Depression as an antecedent to heart disease among women and men in the NHANES I study. National Health and Nutrition Examination Survey.

BACKGROUND: Depression predicts morbidity and mortality among individuals who have coronary heart disease (CHD), and there is increasing evidence that depression may also act as an antecedent to CHD. The studies that have reported a relationship between depression and CHD incidence or mortality either were restricted to men only or analyzed women and men together. The present investigation was conducted to evaluate the differential effect depression may have on CHD incidence and mortality in women and men. RESEARCH METHODS: We analyzed data from 5007 women and 2886 men enrolled in the first National Health and Nutrition Examination Survey (NHANES I) who were free of CHD at the 1982-1984 interview and who had completed the Center for Epidemiologic Studies Depression Scale (CES-D). Participants were evaluated from the 1982 interview date either until the end of the study (1992 interview date) or until the occurrence of a CHD event. Using CHD incidence and CHD mortality (International Classification of Disease, Ninth Revision, codes 410-414) as the outcome variables, Cox proportional hazards regression models were developed to evaluate the relative risk (RR) of CHD incidence and mortality in the depressed women and men separately, controlling for standard CHD risk factors. RESULTS: The women experienced 187 nonfatal and 137 fatal events, compared with 187 nonfatal and 129 fatal events among the men. The adjusted RR of CHD incidence among depressed women was 1.73 (95% confidence internal [CI], 1.11-2.68) compared with nondepressed women. Depression had no effect on CHD mortality in the women (RR, 0.74; 95% CI, 0.40-1.48). The adjusted RR of CHD incidence among depressed men was 1.71 (95% CI, 1.14-2.56) compared with nondepressed men. Depressed men also had an increased risk of CHD mortality compared with their nondepressed counterparts, with an adjusted RR of 2.34 (95% CI, 1.54-3.56). CONCLUSIONS: In this sample, while controlling for possible confounding factors, depression was associated with an increased risk of CHD incidence in both men and women, as well as CHD mortality in men. Depression had no effect on CHD mortality in women.

Exposure to ionizing radiation and risk of cutaneous malignant melanoma. Search for error and bias.

In a 1984 case-control study of cutaneous malignant melanoma (CMM) among workers at the Lawrence Livermore National Laboratory, Austin and Reynolds found an association between reported employment in proximity to ionizing radiation and CMM (odds ratio (OR) = 5.4; 95% confidence interval (CI): 1.4, 20.7). But in a preliminary study in 1981, they found no association between individual radiation dosimetry readings and CMM. We reanalyzed the 1984 Austin-Reynolds case-control data to determine whether error or bias explain the inconsistency between reported employment in proximity to radiation and individual radiation dosimetry readings. Using individual radiation dosimetry readings, we developed an index of occupation-specific radiation. This index was associated with case status (OR = 10.8; 95% CI: 1.4, 85.1). No definitive evidence for influential matched sets, confounding, recall bias, or any other sources of bias was identified. These results suggest that the odds ratio for reported employment in proximity to radiation may be valid.

Influence of the mother's prenatal drug consumption on risk of neuroblastoma in the child.

A case-control study was conducted 1) to determine whether maternal prenatal drug consumption increases the risk of neuroblastoma in the child and, if so, 2) to see if the size of the risk depends on whether the case is inherited or sporadic. Mothers of children with newly diagnosed neuroblastoma (n = 101) were compared with mothers of children newly diagnosed with other forms of childhood cancer (n = 690). Cases and controls were selected from the population of childhood cancer patients at St. Jude Children's Research Hospital, Memphis, Tennessee, between 1979 and 1986. The patients' mothers were interviewed to ascertain their prenatal medication, alcohol, and tobacco consumption patterns. Unconditional logistic regression models were used to adjust for maternal age at birth, patient age at diagnosis, race, social class, exposure to x-ray, miscarriage, and other confounding variables. Adjusted odds ratios were estimated for the total sample and for subgroups that had a higher probability of containing inherited cases. Drugs associated with neuroblastoma case status include diuretics for hypertension (odds ratio (OR) = 4.1, 95 percent confidence interval (CI) 1.0-16.9), tranquilizers (OR = 2.1, 95 percent CI 1.1-4.3), nonprescription pain relievers (OR = 1.9, 95 percent CI 1.1-3.1), and cigarettes (OR = 1.9, 95 percent CI 1.1-3.2). The odds ratios for maternal prenatal drug consumption for the group with inherited cases and the total sample were approximately the same. This equality is inconsistent with predictors based on Knudson's two-stage model of carcinogenesis.

Exploring the order of odds ratios using the bootstrap.

We show the use of the bootstrap resampling method to examine further the order of a series of odds ratios. Specifically, the bootstrap provides a method for estimating the probabilities that one would find in subsequent independent samples from the same population the observed odds ratio rankings. To illustrate this use of the bootstrap, we modelled the responses of 77 white male physicians to an ethical dilemma involving hypothetical patients. Would the physician report positive HIV status to the health department or would he maintain patient confidentiality? To see if a patient's sex, race, or sexual preference would influence the physicians' decisions, each physician received one of eight randomly selected descriptions of a hypothetical patient. To evaluate the initial order of the patient categories, we constructed 1000 bootstrap samples. Black heterosexual males ranked first or second in 92.2 per cent of the bootstrap samples; black homosexual males ranked first, second or third in 88.6 per cent; and white homosexual females ranked sixth or seventh in 82.9 per cent. Thus we would expect to observe these rankings of the categories in a high percentage of subsequent independent samples.

Oxidant stress and glioblastoma multiforme risk: serum antioxidants, gamma-glutamyl transpeptidase, and ferritin.

Case-control studies of serum antioxidants are difficult to interpret, because antioxidants may be altered by the disease under study. However, because glioblastoma multiforme (GBM) is a relatively rare disease, a cohort study would require a large sample observed for many years. In the present case-control pilot study (34 cases and 35 controls), we evaluated the association between serum levels of ascorbic acid (AA) and alpha- and gamma-tocopherol (alpha-T and gamma-T) measured before diagnostic surgery. To control for influence of GBM on serum AA, alpha-T, and gamma-T, we adjusted for oxidant stress indexes (gamma-glutamyl transpeptidase and uric acid) and an acute-phase response index (serum ferritin). When adjusted, AA is inversely related to GBM (p for trend = 0.007). In addition, AA interacts with alpha-T to further reduce GBM risk (test for interaction, p = 0.04). gamma-T is not associated with GBM (p = 0.71). However, gamma-glutamyl transpeptidase (p = 0.004), coenzyme Q (p = 0.01), and ferritin (p = 0.009) are positively and uric acid (p = 0.000) is negatively related to GBM. We conclude that 1) AA and alpha-T are jointly related to GBM after adjustment for GBM-produced oxidant stress and 2) there is a strong association between the presence of GBM and oxidant stress.

Limitations of relative sensitivity in detecting differential misclassification in case-control studies.

Apparent relative sensitivity, based on an investigator's external standard, is the ratio of observed case to control exposure sensitivity. An apparent relative sensitivity different from 1.0 is usually interpreted as evidence for differential misclassification of exposure status. We undertook this investigation to determine the conditions under which an apparent relative sensitivity exceeding 1.0 is actually due to differential misclassification. We also consider whether apparent relative sensitivity correctly quantifies the degree of differential misclassification. To achieve these goals, we derived an algebraic relation involving apparent relative sensitivity, true sensitivities and specificities, true odds ratio, an index of how well the external standard classifies true exposure, and the incidence of the disease among the nonexposed. We found that an apparent relative sensitivity greater than 1.0 correctly indicates differential misclassification when either (1) the investigator's external standard classifies true exposure perfectly, or (2) the investigator's external standard is imperfect, but the true odds ratio equals 1.0, true relative sensitivity is greater than 1.0, and true relative specificity is less than 1.0. We also found that apparent relative sensitivity greater than 1.0 falsely suggests differential misclassification when true relative sensitivity equals 1.0, the investigator's external standard is imperfect, and the true odds ratio is greater than 1.0. Furthermore, even when apparent relative sensitivity correctly detects the presence of differential misclassification, it may misrepresent the degree.

Physician breach of patient confidentiality among individuals with human immunodeficiency virus (HIV) infection: patterns of decision.

To determine whether the sex, race, or sexual preference of a patient infected with immunodeficiency virus (HIV) influences a physician's decision to breach patient confidentiality, Tennessee primary care physicians were mailed a questionnaire containing a case study in which an HIV-infected patient presented a risk to a third party. Eight different descriptions of the sex, race, and sexual preference of the hypothetical patient were distributed randomly among the physicians, one description to each physician. The physicians were asked to decide whether to maintain confidentiality, notify the health department, or inform the patient's partner. Responses of 199 White male physicians were analyzed using an unconditional saturated logistic regression model. The odds ratios for these physicians saying they would send the patient's antibody status to the health department extend from 18.4 (95 percent confidence interval: 1.3, 260.1) for Black homosexual males to .5 (95 percent CI: 0, 11.5) for White homosexual females. The odds ratios for White male physicians saying they would inform the patient's partner range from 7.5 (95 percent CI: .8, 69.2) for Black heterosexual males to 1.0 (reference category) for Black homosexual females. The results suggest that when physicians decide to protect a third party by breaching an HIV-infected patient's confidentiality, their decision may be influenced in some cases by the race, sex, and sexual preference of the patient.

An exploratory analysis of the occupational correlates of large pigmented nevi at Lawrence Livermore National Laboratory.

To examine the association between exposure to occupational factors and the presence on the body of large pigmented nevi greater than 5 mm, data were collected from 110 employees of Lawrence Livermore National Laboratory employed between 1969 and 1980. In this exploratory prevalent case-control study analysis, 38 employees reported having at least one large pigmented nevus and 72 reported none. Occupational factors associated with the presence of large pigmented nevi (adjusted for age and the number of sunburns per year under age 21) were being an engineer (prevalence odds ratio [POR] = 3.20, 95% confidence interval [CI] = 1.0,10.6) or an electrical engineer (POR = 2.56, 95% CI = 0.3,20.3), being hired at Lawrence Livermore National Laboratory before 1962 (POR = 3.07, 95% CI = 1.2,7.7), and having one's skin exposed to rare earth metals (POR = 3.78, 95% CI = 0.9,15.1).

Problems due to small samples and sparse data in conditional logistic regression analysis.

Conditional logistic regression was developed to avoid "sparse-data" biases that can arise in ordinary logistic regression analysis. Nonetheless, it is a large-sample method that can exhibit considerable bias when certain types of matched sets are infrequent or when the model contains too many parameters. Sparse-data bias can cause misleading inferences about confounding, effect modification, dose response, and induction periods, and can interact with other biases. In this paper, the authors describe these problems in the context of matched case-control analysis and provide examples from a study of electrical wiring and childhood leukemia and a study of diet and glioma. The same problems can arise in any likelihood-based analysis, including ordinary logistic regression. The problems can be detected by careful inspection of data and by examining the sensitivity of estimates to category boundaries, variables in the model, and transformations of those variables. One can also apply various bias corrections or turn to methods less sensitive to sparse data than conditional likelihood, such as Bayesian and empirical-Bayes (hierarchical regression) methods.

An exploratory study of environmental and medical factors potentially related to childhood cancer.

To determine whether a general-purpose epidemiologic questionnaire can be used in childhood cancer hospitals to identify associations between environmental factors and the major types of childhood cancer, we report the results of the analysis of the data obtained from such a questionnaire. On admission to St. Jude Children's Research Hospital in Memphis, Tennessee, between 1979 and 1986 a questionnaire was administered to 1,270 mothers of patients diagnosed with childhood cancer. Approximately one-half of the children had acute leukemias (n = 629); the remainder had lymphomas (n = 237) or solid tumors (n = 404). Responses to questions regarding the patients' and parents' environmental and medical histories were compared across nine diagnostic categories. Only 5 of 232 variables remained nominally statistically significant (P less than 0.05) after adjusting for confounding by patient or maternal age, year of birth or diagnosis, patient's race, and social class. The variables identified were length of time the patient was breast-fed (chi 2 = 16.1, P = 0.04); having a garden with fertilizers, herbicides, and pesticides (chi 2 = 17.2, P = 0.03); maternal use of sex hormones during the year before the patient's birth (chi 2 = 18.2, P = 0.02); maternal cigarette consumption (chi 2 = 18.0, P = 0.02); and patient contact with persons with cancer (chi 2 = 20.7, P = 0.01). Despite the large number of patients studied, we identified fewer significant variables than would be expected, on the average, under the null hypothesis. We conclude that the data obtained from a general-purpose epidemiologic questionnaire do not provide a useful overview of the association between exposure to environmental factors and several types of childhood cancer.

Hypotheses concerning roles of dietary energy, cured meat, and serum tocopherols in adult glioma development.

We conducted a pilot study to determine whether: (1) high levels of energy intake increase glioma risk; (2) the previously observed relationship between cured meat consumption and glioma risk can be attributed to confounding by energy intake, and (3) alpha-tocopherol modifies caloric intake and gamma-tocopherol modifies cured meat consumption. We identified 40 age-sex-race matched glioma sets, and obtained serum vitamin C and alpha- and gamma-tocopherol levels for 23 of these sets. Glioma risk increased with quartile of total dietary energy adjusted for fat, protein, and nitrite-containing meat consumption (odds ratio = 1.0, 2.7, 5.8, 8.2; p value for trend test = 0.02). Although positive associations between individual cured meats and glioma risk decreased when adjusted for caloric intake, the small sample size makes it difficult to interpret the results. Serum alpha-tocopherol appeared to modify the effect of calories and serum gamma-tocopherol may modify the effect of cured meat on glioma risk. While the observed interaction is predicted by experimental research, our findings are based on small numbers. Larger studies are needed to further evaluate our preliminary findings.

The influence of exogenous estrogen use on survival after diagnosis of endometrial cancer.

For examination of the effect of prior exogenous estrogen use on survival after diagnosis of endometrial cancer, 244 endometrial cancer cases newly diagnosed at North Carolina Memorial Hospital, Chapel Hill, North Carolina, between 1970 and 1976 were followed until 1982. Estrogen users (n = 46) were younger, had less advanced disease, and were more likely to be nonobese and white than were nonusers (n = 198). The estimated probability of surviving (Kaplan-Meier) five years after diagnosis was 0.89 for users and 0.53 for nonusers. When adjusted for age, grade, stage, obesity, race, and treatment (using the Cox proportional hazards regression model), the survival probabilities throughout the period of observation for estrogen users continued to be higher. The adjusted hazard rate for a nonuser was 2.05 (95% confidence interval (Cl) 0.96-4.39) times that for an estrogen user. The adjusted hazard rate from endometrial cancer only was 4.01 (95% Cl 1.22-13.21) times greater among estrogen nonusers. The more frequent occurrence of endometrial cancer in an earlier stage and grade among estrogen users may not be the sole cause of their lower hazard rate from this disease.

Do health professionals' attitudes interfere with the treatment of depression?

Might the attitudes of health care professionals help to explain why most persons with a depressive disorder do not receive adequate care? To assess this question, the authors surveyed the faculty and staff of a midwestern university. One hundred percent of the social workers who responded found psychotherapy or counseling to be extremely or quite effective in treating persons with a major depressive episode, compared to 55% of the psychologists and 31% of the psychiatrists. For medication, the corresponding figures were 88% of psychiatrists, 64% of psychologists, and 46% of social workers. Many respondents noted problems with interprofessional communication, while most psychiatrists felt that individuals treated by two or more professionals for their depression usually receive poorer care. If future studies indicate that nonmedical therapists who view antidepressants as relatively ineffective are less likely to refer depressed clients for medication evaluation, these findings might help to explain why many depressed individuals who could benefit from medication do not receive it. Concerns about interprofessional communication, as well as psychiatrists' beliefs about the quality of care received by persons treated by more than one professional, might also explain why joint treatment occurs less often than would be desirable. The authors discuss some of the implications that these findings may have for the education of health care professionals.

Presurgical serum albumin levels predict survival time from glioblastoma multiforme.

To determine whether serum albumin levels, before first surgery, predict time until death, 24 glioblastoma multiforme patients newly diagnosed at Ohio State University and Cleveland Clinic Foundation Hospitals between 1993 and 1995 were followed until 1996. Patients with presurgical serum albumin levels below 3.4 mg/dL survived an average (median) of 62 days (95% confidence interval (CI): 34, 135 days) after surgery. Those with serum albumin levels of at least 3.4 mg/dL survived an average of 494 days (95% CI: 241, 624 days). The association between serum albumin level and time until death persists when adjusted for demographic and treatment variables using Cox proportional hazards regression. Adjusted hazard ratios, by quartile of presurgical serum albumin level, are: 1.0, 1.2, 0.1, 0.1 (P-value for trend test = 0.007). In addition to providing a prognostic indicator, presurgical serum albumin levels can be used to evaluate the success of randomization of clinical trials for glioblastoma multiforme therapies. Our findings are consistent with results seen for tumors at other sites. We speculate that our results may be attributable to an association between low serum albumin levels and physiological events associated with angiogenesis.