Environmental tobacco smoke and low birth weight: a hazard in the workplace?

Low birth weight (LBW) increases infant morbidity and mortality worldwide. One well-established risk factor is maternal smoking. Environmental tobacco smoke (ETS) exposure has recently been focused on as another potential risk factor. In this article, we review epidemiologic literature on the effects of ETS on LBW and intrauterine growth retardation (IUGR), the cause of LBW related to maternal smoking. As we consider the feasibility of modifying women's exposure, we focus our discussion on workplace exposure to ETS. The workplace is particularly important to consider because women of child-bearing age are present in the workplace in greater numbers now than ever before. In addition, certain subgroups of working women may be particularly at risk from the effects of ETS on pregnancy because they work in environments with higher exposure or they are more susceptible to its effects. We conclude that there is consistent evidence to relate maternal ETS exposure to an increased risk of adverse pregnancy outcomes and that this association may be generalized to the work environment. In studies with positive findings, infants exposed to ETS antenatally were 1.5-4 times more likely to be born with LBW, but few studies examined LBW. Most studies looked at measures of IUGR. ETS was associated with reductions in birth weight (adjusted for gestational age) ranging from 25 to 90 g. Infants born to women exposed to ETS were generally 2-4 times more likely to be born small-for-gestational age. ETS exposure in the workplace can and should be minimized to protect pregnant women from its adverse effects.

Low birth weight ([LBW] < 2500 g) is the leading cause of infant mortality in the United States. Approximately 7.4% of all births in 1995 were LBW, a proportion that has changed little in the past two decades (1). Smoking is one of the few modifiable risk factors for LBW. There is an abundance of evidence linking maternal smoking to LBW with relative risk estimates in the range of 2-4 (2). Two unfavorable birth outcomes, not mutually exclusive, result in low birth weight: preterm delivery and intrauterine growth retardation (IUGR). The adverse effects of maternal smoking on LBW appear to operate through an effect on IUGR rather than through an effect on preterm delivery. Maternal smoking has consistently been demonstrated to increase the risk of IUGR and to reduce mean birth weight by approximately 150-250 g (2). There is a dose-response relationship between maternal smoking and adverse pregnancy outcomes with a stepwise increase in risk with an increased number of cigarettes smoked per day (2). However, the timing of the exposure (maternal smoking) influences its effects on pregnancy outcomes. The adverse effects on LBW and IUGR are largely limited to smoking in the second half of pregnancy; women who quit smoking by the second half of pregnancy do not have an increased risk of poor outcomes (3)(4)(5). The biologic mechanism by which maternal smoking causes growth retardation has not been definitely established. The current evidence suggests that the impairment of growth associated with maternal smoking results from reduced oxygen flow to the fetus. The maternal blood supply to the placenta is reduced and its oxygen load attenuated by the increased maternal carboxyhemoglobin levels associated with maternal smoking (2,6). This association between maternal smoking and LBW fulfills many of the causal criteria (strength of association, consistency, dose response, reversibility, and biologic plausibility) and has been judged to be causal in the Surgeon General's report (3).
The weight of evidence linking maternal smoking with LBW has led to concern about the effects of environmental tobacco smoke (ETS). In addition, many of the chemicals in cigarette smoke, including nicotine and carbon monoxide, are present in higher concentrations in undiluted sidestream smoke than in the (mainstream) smoke inhaled by the smoker (7,8). Of course, ETS comprises diluted sidestream smoke and exhaled mainstream smoke. Early studies (1960s) reported small effects on birth weight associated with in utero exposure to paternal smoking among nonsmoking mothers (9,10). These studies were limited in their ability to assess exposure and to control for potential confounders. Beginning in the 1980s, as effects of ETS on children were recognized, there has been renewed interest in this issue. Studies of ETS and pregnancy outcomes have generally continued to be based on a woman's report of ETS exposures, although sources of ETS other than the father's smoking have been considered. Recently, a few studies have incorporated biomarkers of exposure, either cotinine (a nicotine metabolite) or nicotine levels in the pregnant or postpartum woman. ETS exposure in pregnant women increases levels of nicotine and cotinine in pregnant women and in their amniotic fluid (11,12). In most studies of ETS and pregnancy, birth weight adjusted for gestational age (a proxy for IUGR) or birth weight alone has been the focus with LBW itself less frequently examined. Based on the studies of maternal smoking, IUGR (assessed by small-for-gestational age [SGA] [e.g., < 5th or < 10th percentile of birth weight for gestation, > 2 SD (standard deviation) below the mean birth weight for gestation] or birth weight adjusted for gestational age) and LBW are indeed the most appropriate outcomes to consider.

Review of ETS-LBW Literature
A review of the published literature on the relationship between ETS and LBW, birth weight, and IUGR (Table 1) was completed using the MEDLINE database (National Library of Medicine, Bethesda, MD) through 1998 and the bibliographies of individual papers. We a priori defined criteria for inclusion of a study to synthesize the current state of knowledge. These criteria were applied to all studies regardless of their findings. Studies were excluded if they did not a) describe clearly how the exposure to ETS was determined (no such studies in fact were found); b) clearly separate active smokers exposed to ETS from nonsmokers exposed to ETS (13); c) consider potential confounders such as socioeconomic status (14)(15)(16)(17)(18), birth weight analysis only (19); d) provide no assessment of statistical significance (10,15,17,20,21); or e) characterize the study population (21,22). Recent studies of ETS generally show an adverse effect of exposure on LBW, birth weight, and IUGR, although as would be expected, the effect sizes are smaller than for maternal smoking. Even in those studies in which the effects were not statistically significant, ETS was consistently associated with small effects in the direction of increased risks for adverse outcomes (reductions in birth weight, increases in risk of LBW or IUGR).
Studies of ETS exposure and birth weight fall into two groups: birth weight adjusted for gestational age (23)(24)(25)(26)(27)(28)(29)(30) or birth weight unadjusted for gestational age (31)(32)(33). Obviously the difference in birth weight will be larger if birth weight is not adjusted for gestational age. More important, when birth weight is adjusted for gestational age, this parameter becomes an estimate of fetal growth as it relates size (weight) with time (gestational age). Without adjusting for gestational age, we cannot determine whether birth weight is reduced among women exposed to ETS as a result of growth restriction or if it is the result of early delivery. Given that the effects of maternal smoking appear to relate only to growth restriction and not to preterm delivery (2,34), we expect a similar effect of ETS exposure, although it is possible that the mechanisms of effect might differ. Furthermore, because study samples often differ in their gestational age distributions, a failure to adjust for gestational age could result in different estimates of the effects of ETS across studies. For this reason alone, the literature may appear inconsistent.
In 4 of the 11 studies examining the effect on birth weight (adjusted and unadjusted for gestational age), the reductions in birth weight were statistically significant ( Figure 1, Table 1). Although not all of the differences in mean birth weight were statistically significant, the estimates of the difference in mean birth weights for 10 of the 11 studies were negative, falling between -25 and -125 g. In the three studies (31)(32)(33)) that examined ETS effects on birth weight unadjusted for gestational age, the effects of ETS were very similar, with point estimates clustering around -105 g. When adjusting for gestational age, the size of the effect was smaller and more variable, ranging between -25 and -87 g. A recently published meta-analysis (23) calculated average birth weight decrements associated with ETS exposure in nonsmoking mothers. Based on studies that adjusted for confounders, the pooled birth weight difference was -28.5 g (95% confidence interval [CI]: -40.8, -16.2). Limiting the meta-analysis to those studies that considered multiple sources of ETS exposure and adjusted for confounders, the pooled birth weight difference was -24 g (95% CI: -39.3, -8.6) (23). Both of these estimates of the birth weight difference are statistically significant. Although this difference in birth weight is not large and may not be clinically significant for any individual infant, the birth weight distribution is shifted down with exposure to ETS.
Such a shift on a population level would lead to increases in LBW, a significant effect of ETS for the population. It is unclear whether the meta-analysis was limited to studies that adjusted for gestational age.
IUGR can be examined using either a continuous measure, such as birth weight adjusted for gestational age, or a dichotomous measure, such as SGA (< 5th or < 10th percentile of birth weight for gestational age, > 2 SD below mean birth weight for gestational age). Many of the studies described in the preceding paragraph and in Figure 1 are in essence studies of IUGR, as birth weight is adjusted for gestational age. In three (24,25,28) of the eight studies that examined ETS effects on birth weight adjusted for gestational age, the effect was statistically significant ( Figure 1, Table 1). In two (30,35) of the five studies with nonsignificant effects of ETS on birth weight, the sample of births was restricted to term (> 37 week) deliveries. The effect on birth weight of any exposure is usually smaller when restricted to this pool of deliveries, as there is less variability in birth weight after 37 weeks gestation. The point estimates of the difference in mean birth weights for all eight of the studies were negative, with the point estimates ranging between -25 and -90 g. Another less frequently used continuous estimator of IUGR is the birth weight ratio.  (24) Matthai (25) Lazzaroni (26) Eskenazi (27) Martinez (28) Zhang (29) Martin ( by first author as given in Table 1. Studies with an asterisk (*) are not adjusted for gestational age.
The numerator is the birth weight observed at a given gestational age, and the denominator is the birth weight expected at a given gestational age (the mean birth weight). Brooke and colleagues (36) examined the effect of ETS on this birth weight ratio measure and found no significant difference in the birth weight ratio related to ETS. This study is difficult to place in context because its methods are quite different from those of other studies in this area; in particular, the birth weight ratio is not a widely accepted estimator of growth retardation.
Four studies examined the effect of ETS on IUGR, using dichotomous measures of SGA ( Figure 2, Table 1). Two (37,38) of the four studies reported strong (relative risk estimates 3-4) and statistically significant associations, notably one of which used a biomarker for exposure (38). The effects of ETS on LBW (a dichotomous variable) have also been less often studied than the effects on birth weight measured as a continuous variable ( Figure 3, Table 1). In three (19,30,39) of the six studies, the odds of LBW were significantly and substantially increased for infants born to women exposed to ETS, although in two of the studies the significant effect was only in a subgroup of the women [women over 30 years of age (39), nonwhite women (19)]. Three of the studies examined LBW only in term infants or in term infants separately (23,30,35); this is a subset of LBW that overlaps with IUGR. (Infants born LBW at term are by definition SGA and growth retarded.) The risk of a LBW infant at term was increased in two studies (23,30), albeit significant at =0.05 for only the Martin and Bracken study (30). Overall, considering findings for both LBW and LBW at term, relative risk estimates for five of the six studies showed an increase in risk. The meta-analysis by Windham (41). Nicotine lies of LBW that adjusted in bodily fluids alone has a very short half-life s was 1.38 (95% CI: (-2-3 hr) and therefore can measure only the most acute exposures (41). However, nicotine levels assessed in hair do not have this same limitation and nicotine levels in e method of determining hair reflect tobacco smoke exposure over the iay account for some of the past few months (43)(44)(45)(46). A biomarker ong these studies; they may assessment of ETS may also be more valid h studies' results are more than self-reported exposure because it can ng a conclusion about the account for differences in exposure not capany of the studies relied on tured by reporting the number of hours one ;s exposure to ETS. Use of is exposed to ETS. For example, differences res rather than biomarkers in ventilation of the environment can affect rential or nondifferential the biologic burden received and would be Nondifferential misclassifi-ignored by a self-reported assessment of expotenuate real associations sure. Compared to self-report, differential birth outcomes. If exposure misclassification would be much less likely Lnderestimated using self-and nondifferential misclassification also rela-,attentuation of the associ-tively less likely. However, biomarkers inte-5 and birth outcomes could grate all sources of exposure and cannot o more difficult to compare separate different sources. Self-reported )sure across studies, as the assessment is crucial if one wishes to study gories were rarely similar. effects of exposure by specific source (e.g., studies have used either home, workplace). or both biomarkers and Among the four studies that used biologic e, primarily two biomarkers measurements rather than self-report, three been assessed in pregnant reported significant associations with adverse a nicotine metabolite; and pregnancy outcomes. Nafstad et al. (38) onent of cigarette smoke. found a significant increased risk of SGA have established cotinine (growth retardation) associated with high lid biomarkers for maternal nicotine levels in hair, categorized in quartg (40)(41)(42). It should be tiles, in nonsmokers. Rebagliato et al. (24) iat bodily fluid (e.g., blood, found a significant reduction in birth weight, nine levels reflect exposure adjusted for gestational age, associated with r period (half-life -17 hr) high cotinine levels, also categorized in quinnot capture chronic expo-tiles, in nonsmokers. This is consistent with 1). Blood levels of nicotine the findings of Haddow et al. (32) who Lrate proxy for the dose of reported that high serum cotinine levels in ed from ETS exposure. nonsmokers were associated with significantly ;tudies of ETS have used decreased birth weight adjusted for gestancentrations because these tional age. Eskenazi et al. (27) also used l found to be valid substi-serum cotinine, classified dichotomously (2-10 ng/mL vs < 2 ng/mL) to assess ETS exposure but found no significant effect on Ahluwalia ( smokers were classified as exposed to ETS In (RR) based on cotinine levels (2-10 ng/mL). This is in contrast to the Rebagliato et al. study 1 weight related to environmental (24) in which the prevalence of exposure in re. Studies labeled by first author the highest quintile was nearly 20% (1.8-14 tudies marked with an asterisk (*) ng/mL) and the Haddow et al. study (32) (24) excluding those over 14 ng/mL as possibly active smokers. The exclusion of women in the 10-14 ng/mL range may also have contributed to the smaller effect size seen in the Eskenazi et al. study (27) compared to the Rebagliato et al. study (24).

Critical Period of Exposure
The timing of the ETS exposure is an important issue addressed explicitly by only one of the studies. Studies of maternal smoking in pregnancy have consistently shown that it is only smoking in the second half of pregnancy that exerts a significantly adverse effect on LBW, IUGR, and birth weight (4,5). Therefore, one might also expect ETS exposures limited to the first half of pregnancy to exert little or no effect. Failure to separate early and late exposures might dilute the estimated effects. There is reason to expect that ETS exposures might change over the course of pregnancy, particularly ETS exposure in the workplace. Women may stop working as their pregnancy progresses and partners and co-workers might reduce their smoking around a woman as she becomes visibly pregnant. Nonsmoking pregnant women might become more active in their attempts to reduce ETS exposures as their pregnancy progresses. The one study (47) that explicitly examined ETS in late pregnancy found an increased risk of LBW, albeit not significant, of workplace ETS exposure in later pregnancy (OR = 1.83) but no substantial increase in risk for workplace ETS exposure overall (OR = 1.21). Although the other studies reviewed did not explicitly examine ETS exposures by gestation, the questions and timing of interviews and/or biomarker assessment offer some clues. Nearly all the interview studies asked women about their exposures generally during pregnancy and did not specify the time period of interest. In two studies, those by Ahluwalia et al. (39) and Martin and Bracken (30), the questions were asked at the first prenatal visit. Neither article provides information on the mean gestational age at first visit. However, even with low income populations in the United States, which comprised the study population, most women obtaining prenatal care do so before the 20th week of pregnancy. A study conducted in Sweden (37) also asked about ETS at the first prenatal visit and noted that the mean gestational age at first visit was 12 weeks. Therefore, these three studies presumably assessed ETS exposure in the first half of pregnancy. The Haddow et al. study (32) relied on serum cotinine measured in the Environmental Health Perspectives * Vol 107, Supplement 6 * December 1999 second trimester and also presumably assessed exposure in the first half of pregnancy. All these studies found strong adverse effects of ETS that might suggest that exposure to ETS did not change during pregnancy. The Martinez et al. study (28), which interviewed families postnatally about paternal smoking and ETS assessment, would be expected to represent the entire pregnancy with a possible bias toward representing exposure in the latter part of pregnancy due to length of recall. In this study, a statistically significant trend of decreasing birth weight with increasing exposure dose was reported. Two of the studies employing biomarkers collected their samples in the third trimester: Eskenazi et al. (27) analyzed cotinine from samples collected at 27-28 weeks gestation and Rebagliato et al. (24) analyzed cotinine from samples collected in the third trimester. The Eskenazi et al. study (27) reported no significant association between high levels of cotinine and birth weight, whereas the Rebagliato et al. study (24) found a significant reduction in birth weight associated with similar levels of cotinine. Nafstad et al. (38) collected hair samples for nicotine analysis in the immediate postpartum period, which would represent the few months of exposure. Presumably, then, the measure of ETS in the Nafstad et al. study (38) would capture exposure in the last trimester of pregnancy. As noted above, this study did find a strong association between nicotine level and risk of SGA.

Outcome Measure
The choice of outcome measure is also important to consider. Overall, there appear to be stronger effects of ETS on LBW and SGA than on birth weight (adjusted or not for gestational age). This difference is informative and may help us to understand how ETS exposure acts. The discrepancies in findings suggest that the reduction in birth weight is not uniform across the full distribution of birth weight. A uniform reduction in birth weight across the distribution of birth weight would result in some increase in LBW or SGA. The size of the effect of ETS exposure on LBW and SGA is often large, however, with ORs of 2 or greater. This effect is more similar to effects seen for maternal smoking during pregnancy, whereas the birth weight effects are much smaller than those for maternal smoking. Perhaps the reduction in birth weight is larger for infants at the lower end of the birth weight distribution. Martin and Bracken (30) also observe that while the decrease in birth weight is relatively small compared to that of direct maternal smoking, the reduction "appears to operate on the low end of the birth weight distribution, thereby increasing risk" of LBW. Perhaps infants born in the lower end of the birth weight distribution are more likely to be exposed to ETS, and several studies have previously demonstrated that women of low socioeconomic status are at increased risk for LBW (48). The stronger effect of ETS on LBW compared to birth weight is also important with regard to infant outcomes. Although reductions in birth weight are associated with increases in infant mortality across the entire birth weight continuum, reductions in birth weight that lead to LBW are much more hazardous. Approximately two-thirds of all infants deaths in the United States in 1995 occurred to infants born with LBW (1). Infants born with LBW are also more likely to have neurodevelopmental problems such as cerebral palsy (49). Therefore, the effects of ETS on LBW are relatively more important in terms of policy than the effects of ETS on birth weight alone.

Confounding
All these studies have considered the potential for confounding of the association between exposure to ETS and birth outcomes. In many of the studies, a wide range of covariates have been included in the final models to produce unconfounded estimates of effect. The obvious potential confounders have been considered. There is some risk, however, that these studies may have controlled for covariates that should not be considered confounders but rather are part of the causal pathway. This may be particularly true of studies considering ETS in the workplace. ETS exposure may indeed be higher in workplaces in which working conditions are more strenuous. Increased ETS levels may be the result of these working conditions and controlling for them in the analysis may remove the very real effects of ETS. This is especially problematic with regard to socioeconomic status. As stated earlier, socioeconomic status is a very strong predictor of LBW risk and also may be related to ETS exposure. The higher ETS exposure levels may be part of the explanation for the increased risk of LBW to women of low socioeconomic status. Therefore controlling for socioeconomic status would mask a real effect of ETS.

Stadstical Power
Given that the effect size is likely small and the prevalence of exposure may vary depending on the population, inadequate statistical power to detect the effects may be a source of inconsistency in the findings reported for ETS and birth outcomes. The only one of the four studies using a biomarker that did not find a significant effect of ETS had an extremely low rate of exposure (less than 10% of women were classified as exposed based on the cotinine levels in the Eskenazi study).

Discussion
LBW increases infant morbidity and mortality worldwide. One well-established risk factor is maternal smoking. ETS exposure has recently been focused on as another potential risk factor. As we consider the feasibility of modifying women's exposure, we have focused our discussion on workplace exposure to ETS. The workplace is particularly important to consider because women of child-bearing age are present in the workplace in greater numbers than ever before. In 1994 (the most recent data available), there were 60 million women in the U.S. labor force and those women made up 46% of the total U.S. labor force. Between the ages of 20 and 44, the peak child-bearing years for women, labor force participation rates exceeded 70% for women (50). In addition, certain subgroups of working women may be particularly at risk from the effects of ETS on pregnancy because they work in environments with higher exposure or are more susceptible to its effects. In 1994, 10 million American women worked in the service industry, 4.3 million worked as machine operators, fabricators, or laborers, and 1.2 million worked in precision production, craft or repair trades (50), workplaces in which ETS may still be a problem.
This discussion evaluates the charges given to the participants in the July 1998 U.S. Occupational Safety and Health Administration workshop on ETS in the workplace and health outcomes. Three specific charges were given in the area of low birth weight: (1A) "What is the dose-response relationship for ETS exposure and LBW?"; (1B) "Are exposures in the workplace in a range of biologic concern?"; (2) "Are there studies of occupational exposure to ETS and LBW?"; (3) "Can results from studies of birth weight and ETS exposure generally be extended to the workplace?". Char 1AA What Is the Dose-Response

Relationship for ETS Exposure and LBW?
A critical examination of studies done to date and with greater weight placed on studies with fewer methodologic limitations leads this reviewer to conclude that there is a consistent and plausible association between ETS and LBW. The recent meta-analysis and qualitative literature review by Windham et al. (23) also comes to the same conclusion, suggesting an average increase in LBW of approximately 38%. In response to the first part of this charge, determination of a dose-response effect, the evidence is somewhat weaker. Few of the studies published appear to have examined this issue. However, unless ETS exposure influences birth weight through a different mechanism than does maternal smoking, we would expect a dose-response pattern similar to what is found for maternal smoking exposure. A caveat to this is that ETS exposures might be constrained to such a limited and low level as to make dose-response effects too subtle to detect and perhaps even of little relevance. Only five of the studies with positive findings (statistically significant associations of ETS exposure with birth outcomes) collected and analyzed the ETS exposure data in such a way as to examine the possibility of dose-response patterns. Both studies using interview data (28,33) and one study using biomarkers of exposure data (24) found some evidence of a dose-response trend. To respond to the latter part of the first charge, we need information on workplace ETS exposure levels for pregnant women. Although most studies report ETS exposure as the number of hours of exposure per day, we can only consider these data as surrogates for exposure levels (51). To determine if workplace exposures are in a range associated with risk, we would need studies that measure biomarker levels in pregnant women as proxies for levels of ETS in the workplace. The most appropriate biomarker of ETS exposure for pregnant women is not certain, as it is unclear which components of cigarette smoke are responsible for the reduction in fetal growth. Nicotine and carbon monoxide are the two components best established as relating to the adverse effect of maternal smoking. In terms of biomarkers, the best established for use in studies of ETS and pregnancy are nicotine and its metabolite, cotinine. Measures of carbon monoxide have not been used in studies of pregnant women. Benowitz (41), in a review of the validity of cotinine as an ETS biomarker, argues that carbon monoxide in the blood is a nonspecific and insensitive marker of ETS exposure. Therefore, although the effects of ETS may operate in part through carbon monoxide, carbon monoxide may be problematic as a best biomarker of ETS exposure in pregnancy.
To establish whether levels of ETS then determine whether the workplace can produce levels of this magnitude. Studies would be needed of nonsmoking pregnant women exposed to ETS only in the workplace. In these studies, cotinine or nicotine levels could be measured across a variety of workplace environments. Furthermore, the fetus may be more or less susceptible to a given level of ETS absorbed by the mother (measured as cotinine or nicotine) due to differences in placental hemodynamics or other interactions with the maternal-uterine-placental environment. Given these challenges, the two most feasible approaches to protect pregnant women would be to either a) rely on standards developed for nonpregnant adults with regard to health outcomes for which more data are available; or b) assume all ETS exposure is unsafe for pregnant women, as the data from two of the best and most recent studies (24,32) indicate that even low levels of maternal cotinine (1-2 ng/mL) are hazardous.  (47,52). It is worth noting that in both these studies, exposure in the home alone did not appear to have adverse effects on pregnancy; but there were effects, albeit small effects, of exposure in the workplace alone. One of the two studies (52) also showed a dose-response trend, with an increasing risk of a SGA infant being born as the number of hours of workplace ETS exposure increased. There are no obvious reasons why such studies could not be generalized. The potential for interactions between ETS and other workplace exposures is the primary factor that would complicate extrapolations of nonoccupational studies. It may be that there are workplace exposures that act synergistically to magnify the effect of ETS on birth weight. These exposures could be other airborne pollutants as well as any of a number of chemical exposures in the workplace. Stress, physical or psychologic, is another factor that could interact with ETS exposure to increase the adverse effects on birth weight. There is one important interaction that has been examined in past studies that may be particularly relevant to pregnant workers and that is the interaction between smoking and maternal age. One study of ETS (39) has shown an interaction between maternal age and ETS exposure such that ETS more than doubled the risk of LBW (OR = 2.4; 95% CI = 1.5, 3.9) for older women (30 years of age) while having no adverse effect on outcomes for younger women (OR = 0.9; 95% CI = 0.8, 1.1). These data are not inconsistent with studies of maternal smoking that show adverse effects for women of all ages but find stronger effects among older women (53). As U.S. women increasingly choose to delay their first birth (54), this interaction may be very important to consider when establishing policies. Poverty might also act to enhance the risk of ETS exposure. Poor women are a vulnerable population with regard to adverse pregnancy outcomes. Women of lower socioeconomic status have lower mean birth weights and higher rates of LBW, IUGR, and infant mortality (55)(56)(57)(58). As discussed earlier, the effect of ETS on birth weight may not be uniform across the birth weight continuum. Even if the effect is uniform, the lower mean birth weight for poor women implies that additional adverse exposures might push birth weight into the more dangerous LBW range. In the workplace, poor women are more likely to hold jobs in which exposure levels are higher (such as in the service industry or in factories) and to be in workplaces in which they have little power to ask for environmental modifications. Poor women are more likely than other women to be undocumented workers, and whether or not legal residents, poor women are more likely to work in unregulated work environments such as garment sweatshops. With the enactment of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (59) (also referred to as TANF, Temporary Assistance to Needy Families), more pregnant women from these disadvantaged groups will be in the workplace with little or no control over their working conditions except for government regulations (60). For all these reasons, ETS exposure is an issue of particular concern with regard to pregnancy outcomes in this population. Poor women are also more likely to be smokers themselves (61) and may be more likely to be exposed to other substances or conditions that increase the risk of LBW.

Conclusion
In summary, ETS exposure appears to have adverse effects on fetal growth parallel to what is seen for maternal smoking, but, as would be expected with less exposure, effects are generally smaller than those for maternal smoking. As a consequence, morbidity and mortality would be expected to be higher for infants born to women exposed to ETS during pregnancy. The workplace is one source of exposure to ETS for pregnant women that can and should be minimized to reduce risk of adverse pregnancy outcomes for working women.