Susceptible subgroups: the challenge of studying interactions.

Abstract

Environmental exposures can be nearly ubiquitous. Although large groups of people are often exposed (and, in the case of ambient air pollution, entire populations are exposed), the risks are usually very small. One might argue that the responsible agent is so innocuous that only very rarely is a case attributable to it. On the other hand, there may be susceptible subgroups within the population that carry the entire burden alone. A paper by Bateson and Schwartz' in this issue examines the effects of air pollution (PM1o) in potentially susceptible subpopulations. These subpopulations included people with previous hospitalization for myocardial infarction, diabetes, congestive heart failure, chronic obstructive pulmonary disease, or conduction disorders. Patients with a history of myocardial infarction or diabetes had a 2to 2.5-fold higher risk of mortality than people experiencing other heart or lung disease. This supports current hypotheses on the biologic mechanisms by which cardiovascular disease can be exacerbated by deposition of particles in the lung. However, the confidence limits around the risks for these subpopu lations are wide and overlapping. The effects on overall mortality in this study are considerably larger than in the National Morbidity, Mortality, and Air Pollution Study (NMMAPS). NMMAPS is based on the 88 largest U.S. metropolitan areas and can be used as a reference estimate for the United States. NMMAPS estimated that a 1-,ug/m3 increase in PM,O increased mortality by O.2%.2 The present study, based in Chicago, found a 1.1% increased mortality. However, the Chicago study was focused on patients with previous records of hospital ization for lung and heart disease and included only persons above age 65.1 The strength of the interaction seen within a disease subgroup depends in part on how well the subgroup has been defined. For example, recent research has shown that diabetes in patients with myocardial infarction was often not well managed a decade ago and received little medical attention. Furthermore, hospital admission for diabetes is most likely when the disease has been badly managed. Thus, the prevalence of diabetes is probably underestimated when relying (as this paper did) on hospital discharge diagnoses. To reduce potential misclassification, detailed individual-level data on exposure as well as on susceptibility should be available. The study of susceptible groups is being extended beyond disease groups to include variations in physiological function defined by genetic polymorphisms. An elegant example can be found in this issue in which Lammer and his colleagues3 investigate the association between smoking and orofacial clefts. The presence of a certain detoxifying enzyme with slower metabolism apparently increased the risk of clefting with mother's smoking. Are genes the answer to the problem of carefully defining susceptible sub