Review : Analysis of Potential Confounding Variables in Epidemiologic Studies of Parental/ Household Smoking and Respiratory Health in Preschool Children

Abstract

In this analytical study, aspects of the design of 41 epidemiologic studies pertaining to the relationship between parental/household smoking and the prevalence of respiratory symp toms and certain illnesses in preschool children (0-5 years old) were examined using a sys tematic protocol developed previously for a similar analysis of epidemiologic studies in school-age or older children. Wide variability was observed from study to study in the consid eration of 21 predetermined potential confounding variables. While the socioeconomic status (SES), family health history, subject's personal health history, family size, age and gender were considered in 50% or more of the studies, indoor pollution (other than gas stove usage), day care use, animal exposures, stress, dampness/cold, season, occupational exposures, nutri tional status of the subject, and maternal smoking during pregnancy were considered infre quently (in < 25% of the studies) or not at all. A wide variation was also observed from study to study in the array of confounders considered. When evaluated on the basis of number of confounders considered per study, only 3 of the 41 papers considered 10 or more of the 21 predetermined variables. The clinical endpoints, usually obtained from parental question naire responses, were validated by medical records or physician examination in only slightly more than 50% (21/41) of the relevant epidemiologic studies. Twelve of the 21 predeter mined confounders were analyzed in detail. There was wide variation in and a lack of stan dardization of the criteria used as indices of SES, family and personal health history, and age. There was also marked variation among the studies in methods used, if any, to adjust for confounding variables. Several of the confounding variables that we looked at appeared to be consistently associated with increased risk of respiratory illness in preschool children, includ ing family and personal health history, young age, male gender, day care use and season. On the other hand, no consistent association was observed for SES, artificial feeding, gas fuel use and indoor pollution, while equivocal or uncertain data in this regard were obtained for dampness/cold and heating/air-conditioning. While no consistent interactions between var ious confounders were observed, the relationship between parental/household smoking and respiratory illness in preschool children appeared to vary inversely related to age. Before epidemiologic associations between parental smoking and respiratory health in preschool children can be concluded to reflect any effects of ETS on the respiratory system of these children, it will be necessary to more thoroughly and more consistently consider the role of a number of potential confounding variables, both individually and in combination. It will also be necessary, before reaching such conclusions, to control and/or adjust for these factors where appropriate, as well as to more consistently validate parental questionnaire-based clin ical endpoints.