Maternal Active and Passive Smoking and Fetal Growth : A Prospective Study in Nagoya , Japan

In order to examine the effect of maternal active and passive smoking on fetal growth, we carried out a population-based cohort study. A self-administered questionnaire was distributed to 15,207 women who notified their pregnancy from April, 1989 to March, 1991. A total of 7,411 mother-singleton infant pairs were analyzed in this study. Paternal smoking status and maternal hours exposed to environmental tobacco smoke (ETS) were used as indicators of passive smoking. Infants born to active smoking mothers were 96 g lighter, on an average, at birth than those born to non-smokers, and the relative risk for intrauterine growth retardation was 1.79 (95% confidence interval (CI) = 1.05-3.04) among active smoking mothers. Infants with smoking fathers weighted 11 g lighter, on an average, than those with non-smoking fathers, and mean birth weight of infants was reduced by 19 g among mothers exposed to ETS. The relative risk for intrauterine growth retardation in non-smoking pregnants with a smoking husband and those exposed to ETS was 0.95 (95% Cl = 0.72-1.26) and 0.95 (95% Cl = 0.71-1.26), respectively. Our findings indicated an adverse effect of maternal active smoking on fetal growth in Japanese pregnant population, but with small influence of maternal passive smoking. J Epidemiol, 2000 ; 10 : 335343


INTRODUCTION
During the past 40 years, many investigators 1-13) have proved that maternal active smoking during pregnancy is hazardous to fetal growth.The effect of passive exposure to environmental tobacco smoke (ETS) on fetal growth is, however, still controversial.
showed that mean birth weight of infants was reduced by 120 g per pack of cigarettes smoked per day by father.Martin et al's.reported that the RR for intrauterine growth retardation (IUGR) was 2.17 among ETS exposed mothers.On the other hand, Peacock et al 33).reported that infants born to mothers with ETS-exposure were insignificantly lighter at birth by 6.7 g, on an average, than those born to non-exposed mothers.
The RR for low birth weight was reported as 0 .99among single full-term live births by Ahlborg et al 28).In Japan, several epidemiological studies 4,6,24,26,29) have assessed the relation between maternal ETS-exposure and birth outcomes.According to Nakamura et al 24) ., the RR for low birth weight was 1.4, though not significantly larger, among non-smoking mothers with smoking husbands, compared with those with non-smoking husbands.Saito Q found that infants with fathers who smoked over 40 cigarettes per day weighted significantly lighter by 111.4g at birth than those with nonsmoking fathers.Ogawa et al 1).reported that mean birth weight was reduced by 10.8 g with the RR of IUGR of 1.0 among mothers exposed to ETS over two hours per day.Among these Japanese studies, Nakamura's study N was only one population-based cohort study, but their study population was not so large (n=3,478) and an indicator of ETS-exposure was only paternal smoking status.In general, many husbands have recently attempted-to avoid exposing-their wives to ETS, and, therefore, an assessment of ETS-exposure only by paternal smoking status may produce a certain degree of misclassification.
We initiated a large-scaled population-based cohort study with two indicators of ETS-exposure, enrolling 15,207 pregnant women in Nagoya, Japan in 1989.This study will allow us a) to assess smoking effect on birth outcomes without mother's recall bias, b) to include the entire pregnant population at the areas studied, minimizing selection bias, and c) to measure ETS-exposure with two indicators: paternal smoking status and maternal hours exposed to ETS.

MATERIALS AND METHODS (1) Study population
In Japan, pregnant women have been obliged to notify their pregnancy with medical certificates to their local government offices.A self-administered questionnaire which inquired lifestyle was distributed to 15,207 pregnants who notified their pregnancy to four health centers in Nagoya from April 1, 1989, to March 31, 1991.Out of these pregnants, 8,624 returned the questionnaires (response rate = 56.7%).In Nagoya, pregnants were also requested to report their birth outcome after delivery by a specified postcard.Of the 8,624 responded pregnants, 7,465 returned the postcards.After excluding 54 multiple pregnancies, 7,411 mother-singleton infant pairs were analyzed in the present study.
(2) Data collection Information obtained from notifications of pregnancy and medical certificates were gestational age and maternal age at notification, values of hemoglobin and blood pressure, and test result by urinalysis at diagnosis of pregnancy.Information obtained from the specified postcard after delivery were maternal parity, infant gender, and gestational age and body weight at birth.Information on maternal height and weight before pregnancy, education and lifestyle including smoking status were obtained by a self-administered questionnaire.
The maternal active smoking status at notification (mean gestational age = 15.3 weeks) was used to classify mothers into smokers, quitted and non-smokers.Non-smokers included women who quitted smoking before their pregnancies.Among non-smoking pregnants, two indicators of ETS-exposure were measured: a) husband's smoking status at notification and b) maternal hours exposed to ETS at home, at workplace or elsewhere.Low birth weight was defined as less than 2,500 g, preterm birth as a gestational age of less than 37 weeks, and intrauterine growth retardation (IUGR) as a score of less than -1.5 standard deviation on the Japanese standard fetal growth curve.
(3) Data analysis Chi-square test, Student t-test or analysis of variance was used when examining the difference between groups subdivided by response to a questionnaire and smoking status (table 1,2 and 3).In calculating mean gestational age and birth weight (table 4, 5), adjustment was performed by analysis of covariance, and multiple logistic regression analysis was applied when obtaining adjusted RR and its 95% confidence interval (CI).In the above analysis, potential risk factors or confounders were adjusted such as maternal age at birth (years), height (cm), body mass index (kg/m2), educational period (years), working status (working, quitted or not working), alcohol intake (drinking, quitted or not drinking), parity (nulliparous or multiparous) and infant gender.All statistical analysis were performed using the Statistical Analysis System 35).

RESULT
Table 1 compares maternal and infant characteristics between pregnants who responded and those not-responded to a self-administered questionnaire.Responded pregnants notified their pregnancy much earlier than not-responded pregnants, and were more frequently nulliparous.No significant difference was, however, noted in mean values of maternal age at notification, hemoglobin and blood pressure, positive rate of urinary glucose / protein at diagnosis of pregnancy, infant gender, mean gestational age or mean body weight at birth.Pregnants who returned a self-administered questionnaire mailed back the postcard after delivery more frequently than not-responded pregnants.
Table 2 details associations with maternal active smoking status.Among 7,411 pregnants who included 65 (0.9%) with missing smoking status, 6,335 (85.5%) were non-smokers, 726 (9.8%) quitted smoking after being aware of their pregnancy (mean gestational age at smoking cessation = 8.8 weeks), and 285 (3.8%) continued to smoke after her being pregnant (average amount smoked = 9.7 cigarettes per day).Compared with non-smoking pregnants, continuously smoking pregnants had shorter educational period, were more frequently continuous alcohol drinkers, with a higher percentage of multiparous parity and of delivering girls.
Table 3 shows maternal and infant characteristics by paternal Table 1.Maternal and infant characteristics among pregnants responded and not-responded to a self-administered questionnaire.
SD: standard deviation NS: Not significant smoking status or ETS-exposure.Among 6,335 non-smoking pregnants who included 36 (0.6%) with missing information on paternal smoking status, 2,630 (41.5%) husbands were nonsmokers, 97 (1.5%)quitted smoking after their wives become pregnant, and 3,572 (56.4%) continued to smoke.Non-smoking pregnants with a husband who continued to smoke were younger, and had shorter educational period than those with a non-smoking husband.
Among 6,335 non-smoking pregnants who included 56 (0.9%) with missing information on ETS-exposure, 3,586 (56.6%) self-reported as those passively exposed to ETS at home, at workplace or elsewhere.
Compared with nonexposed pregnants, ETS-exposed pregnants were younger, more frequently employed, alcohol drinkers and nulliparous, and had shorter educational period.Table 4 summarizes the relationship between maternal active smoking and fetal growth.Infants born to active smoking mothers were 96 g lighter, on an average, with statistical significance than those born to non-smoking mothers.The RR for low birth weight and IUGR were 1.89 (95%CI = 1.09-3.26)and 1.79 (95%CI = 1.05-3.04) in active smoking mothers compared with non-smoking mothers.Among continuously smok-ing mothers, a significant trend toward lower mean birth weight and higher RR of both low birth weight and IUGR were evident as the number of cigarettes smoked per day increased.Mean gestational age and the RR of preterm birth were, however, not significantly different between smoking and non-smoking mothers.Mean birth weight of infants born to smoking-quitted mothers was 42g heavier with statistical significance than that of infants born to non-smoking mothers.Table 5 details confounders-adjusted mean birth weight and gestational age, and the RR for low birth weight, preterm birth and IUGR by paternal smoking status or by maternal ETSexposure among non-smoking pregnants.Adjusted mean birth weight of infants (3,091g) among continuously smoking fathers was 11 g lighter than that of infants (3 ,102g) among non-smoking fathers, though not statistically significant .Mean gestational age was not significantly different between smoking and non-smoking fathers, but among continuously smoking fathers, infants with fathers who smoked more than 20 cigarettes per day were significantly shorter at gestational age than those with fathers who smoked not more than 20 cigarettes .For smoking fathers, adjusted RR was found to be 0 .92(95%CI = 0.71-1.20)for low birth weight, 1.04 (95%CI =

DISCUSSION
A number of investigators 1-13) have reported the adverse effect of maternal active smoking on fetal growth since Simpson 1).According to Kramer s meta-analysis 13), it is evident that mean birth weight of infants is 149 g lighter among active smoking mothers than among non-smoking mothers, with the RR of 2.42 for IUGR.
In our study, birth weight of infants born to active smoking mothers was 96 g lighter, on an average, compared with those born to non-smoking mothers, and the birth weight notably decreased with advancing number of cigarettes smoked per day: being almost doubled RR with statistically significance for low birth weight (RR = 1.89) and IUGR (RR = 1.79).These findings are in essentially parallel with the result of meta-analysis by Kramer 13).
Our RR for preterm birth was slightly smaller among smoking mothers than among non-smoking mothers.This smaller RR may be ascribable to relatively small number of preterm births in our study.Birth weight of infants born to mothers who quitted smoking after being aware of their pregnancy was significantly heavier, compared with those born to non-smoking mothers (3,140 g vs. 3,098 g).This fording may be interpreted as merely due to chance or reflecting a possibility that mothers who quitted smoking were more conscious about their birth-outcomes.It would be warranted to be disclosed by further studies which focus this particular finding.
Previous epidemiological studies on the effects of maternal ETS-exposure on fetal growth are rather controversial as discussed earlier.Some studies 4, 6.17 "23) have shown statistically significant association, though they ranged from studies claiming rather large effects to those showing small effects.In contrast, several other studies 24-33) have reported no significant association.
Possible explanation for this controversy may be as follows.One is the difficulty in accurately measuring ubiquitous ETS-expo-sure.Self-administered questionnaires have been used to assess ETS-exposure in numerous epidemiological studies 1-11,17-33) Validity of self-reported ETS-exposure was examined by comparing the data from smokers in the same household 36-39), with the conclusion that smoking assessment by spouse or other household members by an interview or questionnaire was substantially valid.Validity studies by such biochemical markers as urinary and salivary cotinine have been conducted in recent years 22,40,41).According to Kawachi and Colditz 42) , who have reviewed these biochemical validity studies, concluded that self-reported ETS-exposure moderately correlated with biochemical markers (Pearson's correlation coefficient ranging from 0.2 to 0.5).
Some studies 11,15,19,26,32,33) had examined the effect of ETSexposure measured by biochemical markers on fetal growth, but the findings were still conflicting.Biochemical markers could be regarded as objective and quantitative ETS-indicators, but could only provide ETS-exposure measured within a limited time.To cotinine level, large inter-individual differences in metabolism are inherent 43).In addition, the distribution of cotinine level among non-smoking women who did and did not report ETS-exposure had substantial overlap 33).O'Connor et al 44).evaluated passive smoking exposure by such three methods as individual air monitors, urinary cotinine and questionnaires, and found that exposure measured by air monitors was in "fair" agreement with that by self-administered questionnaires , but in "poor" agreement with that by urinary cotinine.Measurement of ETS-exposure by biochemical marker or by individual air monitor are usually expensive and not practical for use in large-scaled epidemiological studies.This limitation led us to use a self-administered questionnaire in our study.
In a usual questionnaire, different measures have been adopted as an indicator of passive smoking, which might assess intensity or duration.In previous studies, such indicators were mainly either husband's smoking habit 6,14,15,17.20 22, 24,25,30) or maternal hours passively exposed to ETS 16.18 27.29).In our questionnaire, both indicators were included to assess passive smoking.
Another explanation for controversy is confounding by other risk factors, since individuals who exposed to ETS tended to have a less favorable health profile compared with those not exposed 45-48).This indicates the necessity of adjusting potential confounders which are associated with pregnancy outcomes, when analyzing an association by ETS-exposure.This adjustment was readily done in our study.
In our study, mean birth weight of infants was slightly lighter among ETS-exposed mothers with statistical significance than among non-exposed mothers.Among infants with a smoking father, a similar small reduction in birth weight was found, though not with statistical significance.Our RR of ETS-exposure for low birth weight, preterm birth and IUGR was not significantly different from unity when compared with non-exposed mothers; being essentially the same finding as that by Ogawa et al 29).
One possible explanation for our failure to find substantial, if any, effect of passive smoking may be misclassification.It will be comparatively easy for a pregnant to answer her husband's smoking status with number of cigarettes smoked per day.Misclassification, however, might be more likely to occur when pregnants were exposed to ETS outside home or when their husbands attempted to avoid exposing their wives to ETS.In our study, 4.6% of non-smoking pregnants living with nonsmoking husbands self-reported ETS-exposure at home.This finding may be possibly due to smoking by someone in the household other than husband, but may be in part explained by mis-report.Therefore, we analyzed an association of ETSexposure at home limitedly among the pregnants who living with a husband and children only.According to this limitation, only 1.1% of non-smoking pregnants with a non-smoking husband self-reported ETS-exposure at home: being estimated a mis-report rate as approximately 1% in our study.In contrast, 25.0% of non-smoking pregnants living with a smoking husband self-reported no exposure to ETS at home.Assuming that the mis-report rate was similar to the rate of reverse misreport, we could infer that smoking husbands attempted to avoid exposing their wives to ETS with considerable proportion.In addition, 65.9% of non-smoking pregnants who continued to work after pregnancy self-reported ETS-exposure at workplace in our study.When considering these background all together, an assessment of ETS-exposure only by paternal smoking status could still include a certain degree of misclassification.
Use of maternal hours exposed to ETS as an indicator has an advantage.We could assess ETS-exposure outside home and also that at home by someone other than husband as well.However, pregnants might subjectively answer to a question on exposure-duration, and such duration is therefore poorly quantifiable.
Another possible explanation for our failure to find substantial, if any, effect may be the possible influence by uncontrolled confounding factors.A growing number of studies have addressed the issue of systematic difference in health characteristics of individuals exposed to ETS compared with those not exposed.Some studies 45-48) have shown that individuals exposed to ETS tend to have a less favorable health profile compared with those not exposed.Failure to control the effects of potentially confounding factors may usually result in a modest inflation of the relative risk estimates.In our study, major potential confounders associated with pregnancy outcomes were adjusted, but we could not rule out the possible influence by uncontrolled confounding factors such as scioeconomic status, psychosocial stress and others.
We found a similar small reduction in mean birth weight of infants among passive smoking mothers when passive smoking was assessed by either husband's smoking status or mater-nal hours exposed to ETS.This finding is virtually consistent with those obtained by many previous studies, which showed a 10 to 100 g decreased birth weight of infants among ETSexposed pregnants 18.19.21-23,24-33)Our study was a population-based cohort study, and we, therefore, could minimize selection bias as well as recall bias, and could evaluate the effect of ETS-exposure on fetal growth by using two indicators which would reflect different measures of ETS-dose.
Our study might still have some other methodological limitations.First, about half of pregnants, who were requested to complete a self-administered questionnaires, returned questionnaires.This indicates an undeniable possibility that the study was affected by response bias.However, since no substantial differences were noted in pregnancy outcomes between pregnants who returned the questionnaires and those who did not, our findings could generally be applicable to the entire pregnant population in Nagoya.
Second, our data on smoking status and ETS-exposure were those at the time when each pregnant submitted a notification of pregnancy to the health center.If maternal smoking status changed after notification, the effects of maternal active smoking might have been underestimated.In this respect, there is a report by Wang et al 11).who found that maternal active smoking was highly correlated with each other between three intervals of gestation.If paternal smoking habit changed after notification, we might underestimate the effects of ETS-exposure.In this respect, we knew in our study that since only 1.5% of husbands quitted smoking after their wives became pregnant, few husbands changed their smoking habits after their wives' notification of pregnancy to health centers.
In conclusion, our study indicated that maternal active smoking harmfully affected fetal growth, and pregnants, therefore, should quit smoking in order to achieve sound fetal growth.
Our study suggested only the minimal reduction of fetal growth by ETS-exposure during pregnancy, but a slight downward shift in birth weight distribution is believed to be an important public health issue in Japanese pregnant population, since over 50% of pregnants self-reported the presence of ETS-exposure at home, at workplace or elsewhere in Japan .

Table 2 .
Characteristics by maternal active smoking status.

Table 3 .
Characteristics by paternal smoking status or ETS-exposure among non-smoking pregnants.

Table 4 .
Maternal active smoking and fetal growth.

Table 5 .
Passive smoking and fetal growth among non-smoking pregnants..78-1.40)for preterm birth, and 0.95 (95%CI = 0.72-1.26)forIUGR, compared with non-smoking fathers.No uniform trends were observed in adjusted RR for low birth weight, pretemi birth, and IUGR by paternal smoking amount.Adjusted mean birth weight of infants (3,089 g) among nonsmoking pregnants who have been ever exposed to ETS was 19 g lighter, with statistical significance, than that of infants (3,108 g) among non-exposed pregnants, though not with dose-response relationship by hour ETS-exposed.No significant difference in adjusted mean gestational age was found between ETS-exposed and non-exposed pregnants, with adjusted RR of 0.99 (95%CI = 0.75-1.30)for low birth weight, 0.92 (95%CI = 0.68-1.23)for preterm birth, and 0.95 (95%CI = 0.71-1.26)for IUGR.No significant RR was also noted by hour ETS-exposed for low birth weight, preterm birth and IUGR.