Tobacco Documents Online

3 (3) The list of potential confounding variables is long, but is not necessarily complete. Sexual activity has not been considered, for example. P N Lee 17.1.2001

1 10 `2 8 S 17 1 56`t REVIEW 1139 CONFIDENTIAL Subiect ref 8b "Prospective study of exposure to environmental tobacco smoke and dysmenorrhea" C Chen et al Environmental Health Perspectives (2000), 108, 1019-1022 The authors describe the results of a prospective study conducted in Shenyang, China in which 165 newly wed, nonsmoking women intending to get pregnant kept a diary in which they recorded the number of cigarettes smoked at home each day in their presence by household members and whether their menstrual periods were associated with dysmenorrhea, defined as abdominal pain or low back pain for at least two days during the period. The diary was kept for up to a year or until pregnancy occurred. Information on a wide range of factors, including ETS exposure at work (yes/no), was recorded at the start of the study. Based on the cycle ending with the first menstrual period, there were 33 women who reported no cigarettes smoked at home and 132 who reported cigarettes smoked at home. Subdividing these 132 women into three equal groups by mean cigarettes per day (<0.8, 0.8-2.7 and >2.8), the frequency of dysmenorrhea rose with increasing ETS exposure, 6.1 %(2/33) for none, 11.4% (5/44) for low, 15.9% (7/44) for middle and 29.6% (13/44) for high. Formal statistical analysis was based on data for al1625 cycles with complete data, using logistic regression analysis to adjust for the variables district, age, body mass index, education, area of residence, occupation, shift work, perceived stress, occupational exposure to chemical hazards, noise and dust, passive smoking at work and season (i.e. winter, spring, etc.) of the cycle. Generalized estimating equations were used to take account of the fact that the same women contributed data on multiple cycles. Table 3 shows the results of this formal analysis. It can be seen that in both unadjusted and adjusted analysis, risk of dysmenorrhea rose steadily with increasing ETS group, with

2 adjustment somewhat strengthening the association. The adjusted odds ratios were 1.0, 1.1, 2.5 and 3.1 for respectively no, low, middle and high ETS exposure, with the increase per two cigarettes per day estimated as 1.3 (95% CI 1.0-1.8). The significant (p<0.05) association remained evident in various subset analyses, e.g. excluding women who used oral contraceptives or intrauterine devices before the study. Possible limitations of the study mentioned by the authors include lack of biochemical measurements of ETS exposure, relatively small sample size, lack of detailed information on ETS exposure at work, reliance on self-report for data on potential confounding variables and lack of clinical infonnation of unrecognized gynaecological disorders that may produce new dysmenorrhea symptoms. As a result of these they correctly regard their conclusions as only suggestive. A few other comments seem relevant: (1) They also presented results of an analysis of the joint association of ETS at home (no, low, middle, high) and ETS at work (no, yes) which gave odds ratio estimates of 1.3 (low, no), 0.9 (low, yes), 2.5 (middle, no), 2.4 (middle, yes), 3.0 (high, no) and 3.1 (high, yes) relative to the group with neither exposure. This does not indicate any association with workplace ETS exposure at all. But this lack of association with workplace ETS is not brought out in the discussion or abstract. (This analysis actually seems strange - why is there no odds ratio for the group exposed only at work? And why are the pairs of odds ratios for each level of ETS exposure at home so similar, given the substantial sampling variation?) (2) No analysis is presented on the data on ETS exposure at home attempting to answer the question as to whether, for a given woman, dysmenorrhea was more likely if exposure in the previous cycle was relatively high. Such within-person analyses are less susceptible to bias from confounding, though may not be powerful if ETS exposure varies little within person.