Does increased detection account for the rising incidence of breast cancer?

BACKGROUND

The incidence of breast cancer has been increasing over time in the United States.

METHODS

To determine the role of screening in this increase, trends in the incidence of in situ and invasive carcinoma of the breast were evaluated using records of the metropolitan Atlanta SEER program between 1979 and 1986. From a sample of records, evidence of symptoms and mammographic screening prior to diagnosis was recorded.

RESULTS

The average annual age-adjusted incidence of invasive disease rose 29 percent among Whites and 41 percent among Blacks. Incidence increased in all age groups. A trend towards earlier detection of invasive disease was found. Asymptomatic tumors accounted for only 40 percent of the increased incidence among whites and 25 percent of the increased incidence among blacks, with mammography as the principal contributing procedure.

CONCLUSIONS

These data suggest that increased detection accounts for some but not all of the rising incidence of breast cancer in the United States.

Relation of serum testosterone levels to high density lipoprotein cholesterol and other characteristics in men

Although levels of high density lipoprotein (HDL) cholesterol in males decrease during adolescence and after treatment with testosterone derivatives, several studies have reported that levels of HDL cholesterol are positively associated with endogenous levels of testosterone in men. This association was further examined using data collected during 1985 and 1986 from 3,562 white and 500 black men who ranged in age from 31 to 45 years. Black men had higher mean levels of both HDL cholesterol (8 mg/dl) and total testosterone (33 ng/dl) than white men, and positive associations were observed between testosterone and HDL cholesterol levels (r = 0.22, whites; r = 0.26, blacks). In addition, levels of testosterone were related positively to alcohol consumption and cigarette smoking and negatively to age, Quetelet index, and use of beta-blockers. We used stratification and regression analyses to determine if any of these characteristics could account for the positive association between levels of HDL cholesterol and total testosterone. Although controlling for most factors had little influence, adjusting for Quetelet index reduced the strength of the association between levels of testosterone and HDL cholesterol by approximately 30%. These findings suggest that the positive association between levels of testosterone and HDL cholesterol may not be causal. Multivariable analyses that control for obesity and other potentially confounding characteristics should be used in studies that assess the relation of testosterone levels to coronary heart disease.

Commentary: the affected sib-pair method in the context of an epidemiologic study design

The purpose of this commentary is to provide a framework for using the well-known sib-pair methodology in the context of epidemiologic study designs. Using examples from the Pittsburgh family studies of insulin-dependent diabetes mellitus, we illustrate that the sib-pair method can be used in family-based epidemiologic studies. In a cohort study, unaffected relatives of probands ascertained from well-defined populations are followed for disease development. Disease risks are then stratified according to the number of alleles at one or more loci (0, 1, 2) that are identical by descent (ibd) with the proband. In the absence of linkage between the marker locus and the disease locus, disease risks are expected to be identical in the three groups. Measures of relative risk can be computed (with share-0 as baseline group). In a case-control study, relatives of probands that become affected (cases) are compared to a sample of relatives of probands that stay unaffected (controls) with respect to the number of alleles ibd with the proband. Measures of odds ratio can be computed (with share-0 as baseline group). In both cohort and case-control approaches, covariates including other genetic markers and environmental exposures can be evaluated in relation to disease risk and also for evidence of interaction with the specific marker of interest using stratified and multivariate analyses. Family-based epidemiologic studies allow investigators to study, in a single design, the role of environmental factors and specific gene loci in the etiology of diseases.

Extensions to methods of sib-pair linkage analyses

Sib-pair methods provide simple, robust, easily implemented ways to screen for linkage between a marker locus and a suspected disease susceptibility locus. The basic analysis reflects the idea that, in the presence of linkage, siblings who share more alleles at the marker locus should also tend to be concordant for disease. Available sib-pair methods do not lead directly to estimates of risk associated with nongenetic factors, may not account for a variable age-at-onset, or may require that the age-at-onset distribution be known. In this paper, we propose a method for sib-pair linkage analyses that allows for a variable age-at-onset using a logistic model, easily allows modelling of nongenetic factors, reflects the correlation of sibs within a sibship, and allows for nonzero risk in those without the susceptibility genotype. Based on a limited number of simulations, the method has as good or better power than another recently described method that also allows for a variable age-at-onset.

Estimating benefits of screening from observational cohort studies

Analysis and interpretation of observational studies of screening effectiveness is difficult because several biases threaten validity, including the structural healthy screenee bias, length bias, and effects of lead time. Although methods for the analysis of observational studies of screening effectiveness have been proposed, most have limitations such as incomplete control of length bias, or a heavy reliance on distributional assumptions. In this report we present a method for the analysis of observational cohort studies of screening effectiveness. Although developed independently and formulated specifically for estimating benefits of screening, our approach is implied by a more general approach developed previously by Robins. Our approach, in contrast to other available methods, avoids the healthy screenee bias, and length and lead time bias, and allows an empirical approach to analysis that need not depend highly on distributional assumptions. We illustrate application of the approach with analysis of published data from a study of breast cancer screening.

Indirect assessment of confounding: graphic description and limits on effect of adjusting for covariates

Confounding is recognized as a mixing of effects that can lead to spurious conclusions about the association between disease and a putative risk factor. Confounding occurs if an extraneous factor causes disease and is associated with the exposure of interest. Since information on potential confounders may be missing, the investigator may assess confounding indirectly by specifying values for three types of parameters: the prevalence of the covariate in the population, the association between exposure and the covariate, and the effect of the covariate on disease. Qualitative and quantitative arguments suggest that adjustment for a potential confounder may have small effects on the risk ratio, even if the confounder is a strong risk factor. In this report we illustrate graphically the effect that adjustment for a confounder will have on the risk ratio and derive limits for the magnitude of that effect. Our approach allows the investigator to calculate limits for the maximum effect of covariate adjustment, even if only one or two of the relevant parameters can be specified.

Estimation of risk ratios in case-base studies with competing risks

The case-base study is a recently developed modification of the case-control design that leads to risk ratio estimates. Currently available methods for the analysis of case-base studies lead to estimates of the unconditional risk ratio which may be misleading of censoring occurs. In this manuscript, we describe an approach for the analysis of case-base studies that yields conditional risk ratio estimates which remain valid in the presence of censoring. The approach we propose for estimating risk ratios and their standard errors involves a non-iterative procedure.

Illustration of the effects of genotype misclassification on the measurement of familial aggregation in epidemiologic studies

Exposure misclassification is a well-known problem in epidemiologic studies and can considerably dilute relative risk (RR) measures toward unity. A vivid illustration of such misclassification occurs in familial aggregation studies, where "exposure" is defined as a specified genetic relationship to an affected individual (case) or to an unaffected individual (control). Even in the simplest single-gene model, only a fraction of relatives of cases have the disease genotype. Also, if the trait is not fully penetrant, a proportion of relatives of controls can have the genotype. Thus the familial RR measures are subject to misclassification bias if they are interpreted as the relative effect of a susceptible genotype. The effects of this form of "exposure" (or more properly genotype) misclassification on familial RR measures were quantified and applied empirically to known disease-genetic trait associations. Expected RRs in first-, second-, and third-degree relatives were generated and plotted for different genotype RRs, disease, and allele frequencies. In general, familial RRs are substantially lower than genotypic RRs. Even in the case of strong associations such as HLA-B27 and ankylosing spondylitis (RR = 100), RR measures in first-, second-, and third-degree relatives are only 6.0, 3.5, and 2.2, respectively. Such strong misclassification effects may result in considerable reduction of statistical power in family studies. It is suggested that etiologic studies of disease explore directly the role of measurable genetic traits in epidemiologic studies in populations as well as in families.

Limitations of the case-exposure study

The case-exposure study, described in 1983, is a modification of the case-control study that allows estimation of risk ratios without need for a rare-disease assumption. For a fixed population the approach is the same as the case-base or case-cohort study. Two sampling strategies have been described for extending application of the case-exposure approach to stationary populations, although the validity of the extension was not proven. As shown here, the two sampling strategies for stationary populations can provide valid estimates of the risk ratio, but restrictive assumptions related to stationarity may limit application of the strategy. We also show that the case-exposure study depends on the stationarity assumptions in a different, possibly more restrictive way than some other case-control approaches. We emphasize limitations of the case-exposure study that may preclude its use, particularly if the period of risk is long.

Are racial differences in the prevalence of diabetes in adults explained by differences in obesity?

To determine whether the higher prevalence of diabetes found among blacks in the United States is explained by racial differences in obesity, we examined the prevalence of diabetes adjusted for adiposity, education, and income in a cohort of US Army veterans from the Vietnam era. Among 12,558 white men and 1677 black men, aged 30 to 47 years, blacks were more likely than whites to have diagnosed diabetes (adjusted prevalence ratio, 1.9; 95% confidence interval, 1.3 to 2.7). Within every age, adiposity, and socioeconomic stratum, blacks had a higher prevalence of diagnosed diabetes than whites. In a subgroup of veterans for whom fasting serum glucose values were measured, blacks were more likely than whites to have fasting hyperglycemia (fasting serum glucose value greater than or equal to 7.8 mmol/L) (adjusted prevalence ratio, 5.7; 95% confidence interval, 2.7 to 12.0). These data provide evidence that the higher prevalence of diabetes found among blacks is not explained by differences in obesity.

Inappropriate comparisons of incidence and prevalence in epidemiologic research

Several epidemiologists have published papers in major medical journals in which they compare incidence rates and prevalence and use these comparisons to support conclusions regarding questions of major public health importance. Although these papers have been criticized in published correspondence, we believe that continued use and advocacy of such comparisons by some epidemiologists has created the need for a full discussion of this practice. In this commentary, we review basic differences between incidence and prevalence and show that direct comparison of these two measures is inappropriate for conceptual, theoretical, and practical reasons.

Bias arising in case-control studies from selection of controls from overlapping groups

Three examples are used to demonstrate that the selection of controls from categories that overlap can lead to bias. Case-control studies that use friend controls and the practice of age-matching controls to cases within a specified number of years (caliper matching) are examples of the selection bias described in the paper. The bias resulting from the use of friend controls can be large and, since the use of friend controls is common, this source of bias is of considerable practical consequence. The use of friend controls or the selection of controls according to other variables with categories that overlap should be avoided.

Human teratogens, prenatal mortality, and selection bias

Etiologic inferences on human teratogens are usually derived from case-control studies conducted either at birth or in spontaneous abortion series. Because both teratogens and defects may be associated with an increased risk of prenatal mortality, the possibility exists that selection bias may affect etiologic inferences. The authors derive relations between the true odds ratio (OR) relating a teratogen and a defect at the time of the occurrence of the defect and the apparent odds ratios observed in spontaneous abortion series and at birth, as functions of prenatal mortality. Depending on the pattern of interaction between the teratogen and the defect in affecting the rate of prenatal mortality, selection bias may lead to overestimation or underestimation of the true odds ratio. With increasing multiplicative effects on prenatal mortality, changes in selection bias tend to increase the observed odds ratio in spontaneous abortion series but to decrease the observed odds ratio at birth. For a range of rates of prenatal mortality, weak associations between exposures and defects (OR = 0.3-3) may well be due to selection bias; conversely, weak teratogens (OR less than 3) may be missed in case-control studies of defects conducted at birth. Selection bias due to prenatal mortality must be considered in the interpretation of etiologic studies of birth defects.

On the measurement of susceptibility in epidemiologic studies

Although relative effects of risk factors (relative risks) are commonly used in epidemiologic studies of disease, these measures do not provide estimates of the proportion of persons who are "susceptible" to the risk factor. Susceptibility may be defined under a simple sufficient cause model as the underlying factor (or set of factors) sufficient to make a person contract a disease following exposure. The authors derive simple estimates of the proportion of susceptibles in a population based on relative risk, and disease and exposure frequencies. The proportion of susceptibles increases with increasing disease frequency and relative risk but declines at high exposure frequency. For many chronic diseases with a lifetime risk in the range of 1-10 per cent, rare exposures suggest the presence of a large proportion of susceptibles, whereas common exposures suggest fewer susceptibles in the population. The estimation of the proportion of susceptibles is important in the search for genetic and environmental factors that interact with measured risk factors in etiologic studies of disease.

Bias associated with differential hospitalization rates in incident case-control studies

Berkson's bias reflects a statistical phenomenon in which differential hospitalization rates create an exposure distribution among hospitalized cases that differs from that among other cases. Importantly, previous work on Berkson's bias has not explicitly addressed the possibility of excluding prevalent or previously diagnosed cases--exclusions that are key features of many study designs. We indicate that the classically described bias differs from the corresponding bias in studies, such as incidence density studies, in which cases are restricted to those with recent diagnoses. We present methods that may be used to assess the magnitude of Berkson's bias in incidence-density studies. In many, though not all, situations the bias should be small and of little practical concern.

On the measurement of susceptibility to genetic factors

Geneticists usually measure the phenotypic effects of a single gene trait in terms of penetrance and recurrence risks in relatives of affected individuals, while epidemiologists usually compute measures of relative and attributable risks. These concepts can be merged to measure the proportion of individuals "susceptible" to a genetic factor. In the context of a sufficient cause model, susceptibility can be defined as the underlying factor(s) sufficient to make a person develop disease because of the genetic factor in the absence of other causes. The proportion of susceptibles to the genetic factor in the population differs conceptually and often arithmetically from the penetrance of the genotype especially for common diseases with etiologic heterogeneity. For a wide range of disease and allele frequencies, it can be shown that the proportion of susceptibles can be approximated by the risk difference measure (i.e., difference between penetrance with the genotype and penetrance without the genotype). We also apply the concept of susceptibility to estimate familial recurrence of disease due to a genetic factor. This measure of familial recurrence differs conceptually from simple recurrence risk and can be approximated by the familial risk difference measure (i.e., difference between recurrence risks in relatives of cases and relatives of controls) for a wide range of disease and allele frequencies.

Penetrance in the presence of genetic susceptibility to environmental factors

Many single-gene disorders (especially dominant traits) are known to exhibit "incomplete penetrance," a term often criticized as reflecting our ignorance of genetic and environmental factors that may interact with the genotype to determine the ultimate phenotype of the individual. We explore the effects of an environmental factor on penetrance for a simple model of single gene-mediated susceptibility to environmental factors. We show that penetrance of the susceptible genotype increases with increasing frequency of exposure to the factor and the strength of interaction between the factor and the genotype (relative risk). For disorders with disease frequency of 0.001 or less (as seen with many congenital malformations), penetrance is generally low (less than 10%) if the exposure frequency is low (less than 10%) even in the face of strong gene-environment interaction. Such low penetrance can lead to low recurrence risks. Single-gene effects could easily be overlooked in genetic analysis unless specific environmental factors are considered.

Chemical exposures and central nervous system cancers: a case-control study among workers at two nuclear facilities

In a nested case-control study of workers employed between 1943 and 1977 at two nuclear facilities, we evaluated the possible association of primary CNS cancers with occupational exposure to chemicals. Seventy-two white male and 17 white female workers who, according to the information on death certificates, died of primary CNS cancers were identified as cases. For each case, four controls were matched on race, sex, facility at which initially employed (cohort), year of birth, and year of hire. Each job title/department combination was subjectively evaluated for potential exposure to each of 26 chemicals or chemical groups. Statistically significant associations were not found between CNS cancer deaths and any of the 26 chemicals. An increased risk of CNS cancer occurrence was observed among subjects employed for more than 20 yr (OR = 7.0, 95% CI = 1.2,41.1, cases = 9).

An epidemiologic approach to ecogenetics

Although "ecogenetics" seeks to examine genetically mediated differences in susceptibility to environmental agents, researchers often examine the relation between genetic markers and disease without regard to environmental determinants. By using epidemiologic definitions of genotype-environment interaction, it can be shown that the relative risk of disease for the genetic marker is a function of the frequency of exposure to the environmental agent, the strength of interaction between the genotype and the agent, and the specificity of the environmental effect vis-à-vis the genotype. Using examples from the literature, we illustrate under six patterns of genotype-environment interaction that the relative risk associated with the marker can fluctuate markedly. However, with infrequent exposures, the relative risk is close to unity (implying no genetic effect) even in the face of strong genotype-environment interaction. Alternatively, elevated relative risks imply a frequent environmental exposure or a strong pattern of interaction. We suggest that genetic marker-disease associations be evaluated within the context of an epidemiologic study design that considers specific environmental determinants of risk.

Brain cancer and nonoccupational risk factors: a case-control study among workers at two nuclear facilities

In a nested case-control study of nuclear workers, 82 brain cancer cases were compared with 328 matched controls to investigate the possible association with nonoccupational risk factors such as histories of epilepsy or head injury. We observed a moderately strong association between brain cancer occurrence and history of epilepsy (OR = 5.7, 95 per cent CI: 1.0, 32.1), but did not find a positive association with previous head injury (OR = 0.9, 95 per cent CI: 0.2, 4.2).

CNS cancers and radiation exposure: a case-control study among workers at two nuclear facilities

A nested case-control study was conducted among workers employed in two nuclear facilities to investigate the possible association of primary CNS cancers with occupational exposure to radiation from external and internal sources. External radiation monitoring data from film badges were available for 27 cases and 90 matched controls. The radiation dose to the lung from internally deposited uranium was estimated for 47 cases and 120 matched controls from area and personnel monitoring data and was used in analyses in lieu of the brain dose. No association was observed between deaths from CNS cancers and exposure to ionizing radiation from external or internal sources. However, due to the small number of monitored subjects and low doses, a modest association could not be ruled out.

A comparison of estrogen and progesterone receptors in black and white breast cancer patients

After standardization for age and menopausal status, the prevalence of estrogen receptor positivity among 88 White breast cancer patients was about .72 compared with a prevalence of about .54 among Black patients. The prevalence of progesterone receptor positivity was also higher among White than among Black patients, although the magnitude of the difference was smaller. These differences were unchanged after adjustment for tumor size and nodal and distant metastases in addition to age and menopausal status.

Large sample confidence intervals for regression standardized risks, risk ratios, and risk differences

Several methods have been proposed for standardizing risks, risk ratios, and risk differences based on the results of logistic regression. These methods provide an alternative to direct standardization, a particularly useful approach when there are many covariates. In this paper, methods for calculating approximate confidence limits for these standardized measures are presented. A simple example, in which published data are used, illustrates the techniques and allows comparison with confidence limits calculated from the directly standardized risk ratio.

The exposure odds ratio in nested case-control studies with competing risks

A nested case-control study, also known as an ambidirectional study, is a case-control study within a cohort study. Although distortion by competing risks is well-recognized in follow-up studies, the problem has not been as widely appreciated in nested case-control studies. This paper extends previous work concerning the bias associated with competing risks for nested case-control studies. Specifically, the distorting effect of competing risks is illustrated for three methods of control selection. Assuming the proportional hazards model, the authors derived formulas for the bias of the odds ratio when competing risks cannot be ignored. Examples illustrate the magnitude of bias that occurs when the exposure of interest is associated with competing causes of death or withdrawal.

Approximate variance formulas for standardized rate ratios

Some of the techniques which are used to estimate the variance of and confidence intervals for standardized rate ratios either ignore variability of comparison rates or tend to yield confidence intervals which are too narrow when the point estimate is substantially different from one. This paper presents non-iterative, asymptotic formulas for the variance of standardized rate ratios which are applicable when the comparison rates should not be treated as constants, or when the point estimate differs substantially from the null value.

Review: prostate cancer epidemiology

Prostate cancer is common among men in the United States. Factors of possible importance in the etiology of prostate cancer include diet, primarily implicated by ecologic studies of national, regional, and ethnic variation in rates; endocrine function, implicated by the importance of endocrine function in normal prostatic growth and in the treatment of prostate cancer; genetic susceptibility, supported by familial aggregation; some aspect of sexual behavior, suggested by case-control differences in sexual behavior; and occupational exposure, particularly cadmium exposure. Despite the public health importance of prostate cancer, it has received only moderate epidemiologic study; thus the etiologic importance of these and other possible determinants of prostate cancer risk is uncertain [55].

Work-related risk factors for laryngeal cancer

The authors conducted a case-control study to identify employment-related risk factors for laryngeal cancer. Richmond County, Georgia, and the contiguous counties were chosen as the study area because the 1950-1969 laryngeal cancer mortality rate for white males in Richmond County was almost double the corresponding rate for the entire United States, and because a large proportion of the working population was employed in industries for which excess risk among workers had been reported. All persons with newly diagnosed squamous cell carcinoma of the larynx treated in the area hospitals were identified. Potential controls were selected from patients admitted to these hospitals during the same time period as the cases, excluding those patients admitted for cancer, lung disease, or employment-related disease. Eighty-five control subjects were individually matched to 42 cases for sex, age, area of residence, and smoking and alcohol-drinking history. From lifetime employment histories, laryngeal cancer rate ratios were estimated comparing the incidence rate among subjects who had ever worked in an occupation with the incidence rate among subjects who had never worked in that occupation. Among the 10 occupations in which at least 15 subjects had worked, the authors found rate ratio estimates above 3.0 for farmers, textile processors who separated, filtered or dried textile fibers, and for all laborers and maintenance personnel.

Occupational risk for laryngeal cancer

In a case-control analysis, we studied the effects of type of employment on laryngeal cancer risk using the interview data from the Third National Cancer Survey. Effects were measured relative to the risk for those employed in a group of arbitrarily defined industries and occupations with low risk. We excluded females and controlled for age, tobacco use, alcohol use, and race in the analysis. We found ratio estimates above 3.0 for workers in the railroad industry and the lumber industry; and for sheetmetal workers, grinding wheel operators, and automobile mechanics.

Interaction of alcohol and tobacco in laryngeal cancer

Both alcohol and tobacco use are accepted risk factors for laryngeal cancer. The authors used case-control data from previous studies to estimate the value of a previously proposed index of interaction between these two risk factors. In addition to the weighting procedure over exposure categories that was previously proposed for estimating a summary index, they applied maximum-likelihood techniques to facilitate the estimation. Overall, they found moderate synergy between alcohol and tobacco in increasing the risk of laryngeal cancer, in that exposure to both factors increased the risk about 50% more than the increase predicted if the effects of tobacco and alcohol were simply additive.