The impact of ETS exposure on RSV disease in infants and young children is consistent among studies using laboratory confirmation of RSV infection and clinical diagnosis of bronchiolitis or RSV. Among 14 population-based studies that examined the risk of admission to the hospital or ED for RSV disease, 12 showed that at least one type of ETS exposure in each study was associated with a significant adverse outcome in the bivariate or multivariate analysis. The observation that ETS exposure increases the risk that a child will develop RSV disease that will require hospitalization is robust because these studies used different methods (prospective, retrospective, cohort, case–control) in different patient populations (infants who were premature, term, or with compromising conditions), in a variety of countries and cultures. In addition, the evidence suggests that ETS exposure is associated with more severe hypoxia among children hospitalized for RSV [35
], and one study found an increased risk of mortality from bronchiolitis [3
Among studies in premature infants, 5 [9
] of 7 [9
] found ETS exposure to be a significant risk factor in bivariate or multivariate analysis, including 5 of the 6 [9
] studies assessing laboratory-confirmed RSV. One study contradicting this conclusion did not have laboratory confirmation of the diagnosis and relied on a claims database, rather than direct data collection, for ascertainment of the ETS exposure status [27
]. Misclassification of disease status or missed diagnosis may have contributed to the negative findings in this study. The other was a small study of 30 premature infants with BPD on home oxygen therapy, 16 of whom developed RSV [32
The evidence concerning whether ETS exposure increases the risk of mild RSV infection is much less convincing. In a very large study, Carroll et al. found ETS exposure to be associated with only a small increased risk of RSV illness (OR, 1.06) presenting in the outpatient setting [28
]. If this OR represents the true risk, the Holberg [22
] and Marbury [30
] studies, which did not find a risk, would have been underpowered to detect it because of smaller sample sizes. Because nearly all children contract RSV during the first few years of life [1
] and most cases are mild [2
], there may be a ceiling effect. If children with no ETS exposure are almost certain to contract RSV, ETS exposure can increase the risk of mild infection only slightly.
Misclassification of ETS exposure is a major challenge in studying associations of ETS exposure with disease. Misclassification of exposure status produces a bias in the direction of reducing the apparent magnitude of the risk, leading to either an underestimate of the true ETS exposure risk or to a null finding. Misclassification of an infant or child’s ETS exposure comes from researchers’ use of one or a few measures of possible exposure, which actually can come through many avenues. Examples include in utero
exposure through active and passive maternal smoking and postnatal smoking by the mother, father, other individuals living in the home, visitors, and babysitters. ETS exposure outside the home occurs in public places, day care settings, and houses of friends and relatives. No study in this review gathered exposure data for all potential sources. The smoking status of, for example, the parent is a poor proxy for this global exposure as demonstrated by cotinine level studies. Urinary or serum cotinine (a metabolite of nicotine) is an objective measure of ETS exposure but provides information regarding only the previous 48 to 72 hours of exposure [40
]. In one study, infants whose parents reported that they did not smoke had mean cotinine levels that were 80% as high as those for infants with one smoking parent [39
]. Some infants of nonsmoking parents had higher cotinine levels than some infants with two smoking parents [39
Confounding bias is another potential obstacle to determining the association between ETS exposure and serious RSV disease. A few studies found an ETS exposure effect in bivariate analyses but not in multivariate analyses after adjustment for other factors [11
]. Multivariate models that include ETS exposure and its related factors may be difficult to interpret due to potential collinearity. Smoking status has a well-known association with socioeconomic status (SES) [41
], and SES is predictive of ETS exposure in children [42
]. Therefore, SES may be a proxy for global ETS exposure over the early years of life. In addition to SES, smoking status is also predicted by race, educational attainment, and marital status [43
]. Few of the multivariate studies provided a rationale for their selection of control variables. In reporting of future studies, greater details about the multivariate modeling steps may aid in assessment of collinearity when significant bivariate outcomes become nonsignificant in multivariate analysis.
This review has several limitations. The search was limited to studies published from 1990 to April 2009 in the English language. We searched only Pubmed and Embase and did not attempt to locate unpublished studies. The nature of the primary studies precluded a meta-analysis. The large retrospective database analyses included in this study (Table ) all depended on bronchiolitis or RSV disease classification from diagnostic codes or medical record diagnosis. Although RSV is a leading cause of LRTI in infants and children, identifying the etiology of LRTI is not systematically undertaken in EDs or physician offices. Methods and reporting of ETS exposure ascertainment in these studies also varies widely. However, in most studies, data on the child’s ETS exposure level ultimately are gathered from parent or caregiver report, whether through direct data collection for the study or, for example, through retrospective review of the mother’s prenatal health records. Our review highlights the inherent difficulty of accurately assessing global ETS exposure.