personal risk factors what do they mean?

Personal Risk Factors

What Do They Mean?

GUY R. NEWELL, MD, AND VICTOR G. VOGEL, MD, MHS

Epidemiologic studies have identified myriad factors related to cancer risk. Risk can be quantified on the basis of demographic factors, genetics, occupation, medical conditions, and lifestyle. Existing evidence suggests that: (I) individuals at risk often are unaware of their risk; (2) physicians may not know well those factors associated with the highest cancer risk; and (3) methods to reduce risk have been under-applied because of lack of knowledge, lack of funds, or lack of motivation among both patients and physicians. Methods to reduce risk do not follow the usual medical model in that those groups requiring risk-modification intervention usually are symptom free. Evidence indicates that elimination of tobacco use would reduce lung cancer deaths by 83% and substantially reduce the incidence of head and neck malignancies. Broad application of mammographic screening would effect a 30% reduction in breast cancer mortality. Dietary modification could potentially reduce cancer mortality by 30%. An effective program to reduce cancer risk will accomplish the following: (1) develop new and refine existing risk factor measurements to improve identification of individuals at risk; (2) apply risk factor identification and modification more completely to populations known to be at risk; (3) use combinations of risk factors to identify subpopulations at the highest risk; (4) apply existing prevention and screening modalities more broadly and uniformly; (5) identify new screening strategies with improved test perfor- mance and cost effectiveness to maximize screening efficiency; and (6) provide accurate and rapid means of risk assessment and quantification using the best available multivariate estimates of risk. An empha- sis on education of primary care physicians to practice cancer prevention programs focused on personal risk profiles is reasonable since the populations these physicians serve express risk factors which represent 50% of the attributable risk for cancer in the US. A summary of site-specific risk factors which are amenable to intervention by primary care physicians is provided.

Cancer 62:1695-1701. 1988.

HE TWO MAJOR ARMS of cancer prevention are the T identification of those factors which contribute to the cause(s) of cancer, and the action taken in response to this knowledge. Identification of the causes of cancer is the function of the researcher or the alert clinician, whereas control of the causes may be by legislation or, preferably, by voluntary actions taken on the part of informed individuals and health professionals. The practitioners of cancer prevention are the physicians who provide primary health care and the individuals who are willing to accept this care.’

Presented at the American Cancer Society Second National Confer- ence on Cancer Prevention and Detection, Seattle, Washington, June

From the University of Texas M. D. Anderson Cancer Center,

Address for reprints: Guy R. Newell, MD, The University of Texas

Accepted for publication September 24, 1987.

25-21, 1987.

Houston, Texas.

M. D. Anderson Cancer Center, Houston, TX 77030.

Risk Factors

“Risk factors” are those events which contribute to the cause of cancer. Most identified risk factors are environmental, although some risk factors have a genetic predisposition. Webster’s dictionary defines the noun “risk” as “the chance of injury, damage, or loss; danger- ous chance; hazard . . . ; the degree of probability of loss.” As a verb it is defined as “to expose to risk; hazard: as to ‘risk’ one’s life. Syn. see danger.” There are three kinds of risk frequently used in clinical epidemiology (Table 1).

Absolute Risk

The absolute risk is a measure of the occurrence of cancer, either incidence (new cases) or mortality (deaths), in the general population. The factorb) which determine the risk may not be distributed evenly among individuals in the general population; and the factor(s) may not affect individual persons who are exposed to

1695

1696 CANCER October 15 Supplement 1988 Vol. 62

TABLE 1. Major Kinds of Risk

Absolute risk Rate of occurrence (incidence) or deaths (mortality) of disease within the general population

Comparison of occurrence (incidence) or deaths (mortality) among those with a particular risk factor to those without the risk factor

Amount of disease within the population that could be prevented by alteration of risk factor(s)

Relative risk

Attributable risk

the same degree. For example, the increasing incidence and mortality due to lung cancer is a reflection of the number of cigarette smokers in the population because cigarette smoking is the overwhelming risk factor that causes lung cancer. These risk factors either may increase or decrease an individual’s chance of developing or dying from the cancer.

Relative Risk The term “relative risk” (RR) is the measure of risk

more commonly referred to in cancer epidemiology. Relative risk describes those in the general population who have exposures to one or more risk factors. Because of this exposure, this subset of individuals is at increased risk for developing cancer compared to a group of individuals without the exposure. Thus, RR measures the excess chance of developing cancer imposed by a risk factor relative to the risk which may exist without the exposure.

A large RR usually infers a strong association of the risk factor with the cancer. For the association to be causal, it must be linked to the cancer either directly or indire~tly.~-~ As Lilienfeld and Lilienfeld explained, “A causal relationship would be recognized to exist when- ever evidence indicates that the factors form part of the complex of circumstances that increases the probability of the occurrence of disease, and that a diminution of

TABLE 2. Estimated Percentage Contributions of Selected Risk Factors to Annual Deaths*

All Cardiovascular Total cancer disease deaths

Percentage of total

Risk factor deaths 19 51 100

Cigarette smoking 30 30 22 Diet excess 30 30 22 Hazardous worksite 5-10 1-5 5 Alcohol abuse 5-10 5-10 9

13 7 Hypertension - Percentage of deaths

attributable to risk factors 50 50 69

* Adapted from Kottke.*

one or more of these factors decreases the frequency of that di~ease.”~ Most identified risk factors increase the chances of developing cancer. However, some factors such as dietary components may protect against cancer, thus decreasing the risk.

Odds Ratio and Relative Odds The odds ratio (OR) and relative odds (RO) are esti-

mates of RR. These are used in case-control studies of cancer and usually are close approximations to the true RR.

Attributable Risk “Attributable risk” (AR), although not used as com-

monly as the RR, probably is a more important measure of risk for cancer control purposes. The attributable risk is a measure of the amount of disease in the population that would be prevented by alteration of the risk factor(s). For example, a risk factor could convey a very large RR, but if restricted to a few individuals, its alteration would necessarily result in benefit to only a few individuals. On the other hand, a risk factor could convey a small increase in RR, but if it were widely distributed in the population its alteration would prevent much more cancer. Thus, the amount of disease prevented by altering risk factors also is dependent on the prevalence of the disease among the population with which it is associated. This fraction of disease prevent- able by eliminating the risk factor (X 100) has been called the attributable risk percent6 and the etiologic fraction.

Table 2 gives the estimated percentage contribution of the most important risk factors for cancer and cardiovascular disease.’ All cancer makes up 19% of total deaths whereas cardiovascular disease accounts for 5 1 %. Alteration of four risk factors for cancer plus hypertension for cardiovascular disease would result in prevention of 69% of all premature deaths.

Determining Risk Factors

There are two important reasons for identifying and quantitating cancer risks. One is that it contributes to the biologic understanding of cancer. The second reason is that altering or changing these risk factors would re- suit in a decrease of new cases or deaths from cancer.’ Deaths from lung cancer among white men of all ages recently have begun to decline due to a decrease in smoking among this group; widespread use of the Pa- panicolau smear has resulted in a large decline in incidence of and mortality from invasive cancer of the cervix; and application of mammography has resulted in the detection of early breast cancer amenable to curative

No. 8 PERSONAL RISK FACTORS - Newel1 and Vogel 1697

treatment and a demonstrated mortality reduction of at least 30%. Even with this knowledge, the use of mammography for the control of breast cancer is an unreal- ized potential."

Diversity of Risk Factors

There is great variety among the environmental and life-style risk factors associated with cancer.' ' Physical, chemical, and biologic agents all are represented. Sev- eral modes of action are known: genetic mutation, aber- rations during anatomic development, and abnormal responses to physiologic substances. Causal factors can exert their action during the whole span of time: ances- tral times for genetic determinants, in utero, early child- hood, adolescence, young adulthood, and older adulthood. This diversity of risk factors presents both challenge and opportunity. The challenge is to identify and quantitate the diverse factors; the opportunity is that diverse factors should be amenable to varied approaches for their alteration.I2

Site-SpeciJic Risk Factors

In general risk factors are associated with development of one or several types of cancer, rather than with all cancers. This reflects the individuality of most types of cancer in terms of histologic type, tissue of origin, and natural history. For example, the cancer sites for which cigarette smoking conveys an increased risk are not the same sites as those for which diet may exert a protective effect. For purposes of cancer prevention, this may be a fortunate occurrence in that successful preventive activi- ties to reduce smoking and promote a prudent diet could result in the prevention of at least 50% of all cancers. These include cancers of the head and neck, lung and other cigarette-related sites, and cancers of the female breast, large intestine in both males and females, and prostate.

Individuality of Risk Factors

Risk factors are determined by studying groups of individuals exposed to them. Even though each exposed individual may acquire an increased risk, there are no means available to predict whether or not an individual will develop the cancer associated with the risk factor(s). For example, not all who smoke heavily develop lung cancer, although some believe that if every smoker lived long enough, lung cancer would develop in each cigarette smoker.

There is usually a time lag of several years between exposure and development of cancer. Different risk factors manifest themselves in different patterns during the years of exposure, or individuals acquire risks for spe-

6 10 16 20 26 30 36 40 RISK OVER TIME IN YEARS

FIG, 1 . Risk of dying with Iung cancer for white men at 5-year intervals from current age.

cific cancer sites during different ages of the life span. These differences are expressed by different shapes of age-specific probability curves of dying from different cancers a given number of years after one's present age. Such curves reveal information not only about risk, but also important information about cancer biologic fea- tures.

For example, the risk of lung cancer is markedly in- fluenced by increasing age between ages 40 and 60 years, but a less dramatic effect on risk is seen when comparing the 70-year-old group to those a decade younger (Fig. 1). The plateau on the right of the curves for 60-year-olds and 70-year-olds shows that all of the risk for lung cancer already has been acquired in preceding decades. Individuals die from other competing causes, or a protective effect has been present, perhaps mediated by a genetic me~hanism.'~

The graph for white female breast cancer shows much less relative contribution to risk during younger ages, with the largest increase from the fifth to the sixth decades (Fig. 2). Additionally, the risk does not appear to plateau, implying life-long effects from factors other than age (which are well known from epidemiologic studies of breast cancer). The shapes of the curves for breast cancer reflect the large relative contributions to risk from these factors other than age.

A long, progressive increase in risk is seen for prostate cancer among black men (Fig. 3) as they age from 40 to 80 years, and the relative contribution of attained age is more dramatic than for either white male lung cancer or white female breast cancer. This suggests that a factor


- 4 - 1 I

6 10 16 20 26 30 36 40

RISK OVER TIME IN YEARS

FIG. 2. Risk of dying with breast cancer for white women at 5-year intervals from current age.

related to age may be most important to the etiology of prostate cancer in black men.

At some anatomic sites, on the other hand, the chance of cancer developing is so great that individual action is warranted before the cancer actually develops. The prevention of colon cancer by prophylactic colectomy in a young woman with hereditary polyposis of the large intestine is an example of one such risk factor requiring preventive intervention.

d-AGE40

v-AOE60 X = AGE 70

D=AGEM)

- 5 10 16 20 2s ao 36 40

RISK OVER TIME IN YEARS

FIG. 3. Risk of dying with prostate cancer for blacks at 5-year intervals from current age.

Confounding of Risk Factors

Most risk factors do not occur as single entities; rather, they occur as multifactorial risks. This poses a problem in determining which of several factors is most closely associated with the cancer. For example, cigarette smokers also tend to be heavy alcohol drinkers. It is difficult to segregate the individual contributions of tobacco and alcohol to the development of bladder cancer and head and neck cancers. Statistical techniques have been developed which allow more accurate quantitation of individual risk factors when more than one are present.

Logistic Regression

Accurate estimates of multifactorial risks require techniques for multiple risk measurements, or multivariate analyses. Multiple linear regression models are widely used in other disciplines (e.g., economics) to sep- arate the effects of multiple variables. Since epidemiolo- gists estimate risks with measures of probability (RR, OR, RO), techniques that yield individual probabilities for multiple risk factors are desirable. Individual probabilities for each of several risk factors are expressed as ranging between 0 and 1. Multiple logistic regression provides an estimate of risk expressed as a probability. The procedure also provides estimates of the logarithms of the OR which relate disease risk to each particular factor under consideration. The procedure also adjusts for all other risk factors being considered. l4 Techniques for performing logistic regression are well de~cribed,'~ and microcomputer software which carries out the calculations is widely available. l'

Assessing an IndividualS Risk

In spite of suitable statistical procedures and microcomputer software that makes the calculations easy to perform, large data sets for specific cancer sites that allow such multivariate analyses are lacking. On the other hand, the Framingham Heart Study provided a large data set for several heart disease risk factors col- lected within a single large cohort of persons." This allowed for multivariate estimates of known risk factors for heart disease to be obtained for both mortality and morbidity outcomes.

When population-based cohort studies of multiple risk factors for cancer sites become available, similar accurate estimates of site-specific cancer risks will be possible.'* The importance of such estimations is that they allow quantificatian of individual risk in the clinical setting. Programs which automate personal risk appraisal are inexpensive and easy to use" once multivariate risk coefficient estimates become available.

No. 8 PERSONAL RISK FACTORS - Newel1 and Vogel 1699

TABLE 3. One-in-One-Million Risk of Dying as a Result of the Exposure

Exposure (situation) Resulting cause of death

Traveling 700 miles by air Accident Crossing the ocean by air Cancer from cosmic rays Traveling 60 miles by automobile Accident Visiting for 2 mo in Denver Cancer from cosmic rays Living for 2 mo in a stone building Cancer from radioactivity Working for 1.5 wk in a typical factory Accident Working for 3 hr in a coal mine Accident Smoking from 1 to 3 cigarettes Cancer, heart-lung disease Rock climbing for 1.5 min Avcident 20 min being a man 60 yr of age Mortality from all causes

Reprinted with permission from Upton AC. The biological effects of 1982 by Scientific American, low-level ionizing radiation. Copyright

Inc. All rights reserved.

For the optimal practice of prevention, the physician would like to be able to individualize knowledge of risk factors.20,2' Until this is possible, every patient with the risk factor(s) should be considered a candidate for the development of the cancer(s) associated with the factor(s), but it should be made clear to patients that altering a harmful habit or adopting healthy ones does not guarantee escaping the development of cancer.

Assessing Group Risks The maximum potential for effective preventive in-

terventions can be achieved, in part, through factor profiles which quantify risk and facilitate identification of groups at highest risk for disease. Applying secondary prevention to selected populations with a known increase in disease prevalence improves screening test per- formance. For example, the use of screening mammography (with a sensitivity of 80% and a specificity of 99%) in a population of women with a breast cancer prevalence of 60/100,000 results in a predictive value of a positive test of only 0.05. That is, only 5% of women whose mammograms appear abnormal actually will have biopsy-proven breast cancer. Conversely, in a population where the prevalence of breast cancer is 500/ 100,000, the predictive value of a positive mam- mogram increases to 29%.22 Thus, selective screening strategies based on risk factor profiles can improve screening efficiency by maximizing screening test per- formance. Screening costs can be reduced by focu$ng on those individuals at highest risk.23 Early identification of those at greatest risk through risk quantification will have the greatest impact on mortality reduction by focusing modifying interventions where the mortality and morbidity risks are the greatest.24

Individual Perception of Risk It is difficult to conceptualize the magnitude of cancer

risk factors, especially when their deleterious effects may not be manifested for many years in the future. To pro-

TABLE 4. Lifetime Risk of Breast Cancer by Risk Group*

Risk Relative group Risk factors? risk Lifetime risk

Baseline None 1 .o 1/30 (3.3%) Borderline Age younger than 50 yr 1.5 1/20 (5.0%)

Moderate Age younger than 50 yr 2 1/15 (6.7%) one minor factor

+ two minor factors

+ one minor factor

or

Age older than 50 yr

High History of breast cancer 6 1/5 (20.0%)

Mother with breast cancer

or

+ two minor factors

* Data modified from Carter et ~ 1 . ' ~ Minor risk factors: previous cancer other than breast, other relative

with breast cancer, menarche age 10 yr or earlier, nulliparity, first live birth after age 30 yr, menopause after age 55, benign breast disease.

Older than age 50 yr

TABLE 5. Perception of Risk ~ ~~~~

Ranks* given by Annual no. of Group Group Group

Rank/risk deaths I t 2$ 3§

1 Smoking 2 Drinking alcohol 3 Motor vehicles 4 Handguns 5 Electric power 6 Motorcycles 7 Swimming 8 Surgery 9 X-rays

10 Railroads 11 General aviation 12 Large construction 13 Bicycles 14 Hunting 15 Home appliances 16 Fire fighting 17 Police work 18 Contraceptives 19 Commercial aviation 20 Nuclear power 2 1 Mountain climbing 22 Power mowers 23 Scholastic football 24 Skiing 25 Vaccinations 26 Food coloring 27 Food preservatives 28 Pesticides 29 Prescription antibiotics 30 Spray cans

150,000 100,000 50,000 17,000 14,000

3000 3000 2800 2300 1950 1300 1000 1000 800 200 195 160 150 I30 100 30 24 23 18 10 0 0 0 0 0

20 20 4 3 4 6 4 1 3 I 28 1 6 3 2 2 6 16

11 2 17 17 17 20 28 18 8

8 16 14 16 8 I 1 12 27 21 14 30 12 30 12 13 21 11 28 13 19 24 19 9 7 5 14 19 7 5 5

18 26 10 24 29 23 9 21 18

23 10 30 Id 13 9 27 24 22 26 23 29 22 15 15 29 22 27 15 25 25 25 7 26

* Spearman's rank correlation coefficient (RS). t League of Women Voters. $ College students. § Business and professional club members. Reprinted with permission from Upton AC. The biological effects of

low-level ionizing radiation. Copyright @ 1982 by Scientific American, Inc. All rights reserved.


TABLE 6. Risk Quantification by Anatomic Sites Where Patient Education Could Modify Risk

Risk factors Type of Anatomic site known Risk factors modifiable prevention Physician preventive action

Lung Yes Tobacco Head and neck Yes Tobacco, alcohol

Uterine cervix Yes Sexual practices

Breast Yes Diet (possible) Estrogen use Radiation

Colon Yes Diet (probably)

Skin Yes Sun exposure

Testicle Yes Surgical correction of maldescent, inguinal hernia

Primary Primary

Secondary

Primary Secondary Primary? Secondary

Primary? Secondary

Primary Secondary

Primary Secondary

Smoking cessation Smoking cessation Reduce alcohol intake Oral examination Treatment of leukoplakia Modification of risk factors Pap smear Reduce dietary fat, increase fiber Screening

Physician breast exam Mammography BSE instruction

Recognition of family syndromes Reduce dietary fat, increase fiber Screening

Digital exam Fecal occult blood Flexible sigmoidoxopy

Sunscreen Physical examination Recognition of family syndromes Surgery Surveillance TSE instruction

TSE: testicular self-examination; BSE: breast self-examination.

vide reference points against which to compare the magnitude of cancer risk factors, Table 3 lists some fa- miliar events.25 The risk, or chance, of any one of these events occurring is 1 in 1,000,000. A 1% chance of an event occurring is equivalent to 10,000 in 1,000,000. By comparison, if a woman’s risk for developing breast cancer were increased by I%, that would be an increase of 10,000 in a 1,000,000 or 10,000 times greater than any event listed in Table 3. Known breast cancer risk factors can increase a woman’s lifetime breast cancer risk by more than 1%, but the women at risk are unlikely to be aware of this fact (Table 4).

The annual number of deaths from 30 events or “risk factors” are listed in descending rank order in Table 5.25 The 30 events were given to three different groups and the members of the groups were asked to rank the 30 items in the order of importance as contributing causes of death. Among the six leading events, Group 1 named five of six correctly, Group 2 named four correctly, and Group 3 named five correctly. Nuclear power, which was ranked number 20 was named first by Groups 1 and 2; no group included electric power among the first six highest ranked; and no group listed the leading cause of deaths correctly, namely smoking, although it was included among the top six by each group.

Evidently, risks are misperceived by a broad spectrum of societal groups. This misperception of risk may be

one factor that has led to under-application of screening and primary preventive modalities in the population.” Attributable risks for cancer deaths are probably not well understood by the public since Table 5 indicates that the adverse effects of smoking, for example, were badly misjudged by individuals in two of the three re- ported surveys. The public often seeks to identify occu- pational or industrial toxic causes of cancer risk although the best evidence suggests that 10% or fewer of cancer deaths can be attributed to such causes.’ This misperception of risk serves as a barrier to the application of preventive measures. Other bamers to preventive measures (e.g., cost) also are operating, but modification of risk perception possibly is more easily accom- plished.

Table 6 shows those cancer sites for which modifiable risk factors exist if individuals at risk can be shown the relationship of these risks to disease occurrence. Physi- cians must be encouraged to and remunerated for sys- tematically assessing individual patient’s risks, and for intervening with appropriate primary and secondary measures. Physician understanding and use of cancer detection and early interventions could be improved, and evidence suggests that physicians welcome guide- lines for specific preventive interventions.26 Mecha- nisms to increase physician awareness and application of preventive interventions must be sought and imple-

No. 8 PERSONAL RISK FACTORS - Newell and Vogel 1701

mented. In addition, the search for more accurate screening procedures must continue, utilizing new and innovative techn~logies,~’ especially at those anatomic sites where mortality is high, risk factors are known, and the current screening tests are i m p e r f e ~ t . ~ * , ~ ~

REFERENCES

I . Cullen JW, Gritz ER. Behavior modification for cancer prevention. In: Newell GR, ed. Cancer Prevention in Clinical Medicine. New York: Raven Press, 1983; 209-229. 2. Lilienfeld AM. On the methodology of investigations of etiologic

factors in chronic diseases. Am J Epidemiol 1959; 10:41-46. 3. US Surgeon General. Smoking and health. Washington, D.C.:

US Department of Health, Education, and Welfare, 1964; PHS publication no. 1 103. 4. Hill AB. The environment and disease: Association or causation.

Proc R SOC Med 1965; 58:295-300. 5. Lilienfeld AM, Lilienfeld DE. Foundations of Epidemiology.

New York: Oxford University Press, 1980; 292-295. 6. Cole P, MacMahon B. Attributable risk percent in case-control

studies. Br J Prev SOC Med 1971; 25:242-244. 7. Miettinen 0s. Proportion of disease caused or prevented by a

given exposure, trait, or intervention. Am J Epidemiof 1974; 99:325- 332. 8. Kottke TE. Disease and risk factor clustering in the United

States: The implications of public health policy. In: Integration of Risk Factor Interventions. Washington, DC: Office of Disease Prevention and Health Promotion, USDHHS, 1986; publication no. 1987- 185- 970. 9. Doll R, Pet0 R. The causes of cancer. J Natl Cancer Inst 198 1;

10. Howard J. Using mammography for cancer control: An unreal- ized potential. CA 1987; 37:33-48.

1 1. MacMahon B. Overview: Environmental factors. In: Fraumeni JF Jr, ed. Persons at High Risk of Cancer: An Approach to Cancer Etiology and Control. New York: Academic Press, 1975; 285-290. 12. Rennert MP, Telch MJ, Farquhar JW, Killen JD. Single versus

multiple risk factor interventions: An examination of the issues. In: Integration of Risk Factor Interventions. Washington, M?: Office of

66:1191-1308.

Disease Prevention and Health Promotion, USDHHS 1986; publication no. 1987-1 85-970.

13. Hsu TC. Genetic susceptibility to carcinogenesis. Cancer Bull 1986; 38:125-128. 14. Kahn HA. An Introduction to Epidemiologic Methods. New

York Oxford University Press, 1983; 100-120. 15. Breslow NE, Day NE. Statistical Methods in Cancer Research.

New York: Oxford University Press, 1980. 16. Carpenter TE. Epidemiologic modeling using a microcomputer

spreadsheet package. Am J Epidemiol1984; 120:943-95 I. 17. Smith KW, McKinlay SM, Thorington BD. The validity of

health risk appraisal instruments for assessing coronary heart disease risk. Am JPublic Health 1987; 77:419-424. 18. Schecter MT, Miller AB, Baines CJ, Howe GR. Selection of

women at high risk of breast cancer for initial screening. J Chronic Dis

19. Lemeshow S, Pastides H, Arrunin JS, Teres D. Application of microcomputer spreadsheet packages as adjuncts to multiple logistic regression analysis. Am J Epidemiol 1985; 122~710-7 16. 20. Wasson JH, Sox HC, Neff RK, Goldman L. Clinical prediction

rules: Applications and methodological standards. N Engl JMed 1985;

21. Simon R. Importance of prognostic factors in cancer clinical trials. Cancer Treat Rep 1984; 68: 185-192. 22. Cole P, Morrison AS. Basic issues in population screening for

cancer. J Natl Cancer Inst 1980; 64: 1263-1272. 23. Carter AP, Thompson RS, Bourdeau RV et al. A clinically

effective breast screening program can be cost-effective, too. Prev Med

24. Morrison AS. Public health value of using epidemiologic information to identify high-risk groups for bladder cancer screening. Semin Oncol 1979; 6:184-188. 25. Upton AC. The biological effects of low-level ionizing radiation.

Sci Am 1982; 246:41-49. 26. American Cancer Society. Survey of physicians’ attitudes and

practices in early cancer detection. CA 1985; 35: 197-213. 27. Alavanja M, Aron J, Brown C, Chandler J. Cancer risk-assess-

ment models: Anticipated contributions from biochemical epidemiology. JNatl Cancer Inst 1987; 78:633-643. 28. Winawer SJ, Fath RB Jr, Schottenfeld D, Herbert E. Screening

for colorectal cancer. In: Miller AB, ed. Screening for Cancer. New York: Academic Press, 1985; 347-366. 29. Love RR. The efficacy of screening for carcinoma of the pros-

tate by digital examination. Am J Prev Med 1985; 1:36-46.

1986; 39:253-260.

3 13:793-799.

1987; 16~19-34.

personal risk factors what do they mean?

Documents