applying quality control procedures while accurately documenting the entire process. The MySQL relational database system is robust and scalable and has been successfully used by UPitt to store and manipulate complex data sets consisting of millions of records. All data will be imported directly into MySQL tables. The data received from the sites will be transferred directly to MySQL database tables on the UPitt server using a MySQL ODBC driver for Windows. The ODBC method is consistently used by UPitt to populate research databases and it has been thoroughly tested for completeness and accuracy. Extensive edit checks on the newly created tables will be executed to ensure consistency with the source data files. Once data is entered into a MySQL table it is never modified without first storing a copy of the original data in the MySQL table.

The following software will be used to create and maintain the datasets:

Microsoft Excel: Used to import and export data. Microsoft Access: Used to import and export data. Java: Used to develop web-based data access tools. MySQL: The RDBMS. SQL: Query language used to interface to the RDBMS. Used to create, query,

update and maintain data sets and generate reports. Perl: Used to develop web-based reporting tools and to create, query, update and

maintain datasets. Standard Linux tools: Used for access control, backups, data checks, logging, web-

based tool development, report generation and file manipulation.

6.3.6.3. Data Backup

UPitt’s servers use RAID - 5 disk arrays which guards against data loss. Automated daily backups are performed and these backup sets are burned onto DVDs. These backup sets can be used to restore data and software should any data loss occur. In addition, a backup server is used to store the current software and databases. This backup server can be put into place should the main server fail. Any manual modifications to the data involve an audit trail which can be used to revert to an earlier version of a record should an input error occur.

of 125


6.3.6.4. Documentation

The process(es) applied to the data that transform it from source data to the final research database data are documented. In addition any software and/or scripts that are used to create and modify the database data will be internally documented. This internal documentation will include things such as code comments and examples.

6.4. Task 2B – Exposure Modeling and Job Dictionary

Estimation of occupational exposures informs and bolsters the health outcomes data of the occupational epidemiology study. The estimation of exposures relies on two principle sets of information: historic IH data to validate exposure models and information on the manufacturing environment. In this study, exposure reconstruction should draw upon on all available empirical data, including site-specific IH sampling measurements, site-specific historic chemical process data, physiochemical principles and accepted chemical engineering models.

The content and extent of IH data from the 18 visited facilities was identified in Phase 2 of the study. In Phase 3, these data will be collected and abstracted. IH data will also be requested from the facilities considered in Phase 2 that did not meet the selection criteria to be included in Phase 3. These additional data are important for the study to determine between-site and between-country variability of exposure for similar operations. A large portion of the data is available in electronic format, with a manageable remainder in hard copy format that will be copied or scanned. A master exposure database will be constructed for critical elements including contaminant name, concentration measured, date recorded, process area or department where sample was taken and tasks performed. Descriptive statistics will be generated. These data will serve to validate physico-chemical exposure models developed by the research team.

The evaluation of each facility is independent. In order to aggregate the cohort across plants, the determinants of exposure will be adjusted to account for the differences in operations from one facility to another. Facility specific exposure and process related information will be obtained to the extent possible.

General production process information was collected in Phase 2 of the study. In Phase 3, available essential engineering records will be collected and reviewed. These data will be representative of facility changes in operations and layout over space and time. These data may include facility schematics, production records and other types of records and can indicate major engineering/production changes throughout the life of a given facility. Information of ventilation, work processes and personal protective equipment will also be requested.

The integration of facility and production data, complemented by historic IH data, comprises the overall exposure experience of the workforce that will inform the epidemiology study.

of 125


6.5. Task 3 - Cohort Follow-Up

The cohort’s vital status will be determined as of December 31, 2008. The UPitt investigators will utilize customized country-specific vital status tracing protocols, as detailed below.

6.5.1. United States

Study members for US plants with unconfirmed vital status (not known from company-held records to be alive or dead as of the vital status end date (12/31/2008)) will be entered into the standard UPitt two-stage vital status tracing protocol to identify deaths among cohort members (Buchanich et al., 2005).

Phase 1 consists of sending all the names of cohort members not known to be alive to the Social Security Administration (SSA). The SSA identifies living, unknown and deceased vital status. Phase 2 consists of sending subjects identified as having died prior to 1979 to the health department of the state of death to obtain a death certificate. All death certificates will be coded to the underlying cause of death by a U.S. National Center for Health Statistics nosologist using the International Classification of Diseases (ICD) rules in effect at the time of death. This method of coding, combined with the use of standard death rates that are also specific to the corresponding time period, ensure the comparability of cause of death categorization over time.

Subjects identified as unknown or deceased after 1978 will be sent to the National Death Index - Plus (NDI+) to obtain the coded cause of death. NDI-Plus became available to researchers in 1997 and provides the underlying and multiple causes of death for deceased study subjects. These searches will be supplemented, as necessary, by commercial tracing agencies (e.g., Lexis-Nexis).

6.5.2. Austria

Vital status tracing systems in Austria are in electronic form at the pension insurance bureau and Statistik Austria from 1970 to the present. Employees will be matched to the national systems using their unique identification system and date and cause of death will be abstracted for matches. Employees not known to be alive or dead from January 1, 1950 through January 1, 1970 will be manually searched through local registries to determine date and cause of death.

6.5.3. Germany

A data protection contract between the University of Cologne and site plants will be negotiated and signed. Based on this contract, the data protection commissioner will send a list of retired workers with names, date of birth and last address to the University of Cologne for determination of vital status and date and cause of death for deceased persons. These data are then sent back to the data commissioner who imports the data into the HSA,

of 125


where the relevant PID will be matched to the open identifier. The German investigators then merge together the datasets using the PID.

6.5.4. Sweden

Cohort members will be traced using current data lists and previous staff registries, which contain information on time of birth and period of employment. The social security numbers can be determined from plant records or with the help of the Swedish regional social insurance office via the Swedish Central Statistics Agency (SCB) and church records in the respective parishes. Employee social security numbers will be matched to the national death registry for the entire period. Death certificates will then be obtained from the SCB. Employees with foreign citizenship during their employment or those who emigrated will be excluded from evaluation.

6.5.5. United Kingdom

In the UK, the Information Centre for Health and Social Care (IC) can arrange for bona fide researchers to be given copies of death certificates for deaths. The IC is a recently created body that has taken over many of the roles in medical research previously carried out by the Office for National Statistics (ONS). Death certificates will be supplied with underlying causes and multiple causes coded according to the ICD revision in effect at time of death. It will not be necessary to get written permission from each living study subject as regular updates of death are not required for this study. It will be necessary, however, to obtain the approval of a local Research Ethics Committee to obtain the details of earlier deaths.

6.6. Task 4 - Integration of Exposure Assessment Data

This task involves the integration of the exposure assessment data to be developed by UIC. Details of the independent exposure assessment component of the proposed study are provided in Section 5. UPitt investigators will work closely with UIC investigators to coordinate the integration and verification of data from the epidemiological and exposure assessment components of the study.

This will involve linking the job/exposure matrix created by UIC investigators with the individual WHs and epidemiological data compiled by the UPitt investigators (detailed in Sections 5.5.4. and 5.5.5.). UPitt will also serve as a clearing house for all exposure assessment-related study data, and will be responsible for maintaining an inventory and ultimately an archive of all project data.

6.7. Task 5 - Statistical Analyses

6.7.1. Overview of General Approach

As noted, the study sites are highly diverse relative to geographic location, cohort size and cohort entry period (facility start date in most cases). Because of this heterogeneity, we propose to approach the statistical analysis in a site-specific manner, pooling data across

of 125


sites only if warranted by evidence of sufficient homogeneity. The advantage of such diversity from an epidemiological standpoint is the ability to assess the consistency of our findings across the study populations. Efforts will be made to pool data when warranted, however, as this will improve the precision of the mortality risk estimates and increase the statistical power to detect epidemiologically important excess risks.





Part 2 will include multi-variate analyses of internal cohort rates to facilitate the simultaneous adjustment for multiple potential confounding factors and the assessment of effect modification among two or more study factors. In particular, Part 2 will include relative risk regression modeling (based on Cox proportional hazards model) of internal cohort rates.


To provide the most unbiased assessment of lung cancer risk possible from the available study data, we have included in Part 2 a nested case-control study of lung cancer in the AT, SE, UK and US sites. In this study, we will make an intensive effort to obtain the most complete and accurate individual worker-level data on tobacco smoking habits for all cases (deaths) of lung cancer and corresponding groups of non-cases (controls) selected from the remaining cohort members. In the DE sites, we will control for potential confounding by smoking via external adjustment.

While we plan to collect smoking information to the extent possible on all study members, these data may be incomplete. By performing this adjustment in the case-control setting we

of 125


will be more likely to have complete data on smoking as most of the cases and corresponding controls will fall into the later time periods when these data are more complete. The statistical analysis of the case-control data will involve relative risk regression modeling of the matched sets with adjustment for potential confounding by smoking and co-exposures to several known or suspected carcinogens.

The following sections describe details of our statistical analysis plan.

6.7.2. Descriptive Analysis of General Mortality Patterns

6.7.2.1. Construction of Basic Person-Year Arrays

We will examine the total and cause-specific mortality experiences of the site-specific cohorts from the cohort entry date (Table 2) to December 31, 2008. Using an OCMAP-PLUS (Marsh et al., 1998) modified life table procedure, we will jointly classify the person-years at risk contributed by each study member by the primary study factors: study site, race, gender, follow-up period, age group, year of hire, duration (years) of employment (DOE), the time (years) since first employment (TSFE) and worker type (short-term, long-term).

The final categorization of these factor-specific person-year arrays will be determined upon examination of the actual data. For certain sites, we may need to group person-years into broader categories due to the smaller cohort sizes. For workers whose vital status cannot be traced, person-year counts will be accumulated until the last point of verifiable alive status, which is usually the date of termination from employment. Person-years relative to DOE and TSFE will begin on the starting date of the first job in the WH (date of hire) or January 1, 1950 for Reutte, AT employees hired before that date.

6.7.2.2. Analysis by Worker Type

The purpose of our proposed analysis by worker type is to analytically separate the mortality experience of workers with short overall employment histories from those having long-term employment histories. Short-term workers are often more transient and may have behavior and lifestyle characteristics associated with less favorable overall mortality patterns. They may also have unique or higher occupational exposures to the agents of interest due to working in less desirable entry-level positions. By design, this separation by worker type provides the opportunity to examine mortality patterns among long-term workers unconfounded by behavior, lifestyle or exposure patterns unique to short-term workers.

We propose to examine the mortality patterns of short-term workers using two different approaches. While the actual cutpoint used to dichotomize workers into short and long-term categories is essentially arbitrary, a point between one to five years is often used in studies of this type with the ultimate point depending in part on the distribution of the cohort by overall DOE.

of 125


In the first, more conventional prospective approach, we will analyze short-term workers by stratifying the person-years of the total cohort by DOE as noted above, using the short vs. long-term worker cutpoint as the upper bound of the first DOE interval. Here, the person-years of the workers whose total DOE is less than the cutpoint are combined with the short-term portion of person-years contributed by the long-term workers. Because this cohort is primarily an incidence cohort (dates of hire and cohort entry equivalent), the long-term workers will contribute person-years to the short-term category. In effect, this analysis focuses on the mortality experience of all workers during the short-term portion of their employment history. Also, here the short-term workers can serve as the baseline category for the remaining long-term categories.

In the second, retrospective approach to our worker type analysis, we will exclude the short-term portion of the person-years for long-term workers from the short-term workers person-year counts. For long-term workers, person-year counts will begin at the short vs. long-term worker cutpoint. In effect, this analysis of long-term workers resets the starting date for person-year counts from the cohort entry date to the date when the cutpoint is reached. In contrast to the first approach, this analysis focuses on the mortality experience of short-term workers only as a distinct subcohort of the total study population. As such, this group cannot be used as a baseline category for the long-term workers, and we will need to interpret our results considering the fact that they are not entirely prospective in nature.

6.7.3. Procedures to Account for Unknown Race

The EU sites do not routinely collect race nor are national death rates available by race so we will not analyze these sites by race.

We will collect information on race from the US sites but anticipate from our experience with past occupational cohort studies that this variable will be unavailable for some employees. Rather than making arbitrary assumptions about the race of the unknowns or eliminating workers with unknown race from the analysis, we propose to handle this common data deficiency analytically using a procedure unique to the OCMAP-PLUS software (Marsh et al., 1998).

In our analytic approach, called the Proportional Allocation Method (PAM), the person-years and observed deaths accumulated by study members of unknown race are assigned to the white or nonwhite categories in proportion to the distribution of study members with known race. For example, if a cohort with 1,300 person-years included 800 white person-years, 200 nonwhite person-years and 300 person-years with unknown race, 800/1000 or 80% of the 300 unknown race person-years (240) would be assigned to the white category, and 200/1000 or 20% (60) would be assigned to the nonwhite category resulting in an estimated total race-specific person-year distribution of 1040 white and 260 nonwhite.

Also to help reduce bias, the PAM will be performed within strata defined by study site, gender, age group and time period. The resulting total number of assigned white and nonwhite person-years and observed deaths will then represent weighted averages of the stratum-specific assignments. The stratification will not be extended to other study variables such as DOE or TSFE to avoid sparse data problems.

of 125


While the PAM estimation procedure does not enable the imputation of race data for individual study members, it does permit the person-years and observed deaths of study members with unknown race to be included in the grouped external mortality comparisons. This is done under the reasonable assumption that the distribution of whites and nonwhites in a given stratum of workers with unknown race is equivalent to the corresponding distribution based on known race.

6.7.4. Calculation of Internal Cohort Rates, Expected Numbers of Deaths and SMRs

We will compute expected numbers of deaths by multiplying the person-years at risk in the factor-specific categories by the corresponding mortality rates in the appropriate general populations. For the US sites, we will compute expected numbers of deaths using as standard populations the total US and the local study site areas (defined as counties or aggregates of counties) from which the associated workforces were largely drawn.

Our US mortality rate files are available from the earliest observation time (1952 – the start of the Bedford facility) through 2005. Population-weighted county rates will be obtained from the Mortality and Population Data System (MPDS) developed and maintained by the investigators at UPitt (Marsh et al., 2005). MPDS county rates are currently available for cancer categories from 1950 through 2005 and for noncancer categories from 1962 through 2005. For the sites which started operations before 1962 (Bedford - 1952, Fallon - 1960 and Orwell - 1960), we will apply 1962-1964 rates to the1960-1964 period. We anticipate that our US and local rate files will be current through 2007 by the time of this analysis.

For the EU sites, we will compute expected numbers of deaths based on the associated national and regional rates. These requisite rates are available and will be obtained by the country-specific investigators from the appropriate national or regional government vital statistics offices. Modifying our cause of death categories somewhat to account for US-EU differences in cause of death categorizations may be necessary. Our MPDS system is entirely flexible in this regard and will enable us to form compatible US-EU rates.

Because local area death rates usually provide the most valid available mortality comparisons [as they tend to adjust for the social, cultural and economic factors related to disease, including smoking to some extent (Doll, 1985)], the analysis of general mortality patterns will focus primarily on the local county or regional comparisons.

For the sites in Germany, we will use external smoking-specific reference populations, such as those developed in the US by the American Cancer Institute, to control for potential confounding by smoking. For the AT, SE, UK and US sites, we believe that adjusting the lung cancer risks associated with WCCo exposure for potential confounding by smoking can be done more effectively and with greater statistical efficiency in the context of a nested case-control study (described below).

We will express mortality excesses and deficits as Standardized Mortality Ratios (SMRs), that is, the ratio of observed numbers of deaths to expected numbers of deaths. SMRs for total and cause-specific mortality will be computed for the subgroups of the total cohort defined by study site, race (US), gender, age group, follow-up period, year of hire, DOE, TSFE and worker type. Where deemed appropriate via homogeneity analysis, SMRs will also be computed for two or more study sites combined. Appendix B provides the cause of

of 125


death categories and corresponding revision-specific International Classification of Diseases (ICD) Codes that we will use in the mortality analysis.

For our descriptive analyses of SMRs, deviations of the SMR below and above 1.00, indicating deficit and excess mortality risks, respectively, will be identified using exact Poisson probabilities described by Breslow and Day (1987). For all of our analyses, two-tailed p-values less than 0.05 will be deemed statistically significant and those greater than or equal to 0.05 and less than 0.10 will be deemed marginally statistically significant. No formal probability adjustments will be made for the multiple statistical comparisons to be performed in the analysis (although this will be considered in the interpretation of the findings) or for the PAM estimation procedure used to reallocate unknown race person-years and observed deaths.

6.7.5. Analysis of Mortality in Relation to Exposure to WCCo and Other Substances

Using the historical exposure data estimated for WCCo and other agents in the UIC exposure assessment phase of the proposed study, we will use OCMAP-PLUS to compute various time-dependent summary measures of exposure for individual workers. These summary measures, which will serve as the exposure variables in our exposure-response analysis, are described in detail below.

For exposure agents characterized quantitatively or qualitatively, we will compute the basic measure “duration of exposure.” The duration of exposure measure for a given agent (termed Agent_Dur) computed over Nj jobs during exposure period j can be expressed as

where Time i and Exp i represent DOE and exposure level, respectively, of the ith job in the WH during exposure period j. The DOE measure considered in the general mortality analysis can be viewed as a special case of this duration of exposure measure where Exp i = 1 for all jobs in the WH.

For exposure agents measured quantitatively, the time-weighted cumulative exposure (termed Agent_Cum) is defined analogously as

where Exp i is the quantitative exposure value for the ith job.

For quantitatively measured exposure agents, we will also consider a third summary exposure measure, average intensity of exposure. This measure can be computed separately or in conjunction with duration of exposure and/or cumulative exposure. The average intensity of exposure (termed Agent_AIE), computed over Nj jobs during exposure period j, is calculated by dividing Agent_Cum by Agent_Dur or

of 125


Agent_AIE represents the average intensity of exposure during periods when the worker is exposed to that agent. The default exposure period j described in the above expressions is the total employment history (i.e., the time interval from date of hire to date of termination, accounting for gaps such as sick leaves, layoffs, strikes, etc.).

The interpretation of internal rates or SMRs relative to the exposure measures described above relies on the implicit assumption that mortality during a given observation period is related to the duration, cumulative or average intensity of exposure received from hire date up to the point of observation for mortality. For some agents, including WCCo, the actual effectiveness to cause disease may change over time, and may be negligible during parts of the employment history (e.g., for cancers, the time between initiation of tumor growth and death) (Caplan et al., 1984). To enable some adjustment for this changing effectiveness, we may also examine a weighted or lagged form of the above summary measures. The choice of the lagging period depends on known or hypothesized exposure-response and latency (time from first exposure to development of disease) characteristics of the agent and disease under study.

For example, a reasonable lagging period for examining the WCCo exposure-lung cancer relationship is five years, because WCCo exposures received five years before the date of death probably did not impact on the mortality risk (i.e., assuming that on the average, a period of five years elapses between formation of tumor, diagnosis and ultimately death). Lagging can also be used in conjunction with exposure time “windows”, where the “effective” exposure is calculated during a pre-determined window of time that lags the risk period a pre-determined number of years. Lagging with and without exposure time windows was used by UPitt investigators in a recent historical cohort study of man-made vitreous fiber workers (Youk et al., 2001).

In addition to and with the summary measures described above, we will relate mortality risks to the time-dependent variable, time since first exposure the agent(s) of interest (termed Agent_TSFExp). Agent_TSFExp will be computed from the date of first exposure until the first occurrence of the date of death or the end of the study period. Unlike the variable TSFE used in the general mortality analysis, Agent_TSFExp discounts the irrelevant elapsed time from date of hire until the date of the first exposure. For a given agent, the non-time-dependent variables age and calendar time of first exposure will also be considered alone and in combination with the summary exposure measures described above.

6.7.6. Descriptive Analysis

One objective of this analysis is to examine descriptively, using externally standardized (local comparison) SMRs, the relationship between exposure to WCCo and other agents and total and cause-specific mortality, with focus on cancer and emphasis on lung cancer. Although the relative risk regression analysis described below is more appropriate for the analysis of multiple exposures and/or potential confounders, the SMR-based, exposure-related analysis provides descriptive information on observed deaths relative to the same exposure measure categorizations used below in the relative risk regression analysis.

of 125


Using a modified life table algorithm in OCMAP-PLUS, we will jointly classify person-years at risk contributed by each study member by study site, race (US), gender, calendar time period, age group and categories of one or more of the summary exposure measures computed for WCCo and other agents as described above. For a given summary measure and cause of death category, we will categorize deaths using approximate percentiles (e.g., quartiles or quantiles) of the distribution of deaths according to the summary measure. The baseline category will include the person-years of strictly unexposed workers as well as the person-years of exposed workers up to the time of their first exposure. We will compute SMRs for the various categories of the summary exposure measures as described in the general mortality analysis section.

6.7.7. Relative Risk Regression Modeling

Relative risk (RR) regression modeling will be used to investigate the dependence of the internal cohort rates (modeled as time to death) for lung cancer on combinations of the WCCo exposure measures, with adjustment for potential confounding factors and potential co-exposures. Risk sets will be explicitly constructed from the cohort data file with age as the primary time dimension, using the RISKSET program module in OCMAP-PLUS (Marsh et al., 1998). Risk sets will be matched further on gender and year of birth to control for cohort effects. For the exposure-response analyses, the time-dependent exposure measures of interest will be evaluated for each individual at each event time they were at risk (the date of death of the case).

Multiplicative relative risk models of the form λ(t) = λ0(t) exp{x(t)β} will be fit to the internal cohort rates (Cox, 1972; Cox, 1975), and the conditional logistic regression program in STATA (STATA Corp., 2007) will be used to estimate β from the explicitly constructed risk sets. To parallel the descriptive SMR analysis of mortality in relation to exposure, categorized forms of the covariates will be considered. The statistical significance of each main effect (expressed as a global p-value) will be assessed with a likelihood ratio statistic. For the quantitative exposure variables (both WCCo measures and co-exposures) that exhibit a monotonically increasing or decreasing pattern in the parameter estimates, a test for linear trend will be conducted (expressed as a trend p-value). Joint effects of the WCCo exposure measures and the co-exposures as well as effect modifications will be assessed to the extent possible.

To elucidate complexities in possible exposure response relationships, we will attempt to perform analyses relaxing the assumed linearity in dose-response by modeling the effects of exposure using piecewise linear functions (linear splines). Knots will be defined at tertiles (or quartiles) of the lung cancer case distribution for each of the exposures. Splines defined in this manner will be used to assess possible interactions of the exposures with other potential confounders which allow for separate non-linear exposure-response relationships for levels of the confounders.

Additional analyses may also include the assessment of adequate functional forms for exposures using fractional polynomials. All possible exposure response models will be fit including powers of the “primary” exposure, adjusting for the other exposures; the “best” model at each level of complexity will be identified from a direct search based on the residual deviance, and compared to a model assuming a linear functional form.

of 125


To account for potential collinearities among the WCCo exposure and co-exposures (W, WC, Co and C), orthogonal polynomials can be defined based on a modified Gram-Schmidt procedure (Phillips & Esmen, 1999). Orthogonal polynomials can be used to assess the significance of variables with the effect of the collinear variable “removed”. In this way, the contribution of each exposure that is uncorrelated with the other can be determined.

Parallel to the descriptive SMR analyses, 1we will also explore a possible exposure-response relationship between lung cancer and exposure to each exposure of interest utilizing exposure-weighting as an alternative characterization of exposure. Due to the uncertainties in selecting an appropriate exposure-weighting scheme, a range of plausible time lags and unlagged/lagged time windows will be considered.

6.8. Controlling for Confounding by Smoking

The importance of potential confounding by smoking in studies of lung cancer is indisputable. We propose to address the likelihood and extent that confounding by smoking may contribute to the results of the lung cancer analyses in this study in two ways, via external and internal adjustment.

6.8.1. External Adjustment for Confounding by Smoking

For the facilities in Germany, we propose to adjust our lung cancer risk via external adjustment. Without smoking information directly on the workers or from the sites, an estimate of the possible confounding due to smoking can be found using estimates of the smoking prevalence in the WCCo exposed and unexposed workers, the original unadjusted lung cancer risk due to WCCo and an estimate of the lung cancer risk due to smoking in the general population.

We will attempt to identify peer-reviewed studies and occupational exposure surveys providing information on smoking patterns among WCCo-exposed workers that are independent of our epidemiological study. We will also identify comparison smoking prevalence information for the respective general populations of the study using peer-reviewed studies, government reports and other sources.

We will accomplish the external adjustment for smoking using the methods described by Steenland and Greenland (2004). Their Monte Carlo based strategy is an extension of indirect confounding methods proposed by Axelson and Steenland (1988). An unadjusted risk estimate can be adjusted by a bias factor to obtain an externally adjusted risk. This bias factor reflects the possible confounding of the unadjusted risk by a related factor like smoking.

6.8.2. Internal Adjustment for Confounding by Smoking

For the AT, SE, UK and US sites, we will conduct a nested case-control study to obtain smoking information on the cases and controls using an interview with the case or the next-

of 125


of-kin of the case (when the case is deceased). Details regarding the country-specific methods of locating, contacting and interviewing knowledgeable informants are under development and will be finalized as this portion of the Phase 3 study progresses.

6.8.2.1. Matching cases to controls

The nested case-control study of lung cancer can be considered as a special case of the corresponding relative risk regression analysis described above. That is, for each case (death) due to lung cancer identified in the cohort, we will randomly select a group of matched controls from the corresponding risk set formed for the relative risk regression analysis. As noted in that section, the non-case members of each risk set are matched on the event (death) age of the case, gender and year of birth. We will select two controls for each case.

6.8.2.2. Obtaining data on cases and controls

To obtain more detailed or otherwise unavailable information on potential risk factors for the cases and controls, an attempt will be made to locate and interview a knowledgeable informant, ideally the worker himself or a surviving member of the worker’s immediate family (proxy respondent). The potential informant will be found from plant record information or by the same tracing sources used to determine the vital status of the study cohort. To obtain a minimum of two controls for each case, it is estimated that approximately four cohort members per case (or their next of kin, if deceased) will need to be contacted.

Before the contacting and interviewing processes for the case-control study can begin, it will be necessary to obtain approval from the National Death Index to allow access to the NDI-Plus system and the use of personal information on the death certificate (e.g., address of next of kin) for follow-back purposes. As part of securing this approval, it will also be necessary to clear this research proposal through the UPitt Institutional Review Board (IRB), a group that ensures the protection of research subjects. IRB approval will also be required by some state health departments before they will release copies of death certificates, especially for studies involving followback.

Another study protocol issue relates to obtaining the informed consent of study members (or proxy respondents) before they are interviewed for the case-control study (note: consent of individual workers can be waived by the IRB at the cohort level). It is proposed that a country-specific introductory letter be sent to potential informants to announce the purpose and importance of the study and introduce the investigators. To enhance participant response, the letter should indicate that the proposed study is supported by the companies as well as the local union(s). A consent form will be included with the approach letter, requiring informants to indicate their willingness to participate by signing the appropriate line on the form.

Potential informants will be instructed to return the consent form within two weeks of the date of the letter indicating their willingness to participate. At the end of this waiting period, willing informants will be contacted by a professional interviewer for a brief (20-30 minute) telephone interview concerning such items as the subject’s smoking, non-

of 125


occupational and occupational history. To eliminate potential biases, the interviewer will not be aware of case-control status when making the telephone calls. A questionnaire will be mailed, with a pre-addressed postage-paid return envelope, to potential informants with a known address but no available phone number.

The case-control study questionnaire proposed for this study will be based upon the one already used in a nested case-control study of lung cancer among workers with potential exposure to man-made vitreous fibers. We will modify the existing instrument as necessary to capture specific risk factors of interest for WCCo workers.

16.8.2.3. Statistical analysis of nested case-control study

The statistical analysis of the case-control data will be conducted along the lines of the relative risk regression modeling described in above, but will also include the additional data on risk factors that will be obtained from the interviews of cases or their proxy respondent and the matched controls.

6.9. Statistical Power Characteristics

A primary objective of our proposed cohort study is to evaluate mortality risks with emphasis on malignant neoplasms, in particular, lung cancer. Thus, an assessment of the study’s statistical power to detect important cancer excesses is an important consideration. Because cancer in general and specifically lung cancer are both highly prevalent diseases (especially in older males), we are confident that we will obtain an adequate number of cases for sufficient statistical power to detect a true 1.5 or greater mortality risk for all cancers combined and lung cancer. We estimated power using:

1) estimated cohort size of 21,000 individuals (≈8,500 US and ≈12,500 EU)

2) estimated range of average follow-up time of 25-30 years

3) US age-adjusted rates for 1980 white males aged 25+ years (all cancer – 407/100,000; lung cancer – 130/100,000)

Based on these calculations, we estimate between 2,140 and 2,450 all cancer deaths, of which 680-800 will be due to lung cancer.

6.9.1. Cohort study

Based on the above numbers, the power of detecting a true 1.5-fold or larger excess in total cancer and lung cancer deaths is essentially 1.0. We recognize that the statistical power will be less to detect excesses within separate plants or other relevant study subgroups, however, our estimated number of deaths includes a surplus number of deaths that will enable us to achieve good to excellent power in most of the larger subgroups examined.

of 125


6.9.2. Case-control study

For our nested case-control study of lung cancer, the power of detecting a true 1.5-fold or larger excess risk is greater than 95% based on 1:1 case:control matching and assuming the proportion of controls exposed is 30%. The large number of lung cancer deaths we expect to observe in the cohort study includes a surplus number of deaths sufficient for maintaining good to excellent statistical power while accounting for the additional sample size requirements needed to assess confounding and effect modification.

6.10. Strengths and Weaknesses of the Proposed Study

The strengths of the proposed study can be summarized as follows:

1. The study represents the joint efforts of a nationally and internationally recognized occupational health research team with more than 30 years of experience designing and conducting historical cohort and case-control studies of the types proposed herein, including the largest occupational cohort, incidence and case-control study of brain cancer ever conducted.

2. The UIC and UPitt investigators share a long history of highly successful and productive collaborative research, as evidenced by the many jointly authored peer-reviewed publications cited in their respective curricula vitae.

3. The proposed historical cohort study design and nested case-control study of lung cancer, which will enable analytic adjustments for smoking and co-exposures to known or suspected carcinogens, will provide the best available estimates of total and cause-specific mortality risks among workers, overall, and in relation to occupational WCCo exposure.

4. A particular strength of this study is the ability to characterize the working history of study members relative to WCCo exposure and co-exposures to other known or suspected carcinogens. The UIC exposure assessment team plans to use a variety of statistical and other estimation methods to assign meaningful and scientifically defensible exposure estimates to workers over time.

5. Based on our estimates, the statistical power of the proposed cohort study to detect epidemiologically important (1.5-fold or greater) overall excesses in all cancers combined and in lung cancer is essentially 1.0. The study affords a sufficiently long time period for potential WCCo exposure and a sufficiently long observation period to observe cancer outcomes in relation to WCCo exposure. Therefore, our proposed study is able to detect a true increased risk or to conclude that there is no increased risk if one is not detected. Additionally, because the number of expected cancers is so large, the power necessary to detect 1.5 fold or greater risks in larger subgroups of the cohort will also be in the good to excellent range.

6. Also based on our independent estimates, the power of the nested case-control study of lung cancer to detect epidemiologically important (1.5-fold or greater) excesses in relation to key study factors is in the excellent range (95% or greater).

of 125


This focused sub-study will also allow adjustment for potential confounding by smoking in addition to co-exposures to known or suspected carcinogens.

7. The addition of Co production companies to the study would be a large advantage of the Phase 3 study over previous studies of WCCo workers. However, the inability to include subjects with Co-only exposure does not represent a fatal flaw to the main study design. The Phase 3 WCCo worker epidemiology study will be comprehensive, scientifically sound and far superior overall to the existing epidemiology studies. The limited orthogonal contrasts for Co and WC simply suggests that we may be unable to sort out completely the contribution of each exposure alone if increased lung cancer mortality risks are observed among subjects with combined exposures.

8. The historical cohort study can form the basis for ongoing mortality surveillance of the workers with potential occupational exposure to WCCo.

The weaknesses of the proposed study can be summarized as follows:

1. Although it appears from the Phase 1 and Phase 2 studies that much of the data necessary to adequately address the question regarding WCCo exposure and lung cancer exists, we may find that some records are unavailable. We believe, however, that we have taken every reasonable approach to resolve or work around these limitations in designing the proposed study.

2. Although the cohort study has good to excellent statistical power to detect important excesses in mortality from all cancers combined and lung cancer in the total cohort and its larger subgroups, the power will be less to detect excesses in smaller subgroups of the total study population, such as factor-specific groups within study site. The same feature will also apply to the case-control study.

3. Because a large and diverse number of agents are used or produced at the plants, it may not be possible to identify the specific etiologic agent or combination of agents if an overall excess in cancer mortality is found. This is a limitation of all epidemiology studies of this type that attempt to relate mortality outcomes to specific chemical exposures. We have, to the extent possible, designed all aspects of our study to offset or mitigate this limitation.

4. The use of external adjustment for confounding by smoking in the German sites has some limitations. Although we will attempt to estimate the patterns of smoking among WCCo exposed workers from information available in the peer-reviewed literature, there remains the chance that these estimates are not truly representative of the smoking prevalence in the worker cohorts. The smoking rates will most likely be based only on one time period which would not reflect temporal changes in smoking habits which can occur, particularly over the long duration of an historical cohort study. We may also be limited by the extent to which detailed smoking information is available in the literature.

On balance, we believe that the strengths of the proposed study outweigh its weaknesses. We are also confident that our investigation will produce scientifically sound, meaningful

of 125


and defensible results, and will provide a significant contribution to the body of knowledge concerning the health implications of exposure to WCCo.

6.11. Qualifications of the Research Teams

6.11.1. University of Pittsburgh

For more than 45 years, the UPitt, Department of Biostatistics (BIOS) has been one of the leading academic centers of occupational and environmental health research in the United States. BIOS faculty have been active in the development and application of biostatistical methods to study potential health effects of workplace exposures in a variety of industrial settings.

The Center for Occupational Biostatistics and Epidemiology (COBE) was established in February 2008 as a specialty research center within the Department of Biostatistics (BIOS) in UPitt’s Graduate School of Public Health (GSPH). The mission of the COBE is to build further upon existing departmental strengths in occupational biostatistics and epidemiology, to enhance collaborative research across departments and schools at UPitt, to promote both national and international recognition of these fields of strength and to increase opportunities for external collaboration and programmatic funding.

Gary M. Marsh, Ph.D., F.A.C.E., Professor of Biostatistics is the Director of the COBE; Jeanine M. Buchanich, M.Ed., Ph.D., Research Assistant Professor of Biostatistics is Deputy Director of Epidemiology and Ada O. Youk, Ph.D., Research Assistant Professor of Biostatistics is Deputy Director of Biostatistics. COBE research and administrative staff include two master’s level biostatisticians, two master’s level information science specialists, one computer programmer, two research specialists and four technical/clerical support staff including graduate student researchers and a professional telephone interviewer.

The UPitt group has conducted occupational studies to investigate the long-term health effects of exposure to such agents as man-made mineral fibers, formaldehyde, acrylamide, acrylonitrile, arsenic, petrochemicals, aromatic amines and pharmaceuticals. They have also applied their expertise in occupational epidemiological research to environmental epidemiologic studies of communities exposed to industrial pollutants or to hazardous waste site materials.

Currently, the UPitt group is conducting an historical cohort study of nearly a quarter million jet engine manufacturing workers for the Pratt & Whitney Company, a mortality surveillance system for the Owens Corning Company and an historical cohort study of pharmaceutical workers for the Eli Lilly Company. The Pratt &Whitney study is a collaborative effort with the Department of Neuro-Oncology within UPMC and UIC.

6.11.2. University of Illinois at Chicago

UIC’s Environmental and Occupational Health Sciences (EOHS) Division is housed within the School of Public Health and its goal is to protect the environment and improve the

of 125


health of workers and the general public. The Illinois Occupational and Environmental Health and Safety Education and Research Center (Illinois ERC) was established in 1977 as one of the first National Institute of Occupational Safety and Health (NIOSH) Educational Resource Centers in the US.

The Division has a nationally recognized program in IH that is accredited by the Accreditation Board for Engineering and Technology. Within the division, the research carried out by the occupational epidemiology team includes mathematical modeling, engineering, industrial and environmental field studies and general occupational hygiene. In addition to the mathematical, engineering, aerosol physics and industrial hygiene expertise, the team can draw upon the medical expertise within the division. Currently, the UIC team is engaged in several projects, three of which are in collaboration with the University of Pittsburgh team. Before the formation of the UIC team, the collaboration in large scale industrial epidemiology studies between the UPitt and UIC principal investigators started in 1975 and has continued more or less unbroken ever since.

Nurtan A. Esmen, Ph.D., FAIHA, FRSH, is a Professor of EOHS; Steven E. Lacey, Ph.D., CIH, CSP, is an Assistant Professor of EOHS and is Director of the Illinois ERC Industrial Hygiene Program. The UIC team also includes a project manager and a senior research scientist.

of 125


7. Phase 3 Study - Administrative Details

7.1. Project Structure

UPitt will serve as the main contractor and will provide the epidemiology and biostatistics component of the proposed cohort mortality study of workers potentially exposed to WCCo. UIC will serve as a subcontractor to UPitt and will direct the exposure reconstruction component. The work of the country-specific investigators and the data processing subcontractor will also be overseen by UPitt.

7.2. Period of Performance

The work proposed herein for the Phase 3 study is anticipated to take approximately four to five years to complete. Part 1 of the Phase 3 study is anticipated to begin in November 2008 or following receipt of pending funds from the State of Pennsylvania Department of Health (PADOH). Subsequent parts will be started as funding becomes available.

7.3. Investigators

The UPitt contract will be directed by Gary M. Marsh, Ph.D., Professor of Biostatistics, who will be Principal Investigator. Jeanine M. Buchanich, Ph.D., Research Assistant Professor of Biostatistics and Ada O. Youk, Ph.D., Research Assistant Professor of Biostatistics will be Co-Investigators with Dr. Marsh. The UIC subcontract will be directed by Nurtan Esmen, Ph.D., Professor of Environmental and Occupational Health Sciences, and Steven Lacey, Ph.D., Assistant Professor of Environmental and Occupational Health Sciences, will be Co-Investigator with Dr. Esmen. The investigators’ biosketches investigators are available in Appendix C.

7.4. Progress Reports

Reports will be issued to the sponsor annually and at the end of the project. Investigators will also prepare manuscripts for publication in peer-reviewed journals throughout the project period.

7.5. Data Confidentiality

The Principal Investigators will maintain the confidentiality of all hard copy and electronic records, assuring that information is kept in locked files (and password protected electronic media files) and that persons working with these records are made aware of their

of 125


confidentiality. A Confidentiality Statement Form will be executed and kept on file for each individual working with these records. All UPitt staff will also sign the required NDI-Plus Supplemental Assurance Form to ensure the confidentiality of NDI-Plus vital status and cause of death data.

According to university guidelines, upon funding of the proposed research project this protocol will be submitted to the respective IRB’s of UPitt and UIC.

7.6. Quality Assurance and Quality Control Features of Proposed Study

7.6.1. Good Epidemiology Practices Guidelines

During the preparation of this proposal and the performance of the contract, UPitt and UIC have/will adhered/adhere strictly to the Good Epidemiology Practices Guidelines (GEP) described by Cook (1991). The GEP provides guidance regarding protocol development, reporting, quality assurance of data maintenance and documentation of analytic procedures.

7.6.2. Metrics Used to Ensure Data Quality and to Determine Progress

Standard UPitt data checks will be used throughout the proposed study to ensure the integrity of all study data. This will include a formal check of the completeness and accuracy of the cohort enumeration and vital status tracing, as well as comprehensive audits of all primary study data coding and keying. For example, the coding and key-entering of all primary study data will be 100% by a second data coding clerk. UPitt staff will strive to achieve cohort completeness, follow-up and cause of death ascertainment rates of at least 95%.

of 125


8. Project Budget Justification

The proposed Phase 3 Study is anticipated to take approximately four to five years to complete. Part 1 of the Phase 3 study is anticipated to begin in November 2008 or following receipt of pending funds from the PADOH. Subsequent parts of Phase 3 will be started as additional funding becomes available.

The PADOH has stated their intention to provide UPitt with $670,000 in direct costs for Part 1 of Phase 3 of the full epidemiology study and they have received a detailed application from UPitt for these funds. The initiation of Phase 3 Part 1 is dependent upon the receipt of the funds. Subsequent parts of Phase 3 are contingent upon procuring additional funding sources and, at that time, the specific tasks and detailed budgets will be provided to the sponsor.

We estimate that the total costs of the Phase 3 study will be in the $4-5 million range (direct costs), depending on the total number of manufacturing sites included.

of 125


References Cited

Axelson O, Steenland, K. Indirect Methods of Assessing the Effects of Tobacco Use in Occupational Studies. American Journal of Industrial Medicine 1988:105-118, 1988.

Breslow NE, Day NE. The Design and Analysis of Cohort Studies. In: Statistical Methods in Cancer Research, Vol. II, International Agency for Research on Cancer, IARC Scientific Publications No. 82, Lyon, France, 1987.

Buchanich J, Dolan D, Marsh G, et al. Under ascertainment of deaths using Social Security records: A recommended solution to a little-known problem. American Journal of Epidemiology..162:193-194, 2005.

Caplan RJ, Marsh GM, Enterline PE. A generalized effective exposure modeling program for assessing dose-response in epidemiologic investigations. Comput Biomed Res. 16:587-596, 1984.

Cook RR: Overview of good epidemiologic practices. J Occup Med. 33(12):1216-20, 1991.

Cox DR. Regression models and life tables (with discussion). J R Stat Soc. 34B:187-220, 1972.

Cox DR. Partial likelihood. Biometrika 62:269-276, 1975.

Doll R: Occupational Cancer: A Hazard for Epidemiologists. Inter J Epidemiol. 14: 22-31, 1985.

Esmen NA: Retrospective industrial hygiene surveys. Amer. Ind. Hyg. Assoc. J. 40:58-65, 1979.

Esmen NA: Exposure estimation in four major epidemiologic studies in acrylonitrile industry Scand. J. Work Environ Health. 24:Sup2:63-70, 1998.

Esmen NA, Kennedy KJ, Hall TA, Phillips ML, Marsh GM. Classification of Worker Exposures. Chemico-Biological Interactions. 166:245-253, 2007a.

Esmen NA, Hall TA, Phillips ML, Marsh GM. Chemical process based reconstruction of exposures for an epidemiological study: I. Theoretical and Methodological issues. Chemico-Biological Interactions, 166:254-263, 2007b.

Esmen NA, Hall TA, Phillips ML, Jones EP, Basara H, Marsh GM, Buchanich JM. Chemical process based reconstruction of exposures for an epidemiological study: II. Estimated exposures to Chloroprene and Vinyl Chloride. Chemico-Biological Interactions, 166:264-276, 2007c.

Hogstedt, C. and Alexandersson, R.. Mortality among hard metal workers. Arbete Hälsa, 21, 1-26, 1990.

of 125


Lasfargues, G., Wild, P., Moulin, J.J., Hammon, B., Rosmorduc, B., Rondeau du Noyer, C., Lavandier, M. and Moline, J.J. Lung cancer mortality in a French cohort of hard-metal workers. Am. J. Ind. Med., 26, 585-595. 1995.

Marsh GM, Youk AO, Stone RA, Sefcik S, Alcorn CW: OCMAP-PLUS, A New Program for the Comprehensive Analysis of Occupational Cohort Data. J Occup Environ Med. 40:351-362, 1998.

Marsh GM, Ehland J, Sefcik S, Alcorn C. Mortality and Population Data System (MPDS). Pittsburgh, PA: University of Pittsburgh (Department of Biostatistics Technical Report), 2005.

Moulin JJ, Wild P, Mur JM, Fournier-Betz M, Mercier-Gallay M. A Mortality Study of Cobalt Production Workers: An Extension of the Follow-Up. American Journal of Industrial Medicine. 23:281-288, 1993.

Moulin, J.J., Wild, P., Romazini, S., Lasfargues, G., Peltier, A., Bozec, C., Deguerry, P., Pellet, F. and Perdrix, A. Lung cancer risk in hard metal workers. Am. J. Epidemiol., 148, 241-248. 1998.

Mur JM, Moulin JJ, Charruyer-Seinerra MP, Lafitte J. A Cohort Mortality Study Among Cobalt and Sodium Workers in an Electrochemical Plant. American Journal of Industrial Medicine. 11:75-81, 1987.

Phillips ML, Esmen NA. Computational Method for ranking task-specific exposures using multi-task time-weighted average samples. Ann Occup. Hyg. 43:201-213, 1999.

Phillips ML, Esmen NA, Costantino J. The Reliability of Multiple Regression and an Alternative Method for Extracting Task-Specific Exposure Estimates from Time-Weighted Average Data. App. Env. and Occup. Hyg. 16:1-10, 2001.

Schell J.D., MacNair D.J. and Tollerud, D.J.. Epidemiological investigation feasibility study prepared for the International Tungsten Industry Association. 2006.

STATA Statistical Software: Release 6.0. College Station, TX, STATA Corp., 2007.

Steenland K, Greenland, S. Monte Carlo Sensitivity Analysis and Bayesian Analysis of Smoking as an Unmeasured Confounder in a Study of Silica and Lung Cancer. American Journal of Epidemiology 160:384-392, 2004.

Wild, P., Perdix, A., Romazini, S., Moulin, J.J. and Pellet, F. Lung cancer mortality in a site producing hard metals. Occup. Environ. Med., 57, 568-573. 2000.

Youk AO, Marsh GM, Stone RA, Buchanich JM, Smith TJ. Historical Cohort Study of U.S. Man-Made Vitreous Fiber Production Workers III: Analysis of Exposure-Weighted Measures of Respirable Fibers and Formaldehyde in the Nested Case-Control Study of Respiratory System Cancer. J Occ Env Med 43(9): 767-778, 2001.

of 125


Appendix A

Phase 2 Survey Instrument

of 125


TUNGSTEN INDUSTRY PHASE TWO STUDY University of Pittsburgh Epidemiology Component

1. In what year did plant operations begin?

2. How many employees:

a. Currently?

b. At plant start up?

c. Peak employment period(s)?

d. When was (were) the peak employment period(s)?

e. How much turnover is there (i.e., on average, what percentage of the workforce terminates employment each year)?

3. Are detailed work history records available for all employees since the year that plant operations began?

a. If no, what type of employee records may be missing (terminated, retired)?

b. If no, for what years might records be missing?

4. In what format are the records?

a. Does the format vary by time period?

b. What format(s) is/are used for the time periods identified?

5. Do records have the following information?

a. Complete name?

b. Social security number (US)/Other unique ID (non-US)?

i. If non-US is answered “no”, is a corresponding identifying number used in this country for vital status tracing?

ii. What is this number called and how is it used?

c. Date of birth?

d. Gender?

e. Race or ethnicity?

of 125


University of Illinois at Chicago Exposure Reconstruction Component

1. What products or materials has the plant produced?

2. Has the plant ever produced or used the following materials?

a. Tungsten

b. Tungsten carbide

c. Other carbides (e.g., titanium, tantalum, niobium)

d. Cobalt

e. Nickel

f. Chromium

g. Powdered metals

h. Any other materials not mentioned

3. Were any of the following processes ever performed at the plant?

a. Metal powder production

b. Mixing, granulating, palletizing, green shaping, or pressing of metal powder

c. Hardmetal finishing, grinding, or sharpening

d. Welding, brazing, or sintering

e. Foundry operations

f. Surface treatment (e.g., plasma spray, chemical vapor deposition)

g. Any other processes not mentioned

4. Was industrial hygiene monitoring (e.g., air sampling) ever performed at the plant?

a. If yes, are these records accessible?

4. Does the plant maintain any engineering process records?

a. If yes, what types of information do the records contain?

of 125


5.

Appendix B

ICDA Defined Cause of Death Categories

of 125


Appendix ACauses of Death Categories and Revision-Specific International Classification

of Disease (ICD) Codes Used in Mortality Analysis

LabelCause of Death Category

6th & 7th Revision(1950-67)

8th Revision(1968-78)


10th Revision(1999+)

01 All Causes of Death 001-999 000-999 001-999 A00-Y89

02 Tuberculosis 001-019 010-019 010-018 A15-A19

03 All Malignant Neoplasms 140-205 140-209 140-208 C00-C97

04 Buccal Cavity and Pharynx 140-148 140-149 140-149 C00-C14

05 Digestive Organs and Peritoneum 150-159 150-159 150-159 C15-C26, C48

06 Esophagus 150 150 150 C15

07 Stomach 151 151 151 C16

08 Large Intestine 153 153 153 C18

09 Rectum 154 154 154 C20-C21

10 Biliary Passages and Liver Primary 155 155,156 155,156 C22,C24

11 Pancreas 157 157 157 C25

12 All Other Digestive 152,156,158,159 152,158,159 152,158,159 C17,C19, C23, C26,C48

13 Respiratory System 160-164 160-163 160-165 C30-C39

14 Larynx 161 161 161 C32

15 Bronchus, Trachea, Lung 162,163 162 162 C33-C34

16 All Other Respiratory 160,164 160,163 160,163,164,165 C30-C31, C37-C39

17 Breast 170 174 174,175 C50

18 All Uterine (female only) 171,172-174 180,181,182.0 , 182.9 179,180,181,182 C53-C55

19 Cervix (female only) 171 180 180 C53

20 Other Female Genital Organs (female only) 175,176 183-184 183-184 C51-C52,

C56-C58

21 Prostate (male only) 177 185 185 C61

22 Testis and Other Male Genital Organs (male only) 178,179 172.5,173.5,186

187 186,187 C60,C62-C63

23 Kidney 180 189.0,189.1,189.2 189.0,189.1,189.2 C64-C65

24 Bladder and Other Urinary Organs 181 188,189.9 188,189.3,189.4,189.8, 189.9 C66-C68

25 Malignant Melanoma of Skin 190 172.0-172.4172.6-172.9 172 C43

26 Eye 192 190 190 C69

27 Central Nervous System 193 191,192 191,192 C70-C72

of 125







28Thyroid Gland and Other

Endocrine Glands and RelatedStructures

194,195 193,194 193,194 C73-C75

29 Bone 196 170 170 C40-C41

30 All Lymphatic and Hematopoietic Tissue 200-205 200-209 200-208 C81-C96

31 Hodgkins Disease 201 201 201 C81

32 Non-Hodgkins Lymphoma 200, 202, 205 200, 202 200, 202.0, 202.1, 202.8, 202.9

C82, C83.0-C83.8, C84, C85.1-C85.9

33 Leukemia and Aleukemia 204 204-207 204-208 C91-C95

34 All Other Lymphatic and Hematopoietic Tissue 203 203, 208, 209

202.2, 202.3, 202.4, 202.5, 202.6, ,203

C88, C90, C96

35 All Other Malignant Neoplasms 165,191,197-199171,173.0-173.4

173.6-173.9195-199

171,173 , 195-199

C44-C47, C49,C76-C79,

C80, C97

36 Benign Neoplasm 210-239 210-239 210-239 D10-D36

37 Diabetes Mellitus 260 250 250 E10-E14

38 Cerebrovascular Disease 330-334 430-438 430-438 I60-I69

39 All Heart Disease 400-402,410-443390-398,400.1400.9,402,404

410-414,420-429

390-398,402,404410-429

I00-I02,I05-I09, I11, I13-I14,

I20-I28,I30-I52

40 Rheumatic 400-402,410-416 390-398 390-398 I00-I02,I05-I09

41 Ischemic 420,422.1 410-414 410-414 I20-I25

42 Chronic Disease of Endocardiumand Other Myocardial Insufficiency 421,422.0,422.2 424,428 424,428 I33-I41

43 Hypertension with Heart Disease 440-443 400.1,400.9,402404 402,404 I11,I13

44 All Other Heart Disease 430-434 420-423425-427,429

415-417,420-423425-427,429

I26.0, I27-I28, I30-I32, I42-I43,

I44-I52

45 Hypertension w/o Heart Disease 444-447 400.0,400.2400.3,401,403 401,403,405 I10, I12, I15

46 Nonmalignant Respiratory Disease 241,470-527 460-519 460-519 J00-J99

47 Influenza and Pneumonia 480-483,490-493 470-474,480-486 480-487 J10-J18

48 Bronchitis, Emphysema, Asthma 501,502,527.1,241 490-493 490-493 J40-J46

49 Bronchitis 501,502 490,491 490,491 J40-J42,J44

50 Emphysema 527.1 492 492 J43

51 Asthma 241 493 493 J45-J46

of 125







52 Other Nonmalignant Respiratory 470-475,500,510-527.0,527.2

460-466500-519

460-466 ,470-478,

494-496,500-519

J00-J06,J20-J22,J30-J39, J47,

J60-J70,J80-J86, J90-J99

53 Ulcer of Stomach and Duodenum 540,541 531-533 531-533 K25-K27

54 Cirrhosis of Liver 581 571 571 K70,K74

55 Nephritis and Nephrosis 590-594 580-584 580-589 N00-N29

56 All External Causes of Death 800-999 800-999 E800-999 V01-Y89

57 Accidents 800-962 800-949 E800-949 V01-X59

58 Motor Vehicle Accidents 810-835 810-823 E810-825 V01-V99

59 All Other Accidents 800-802,840-962 800-809,824-949 E800-807,E826-949 W00-X59

60 Suicides 963,970-979 950-959 E950-959 X60-X84

61 Homicides and Other External 964,965,980-999 960-978980-999

E960-978E980-999

X85-Y36,Y40-Y89

62 All Other Causes

020-138,206-207,240,242-254,270-326,340-398,450-468,530-539,542-580,582-587,600-

795

000-009020-136240-246251-389440-458520-530534-570572-577590-796

001-009020-139240-246251-389440-459520-530534-570572-579590-799

A00-A09,A20-B19,B25-B99,D00-D09,D37-D89,E00-E07,E15-G99,H00-H99,I70-I99,

K00-K23,K28-K67,K71-K73,K75-K93,L00-L99,

M00-M99,N30-R99

63 Acquired Immunodeficiency Syndrome (AIDS) (from 1987) not applicable not applicable 042-044, 795.8 B20-B24

of 125


Appendix C

Biosketches of Investigators

of 125


BIOGRAPHICAL SKETCHProvide the following information for the key personnel in the order listed for Form Page 2.

Follow the sample format for each person. DO NOT EXCEED FOUR PAGES.

NAME

Gary M. Marsh, Ph.D., F.A.C.E.

POSITION TITLEProfessor of BiostatisticsDirector, Center for Occupational Biostatistics and Epidemiology

EDUCATION/TRAINING (Begin with baccalaureate or other initial professional education, such as nursing, and include postdoctoral training.)

INSTITUTION AND LOCATION DEGREE(if applicable)

YEAR(s) FIELD OF STUDY

University of Pittsburgh, Pittsburgh, PA B.S. 1973 MathematicsUniversity of Pittsburgh, GSPH M.S. (Hyg.) 1974 BiostatisticsUniversity of Pittsburgh, GSPH Ph.D. 1977 Biostatistics

A. Positions and Honors

Positions and Employment1978 – 84 Assistant Professor of Biostatistics, Univ. of Pittsburgh, Graduate School of Public

Health (GSPH)1983 – 92 Assistant Director, Center for Environmental Epidemiology Univ. of Pittsburgh, GSPH

1984 Faculty, Graduate Summer Session in Epidemiology, Univ. of Minnesota, School of Public Health

1984 – 91 Associate Professor of Biostatistics, Univ. of Pittsburgh, GSPH1991-date Professor of Biostatistics, Univ. of Pittsburgh, GSPH2007 Interim Chair, Dept. of Biostatistics, Univ. of Pittsburgh, GSPH2008-date Director, Center for Occupational Biostatistics and Epidemiology, Univ. of Pittsburgh,

GSPH

Other Experience and Professional Memberships11974-date American Statistical Association

-Secretary, Vice President, President-Pittsburgh Chapter, 1979-82 -National Council Representative, 1981-1982

1974-date Biometric Society1978-date Society for Occupational and Environmental Health

-National Governing Council, 1986-19891979-date Society for Epidemiological Research1986-date Pennsylvania Public Health Association

-Member, Board of Directors, 1989-921988-date International Society for Environmental Epidemiology1996-date International Commission on Occupational Health1997-date Fellow, American College of Epidemiology (F.A.C.E.)

Honors1973 B.S. Mathematics, cum laude1981 Adolf G. Kammer Merit in Authorship Award for Best Publication in Field of

Occupational Health, Principal Author1994 Outstanding Teacher Award, University of Pittsburgh GSPH1997 Fellow, American College of Epidemiology1999 50 at 50 Award, selected as one of the 50 outstanding contributors in the field of

public health in the 50 year history of the Graduate School of Public Health

of 125


2006, 08 University of Pittsburgh Innovator Award 2006, 2008

6. B. Selected Peer-Reviewed Publications (Publications selected from 127 peer-reviewed publications)1. MARSH GM, Ehland J, Paik M, Preininger M, Caplan R: OCMAP/PC: A User Oriented

Cohort Mortality Analysis Program for the IBM PC. The American Statistician 40:308-309, 1986.

2. MARSH GM: A Strategy for Merging and Analyzing Work History Data in Industry-wide Occupational Epidemiology Studies. American Industrial Hygiene Association Journal, 48:414-419, 1987.

3. MARSH GM, Costantino JP, Lyons EE, Logue JN, and Fox JM: Health Effects of Exposure to the Drake Chemical Company Superfund Site: Morbidity Patterns Among Former Employees. Journal of Environmental Health 50:389-394, 1988.

4. MARSH GM, Co-Chien H, Rao BR, and Ehland J: OCMAP: Module 6-A New Computing Algorithm for Proportional Mortality Analysis. American Statistician 43:127-128, 1989.

5. MARSH GM, Callahan C, Pavlock D, Leviton LC, Talbott E, Hemstreet G: A Protocol for Bladder Cancer Screening and Medical Surveillance Among High Risk Groups: The Drake Health Registry Experience. Journal of Occupational Medicine 32:881-886, 1990.

6. MARSH GM, Leviton LC, Talbott E, Callahan C, Pavlock D, Hemstreet G, Logue JN, Fox J, Schulte P: The Drake Chemical Workers Health Registry Study: I. Notification and Medical Surveillance of a Group of Workers at High Risk of Developing Bladder Cancer. American Journal of Industrial Medicine 19:291-302, 1991.

7. MARSH GM, Enterline PE and McCraw D: Mortality Patterns Among a Cohort of Petroleum Refinery and Petrochemical Plant Workers. American Journal of Industrial Medicine 19:29-42, 1991.

8. MARSH GM, Day R: A Model Standardized Risk Assessment Protocol for Use with Hazardous Waste Sites. Environmental Health Perspectives 90:199-208, 1991.

9. MARSH GM, Stone RA, Henderson V: A Reanalysis of the National Cancer Institute Study on Mortality Among Industrial Workers Exposed to Formaldehyde. Journal of Occupational Medicine 34:42-44, 1992.

10. MARSH GM, Stone RA, Esmen NA, Henderson VL: Mortality Among Chemical Plant Workers Exposed to Formaldehyde and Other Substances. Journal of the National Cancer Institute 86:384-385, 1994.

11. MARSH GM, Stone RA, Youk AO, Smith TS, Quinn MM, Henderson VL, Schall LC, Wayne LA, Lee KY: Mortality Among United States Rock Wool and Slag Wool Workers: 1989 Update. Journal of Occupational Health and Safety - Australia and New Zealand, 12(3):297-312, 1996.

12. MARSH GM, Stone RA, Esmen NA, Gula MJ, Gause CK, Petersen NJ, Meaney FJ, Rodney, S, Prybylski D: A Case-Control Study of Lung Cancer Mortality in Six Gila Basin, Arizona Smelter Towns. Environmental Research 75:56-72, 1997.

13. 1MARSH GM, Sefcik S, Alcorn C, Youk AO: OCMAP-PLUS, A New Program for the Comprehensive Analysis of Occupational Cohort Data. Journal of Occupational and Environmental Health 40:351-362, 1998.

14. MARSH GM, Stone RA, Esmen NA, Gula MJ, Gause CK, Petersen NJ, Meaney FJ, Rodney S, Prybylski D: A Case-Control Study of Lung Cancer Mortality in Four Rural Arizona Smelter Towns. Archives of Environmental Health 53:15-28, 1998.

15. MARSH GM, Lucas L, Youk AO, Schall LC: Mortality Patterns among Workers Exposed to Acrylamide: 1994 Follow-Up. Occupational and Environmental Medicine, 56:181-190, 1999.

16. MARSH GM, Youk AO, Collins J: A Reevaluation of Lung Cancer Risk in the NCI/NIOSH Acrylonitrile Cohort Study. Scandinavian Journal of Work, Environment and Health 27:5-13, 2001.

17. MARSH GM, Youk AO, Stone, RA, Buchanich JM, Gula MJ, Smith TJ, Quinn MM: Historical Cohort Study of U.S. Man- Made Vitreous Fiber Production Workers. I. 1992 Fiber Glass Cohort Follow-Up- Initial Findings. Journal of Occupational and Environmental Medicine 43:741-756, 2001.

of 125


18. MARSH GM, Buchanich JM, Youk AO: Historical Cohort Study of U.S. Man-Made Vitreous Fiber Production Workers. VI.Respiratory System Cancer SMRs Adjusted for the Confounding Effect of Cigarette Smoking Journal of Occupational and Environmental Medicine 43:803-808, 2001.

19. 1MARSH GM, Youk AO, Stone RA, Buchanich JM, Gula MJ, Smith TJ, Churg A, Colby T: Does Fiber Glass Pose a Respiratory Cancer Risk in Man? Findings from the Latest Update of the U.S. Cohort Study of Man-Made Vitreous Fiber Workers. Annals of Occupational Hygiene 46(Supp.1): 110-114, 2002.

20. MARSH GM, Gula MJ, Youk AO, Cassidy LD: Bladder Cancer among Chemical Workers Exposed to Nitrogen Products and Other Substances. American Journal of Industrial Medicine 42:286-295, 2002.

21. MARSH 1GM, Gula MJ, Roggli V, Churg A: The Role of Smoking and Asbestos Exposure in a Questionable Case of Mesothelioma. Industrial Health, 41:332-334, 2003.

22. MARSH 1GM, Youk AO. Reevaluation of Mortality Risks from Leukemia in the Formaldehyde Cohort Study of the National Cancer Institute. Regulatory Toxicology and Pharmacology, 40:113-124, 2004.

23. MARSH Marsh GM, Youk AO, Buchanich JM, Esmen NA. Mortality Patterns among Workers in a U.S. Pharmaceutical Production Plant. Annals of Epidemiology, 15:112-122, 2005.

24. MARSH 1GM, Youk AO. Reevaluation of Mortality Risks from Nasopharyngeal Cancer in the Formaldehyde Cohort Study of the National Cancer Institute. Regulatory Toxicology and Pharmacology, 42:275-283, 2005.

25. MARSH GM, Gula MJ. Employment as a Welder and Parkinson’s Disease among Heavy Equipment Manufacturing Workers. Journal of Occupational and Environmental Medicine, 48:1031-1045, 2006.

26. MARSH GM, Youk AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality Patterns among Industrial Workers Exposed to Chloroprene and Other Substances: I. General Mortality Patterns. Chemico-Biological Interactions 166:285-300, 2007.

27. MARSH GM, Youk AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality Patterns among Industrial Workers Exposed to Chloroprene and Other Substances: II. Mortality in Relation to Exposure. Chemico-Biological Interactions 166:301-316, 2007.

28. MARSH GM, Youk AO, Morfeld P. Mis-Specified and Non-Robust Mortality Risk Models for Naspharyngeal Cancer in the National Cancer Institute Formaldehyde Worker Cohort Study. Regulatory Toxicology and Pharmacology 47:59-67, 2007.

29. MARSH GM, Youk AO, Buchanich JM, Kant IJ, Swaen G. Mortality among Workers Exposed to Acrylamide: Updated Follow-up. Occupational and Environmental Medicine, 49:82-95, 2007.

30. MARSH GM, Youk AO, Buchanich JM, Erdal S, Esmen NA. Work in the Metal Industry and Nasopharyngeal Cancer Mortality among Formaldehyde-Exposed Workers. Regulatory Toxicology and Pharmacology, 48:308-319, 2007.

31. MARSH GM, Buchanich J, Youk A, Cunningham M, Kennedy K, Lacey S, Hancock R, Esmen N. Long-Term Health Experience of Jet Engine Manufacturing Workers: I. Mortality from Central Nervous System Neoplasms. Journal of Occupational and Environmental Medicine 50:1099-1116, 2008.

32. MARSH GM, Buchanich J, Youk A, Cunningham M, Kennedy K, Lacey S, Hancock R, Esmen N. Long-Term Health Experience of Jet Engine Manufacturing Workers: II. Total and Cause-Specific Mortality excluding Central Nervous System Neoplasms. Journal of Occupational and Environmental Medicine 50:1117-1129, 2008.

of 125




NAME

Jeanine M. Buchanich

POSITION TITLE

Research Assistant Professor




University of Pittsburgh College of Arts & Sciences BS 1993 Psychology University of Pittsburgh School of Education MEd 1996 School Counseling

University of Pittsburgh Graduate School of Public Health MPH 1998 Epidemiology

University of Pittsburgh Graduate School of Public Health PhD 2007 Epidemiology

B. Positions and Honors

Positions and Employment

2008-date University of Pittsburgh Research Assistant Professor

Graduate School of Public Health

2000-2008 University of Pittsburgh Research Specialist VGraduate School of Public Health

1998-2000 University of Pittsburgh Research Specialist IVGraduate School of Public Health

11996-1998 University of Pittsburgh Research Specialist IIIGraduate School of Public Health

Other Experience and Professional Memberships

11999-date Society for Epidemiological Research2006-date Society for Neuro-Oncology2008-date Deputy Director, Epidemiology, Center for Occupational

Biostatistics and Epidemiology, Graduate School of Public Health, University of Pittsburgh

Honors

11989-1993 Chancellor’s Scholar, Full Scholarship, University of Pittsburgh1990-1993 Golden Key National Honor Society1993 B.S., Cum Laude1995-1996 Chi Sigma Iota National Honor Society in Counseling2008- Who’s Who Among Executives and Professionals

7. B. Selected Peer-Reviewed Publications

1. Schall LC, Buchanich JM, Marsh GM, Bittner G: Utilizing Multiple Vital Status Tracing Services is Necessary for Attaining Optimal Mortality Follow-up in Cohort Studies. Annals of Epidemiology, 11:292-296, 2001.

of 125


12. Buchanich JM, Marsh GM, Youk AO: Historical Cohort Study of U.S. Man-Made Vitreous

Fiber Production Workers. V. Tobacco Smoking Habits. Journal of Occupational and Environmental Medicine 43:793-802, 2001.

3. 1Marsh GM, Youk AO, Buchanich JM, Cassidy LD, Lucas LJ, Esmen NA, Gathuru I. Pharyngeal Cancer Mortality among Chemical Plant Workers Exposed to Formaldehyde. Toxicology and Industrial Health, 18:257-268, 2002 (actual publication date January 2004).

4. Stone RA, Youk AO, Marsh GM, Buchanich JM, Smith TJ, Quinn MM: Historical Cohort Study of U.S. Man-Made Vitreous Fiber Production Workers IX: Summary of 1992 Mortality Study Follow-Up and Analysis of Respiratory System Cancer among Female Workers. Journal of Occupational and Environmental Medicine 46(1):55-67, 2004.

5. Dolan D, Youk AO, Marsh GM, Buchanich JM: A 50-year Historical Cohort Mortality Study of Workers in a Pharmaceutical Plant. Journal of Occupational and Environmental Medicine 46(2):161-166, 2004.

6. Marsh GM, Youk AO, Esmen NE, Buchanich JM. Mortality patterns among workers in a U.S. pharmaceutical production plant. Annals of Epidemiology 15:112-122, 2005.

7. Cassidy LD, Buchanich JM, Guice K. Comparison of Injury Severity Scoring Systems to Identify Most Relevant Variables for Inclusion in a National Trauma Registry for Children. Journal of Registry Management 32(1):4-10, 2005.

8. Buchanich JM, Dolan DG, Marsh GM, Madrigano J. Under-Ascertainment of Deaths Using Social Security Records: A Recommended Solution to a Little-Known Problem. American Journal of Epidemiology, 162(2):193-4, 2005.

9. Marsh GM, Youk AO, Buchanich JM, Swaen G, Kant IJ. Mortality Patterns Among Workers Exposed to Acrylamide: Updated Follow-up. Journal of Occupational and Environmental Medicine, 49(1):82-95, 2007.

10. Esmen NA, Hall TA, Phillips ML, Jones EP, Basara H, Marsh GM, Buchanich JM. Chemical process based reconstruction of exposures for an epidemiological study: II. Estimated exposures to Chloroprene and Vinyl Chloride. Chemico-Biological Interactions, 166: 264-276, 2007.

11. Hall TA, Esmen NA, Jones PE, Basara H, Phillips ML, Marsh GM, Youk AO, Buchanich JM, Leonard R. Chemical Process Based Reconstruction of Exposures for an Epidemiological Study: III. Analysis of Industrial Hygiene Samples. Chemico - Biological Interactions, 166: 277-284, 2007.

12. Marsh GM, Youk AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality Patterns among Industrial Workers Exposed to Chloroprene and Other Substances I. General Mortality Patterns Chemico-Biological Interactions, 166: 285B300, 2007.

13. Marsh GM, Youk AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality Patterns among Industrial Workers Exposed to Chloroprene and Other Substances II. Mortality in Relation to Exposure Chemico-Biological Interactions, 166: 301-316, 2007.

14. Leonard RC, Lineker GA, Kreckmann KH, Marsh GM, Buchanich JM, Youk AO. Comparison of standardized mortality ratios (SMRs) obtained from use of a reference population based on a company-wide registry cohort to SMRs calculated against local and national rates. Chemico-Biological Interactions, 166: 317-322, 2007.

15. Marsh GM, Youk AO, Buchanich JM, Erdal SE, Esmen NE. Work in the Metal Industry May Help Explain Nasopharyngeal Cancer Mortality Excess among Workers Exposed to Formaldehyde. Regulatory Toxicology and Pharmacology, 48: 308-319, 2007.

of 125


16. Lippert JF, Lacey SE, Kennedy KJ, Esmen NA, Buchanich JM, Marsh GM. Magnetic field exposure in a non-destructive testing operation. Archives of Environmental and Occupational Health, 62: 187-193, 2007.

17. Buchanich JM, Cassidy L, Guice K. National Trauma Registry for Children Project: A National Hospital Survey. Journal of Registry Management, 35:12-17, 2008.

18. Marsh GM, Buchanich JM, Youk AO, Cunningham M, Lieberman F, Kennedy KJ, Lacey SE, Hancock RP, Esmen NA. Long-Term Health Experience of Jet Engine Manufacturing Workers: I. Mortality from Central Nervous System Neoplasms. Journal of Occupational and Environmental Medicine, 50:1009-1016, 2008.

19. Marsh GM, Buchanich JM, Youk AO, Cunningham M, Lieberman F, Kennedy KJ, Lacey SE, Hancock RP, Esmen NA. Long-Term Health Experience of Jet Engine Manufacturing Workers: II. Total and Cause-Specific Mortality excluding Central Nervous System Neoplasms. Journal of Occupational and Environmental Medicine, 50:1017-1029, 2008.

20. Cassidy LD, Cunningham MA, Buchanich JM, Marsh GM, Guice K. Proposed Sampling Plan for a National Trauma Registry for Children. Journal of Registry Management, 35:113-119, 2008.

21. Buchanich JM, Songer TJ, Cassidy LD, Marsh GM, Ford HR. A Clinical Decision Making Rule for Mild Head Injury in Infants & Toddlers. Saarrbrucken: VDM Verlag Dr. Muller, 2008.

of 125




NAME

Ada O. Youk, Ph.D.

POSITION TITLE

Research Assistant Professor of Biostatistics




Duquesne University, Pittsburgh, PA B.S. 1988 MathematicsUniversity of Pittsburgh M.A. 1990 BiostatisticsUniversity of Pittsburgh, GSPH Ph.D. 1996 Biostatistics

C. Positions and Honors

Positions and Employment11988-90 Teaching Assistant, University of Pittsburgh, Department of Mathematics1990-96 Graduate Student Researcher, University of Pittsburgh, Department of

Biostatistics1996-02 Research Associate, University of Pittsburgh, Department of Biostatistics2002- Research Assistant Professor, University of Pittsburgh, Department of

Biostatistics

Other Experience and Professional Memberships11995-date Member, Biometric Society (ENAR)2004 Local Arrangements Chair, Biometric Society (ENAR)1996-date Member, American Statistical Association (ASA)

Honors11984-88 Partial Competitive Scholarship, Duquesne University1986 Phi Eta Sigma National Honor Society1988 B.A., Cum Laude1988 National Collegiate Mathematics Award1988 Who’s Who Among American Colleges and Universities1988-90 Full Tuition Scholarship & Assistantship, University of Pittsburgh,

Mathematics1990-96 Full Tuition Scholarship & Assistantship, University of Pittsburgh,

Biostatistics1997 Delta Omega National Honor Society in Public Health1997 Student of the Year, American Statistical Association, Pittsburgh

Chapter2004-2005 Delta Omega National Honor Society in Public Health, President2006 University of Pittsburgh Innovator Award2007 Who’s Who Among American Educators

8. B. Selected Peer-Reviewed Publications (Publications selected from 49 peer-reviewed publications)

Talbott EO, YOUK AO, McHugh K, Shire J, Murphy B. Mortality among the residents of the Three Mile Island area. Environmental Health Perspectives 2000;108(6):1-8.

Jain A, DiMartini A, Kashyap R, YOUK A, Gray W, Rohal S, Fung J. Long-term Follow-

of 125


up after Liver Transplantation for Alcoholic Liver Disease under Tacrolimus. Transplantation 2000; 70(9):1335-1342.

Marsh GM, YOUK AO. Absence of Cancer Risks among Workers Exposed to Acrylonitrile: A Reanalysis of the NCI/NIOSH Cohort Study Data. Scan J Work, Environ and Health. 2001; 27(1):5-13.

Marsh GM, YOUK AO, Stone RA, Buchanich JM, Gula MJ, Smith TJ, Quinn MM. Historical Cohort Study of U.S. Man-Made Vitreous Fiber Production Workers: I. 1992 Fiber Glass Cohort Follow-Up - Initial Findings. J Occ Env Med 2001; 43(9): 741-756.

YOUK AO, Marsh GM, Stone RA, Buchanich JM, Smith TJ. Historical Cohort Study of U.S. Man-Made Vitreous Fiber Production Workers III: Analysis of Exposure-Weighted Measures of Respirable Fibers and Formaldehyde in the Nested Case-Control Study of Respiratory System Cancer. J Occ Env Med 2001;43(9): 767-778.

Stone RA, YOUK AO, Marsh GM, Smith TJ, Quinn MM, Buchanich JM. Historical Cohort Study of U.S. Man-Made Vitreous Fiber Production Workers IV: Quantitative Exposure-Response Analysis of the Nested Case-Control Study of Respiratory System Cancer. J Occ Env Med 2001;43(9): 779-792.

Marsh GM, YOUK AO, Stone RA, Buchanich JM, Gula MJ, Smith TJ, Churg A, Colby T. Does Fiber Glass Pose a Respiratory System Cancer Risk in Humans? Latest Findings from the US Cohort and Nested Case-control Studies. Ann Occup Hyg 2002;46(1):1-5.

Marsh GM, Gula MJ, YOUK AO, Schall LC: Bladder Cancer among Chemical Workers Exposed to Nitrogen Products and Other Substances. American Journal of Industrial Hygiene, 2002;42:286-295.

Cassidy LD, YOUK AO, Marsh GM. The Drake Health Registry Study: Cause-Specific Morality Experience of Workers Potentially Exposed to Beta-naphthylamine. American Journal of Industrial Medicine, 2003; 44:282-290.

Bloemen LJ, YOUK AO, Bradley TD, Bodner KM, Marsh GM. Lymphohematopoietic Cancer Risk among Chemical Workers Exposed to Benzene. Occupational and Environmental Medicine, 2004; 61(3):270-274.

YOUK AO, Stone RA, Marsh GM. A Method for Imputing Data in Longitudinal Studies. Annals of Epidemiology, In press.

Starr TB, Gause CG, YOUK AO, Stone RA, Marsh GM, Collins JJ. A risk assessment for occupational acrylonitrile exposure using epidemiology data. Risk Analysis, In press.

Stone RA, YOUK AO, Marsh GM, Buchanich JM, Smith TJ. Historical Cohort Study of US Man-Made Vitreous Fiber Workers IX: Analysis of Respiratory System Cancer Among Female Workers. JOEM, 2004; 46(1):55-67.

Dolan DG, YOUK AO, Marsh GM, Buchanich JM. A 50-Year Historical Cohort Mortality Study of Workers in a Pharmaceutical Plant. JOEM, 2004;46(2):161-166.

Marsh GM, YOUK AO. Reevaluation of Mortality Risks from Leukemia in the Formaldehyde Cohort Study of the National Cancer Institute. Regulatory Toxicology and Pharmacology, 2004; 40:113-124.

of 125


Wright CE, Zborowski JV, Talbott EO, McHugh-Pemu K, YOUK AO. Dietary Intake, Physical Activity, and Obesity in Women with Polycystic Ovary Syndrome. International Journal of Obesity & Related Metabolic Disorders, 2004; 28(8):1026-1032.

Marsh GM, YOUK AO, Buchanich JM, Esmen N. Mortality Patterns among Workers in a US Pharmaceutical Production Plant. Annals of Epidemiology, 2005; 15(2):112-122.

Marsh GM, YOUK AO. Reevaluation of Mortality Risks from Nasopharyngeal Cancer in the Formaldehyde Cohort Study of the National Cancer Institute. Regulatory Toxicology and Pharmacology, 2005; 42:275-283.

Bromberger JT, Kravitz H, Wei HL, Brown C, YOUK AO, Matthews K. History of Depression and Women’s Current Health and Functioning During Midlife. General Hospital Psychiatry, in press.

Ramos RG, Talbott EO, YOUK AO, Karol MH. Community Urbanization and Hospitalization of Adults for Asthma, Journal of Environmental Health, 2006; 68(8):26-32.

Marsh GM, YOUK AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML, Mortality patterns among industrial workers exposed to chloroprene and other substances: I. General mortality patterns, Chem.-Biol. Interact., 2007; 166:285-300.

Marsh GM, YOUK AO, Buchanich JM, Cunningham M, Esmen NA, Hall TA, Phillips ML. Mortality patterns among industrial workers exposed to chloroprene and other substances: II. Mortality in relation to exposure, Chem.-Biol. Interact., 2007;166: 301–316.

Marsh GM, YOUK AO, Morfeld P. Mis-specified Mortality Risk Models for Naspharyngeal Cancer in the National Cancer Institute Formaldehyde Worker Cohort Study. Regulatory Toxicology and Pharmacology, 2007; 47:59-67.

Leonard RC, Lineker GA, Kreckmann KH, Marsh GM, Buchanich J, YOUK AO. Comparison of Standardized Mortality Ratios (SMRs) Obtained from Use of a Reference Population Based on Company-Wide Registry Cohort to SMRs Calculated against Local and National Rates. Chem-Biol Interact, 2007; 166:317-322.

Marsh GM, YOUK AO, Buchanich JM, Kant IJ, Swaen G. Mortality Patterns Among Workers Exposed to Acrylamide: Updated Followup. Occupational and Environmental Medicine, 2007; 49:82-95.

Hall TA, Esmen NA, Jones PE, Basara H, Phillips ML, Marsh GM, YOUK AO, Buchanich JM, Leonard R. Chemical Process Based Reconstruction of Exposures for an Epidemiological Study: III. Analysis of Industrial Hygiene Samples. Chemico-Biological Interactions, 2007;166: 277-284.

Marsh GM, YOUK AO, Buchanich JM, Erdal S, Esmen NA. Work in the Metal Industry May Help to Explain Nasopharyngeal Cancer Mortality Excess among Workers Exposed to Formaldehyde. Regulatory Toxicology and Pharmacology, 2007; 48: 308-319.

1Bromberger JT, Kravitz HM, Matthews K, YOUK AO, Brown C, Feng W. Predictors of First Lifetime Episodes of Major Depression in Midlife Women. Psychological Medicine, in press.

of 125


Marsh GM, Buchanich JM, YOUK AO, Cunningham M, Lieberman F, Kennedy KJ, Lacey SE, Hancock RP, Esmen NA. Long-Term Health Experience of Jet Engine Manufacturing Workers: I. Mortality from Central Nervous System Neoplasms. JOEM, in press.

Marsh GM, Buchanich JM, YOUK AO, Cunningham M, Lieberman F, Kennedy KJ, Lacey SE, Hancock RP, Esmen NA. Long-Term Health Experience of Jet Engine Manufacturing Workers: II. Total and Cause-Specific Mortality excluding Central Nervous System Neoplasms. JOEM, in press.

of 125


BIOGRAPHICAL SKETCHProvide the following information for the key personnel in the order listed on Form Page 2.

Photocopy this page or follow this format for each person.

NAME

Nurtan Alan Esmen

POSITION TITLE

Professor


INSTITUTION AND LOCATIONDEGREE

(if applicable) YEAR(s) FIELD OF STUDY

Northeastern University, Boston, MAUniversity of Pittsburgh, Pittsburgh, PAUniversity of Pittsburgh, Pittsburgh, PA

B.Sc.M.Sc.Ph.D.

196519661970

EngineeringChem & Air Eng.Chem / Air Eng & Math

Professional Experience:Fellow: American Industrial Hygiene Association & Royal Society for the Promotion of HealthUniversity of Illinois (Chicago) – Professor (2003 – to date); University of Oklahoma – Professor (1995 – 2003) and Chair (1997 - 2002) of Occup. & Environ. Health; Esmen Research - Senior Partner & Principal Scientist (1990 - 1995); University of Pittsburgh – Professor (1980 – 1993), Program Director of Industrial Hygiene (1980 – 1990) Associate Professor of Industrial Hygiene Engineering (1975 – 1980); Owens-Corning Fiberglas - Principal Scientist (1974 – 1975); University of Delaware - Assistant Professor of Environmental Engineering (1970 – 1974); University of Pittsburgh - Senior Research Staff (1965 – 1970) Harvard School of Public Health - Research Staff (1960 – 1965)

PUBLICATIONS:(More than 150 articles and abstracts in Occupational. Hygiene, Aerosol Physics, Epidemiology and related subjects. Only selected recent publications after 2000 are listed.)

JF Lippert, S.E. Lacey, K.J. Kennedy. N.A. Esmen, JM Buchanich and G,M, Marsh : Magnetic Field Exposure in a Nondestructive Testing Operation, Archives of Environmental and Occupational Health 62:187-193 (2008)TM Sorahan, , Pang, D., Esmen, N. A. and Sadhra, S: Urinary Concentrations of Toxic Substances: An Assessment of alternative approaches to adjusting for specific gravity, J Occup. Env, Hyg, 5:721-723 (2008).GM Marsh Buchanich JM, Youk, AO, Cunnigham MA, Lieberman FS, Kennedy, KJ, Lacey, SE, Hancock, RP, Esmen, NA : Long-Term Health Experience of Jet Engine Manufacturing Workers: I. Mortality From Central Nervous System Neoplasms J Occup Environ Med. 50:1099–1116 (2008)GM Marsh, Buchanich JM, Youk, AO, Cunnigham MA, Lieberman FS, Kennedy, KJ, Lacey, SE, Hancock, RP, Esmen, NA : Long-term Health Experience of Jet Engine Manufacturing Workers: II. Total and Cause-Specific Mortality Excluding Central Nervous System Neoplasms J Occup Environ Med. 50:1117–1129 (2008)GM Marsh, AO. Youk, JM. Buchanich, S. Erdal, NA. Esmen: Work in the Metal Industry May Help Explain Nasopharyngeal Cancer Mortality Excess among Workers Exposed to Formaldehyde Reg Tox and Pharm. 48:308-319 (2007).NA Esmen, KJ Kennedy, TA Hall, ML. Phillips, GM Marsh: Classification of Worker Exposures, Chemico-Biological Interactions, 166:245-253(2007)NA Esmen, TA Hall, ML Phillips, GM Marsh: Chemical process based reconstruction of exposures for an epidemiological study: I. Theoretical and Methodological issues, Chemico-Biological Interactions, 166:254-263(2007)NA Esmen, TA Hall, ML Phillips, EP Jones, H Basara, GM Marsh, JM Buchanich, Chemical process based reconstruction of exposures for an epidemiological study: II. Estimated exposures to Chloroprene and Vinyl Chloride, Chemico-Biological Interactions, 166:264-276(2007)TA Hall, NA Esmen, EP Jones, H Basara, ML Phillips, GM Marsh, AO Youk, JM Buchanich, RC Leonard,, Chemical process based reconstruction of exposures for an epidemiological study: III. Analysis of Industrial Hygiene Samples Chemico-Biological Interactions, 166:277-284(2007)GM Marsh, AO Youk, JM Buchanich, M Cunningham, NA Esmen, TA Hall, ML Phillips: Mortality patterns among

of 125


industrial workers exposed to chloroprene and other substances: I. General mortality patterns, Chemico-Biological Interactions, 166:285-300(2007)GM Marsh, AO Youk, JM Buchanich, M Cunningham, NA Esmen, TA Hall, M.L. Phillips, Mortality patterns among industrial workers exposed to chloroprene and other substances: II. Mortality in relation to exposure, Chemico-Biological Interactions, 166:301-316 (2007)GM Marsh, AO Youk, JM Buchanich, S Erdal, NA Esmen: Work in the Metal Industry May Help Explain Nasopharyngeal Cancer Mortality Excess among Workers Exposed to Formaldehyde Reg Tox and Pharm. 48:308-319 (2007).GA Day, G. A., AB Stefaniak, A.B.,Hoover,M. D., Dickerson,R.M., Peterson,E. J., Esmen, N. A., Scripsick, R. C.: "Bioavailability of Beryllium Oxide Particles: An In Vitro Study in the Murine J774A.1 Macrophage Cell Line Model" Experimental Lung Research. 31: 341-360 (2005)Johnston, KL, Phillips, ML, Esmen, NA, and Hall, TA: Evaluation of an Artificial Intelligence Program for Estimating Occupational Exposures: Annals of Occ. Hyg. 49:147-153 (2005)Marsh, GM, Youk, AO, Esmen, NA and Buchanich, JM: Mortality Patterns among Workers in an U.S. Pharmaceutical Production Plant Annals of Epidemiology 15:112–122(2005).Phillips, ML, Esmen, NA, Hall, TA, and Lynch, RL: Determinants of Exposure to Volatile Organic Compounds in Four Oklahoma Cities J. Exp. Anal. and Environ. Epid. 15: 35-46 (2005)Sorahan, T., and Esmen, N. A.: "Lung Cancer Mortality in UK nickel-cadmium battery workers, 1947-2000" Occup. Env. Medicine 61:1008-1016 (2004)Johnson, D. L. and Esmen, N. A.: Method induced exposure misclassification for a respirable dust sampled using ISO/ACGIH/CEN criteria. Ann Occup Hyg. 48:13-20, (2004)Esmen, N.A. " Benzene exposure assessments: past practices and future prospects" Proceedings of Workshop on Leukemia Risks in Relation to Benzene Exposure. TM Sorahan (Edit) Institute of Petroleum, London p15-27(2003).Marsh, G. M., Youk, A. O., Buchanich, J.M., Cassidy, L.D., Lucas, L.J., Esmen, N. A., Gathuru, I.: Pharyngeal cancer mortality among chemical plant workers exposed to formaldehyde Toxicology and Industrial Health. !8:257-268 (2002)Phillips, M. L., Esmen, N. A., Wang, D. and Hall, T. A.: Temporal and Spatial Variation in Monitoring of Urban Air Pollutants In: Fate and Transport of Chemicals in the Environment: Impacts, Monitoring, and Remediation (R. L. Lipnick, R. P. Mason, M. L. Phillips, & C. U. Pittman Eds.) American Chemical Society Monograph (2002)Clinkenbeard, R. E., England, E.C., Johnson, D.L., Esmen, N. A., Hall, T.A., and Carlton, G. “Comparison of the IOM inhalable aerosol sampler and a modified 37-mm cassette during aircraft corrosion control maintenance” App. Env. and Occup. Hyg. 17:622-627(2002).Esmen, N.A., Johnson, D. L., and Agron, G. M.: ”The Variability of Delivered Dose of Aerosols with the Same Respirable Concentration but Different Size Distributions” Ann Occup. Hyg. 46:401-408(2002)Collins, J. J., Ness, R., Tyl, R. Krivanek, N., Esmen, N. A., Hall, T. A., : A review of adverse pregnancy outcomes and formaldehyde exposure in human and animal studies Regulatory Toxicology and Pharmacology 34:17-34 (2001)Esmen, N. A. and Johnson, D. L.: Simulation Analysis of Inhalable Dust Sampling Errors Using a Multi-Component Error Model Aerosol Science and Technol. 35:824-828 (2001)Phillips, M. L., Hall, T. A. Esmen, N. A., Lynch, R. A. and Johnson, D. L. “Use of global positioning system technology to track subjects’ location during environmental exposure sampling. J. Exposure Analysis and Environ. Epid. 11: 207-215 (2001).Phillips . M. L., Esmen, N. A., Costantino, J.: The Reliability of Multiple Regression and an Alternative Method for Extracting Task-Specific Exposure Estimates from Time-Weighted Average Data App. Env. and Occup. Hyg. 16:1-10(2001)Collins, J. J., Esmen, N. A. and Hall, T. A. : A Review and Meta-analysis of Formaldehyde Exposure and Pancreatic Cancer Amer. J. Indus. Med. 39:336-345.(2001)Johnson, D. L., Esmen, N. A., Carlson, K.D., Pearce, T. A and Thomas, B. N.: Aerodynamic behavior of lipid microtubule aerosols. Journal of Aerosol Science 31:181-188 (2000) Esmen, N. A. and Hall, T. A.: Theoretical investigation of the interrelationship between stationary and personal sampling in exposure estimation App. Env. and Occup. Hyg. 15:114-119 (2000).Esmen, N. A.: Multicomponent error model for mass measurement based size fractionating aerosol samplers App. Env. and Occup. Hyg. 15:72-79 (2000).

of 125


9. BIOGRAPHICAL SKETCHProvide the following information for the key personnel in the order listed on Form Page 2.

Follow this format for each person. DO NOT EXCEED FOUR PAGES.

NAME

Lacey, Steven Edward

POSITION TITLE

Assistant ProfessorEDUCATION/TRAINING (Begin with baccalaureate or other initial professional education, such as nursing, and include postdoctoral training.)

INSTITUTION AND LOCATION DEGREE(if applicable) YEAR(s) FIELD OF STUDY

Texas Tech University, Lubbock, TX BS 1991-1995 ZoologyTexas A&M University, College Station, TX MS 1997-2000 Industrial HygieneUniversity of Illinois at Chicago, Chicago, IL PhD 1999-2002 Env & Occ Health ScienceJohns Hopkins University, Baltimore, MD Postdoctoral Fellow 2002-2003 Env Health Engineering

A. Positions and Honors

Positions and Employment1997-1998 Emergency Medical Technician, Texas A&M University, College Station, TX1998 Industrial Hygienist, Amoco Corporation, Houston, TX1998-1999 Teaching Assistant, Texas A&M University, College Station, TX1999 Industrial Hygienist, Albemarle Corporation, Houston, TX1999-2002 Research Assistant, Great Lakes Center for Env & Occ Safety and Health, Chicago, IL2000-2002 Teaching Assistant, University of Illinois at Chicago, Chicago, IL2002-2003 Postdoctoral Fellow, Johns Hopkins University, Baltimore, MD2004-2008 Assistant Research Professor, University of Illinois at Chicago, Chicago, IL2008-present Assistant Professor, University of Illinois at Chicago, Chicago, IL

Director, Industrial Hygiene Program & Occupational Safety Program Certified Industrial Hygienist #16269Certified Safety Professional #19753

Professional Memberships2003-present American Industrial Hygiene Association, Full Member

2004-2005, AIHA Engineering Committee Chair2005-2006, AIHA Student & Early Career Professionals Committee Chair2008-present, AIHA Board of Directors

2006-present International Society of Exposure Analysis, Member2006-present American Conference of Governmental Industrial Hygienists, Member2006-present International Occupational Hygiene Association, Affiliate Member2006-present British Occupational Hygiene Society, Member2008-present American Society of Safety Engineers, Member

Honors1995 Texas Tech University Biological Science Honor Student1998 AIHA, Deep South Section, Fred S. Venable Scholarship

of 125


1999-2002 NIOSH Industrial Hygiene Traineeship, University of Illinois at Chicago2001 Stockhausen Corporation Scholarship for Industrial Hygiene2002 Best Poster on Exposure Modeling, American Industrial Hygiene

Conference & Exposition, Graduate Student Section, San Diego, CA2004 AIHA Outstanding Committee Award, Engineering Committee2006 AIHA Outstanding Committee Award, Student & Early Career Professionals Committee2006 President’s Award, AIHA Student & Early Career Professionals Committee

B. Publications and Recent Abstracts

Publications1. Marsh G, Buchanich J, Youk A, Cunningham M, Lieberman F, Kennedy K, Lacey S, Hancock R, Esmen N. Long-term health experience of jet engine manufacturing workers: I. Mortality from central nervous system neoplasms. J Occ Env Med 50:1099-1116 (2008).

2. Marsh G, Buchanich J, Youk A, Cunningham M, Lieberman F, Kennedy K, Lacey S, Hancock R, Esmen N. Long-term health experience of jet engine manufacturing workers: II. Total and cause-specific mortality excluding central nervous system neoplasms. J Occ Env Med 50:1117-1129 (2008).

3. Kovalsky A, Lacey S, Kaphle U, Vaughn J. Risk perception and water purification practices for water- borne parasitic infections in remote Nepal. Trop Doct 38:229-231 (2008).

4. Lippert, J., Lacey, S., et al. Magnetic field exposure in a nondestructive testing operation. Arch Env Occ Health 62:187-193 (2007).

5. Lacey, S., Forst, L., et al. Eye injury in migrant farm workers and suggested hazard controls. J Ag Safety Health 13:259-274 (2007).

6. Lacey, S., Conroy, L., et al. Dust emission rates from food processing. Ann Agric Environ Med 13:251-257 (2006).

7. Lacey, S., Conroy, L., et al. Personal dust exposures at a food processing facility. J Agromed 11:49-58 (2006).

8. Forst, L., Noth, I., Lacey, S., et al. Barriers and benefits of protective eyewear use by Latino farm workers. J Agromed 11:11-17 (2006).

9. Forst, L., Lacey, S., et al. Effectiveness of community health workers for promoting use of safety eyewear by Latino farm workers. Am J Ind Med 46:607-613 (2004).

Recent Abstracts1. Lacey, S., Horvatin, S., Kennedy, K., et al. Examination of worker ability to recall personal occupational history. Abstract and presentation at British Occupational Hygiene Society Conference, Bristol, UK, May 2008.

2. Lacey, S., Esmen, N., Palmer, J., Kennedy, K. Reconstruction of exposures to metalworking fluids using

of 125


industrial time studies. Abstract and presentation at American Industrial Hygiene Conference and Exposition, Philadelphia, PA, June 2007.

3. Lacey, S., Lippert, J., Esmen, N., Marsh, G. Electromagnetic field exposure in a non-destructive testing operation. Abstract and presentation at American Industrial Hygiene Conference and Exposition, Philadelphia, PA, June 2007.

4. Hancock, R., Esmen, N., Lacey, S., et al. Reconstruction of exposures in high temperature alloy surface grinding. Abstract and presentation at American Industrial Hygiene Conference and Exposition, Philadelphia, PA, June 2007.

6. Esmen, N., Hancock, R., Lacey, S., Kennedy, K. Exposure to aerosol generated in high temperature alloy surface grinding. Abstract and presentation at British Occupational Hygiene Society Conference, Glasgow, UK, April 2007.

7. Esmen, N., Lacey, S., Kennedy, K., Hancock, R. Exposure reconstruction strategy to examine glioblastoma multiforme risk among a cohort of jet engine manufacturing workers. Abstract and presentation at Society for Neuro-Oncology Annual Scientific Meeting, Orlando, FL, November 2006.

8. Lacey, S., Espinosa, R., et al. Development of a geospatial time dependent information system for industrial hygiene. Abstract and presentation at American Industrial Hygiene Conference and Exposition, Chicago, IL, May 2006.

9. Esmen, N., Lacey, S., et al. Validation of near-field vapour exposure prediction equations for unventilated chemical industrial processes. Abstract and presentation at British Occupational Hygiene Society Conference, Newcastle, UK, April 2006.

10. Lacey, S., Esmen, N., et al. Empirical observations on exposure rates with its implications in exposure reconstruction. Abstract and presentation at International Society for Exposure Analysis Conference, Tucson, AZ, November 2005.

of 125

consulting info with logo template etc in header and footer · web viewtungsten carbide with a...

Documents