External comparisons with the case-cohort design

The case-cohort design can be an economic alternative to the standard cohort design. Prentice (Biometrika 1986;73:1-11) showed how the case-cohort design can be used to obtain relative risk estimates for comparisons within the cohort being studied. In this paper, the authors consider ways in which the case-cohort design can be used for comparing risk in exposure groups within the cohort to the risk in an external population. The problem reduces to estimating the number of expected cases at each exposure level in the total cohort, when exposure status is available only for members of a subcohort, i.e., a random sample of the total cohort. The authors describe theoretical and empirical properties of several variations of the design and analysis of case-cohort studies. Empirical properties were examined by replicating the selection of the subcohort in a study of second cancer risk after chemotherapy for a first cancer. Use of a case-cohort design in that study would have saved five-sixths of the cost of gathering covariate information at the price of only an 11% loss in efficiency relative to a full cohort study.     

A cohort study of tobacco use,diet, occupation,and lung cancer mortality

In 1966, a cohort of White males aged 35 or over, who were policy-holders with the Lutheran Brotherhood Insurance Society (United States), completed a mail questionnaire on tobacco use, diet, and demographic characteristics. During the 20 years of follow-up, 219 lung cancer deaths occurred. Besides the strong relationship with cigarette smoking, we observed an effect on lung cancer risk among current users of cigars or pipes who were nonsmokers of cigarettes (relative risk [RR]=3.5, 95 percent confidence interval[CI]=1.0–12.6) or who were past/occasional users of cigarettes (RR=2.7, CI=1.4–5.3). In addition, elevated risks (from 1.5 to 2.6) of lung cancer were found among craftsmen and laborers, with the highest risks among subjects who worked in the mining or manufacturing industry. No association between current (as of 1966) use of beer or hard liquor and lung cancer was observed, although past users were at elevated risk. An inverse association between lung cancer and intake of fruits was observed, and risks of lung cancer were lower among persons in the highest dietary intake quintiles of vitamins A and C. Except for oranges, however, none of the inverse associations with fruits or dietary nutrients had statistically significant trends. The findings from this cohort study add to the evidence of an adverse effect of cigar/pipe smoking and possibly protective effect of dietary factors on lung cancer risk.     

Determination of the risk of a 2d cancer in patients treated for a first cancer

A paradoxical effect of radiotherapy and chemotherapy for cancer is that some of these treatments can themselves cause new cancers. Most epidemiologic methods can be applied successfully to the investigation of this problem and this paper reviews various approaches that have already been used by various researchers. The authors first review the more traditional methods, i.e., cohort and case-control studies and they then describe designs that have been proposed more recently, such as case-cohort studies. A distinction is established between internal comparisons, carried out within the study population, and external comparisons, in which a general population external to the population under study is used as the reference category. This presentation is mainly aimed at investigators using tumor registry data. However, the general principles formulated here are easily generalized to contexts other than that of registries.     

BRCA1 and sex ratio

Two recent papers suggest distorted sex and transmission ratios associated with BRCA1 mutations. If real, these would provide novel insights into the normal biological function of this gene and have implications for genetic epidemiologic methods used to estimate penetrance. We addressed these observations in two settings: offspring of 283 mutation carriers and 471 mutation negative subjects from BRCA1/2 mutation-positive families with multiple cases of breast and ovarian cancer (NCI families); and relatives of 115 BRCA1/2 mutation carriers from the Washington Ashkenazi Study (WAS). The male:female ratio was below one in both BRCA1 (0.85, 95% CI 0.7-1.1 in NCI families; 0.90, 95% CI 0.6-1.4 in WAS) and BRCA2 families (0.77, 95% CI 0.5-1.3 and 0.80, 95% CI 0.5-1.2, in the NCI and WAS study groups, respectively). None of the sex ratios deviated significantly from one, and there was no significant difference between BRCA1 and BRCA2 families. The reduced sex ratio was due largely to the offspring of males, a distortion that is probably an artifact of ascertainment biases. Among adult daughters without breast or ovarian cancer born to mutation carriers, as expected, fewer than 50% were mutation carriers (39% in BRCA1 families and 44% in BRCA2 families). It is difficult, due to ascertainment biases, to draw firm conclusions regarding sex ratios in studies of a sex-limited phenotype. Nonetheless, these observations do not support the idea that BRCA1 mutation carriers have a lower ratio of male offspring than BRCA2 mutation carriers.     

Residential exposure to magnetic fields: an empirical examination of alternative measurement strategies

OBJECTIVES: To investigate the impact of measuring a single home then imputing information from another home among subjects who lived in two homes in a subset of the National Cancer Institute/Children's Cancer Group (NCI/CCG) investigation of residential exposure to magnetic fields and risk of childhood leukaemia. METHODS: Each subject's summary time weighted average (TWA) exposure was derived from measurements of two homes, weighted by the fraction of the reference period lived in the residence. The three cost efficient field work strategies examined were measuring: (a) the longer lived in home; (b) the currently lived in home; and (c) the former lived in home. Two different methods were used for imputing the missing values: (a) control mean imputation, (b) status specific mean imputation. The subject's summary exposure to magnetic fields estimated with each approach was compared with the subject's TWA calculated from measurements in both homes. The association between estimated exposure to magnetic fields and the risk of leukaemia under different approaches was examined with unconditional logistic regression analysis. RESULTS: The Pearson correlation coefficient between the two measurements within subjects was 0.31 (p < 10(-4), indicating a lack of independence of measurements. Differences were found between mean exposures in current and former homes of cases, and between longer and shorter lived in homes of controls. All methods with measurements from one of the homes in conjunction with imputation of measurements for the second home led to marked attenuation of risk estimates at the highest exposure category, particularly when measurements from current homes were used and those from former homes were imputed. CONCLUSION: Results argue against attempting to estimate lifetime magnetic field exposure from imputed values derived from current residences to fill in gaps caused by unmeasured residences previously lived in.     

Are children living near high-voltage power lines at increased risk of acute lymphoblastic leukemia?

In the National Cancer Institute/Children's Cancer Group case-control study of childhood acute lymphoblastic leukemia (1989-1993), living in a home with a high-voltage wire code was not associated with disease risk. To further investigate risk near power lines, the authors analyzed distance to transmission and three-phase primary distribution lines within 40 m of homes and created an exposure index of distance and strength of multiple power lines (408 case-control pairs). Neither distance nor exposure index was related to risk of childhood acute lymphoblastic leukemia, although both were associated with in-home magnetic field measurements. Residence near high-voltage lines did not increase risk.     

Bias despite masked assessment of clinical outcomes when an outcome is defined as one of several component events.

Hypothetical examples are presented that show that bias can result from misclassification of clinical outcomes defined as one of several events, even though the assessment of each component event was masked. In one example, the effect of the overall misclassification is to make a treatment that reduces the risk appear to increase it; in another example, misclassification causes the overall treatment effect to appear stronger than it actually is. These anomalies are due to the fact that the misclassification of the overall outcome can be differential, even though the misclassification of the individual components is nondifferential.     

Response rates in a case-control study: effect of disclosure of biologic sample collection in the initial contact letter

PURPOSE: Participants in epidemiology studies are often asked to complete an interview and to provide biospecimens. In a population-based case-control study of kidney cancer involving an interview and optional biospecimens, we examined whether mentioning the biospecimens in the initial contact letter adversely affects willingness to be interviewed. METHODS: Eligible cases (n = 434) and controls (n = 775) in Detroit were alternately assigned to receive one of two versions of the contact letter. Both explained that the study involves an interview with 100 dollars compensation plus an optional component with additional compensation; only one disclosed that the optional component involved biospecimens. RESULTS: There were no meaningful differences between the groups in willingness to be interviewed. However, among 303 cases and 351 controls already interviewed, the proportion providing biospecimens was higher in the fully informed group: for blood, the differences were 10.8 (95% CI, 2.0, 19.5) for cases and 6.7% (95% CI, -1.7, 15.1) for controls. Findings were similar for saliva. CONCLUSIONS: In a study involving an interview and optional biospecimens, informing people about the samples in the contact letter seems preferable to a non-specific reference to a second study component. Both approaches yielded similar interview participation rates, but biospecimen participation rates were higher among those informed about the samples in the contact letter.