The association between maternal active smoking during pregnancy and the risk of having a smaller baby has been well established since the 1960s based on epidemiologic studies [1
]. The magnitude of this effect is, on average, a reduction of 250 grams in birth weight of babies born to smokers compared to babies born to non-smokers [4
]. One large cohort study of more than 250,000 births found a 320 gram reduction in mean birth weight of infants whose mothers smoked compared to infants whose mothers did not smoke [5
]. The relative risks of low birth weight with maternal prenatal smoking have ranged from 1.5 to 3.5, and an exposure-response relationship between increasing amount smoked and higher relative risk is seen in numerous studies [4
]. The association between exposure to secondhand smoke (SHS) or environmental tobacco smoke (ETS) and birth weight has also been examined in epidemiologic studies, with most of these studies published within the past three decades. The body of literature includes both statistically significant and non-significant findings, but does provide some support for a causal association [6
]. Relative risks and odds ratios for delivering a smaller baby for those with prenatal SHS exposure compared to no or low exposure range from 0.5 (95% CI: 0.13, 1.69) [9
] to 2.31 (95% CI: 1.06, 4.99) [10
], and estimated absolute reductions in mean birth weight between SHS-exposed and unexposed groups range from 3 grams [11
] to 228 grams [12
]. These ranges overlap somewhat with the range of measures seen with active maternal smoking, although this is expected given that it is difficult to distinguish between non-smokers highly exposed to SHS and light active smokers, with cotinine levels in these two groups being similar. It is clear, however, that the findings for prenatal SHS exposure are smaller in magnitude than findings for active maternal smoking.
Four meta-analyses have estimated the magnitude of the effect of prenatal SHS exposure to be a mean deficit of 31–60 grams in infant birth weight [13
]. The magnitude of this effect is weaker by a factor of 4 to 8 compared to the average 250 gram reduction in infant birth weight with active maternal smoking. Pooled odds ratios for low birth weight with SHS exposure were also reported in four meta-analyses [14
]. These pooled overall risk estimates ranged from 1.16 to 1.60. Table 1
summarizes the five meta-analyses published over the past decade. The highest reported pooled odds ratio, 1.60, was published in Chinese [17
] and the meta-analysis included both Chinese and English language studies. Liu and Chen [17
] also reported pooled odds ratios for low and high levels of SHS exposure compared to no exposure: 1.53 (95% CI: 1.14, 2.04) and 2.53 (95% CI: 1.46, 4.36), respectively. These estimates overlap with the range of odds ratios reported for active maternal smoking (1.5 to 3.5 [4
]), although the magnitude of effect for SHS exposure is likely smaller than for active maternal smoking during pregnancy.
Meta-analyses of the impact of prenatal SHS exposure on infant birth weight.
Meta-analyses of the impact of prenatal SHS exposure on infant birth weight.
|Citation||No. studies included||Deficit in birth weight||Odds ratiofor low birth weight|
et al. 1998 ||11||31 grams (95% CI: 19, 44)||-----|
et al. 1999 ||19||31 grams (95% CI: 20.4, 41.6)||-----|
|3||-----||1.38 (95% CI: 1.01, 1.87)|
et al. 2008 ||17 prospective||33 grams (95% CI: 15.7, 51.3)||-----|
|27 retrospective||40 grams (95% CI: 25.8, 54.4)||-----|
|9 prospective||-----||1.32 (95% CI: 1.07, 1.63)|
|17 retrospective||-----||1.22 (95% CI: 1.08, 1.37)|
|Liu and Chen 2009  *||26||-----||1.60 (95% CI: 1.25, 2.05)|
et al. 2010 ||76||60 grams (95% CI: 39, 80)||1.16 (95% CI: 0.99, 1.36)|
The ability to detect a smaller magnitude of effect of prenatal SHS exposure on infant birth weight can be improved when exposure prevalence is higher, as it is in China. The prevalence of smoking among men, and consequently the prevalence of SHS exposure among women, is significantly higher in China than in the U.S. For this reason, and the fact that birth outcomes are of particular interest in a country where the one-child-per-couple family planning policy focuses attention on prenatal and newborn health, we examined the impact of prenatal SHS exposure on infant birth weight in northeast China. The primary aim of our study was to determine the difference in mean birth weight between infants whose mothers were exposed to SHS and infants whose mothers were not exposed in the Chinese cities of Beijing and Changchun (the capital of Jilin Province). A secondary aim of our study was to better characterize the SHS exposure profile in China, an important task in itself given the evidence for multiple adverse health effects associated with SHS [6
] and the gender difference in smoking prevalence in China—63% among men and 4% among women [18
]. This is in contrast to the adult smoking prevalences in the United States—24% among men and 18% among women [19
]. These differences in smoking prevalence rates, and inter-country variations in public policies on cigarette smoking [20
], suggest that SHS exposure patterns in China may differ from the U.S. and other developed countries.
Overall, we found no association between prenatal SHS exposure and birth weight among babies after taking into account the effects of known predictors of birth weight. Few studies in the published literature examine prenatal SHS exposure and infant birth weight in China. Five such studies conducted in China can be found through PubMed—three in English [29
] and two in Chinese with English abstracts [32
]. In a cross-sectional study of 1,058 infants, Chen et al.
] reported a decrease of 11 grams in the average birth weight of exposed infants (3,242 grams, SD = 557) compared to unexposed infants (3,253 grams, SD = 430). Pan [32
] found a 1.7 odds ratio (calculated 95% CI: 0.69, 4.10) for small-for-gestational-age (SGA) with husbands smoking at home compared to unexposed (n = 253). In a cross-sectional study of 1,785 infants, however, Zhang and Ratcliffe [30
] found a slight increase of 32 grams in average birth weight among infants whose fathers smoked a pack a day or more (3,213 grams, 95% CI: 3,025, 3,401) compared to fathers who did not smoke (3,191 grams, 95% CI: 2,995, 3,367). In a prospective study of 1,388 mothers who self-reported SHS exposure during pregnancy, mean birth weight of infants among exposed mothers was 30.5 grams less (p
= 0.2327) than the mean birth weight of infants whose mothers were not exposed [31
]. These four studies were conducted in or near Shanghai, while the matched case-control study of 310 infants by Han et al.
] was conducted in Beijing. The Beijing study reported the only statistically significant finding—an odds ratio of 3.42 (95% CI: 1.44, 8.14; p
< 0.01) for SGA with SHS exposure during pregnancy. In our study, average infant birth weight was actually lowest among the unexposed group compared to every other category of exposure level as measured by cigarettes per day smoked by the husband. The average birth weight does decrease, however, with increasing level of SHS exposure among the exposed groups, suggesting possible misclassification of the self-reported unexposed group.
Stratification by city revealed a slight positive and statistically significant association between self-reported SHS exposure and birth weight in Beijing, with exposed women having slightly heavier newborns than unexposed women, while no statistically significant associations were found among the Changchun study participants. The main contribution to SHS exposure in this population was from the home environment, and specifically the husband smoking in the home, although self-reported SHS exposure among women in this study was not consistent with male smoking rates and did not follow the same age-prevalence pattern for male smokers, as described in a report of a 1996 national survey conducted in China [18
]. In our study population, 75% of the women denied any SHS exposure from the husband smoking at home and 58% reported no exposure from any sources (i.e.
, home, work, public places), yet the 1996 national survey finds that 64% of urban men in China are active smokers. Another national study conducted in China in 2000–2001 reported an active smoking prevalence of 54.5% among urban men [34
]. This inconsistency between self-reported SHS exposure prevalence in our study and reported active smoking rate among urban men in China may reflect underreporting of exposure in our study.
Underreporting of SHS exposure during pregnancy may have occurred due to social desirability response bias, a phenomenon that has been examined and reported in other areas of research [35
]. Such bias has been defined as a tendency to provide responses that deny socially undesirable traits [37
], or that are consistent with societal beliefs [36
]. Given the family planning policy in China, there is high social and psychological value ascribed to having a healthy (and large) baby. At the time of this study, Beijing had an anti-smoking public health campaign, as well as provided prenatal care that incorporated education about SHS exposure. The public health campaign, which began in 1987, included television and newspaper coverage of anti-smoking messages, banning smoking in public places (such as the Beijing Railway Station), and banning cigarette advertising within the city [38
]. The national Chinese Association on Smoking and Health (CASH) was established in 1990 (now the Chinese Association on Tobacco Control), with a branch office in Beijing. CASH was charged with three tasks: (1) Prevent and stop smoking among young people; (2) Develop policies, laws and regulations for tobacco control; and (3) Educate and broadcast messages about the harmful effects of tobacco [38
]. Changchun did not have a CASH branch office, and did not have similar programs in place. Thus, mothers in Beijing with smaller babies may have underreported their SHS exposure during pregnancy, resulting in the finding of heavier babies among those exposed in Beijing (but not in Changchun) compared to those not exposed.
Another indication of possible exposure misclassification is that exposure prevalence and exposure duration in our study population decreased with increasing maternal age, while in the 1996 national survey population smoking prevalence among men in China increased with increasing age. Assuming that the ages of husband and wife are correlated, the reversal of age-prevalence patterns of male smoking in the survey population and SHS exposure in the study population is a notable difference. This may be driven in part by paternal education level: 53% of women 35 years and older had husbands with a university education or higher, whereas only 35% of women under 25 years of age had husbands with a university education (p < 0.0001). Similarly, 26% of younger women and 49% of older women had a university education. Education level, however, did not significantly modify the effect of maternal age on infant birth weight after adjusting for the known confounders, likely due to the younger women not reaching the age to have attained a university degree (48% of the youngest age category were <23 years of age).
In the 1996 national survey, 53.5% of nonsmokers reported SHS exposure for at least 15 minutes per day on more than one day per week [18
]. The 2000–2001 national study reported SHS exposure among approximately 53% of nonsmoking urban women [34
]. In our study, only 37.8% of new mothers reported SHS exposure for at least 15 minutes per day at least one day per week during their pregnancy. Prevalence of exposure from the various sources was also higher in the two nationally representative populations than in our study population. Among the 1996 survey sample, 71% reported exposure in the home and 25% in the workplace (all respondents, age 15–69 years). In the 2000–2001 study, 57.5% reported home exposure and 33.8% reported workplace exposure (nonsmoking women 35–44 years of age). Among our study population of new mothers, 30% reported exposure in the home and 17% in the workplace (32% among those who worked during pregnancy). The much higher overall and source-specific exposure prevalences in the national population samples suggest exposure misclassification among our study participants.
The sensitivity analysis results showed no change in our study findings when up to 45% of the total unexposed group was reclassified as exposed, but did change our findings when women with no reported home exposures were excluded from the analysis. This sensitivity analysis showed a reduction in infant birth weight, which may reflect a greater degree of self-reported exposure misclassification from home sources than from other sources. In an effort to minimize recall bias, the new mothers were asked about their prenatal SHS exposure within a week after giving birth. Since the findings suggest, however, that bias and exposure misclassification are likely, the possibility should be considered that an awareness of the potential negative effects of SHS influenced the accuracy of self-reported exposure during pregnancy.
The exposure indicators in this study were limited and measures of nicotine and its metabolite, cotinine, were not attainable. The mother’s self-reported exposure duration is clearly a crude estimate of actual exposures. In addition, misclassification is suggested by the discrepancies in the prevalence of SHS exposure in this study population compared to other study populations in China and the two national estimates of exposure in China. If a true association between higher prenatal SHS exposure and lower infant birth weight exists in our study population, then underreporting of SHS exposure by the mothers could have resulted in our null findings. However, the crude measures of summed daily SHS exposure duration do not account for the possibility that exposure sources may have overlapped in time, and thus may overrepresent the actual hours per day in which a woman was exposed. Such misclassification of exposure would bias the association towards the null hypothesis. Another limitation of our study is the generalizability of our findings, since the study population is exclusively urban. In addition to demographic differences between urban and rural populations, smoking and SHS exposure rates are known to be somewhat higher in rural than urban populations in China [18
The advantage of this current study design, however, is the inclusion of SHS from all sources. Zhang and Ratcliffe [30
] had data only on the number of cigarettes smoked per day by the father at the time of the study (not during the pregnancy), while Chen et al.
] had data on the number of cigarettes smoked per day by the father and other family members in the home. Pan [32
] also included workplace SHS exposure, which was shown in this current study to be an important source of SHS (44% of total exposure time among those who worked during pregnancy). Since more than half the women in the current study worked during their pregnancy, workplace exposures are non-trivial.
Despite the variations in SHS exposure prevalences found in this study and the other studies in China, exposure remains more prevalent in China than in the U.S. Table 10
summarizes the published studies (in English) reporting SHS exposure prevalence among women in China. The reported prevalences range from 51% [31
] to 75% [40
], with findings from the more recent studies in the higher end of the range. In comparison, data from the 1999–2000 National Health and Nutrition Examination Survey (NHANES) show that 52.5% of the non-smoking population in the U.S. was exposed to SHS (defined as serum cotinine level ≥0.05 ng/mL) [19
]. This population exposure prevalence dropped to 40.1% in the 2007–2008 NHANES [19
]. In that same time frame, SHS exposure among non-smoking women in the U.S. dropped from 47.5% to 37.4% [19
]. One U.S. cohort study looked specifically at SHS exposure during pregnancy and found that 30% of mothers reported exposure in the three months before conception through pregnancy [41
]. Both home and work sources of exposure were accounted for in this multi-site population of 4,667 controls enrolled in a national birth defects study from 1997 to 2003.
The estimated incidence of low birth weight (standardly defined as <2,500 grams) in urban China is 4.2% [42
] while the mean birth weight is approximately 3,300 grams [42
]. The incidence of low birth weight in this urban study population was notably lower at 1.2% and the mean birth weight was slightly higher at 3,480 grams. The better birth weight outcomes in our study may be due to the study population not being representative of urban China, since Changchun is a provincial capital and Beijing is the national capital.
Published studies reporting SHS exposure prevalence among women in China.
Published studies reporting SHS exposure prevalence among women in China.
|Citation and Location||Study design||Study population||Exposure measures and indicators||Findings|
et al. (1989)  Shanghai||Cross-sectional||1,058 births to non-smoking women, 1981||Self-report retrospectively||72% had a smoking family member in the home|
|Zhang and Ratcliffe (1993)  Shanghai||Control group in case-control study of birth defects||1,785 births to non-smoking women, 1986–1987||Self-report retrospectively||58% exposed to paternal smoking|
et al. (1999) ||Nationally representative survey||120,298 selected from cluster random sampling, 1996||Self-report||More than 60% of non-smoking women of child-bearing age (25–50 years) exposed at least 15 minutes daily more than 1 day per week|
et al. (2000)  Guangzhou||Cross-sectional||1,449 never-smoking pregnant women, 1996−1997||Self-report at time of enrollment (during pregnancy)||60.2% of women had smoking husbands and 71% of these women had SHS exposure at home; 33% of women with non-smoking husbands had SHS exposure at home|
et al. (2004) ||Cross-sectional nationally representative sampling ||15,540 adults, 2000−2001||Self-report||51.3% of non-smoking women had SHS exposure at home|
et al. (2007)  Anqing||Cohort of textile mills workers||1,388 new mothers with births, 1996−2000||Prospective self-report||51% exposed to SHS at home during pregnancy|
et al. (2008)  Shanghai||Cross-sectional||701 never-smoking new mothers with infants, 2005−2006||Retrospective self-report||41.9% exposed to SHS during pregnancy; 55.9% exposed to SHS 3 months before pregnancy|
et al. (2009)  Sichuan, Jiangxi and Henan||Cross-sectional survey in 6 counties||8,142 non-smokers, 2004||Self-report||71% of non-smoking women exposed at least 15 minutes daily at least 1 day per week|
et al. (2010)  Sichuan||Cross-sectional ||1,181 randomly selected never-smoking pregnant women with smoking husbands, 2008||Self-report at time of enrollment (during pregnancy); subset with hair nicotine analysis||75.1% exposed at least 15 minutes daily more than 1 day per week|
et al. (2010)  ||Nationally representative survey (Global Adult Tobacco Survey)||13,354 selected from cluster random sampling, Dec 2009−Mar 2010||Self-report||71.6% of women exposed to SHS|
et al. (2010)  |