A case-control study of prosthetic implants and selected chronic diseases

We examined the association between prosthetic nonbreast implants and selected malignant neoplasms, connective tissue disorders, and neurologic diseases. We conducted a case-control study from an insurance claims database. We selected controls who had diseases for which no association with implants have been claimed or reported. Data were analyzed using both tabular and polytomous regression analysis methods, including methods to account for the large number of comparisons. All analyses exhibited positive associations between implants (both silicone and metal) and neurologic conditions, especially idiopathic progressive neuropathy and Meniere syndrome, as well as the expected associations with arthritic conditions. There also was an unexpected negative association between metal implants and brain tumors. In conclusion, further studies of prosthetic implants and neurologic diseases appear warranted. These studies should obtain medical histories to control for possible confounding effects of drug treatments associated with implant surgery.

Absence of confounding does not correspond to collapsibility of the rate ratio or rate difference

Miettinen and Cook (Am J Epidemiol 1981; 114:593-603) showed that absence of confounding does not imply collapsibility of the odds ratio; that is, the crude odds ratio need not equal a common stratum-specific odds ratio even if the exposed and unexposed study groups have the same distribution of risk factors. Less well known is that absence of confounding does not correspond to collapsibility of the person-time rate ratio or rate difference. For example, two study groups can have the same distribution of all risk factors and yet the crude rate ratio need not equal a common stratum-specific rate ratio. The present paper provides an example and explanation of this phenomenon. The discrepancy between nonconfounding and collapsibility in rate comparisons arises when person-time is a post-exposure variable whose distribution can be altered by the effects of exposure and other risk factors.

Relationship between early side effects and therapeutic effects of clomipramine therapy in obsessive-compulsive disorder

Early adverse effects of a drug may be a manifestation of individual differences in drug metabolism or of different pathologic processes. These differences may influence therapeutic responsiveness. Using data from Ciba-Geigy's multicenter 10-week clinical trial, we studied the relationship between early side effects and subsequent therapeutic response to clomipramine (CMI) in obsessive-compulsive disorder. We used tabular analyses and multiple regression to evaluate associations between early complaints and change in score on the Yale-Brown Obsessive-Compulsive Scale. We also evaluated whether early complaints were drug related (i.e., true side effects). It appeared that dry mouth, constipation, dizziness, insomnia, male impotence, nervousness, palpitation, and tremor reported during the first 4 weeks were predictive of good response to CMI. Myoclonus and tinnitus appeared weakly associated with treatment success. Most of these complaints were reported more by the CMI group than the placebo group, and more during CMI treatment than before. The more common complaints may reflect an individual's ability to metabolize CMI appropriately so that adequate therapeutic blood levels are attained. The less common complaints may reflect a sensitivity to CMI's serotonergic actions.

Simulation study of hierarchical regression

Hierarchical regression - which attempts to improve standard regression estimates by adding a second-stage 'prior' regression to an ordinary model - provides a practical approach to evaluating multiple exposures. We present here a simulation study of logistic regression in which we compare hierarchical regression fitted by a two-stage procedure to ordinary maximum likelihood. The simulations were based on case-control data on diet and breast cancer, where the hierarchical model uses a second-stage regression to pull conventional dietary-item estimates toward each other when they have similar levels of food constituents. Our results indicate that hierarchical modelling of continuous covariates offers worthwhile improvement over ordinary maximum-likelihood, provided one does not underspecify the second-stage standard deviations.

Confounding and exposure trends in case-crossover and case-time-control designs

As with ordinary studies, both case-crossover and case-time-control studies can suffer from confounding, including confounding by indication. In a case-crossover analysis, confounding by fixed (constant) characteristics is eliminated by pairing of cases to themselves, at the possible cost of introducing bias due to time trends in exposure. A case-time-control analysis can correct case-crossover results only for bias due to such time trends. If an uncontrolled confounder (such as disease severity) is present, the use of time controls can introduce new confounding, and the case-time-control results may end up either more or less confounded than the ordinary and case-crossover results. The relative confounding in the different approaches depends on details of the relations among the unmeasured confounder, the study exposure, the study disease, and any trend in these variables or their effects. Like an ordinary study, a case-time-control study must assume absence of unmeasured confounders, whether fixed or time-varying. Like a case-crossover study, it must also assume absence of carryover effects and can be more prone to misclassification bias than an ordinary study.

Evidence for recent growth of the HIV epidemic among African-American men and younger male cohorts in Los Angeles County

To estimate the recent course of the human immunodeficiency virus type 1 (HIV) epidemic among men within birth cohorts, ethnic groups, and HIV-risk groups in Los Angeles County, backcalculation methods were combined with log-linear models and census data to reconstruct HIV incidence in subgroups from AIDS surveillance data. Results were compared with directly measured HIV seroprevalence in public sexually transmitted disease (STD) clinics in Los Angeles. Models of HIV incidence indicate that the initial epidemic pattern among men who have sex with men, including a decline in incidence since the mid-1980s, does not apply to all post-1960 birth cohorts. Later peaks were observed in younger birth cohorts and among injection drug users, especially among African-American men, with no evidence of a peak before the 1990s among men born after 1960. Our results indicate that HIV continued to spread near peak rates into the 1990s among younger birth cohorts, especially among young African-American men who have sex with men. Because of the lengthy incubation period from HIV infection to AIDS incidence, our results imply that the AIDS epidemic has not yet peaked in these cohorts and may continue to grow through the present decade in several subgroups. The large variation in HIV incidence and prevalence across birth cohorts and other subgroups needs to be addressed in future community intervention plans.

Historical HIV incidence modelling in regional subgroups: use of flexible discrete models with penalized splines based on prior curves

This paper presents an approach to back-projection (back-calculation) of human immunodeficiency virus (HIV) person-year infection rates in regional subgroups based on combining a log-linear model for subgroup differences with a penalized spline model for trends. The penalized spline approach allows flexible trend estimation but requires far fewer parameters than fully non-parametric smoothers, thus saving parameters that can be used in estimating subgroup effects. Use of reasonable prior curve to construct the penalty function minimizes the degree of smoothing needed beyond model specification. The approach is illustrated in application to acquired immunodeficiency syndrome (AIDS) surveillance data from Los Angeles County.

Impact of model-form selection on the accuracy of rate estimation

A key assumption underlying the use of model-based estimates in epidemiology is that the structural-model form is an adequate mathematical description of the dependence of disease occurrence on exposures and covariates (that is, the model form is correctly specified). If this assumption is violated, model-based point estimators and variance estimators may be biased, standard confidence intervals may be invalid, and inferences derived from these estimators may be incorrect. In practice, the true structural-model form is usually unknown, and investigators frequently use their data to help select a model form. We conducted a simulation study to examine the impact of model-form selection on the accuracy of rate estimation in cohort-study situations resembling those found in environmental and occupational epidemiology. For the situations we examined, the increase in variance produced by using model-form selection was often more than offset by the corresponding reduction in bias, sometimes resulting in a dramatic increase in accuracy. Model-form selection was observed to be most beneficial relative to no selection when effects were stronger, the sample size was larger, and the candidate model forms included the true model form or allowed the model to more closely approximate the true model form. It was least beneficial when effects were weak and the sample size was small, even if the candidate model forms included the true model form.

A critical look at methods for handling missing covariates in epidemiologic regression analyses

Epidemiologic studies often encounter missing covariate values. While simple methods such as stratification on missing-data status, conditional-mean imputation, and complete-subject analysis are commonly employed for handling this problem, several studies have shown that these methods can be biased under reasonable circumstances. The authors review these results in the context of logistic regression and present simulation experiments showing the limitations of the methods. The method based on missing-data indicators can exhibit severe bias even when the data are missing completely at random, and regression (conditional-mean) imputation can be inordinately sensitive to model misspecification. Even complete-subject analysis can outperform these methods. More sophisticated methods, such as maximum likelihood, multiple imputation, and weighted estimating equations, have been given extensive attention in the statistics literature. While these methods are superior to simple methods, they are not commonly used in epidemiology, no doubt due to their complexity and the lack of packaged software to apply these methods. The authors contrast the results of multiple imputation to simple methods in the analysis of a case-control study of endometrial cancer, and they find a meaningful difference in results for age at menarche. In general, the authors recommend that epidemiologists avoid using the missing-indicator method and use more sophisticated methods whenever a large proportion of data are missing.

Clomiphene citrate and neural tube defects: a pooled analysis of controlled epidemiologic studies and recommendations for future studies

OBJECTIVE

To estimate the degree to which neural tube defects (NTDs) are associated with periconceptional clomiphene citrate (CC) exposure in controlled epidemiologic studies, to investigate the consistency of study findings with respect to this association, and to identify key problems that future studies should address.

DESIGN

Pooled analysis of 10 epidemiologic studies.

SETTINGS

Hospitals and clinics.

PATIENTS

Women undergoing treatment for infertility.

INTERVENTIONS

Oral administration of CC.

MAIN OUTCOME MEASURE

Prevalence ratio for NTDs.

RESULTS

Ten controlled epidemiologic studies were identified that supplied sufficient data on CC and NTDs for inclusion. The estimated ratio of NTD prevalence among CC-exposed versus unexposed pregnancies ranged from 0.55 to 5.73 among the studies, but the variation was compatible with random fluctuation. The estimated summary prevalence ratio was 1.08, with 95% confidence limits of 0.76 and 1.51.

CONCLUSION

This analysis indicates that an elevation in NTD risk due to CC cannot be ruled out, but any such elevation seems likely to be less than twofold, and there may be no elevation at all. Future studies should be designed to avoid several methodological problems not addressed in studies to date.

Standardized estimates from categorical regression models

We consider the problem of interpreting categorical regression models, such as the polytomous logistic model, the continuation-ratio model, the stereotype model, and the cumulative-odds model. We present a method to convert categorical regression coefficients into estimates of standardized fitted probabilities, probability differences and probability ratios. We use a delta-method approach to estimate standard errors. We then present a small simulation study to compare different transforms for setting confidence limits, and provide an illustration of our approach in an observational study of drug therapy of polymyositis.

Guidelines for application of meta-analysis in environmental epidemiology. ISLI Risk Science Institute

The use of meta-analysis in environmental epidemiology can enhance the value of epidemiologic data in debates about environmental health risks. Meta-analysis may be particularly useful to formally examine sources of heterogeneity, to clarify the relationship between environmental exposures and health effects, and to generate information beyond that provided by individual studies or a narrative review. However, meta-analysis may not be useful when the relationship between exposure and disease is obvious, when there are only a few studies of the key health outcomes, or when there is substantial confounding or other biases which cannot be adjusted for in the analysis. Recent increases in the use of meta-analysis in environmental epidemiology have highlighted the need for guidelines for the application of the technique. Guidelines, in the form of desirable and undesirable attributes, are presented in this paper for various components of a meta-analysis including study identification and selection; data extraction and analysis; and interpretation, presentation, and communication of results. Also discussed are the appropriateness of the use of meta-analysis in environmental health studies and when meta-analysis should or should not be used.

Dose-response and trend analysis in epidemiology: alternatives to categorical analysis

Standard categorical analysis is based on an unrealistic model for dose-response and trends and does not make efficient use of within-category information. This paper describes two classes of simple alternatives that can be implemented with any regression software: fractional polynomial regression and spline regression. These methods are illustrated in a problem of estimating historical trends in human immunodeficiency virus incidence. Fractional polynomial and spline regression are especially valuable when important nonlinearities are anticipated and software for more general nonparametric regression approaches is not available.

Pain management: a review of the uses of epidural analgesia

Epidural analgesia can be a valuable method of pain control in a variety of situations. It can facilitate early mobilisation and enhance recovery from surgery. This article describes the physiology of epidural pain relief and the nurse's role in monitoring patients effectively, recognising complications and minimising risk.

A meta-analysis of body mass index and risk of premenopausal breast cancer

Increased body mass index (BMI) has been found to be associated with elevated risk of postmenopausal breast cancer. Whether BMI is related to premenopausal breast cancer has not yet been established. We performed a meta-analyses of data from 23 studies that provided information on BMI and incidence of premenopausal breast cancer. Overall, the data support a modest inverse association. For a BMI difference of 8 kg per m2, that is, the difference between a thin person and someone who is morbidly obese, the random effects estimate of the rate ratio from the four cohort studies was 0.70 [95% confidence interval (CI) = 0.54-0.91], and the random effects estimate of the odds ratio from the 19 case-control studies was 0.88 (95% CI = 0.76-1.02). Because of substantial heterogeneity among the study-specific estimates, however, we also examined the influence of certain aspects of study design. Case-control studies with community controls had a more inverse association, whereas case-control studies that interviewed cases shortly after diagnosis applied the same exclusion criteria to cases and controls, or with confounder adjustment beyond age had a more positive association between BMI and breast cancer. Possible reasons for the discrepancies among the case-control studies are discussed.

Endometrial cancer risk after discontinuing use of unopposed conjugated estrogens (California, United States)

To examine the decline in risk of endometrial cancer after discontinuation of use of conjugated estrogens, we conducted a case-control study in a prepaid health plan. We identified 318 patients who had endometrial cancer but had no history of bilateral oophorectomy and had been in the Southern California (United States) Kaiser Foundation Health Plan for more than 10 years. For each patient, one or two control members were selected, 599 in all, matched for age and duration of membership at the time of cancer detection and who had had neither hysterectomy nor bilateral oophorectomy. A history of prescriptions for conjugated estrogens and of potential confounders was obtained for each subject by reviewing outpatient medical records. Rate ratios (RR) contrasting users with nonusers were estimated by time of latest prescription. We found that estrogen-induced risk of endometrial cancer decreases rapidly as the estrogen-free interval increases. The RR estimates, adjusted for duration of use and potential confounding factors, declined from 5.0 for those receiving their latest prescription within 24 months (95 percent confidence limits [CL] = 2.6-9.8), to 1.8 for those receiving their latest prescription within 24 to 48 months (CL = 0.9-3.7), to values near one for each latest prescription interval earlier than 48 months ago (P for trend = 0.00004). For those who used conjugated estrogens extensively (five or more prescriptions, five to 10 years ago), the RR estimate declined from 5.1 for those whose latest prescription was within two years to 0.6 yr for those whose latest prescription was four to five years previously (P for trend = 0.05).

Hierarchical regression for epidemiologic analyses of multiple exposures

Many epidemiologic investigations are designed to study the effects of multiple exposures. Most of these studies are analyzed either by fitting a risk-regression model with all exposures forced in the model, or by using a preliminary-testing algorithm, such as stepwise regression, to produce a smaller model. Research indicates that hierarchical modeling methods can outperform these conventional approaches. These methods are reviewed and compared to two hierarchical methods, empirical-Bayes regression and a variant here called "semi-Bayes" regression, to full-model maximum likelihood and to model reduction by preliminary testing. The performance of the methods in a problem of predicting neonatal-mortality rates are compared. Based on the literature to date, it is suggested that hierarchical methods should become part of the standard approaches to multiple-exposure studies.