Smoking and miscarriage risk

In 2006, George et al1 reported the results of a population-based case–control study, investigating the relationship between tobacco smoke and miscarriage. The participating women (463 cases and 864 controls) were classified according to plasma cotinine levels as unexposed to tobacco smoke (cotinine,<0.1 ng/mL), exposed to environmental tobacco smoke (ETS) (cotinine, 0.1–15.0 ng/mL), or active smokers (cotinine, >15.0 ng/mL). Odds ratios (ORs) and 95% confidence intervals (CIs) for miscarriage were calculated using unexposed women as reference category. In these analyses, maternal age was considered as a confounder; adjusted ORs were 1.67 (95% CI 1.17–2.38) for women exposed to ETS and 2.11 (1.36–3.27) for active smokers. We report a reanalysis aimed at assessing the role of maternal age in the relationship between plasma cotinine levels and miscarriage, using aggregated data provided by George et al1 (Table 1 of their article). In the original publication, plasma cotinine levels were converted to tobacco exposure categories and interpreted as an ordinal variable; we instead used plasma cotinine categories (<0.1; 0.1–15; >15 ng/mL) as a quantitative variable. In line with the results presented by George et al, the Mantel-Haenszel OR estimate for a 1-unit increase in plasma cotinine category, controlled for maternalage classes was 1.46 (1.25–1.71). However, this analysis showed signs of heterogeneity (P value for test of homogeneity of ORs across maternal age strata 0.054). To better investigate maternal age as an effect modifier, we also conducted an analysis restricted to subjects with lower plasma cotinine levels (women not exposed to tobacco smoke) or higher levels (active smoker women) to minimize possible misclassification.2 Using this population subset, we compared a logistic regression model having interaction terms between maternal age classes and tobacco smoke exposure (the Table), and a model without interaction. The P value for the likelihood ratio test was 0.046, confirming the possible contribution of the interaction terms to the relationship between cotinine levels and risk of miscarriage. Using the logistic regression model with interaction terms, we calculated stratum- specific ORs of miscarriage for active smokers compared with women not exposed to tobacco smoke: 3.12 (1.45–6.74) for women aged >24; 0.97 (0.52–1.82) for women aged 25–29; 2.55 (1.43– 4.56) for women aged 30–34; and 2.76 (1.22– 6.24) for women aged >35. Apparently, miscarriage risk for women aged 25–29 was not associated with tobacco smoke exposure, whereas all other age classes presented a higher risk of miscarriage in presence of active smoking. Although this finding might seem to be implausible, age-related variations in metabolism could influence the effects of tobacco smoke on miscarriage risk. In his commentary, Bracken2 speculated that if some of the carcinogens present in tobacco smoke are also teratogenic, then cytochrome metabolism (implicated in the activation of many carcinogens) might change the risk for miscarriage with ETS exposure. Some data in the scientific literature support the hypothesis that cytochrome activity varies with age. Kimet al3 reported that levels of 4 metabolites of benzene (E,E-muconic acid, SPMA, phenol, and hydroquinone) diminished with age at rates of 1%–2% per year of life. Our reanalysis suggests a possible role of maternal age in determining tobacco smoke effect during pregnancy; treating age as a confounder might hinder the appreciation of the true magnitude of damage due to ETS and to active smoking.