September 1, 2006

CLUSTERING OFCLUSTERING OF PRIMARY SCLEROSING CHOLANGITIS NEAR TOXIC PRIMARY SCLEROSING CHOLANGITIS NEAR TOXIC

WASTE SITESWASTE SITES

Presented by Joseph A Odin, M.D., Ph.D.

Assistant Professor of Medicine

Mount Sinai School of Medicine

September 1, 2006

I have nothing to disclose and my presentation will not include discussion of off-label/investigative use or application of a product or device

September 1, 2006

Pathogenesis of Primary Sclerosing Pathogenesis of Primary Sclerosing CholangitisCholangitis

Primary Primary Sclerosing Sclerosing Cholangitis Cholangitis

Genetic

Susceptibility

Environmental

Factors

Immune

Dysregulation

Animal

Models

September 1, 2006

PSC is prevalence poorly studiedPSC is prevalence poorly studied

– Boberg et al, 2001 report the prevalence of PSC is increased in Scandinavia, but the cause is unclear.

– Very difficult to know if prevalence differences with regard to race, familial incidence, geography are environmental, societal, or genetic in origin.

– In animal models, toxin exposure may induce PSC-like peri-biliary inflammation and fibrosis.

September 1, 2006

PSC-like Animal ModelsPSC-like Animal Models

– Direct toxin-mediated biliary injury or murine GVHD models result in PSC-like lesions.

– Mdr2(-/-) knockout mice- leakage of bile through disrupted tight junction leads to a pro-inflammatory/fibrotic cascade. No known mutations of the human homologue of Mdr2 (MDR3).

– Dextran sulfate sodium (DSS) treated CFTR(-/-) KO mice develop colitis and biliary inflammation.

– Murine bacterial overgrowth or LPS in immuno-deficient mice may cause peribiliary inflammation.

Vierling, JM, 2003. Liver Immunology. Review.

September 1, 2006

Reprinted from Selmi et al, 2004, Epidemiology and pathogenesis of primary biliary cirrhosis

Autoimmune diseases have been associated with toxins or xenobiotics

September 1, 2006

Studies of disease prevalence have shown Studies of disease prevalence have shown geographic variability in PBCgeographic variability in PBC

Author(s), year Geo. clustering #cases Site Factor

Triger, 1980 Yes 34 Sheffield, U.K. water?

Hamlyn et al, 1983 No (seasonal?) 117 northeast England (sunlight)

Triger et al, 1984 Yes 552 Western Europe none

Borda et al, 1989 Yes 50 Navarra, Spain none

Danielsson et al,1990 Yes 111 northern Sweden none

Myszor et al, 1990 No 347 northeast England

Witt-Sullivan et al, 1990 No 225 Ontario, Canada

Prince et al, 2001 Yes 770 northeast England urban

Studies of immigrants suggest environment affects the prevalence of PBC Studies of immigrants suggest environment affects the prevalence of PBC (Watson et al, 1995 & Anand, 1996)(Watson et al, 1995 & Anand, 1996)

September 1, 2006

Pathogenesis of Pathogenesis of Cholestatic Liver DiseaseCholestatic Liver Disease

• Certain toxins are known to cause secondary

sclerosing cholangitis in individuals, but naturally

history may be different than PSC.

• Question: Do unrecognized environmental toxins

trigger PSC or PBC in genetically susceptible

individuals and/or promote disease progression?

September 1, 2006

Possible Approaches to Identifying Possible Approaches to Identifying Pathologic ToxinsPathologic Toxins

• Epidemiological studies– Questionnaires regarding exposures– Prevalence/Cluster analysis and geographic

information systems (GIS) technology

• Further studies in animals exposed to toxins

• Tissue analysis– Accumulated toxins may be detected in adipose

tissue– exposure upregulates expression of certain

genes (e.g. MDR, cytochrome p450 genes).

September 1, 2006

What is GIS?What is GIS?

• Geographic Information Systems (GIS)- “a structural approach to collecting, archiving, analyzing, manipulating, and displaying data having one or more spatial components, using a combination of personnel, equipment, computer software, and organizational procedures.” (National Research Council)

• “Layers” of spatial data are combined on the same projection or map. Often unsuspected patterns that escape detection in tabular reports are identified by projection. (e.g. higher cervical cancer rates in mountainous rural areas and observation of increased lung tumors in port cities helped establish asbestos exposure in shipyards as a risk factor).

• For health studies the combined layers may include a digital photograph of the area under study, administrative boundaries (e.g. postal codes), environmental layers (e.g. contaminated areas), facility locations (e.g. hospitals), subject locations (e.g. patient residences).

• Each of these spatial layers can be linked by a unique identifier to attribute data (i.e. clinical data or demographic data). This linkage gives GIS its analytical power.

September 1, 2006

GIS STEPSGIS STEPS

1. Organize a team of individuals committed to the study including clinicians, epidemiliogists, data entry specialists and spacial statistician.

2. Identify study subjects and environmental points of interest and collect address locations.

3. Geocoding of address information-conversion to longitude and latitude. Best if done immediately when address supplied since administrative boundaries change over time. Collecting all lifetime addresses preferable.

4. Mapping of prevalence rates in different areas (e.g. postal codes)- adjust rates to control for potential confounders such as differences in age, gender, and race distribution among different areas. Confidence levels for each rate should be included.

5. Pattern analysis-spatial statistical methods may be helpful in providing a quantitative answer if not obvious from mapping.

September 1, 2006

Step 1- A team approachStep 1- A team approach

• Interested clinicians- Aftab Ala and Nancy Bach

• Spatial statistician- Sylvan Wallenstein

• Epidemiologist/data entry specialist- “Carmen Stanca”

• Substitute- Joseph Odin

September 1, 2006

• stringent case inclusion criteria;

• definition of date of disease onset;

• well-defined study period, area and population;

• multiple case finding methods;

• rigorous tracing of all possible cases.

Metcalf, J. & James, O., Semin Liver Dis, 1997

Step 2. Recommended guidelines for identifying subjects

September 1, 2006

Step 3. Address geocoding-converting Step 3. Address geocoding-converting addresses into map locationsaddresses into map locations

Software loaded with georeferenced files can automatically convert appropriately formatted street addresses if available into specific longitudes and latitudes. Confidentiality must be maintained however.

Data Quality Issues:

• Who- best if patient provides address as opposed to insurance data.

• What- street mail address of residence preferred since offers the smallest point of reference.

• When- address at diagnosis usually best for most studies. An important point is to be consistent and note each factor along with the address

data. Spot checks of data accuracy essential.

September 1, 2006

Available U.S. Liver Disease Clinical Available U.S. Liver Disease Clinical DatabasesDatabases

1. Individual medical center patient databases.

2. Independent laboratory records.

3. Veterans hospitals records.

4. OPTN (Organ Procurement and Transfer Network) data on

patients listed for liver transplantation.

5. Nascent international multi-center PSC and PBC registries.

September 1, 2006

OPTN databaseOPTN database

• OPTN data limits referral bias as compared to

individual medical centers.

• Demographic information restricted to time of listing

and is not updated.

• Limited subset of patients are listed based on clinical

and non-clinical status.

• Accuracy of data is uncertain.

September 1, 2006

OPTN databaseOPTN database

• 102 individuals residing in the NY metropolitan area with a diagnosis of PSC and 127 individuals with PBC were listed in the OPTN database between 1995 and 2003.

September 1, 2006

Long Natural History May Distort Long Natural History May Distort Analysis of Subject Spatial DataAnalysis of Subject Spatial Data

Progression(diagnosis)

10-30 yrs?10-30 yrs?

Cirrhosis(listing)

Normal liver(disease onset)

Carcinoma/DeathCarcinoma/Death

TransplantationFigure adapted from SL Friedman, MSSM

September 1, 2006

Step 4. Prevalence Data AdjustmentStep 4. Prevalence Data Adjustment

• Demographic data available for all listed PSC and PBC patients living in New York State and US census data for individual zip codes was used to standardize expected prevalence rates for each New York City zip (postal) code.

• Observed/stdexpected prevalence rates were mapped for each zip code with darker colors indicating higher rates.

• There was no difference in confidence levels for each value.

September 1, 2006

Calculating and Standardizing Calculating and Standardizing Prevalence Ratios by Zip CodePrevalence Ratios by Zip Code

• The expected prevalence of patients listed for transplant with PBC or PSC was based on overall data for UNOS region 9 (New York State and Vermont) from 2000 to 2004.

• Using the 2000 US census data for each zip code, we were thus able to correct for differences between zip codes with regard to population, age, gender, and race and standardize the expected prevalence rates (stdE).

Asian2%

Black7%Other

5%

White86%

65yr+over13%

<18yr0%

18-34yr19%

35-49yr55%

50-64yr13%

Male

14%

Female

86%

September 1, 2006

Step 4. PBC and PSC cluster differentlyStep 4. PBC and PSC cluster differently

PBCPBC PSC PSC PBC PBC PSC PSC

Man

hat

tan

/Bro

nx

Sta

ten

Isl

and

Bro

okly

n

Q

uee

ns

September 1, 2006

Identified Toxic Sites in New Identified Toxic Sites in New York CityYork City

Courtesy of the NY DEC

This NY website also provides address information for each of these sites, which is more accurate than the locations shown on the map.

New Jersey

BIGGER TRIANGLES Brooklyn

Queens

Bronx

Man

hatt

anStaten Is.

Boroughs/CountiesSuperfund Sites

Nas

sau

Adapted from the New York Department of Environment 2004

Superfund site locations matched best with high disease prevalence areas

WestchesterWestchester

September 1, 2006

Superfund Toxic Waste Sites (SFS)Superfund Toxic Waste Sites (SFS)

•SFS are sites of public health hazards that have been designated for immediate remedial action.

•A recent NYC Department of Health study showed an increased cancer incidence (e.g. breast and lung cancer) in neighborhoods surrounding NYC SFS.

•Vine et al, 2000, demonstrated deleterious effects on the immune system of those living near a SFS.

September 1, 2006

Predominant SFS ToxinsPredominant SFS Toxins

•Heavy Metals e.g. mercuryHeavy Metals e.g. mercury

•Methyl chloride

•Lacquer

•Solvents

•Trichloroethane

•Xylene

•Halogenated solventsHalogenated solvents

•PCE 1,1,2,2-tetrachlorethylene (dry cleaning industry, household detergents)

•PCBPolychlorinated biphenyls (formely used in hydraulic systems, plasticizer, textile, transformers)

•Tetrachloroethane

September 1, 2006

HypothesisHypothesis

The prevalence of PSC and PBC patients listed for transplantation are increased near Superfund sites.

September 1, 2006

Step 5. Statistical AnalysisStep 5. Statistical Analysis

(i) Comparison of prevalence between “grouped” zip codes with and without

SFS

(ii) Comparison of prevalence among boroughs and SFS density in each borough

(iii) Validated computer analysis (SaTScan): a) ‘global’ clustering of patients b) ‘focused’ clustering of patients near

SFS

September 1, 2006

Non-SFS ‘cluster’

SFS ‘cluster’

Grouped zip code comparisonGrouped zip code comparison

SFS

Each zip code and its adjacent zip codes were considered a group or ‘cluster’

September 1, 2006

The median std prevalence ratio of PBC is The median std prevalence ratio of PBC is significantly higher in SFS clusterssignificantly higher in SFS clusters

CLUSTERS WITHOUT

SFS

CLUSTERS WITH

SFSP

PBC 0.51 0.94 0.001

PSC 0.28 0.28 0.572

p: significance values for Mann-Whitney U test, 2-tailed Std prev ratio: observed/std expected prevalence

September 1, 2006

Staten Island has the highest prevalence of PBC Staten Island has the highest prevalence of PBC & the highest density of SFS& the highest density of SFS

NYC BoroughStd Prev Ratio (rank) #SFS/100,000/sq mile (rank)

Manhattan 0.66 (5) 0.0211 (5)

Brooklyn 0.91 (3) 0.0411 (4)

Queens 0.88 (4) 0.0561 (2)

Bronx 1.02 (2) 0.0415 (3)

Staten Island 1.54 (1) 0.3150 (1)

September 1, 2006

SaTScanSaTScanTMTM

METHODMETHOD

•Spatial analysis of the case distribution was conducted using a cluster detection spatial scan statistic, SaTScan v5.0 adjusting for the underlying background population.

•Longitudes and latitudes are used

•Global analysis-enter only patient lat & long (zip code center)

•Focused analysis-enter patient and SFS long & lat

Bernoulli and Poisson Models:Kulldorff M. A spatial scan statistic. Communications in Statistics: Theory and Methods,26:1481-1496,1997

September 1, 2006

PSC clusters

Global and focused cluster analysis of PSC-OLT patients revealed statistically significant clusters:

• one that encompassed all of Staten Island (p=0.050)• a cluster in Nassau County, N.Y. that is near a known SFS

(not shown).• a cluster in Chicago, Illinois. We have started investigating

Chicago given higher numbers of Scandinavians in that city.

The cluster encompassing Staten Island encircles too many SFS to identify any specific toxin. The toxin at the Nassau County SFS included only tetrachloroethylene (PCE). However, as in Staten Island, a large county-wide active garbage disposal site is also present in this zip code.

September 1, 2006

BIGGER TRIANGLES Brookl

yn

Queens

Bronx

Man

hat

tan

Staten Is.

New Jersey

Nass

au

Global Clusters

Active Solid Waste SiteSuperfund Sites

6

6

4

2

2

10 km10 km

1919

77

1010

77

1010

Mount Sinai Clusters

Two global clusters identified by SatScan overlap with Mount Sinai PBC patient clusters

September 1, 2006

•SFS were present within 5 out of 6 clusters, but the clusters were not statistically significant.

PBC clusters identified by global PBC clusters identified by global analysis were not statistically significantanalysis were not statistically significant

September 1, 2006

PBC SFS-focused analysis identified PBC SFS-focused analysis identified two statistically significant clusterstwo statistically significant clusters

• Statistically significant SFS-focused clusters

1. Westchester (10595, 10532) r=2.67 km, p<0.05

2. Staten Is (10312, 10308) r=3.69 km, p=0.05

September 1, 2006

Staten Island SFS ToxinsStaten Island SFS Toxins

Number of organic compounds including:• polychloroethane (PCE), predominately• Heavy metals e.g. mercury• Solvents• Lacquer

Exposure to toxins by aerosol distribution more likely since groundwater not used in New York City. Toxins may attach to particulate matter for wider distribution.

September 1, 2006

Fresh Kills - Staten IslandFresh Kills - Staten IslandThe largest land fill in the worldThe largest land fill in the world

Courtesy of the NYC DEC 2003

September 1, 2006

Related Animal StudiesRelated Animal Studies

• We have begun to study hepatic changes in mice exposed to particulate matter air pollutants 5 hours per day/ 5 days per week for variable periods.

• Early results indicate that exposure for 2+ months significantly increases hepatic inflammation with a trend towards increased fibrosis, but cholestatic changes have not been observed.

• Different mice strains may yield different results.

September 1, 2006

SummarySummary

• The overall prevalence of only PBC patients listed for transplant is increased in zip codes near NYC SFS.

• The increased prevalence of PSC and PBC patients listed for transplantation in Staten Island corresponds to its high density of SFS

• Statistically significant clustering of both PSC and PBC patients listed for transplantation occurs near SFS.

September 1, 2006

LimitationsLimitations

• Study population-Cases that never progress beyond early stage disease are excluded.

-Economics may affect both SFS location and who is transplanted. -Relatively small number of cases limited the study’s power.

-The accuracy of OPTN database is unknown.

• Using zip codes -Exact addresses are not available from OPTN. -Only the zip code at the time of listing for transplant is saved.

• Migration -Length of time residing in given zip code prior to listing is unknown.

-In NYC about 50% were living in the same house in 2000 as in 1995.

September 1, 2006

ConclusionConclusion

Exposure to toxic wastes may be one of the environmental factors that plays a role

in the pathogenesis of PSC and PBC

September 1, 2006

ACKNOWLEDGEMENTSACKNOWLEDGEMENTS

This work was supported in part by Health Resources and Services Administration contract 231-00-0115. The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government

ARTZT FAMILY PBC FOUNDATION

September 1, 2006

September 1, 2006

GLOBAL SaTScanGLOBAL SaTScanTMTM ANALYSIS ANALYSIS

• Most likely PBC clusters– SI (10312, 10308, 10309) r=3.44 km, p=0.17

FOCUSED SaTScanTM ANALYSIS

• Most likely PBC clusters– SI (10312, 10308, 10309) r=3.69 km, p=0.05

September 1, 2006

PBC-Staten IslandPBC-Staten Island

MOST LIKELY

PBC CLUSTER

September 1, 2006

Large geographical variations in disease (PBC) frequency, both between and within studies, tantalizingly suggest the presence of as-yet-unidentified risk factors. This should be further followed up with new analytical epidemiological studies.

September 1, 2006

Prevalence May Vary Locally Due to Prevalence May Vary Locally Due to Genetic or Environmental FactorsGenetic or Environmental Factors

1) Genetic predisposition

- Family and twin studies support a link

2) Environmental factors

- Differing water supplies has been linked to local variability of PBC.

(Triger et al. Br Med J. 1980)

- Cross-reactivity of PBC autoantibodies with microbial protein epitopes and with autoantigen modified by environmental chemicals (xenobiotics)

September 1, 2006

XenobioticsXenobiotics

• Xenobiotics are foreign chemicals that may alter defined self-proteins, inducing a change in the molecular structure of the native protein sufficient to induce an immune response

• Association of autoimmune diseases with xenobiotics. Carpenter D.O et al. Incidence of endocrine disease among

residents of New York areas of concern. Environ Health Perspect 2001.

• Many xenobiotics are metabolized in the liver

September 1, 2006

PBC in NYCPBC in NYC An increased number of PBC cases noted in Staten Island. Referral An increased number of PBC cases noted in Staten Island. Referral

bias?bias?

Staten IslandStaten Island

ManhattanManhattanBrooklynBrooklyn

QueensQueens

BronxBronx

September 1, 2006

Staten Island is home to a huge garbage dump and Staten Island is home to a huge garbage dump and numerous other toxic sitesnumerous other toxic sites

September 1, 2006

Potential geoclustering of Mount Sinai Potential geoclustering of Mount Sinai patients (212) with AMA+ PBCpatients (212) with AMA+ PBC

10 km10 km

1919

77

1010

77

44

1010