Low birthweight and preterm birth rates 1 year before and after the Irish workplace smoking ban

Abstract

Objective  It is well-established that maternal smoking has adverse birth outcomes (low birthweight, LBW, and preterm births). The comprehensive Irish workplace smoking ban was successfully introduced in March 2004. We examined LBW and preterm birth rates 1 year before and after the workplace smoking ban in Dublin.

Design  A cross-sectional observational study analysing routinely collected data using the Euroking K2 maternity system.

Setting  Coombe University Maternal Hospital.

Population  Only singleton live births were included for analyses (7593 and 7648, in 2003 and 2005, respectively).

Methods  Detailed gestational and clinical characteristics were collected and analysed using multivariable logistic regression analyses and subgroup analyses.

Main outcome measures  Maternal smoking rates, mean birthweights, and adjusted odds ratios (ORs) of LBW and preterm births in 2005 versus 2003.

Results  There was a 25% decreased risk of preterm births (OR, 0.75; 95% CI, 0.59–0.96), a 43% increased risk of LBW (OR, 1.43; 95% CI, 1.10–1.85), and a 12% fall in maternal smoking rates (from 23.4 to 20.6%) in 2005 relative to 2003. Such patterns were significantly maintained when specific subgroups were also analysed. Mean birthweights decreased in 2005, but were not significant (P = 0.99). There was a marginal increase in smoking cessation before pregnancy in 2005 (P = 0.047).

Conclusions  Significant declines in preterm births and in maternal smoking rates after the smoking ban are welcome signs. However, the increased LBW birth risks might reflect a secular trend, as observed in many industrialised nations, and merits further investigations.

Introduction

Low birthweight (LBW) and preterm birth (birth prior to 37 weeks of gestation) are important risk factors for perinatal morbidity and mortality, and are associated with an increased risk of cardiovascular disease in adulthood.^1,2 Active maternal smoking during pregnancy is an important modifiable risk factor for LBW and preterm birth in Western countries.^3,4 Such risks are also evident in mothers using smokeless products.⁵ Birthweight in the offspring of smoking mothers is 150–250 g lower than in the offspring of nonsmoking mothers.^3,4 It is also well-established that there is a two-fold increased risk of LBW babies in mothers actively smoking during pregnancy, when compared with nonsmoking mothers.^6,7 A recent cohort study showed that active prenatal maternal smoking increases the risk of preterm births by almost 36%.⁸ The interesting hypothesis of a ‘smoking paradox’ around the issue of pregnancy-induced-hypertension (PIH)/pre-eclampsia also suggests a ‘protective’ effect of maternal smoking on PIH/pre-eclampsia.^9,10

Ireland introduced a comprehensive workplace smoking ban in March 2004.¹¹ The legislation was intended to protect workers from environmental exposure. Positive health effects postsmoking ban have been reported.^11,12 However, it is unclear whether the smoking ban had any impact on the maternal smoking prevalence during pregnancy. Although there are mixed reports on whether maternal second-hand smoke (SHS) exposure during pregnancy has an adverse outcome on birthweight,^13,14 the 2006 US Surgeon General report concludes that passive smoking through exposure to SHS in the workplace or at home has indeed a negative impact on birthweight and gestational age at birth.¹⁵ Also, recent studies have shown that SHS exposure during pregnancy is associated with lower mean birthweights, namely 25–75 g lower in the offspring of mothers exposed to SHS during pregnancy,^13,16 as well as with nonsignificant increased risks in both LBW and preterm births.¹⁷ However, such evidence following the introduction of a comprehensive workplace smoking ban is not available to date.

With this information in mind, this study looked at the following objectives using hospital-based data from a tertiary referral maternity hospital in Dublin.

• • 
  To determine whether there was a significant change in active maternal smoking prevalence during pregnancy after the workplace smoking ban.
• • 
  To estimate whether there were significant changes in LBW and in preterm birth rates 1 year before and after the Irish workplace smoking ban.

Methods

The Coombe Women and Infants University Hospital is a large teaching hospital located in central Dublin. As a centre of clinical excellence it provides the most extensive women’s healthcare service in Ireland, with more than 7800 babies born each year: 7854 and 7930 infants were delivered in 2003 and 2005, respectively. This cross-sectional observational study used routinely collected data from this University Hospital Clinical database system, the Euroking K2 maternity system, which validates mandatory fields for missing data. During the study period, 15 241 singleton live-birth babies (96.6%) with complete birthweight data were analysed (7593 and 7648 in 2003 and 2005, respectively), and the remaining were mostly multiple births. Only 34 of the 15 241 mothers with singleton live births had no information on smoking behaviour. All 15 241 mothers were included in the multivariable analyses for the two main health outcomes (LBW and preterm birth rates).

Main outcomes

Maternal smoking rates, gestational-age-adjusted mean birthweights, and the risks of LBW and preterm births in 2005 vis-à-vis 2003. LBW babies were defined as those weighing less than 2.5 kg, and babies born before 37 weeks of gestation were classed as preterm births.

Covariates

Detailed socio-demographic, gestational and clinical characteristics of pregnant mothers were ascertained using data completed at the time of the delivery. All babies were weighed uniformly in a single centre. In addition to collecting data on babies (dates of births, sex and birthweights) and maternal socio-economic and demographic factors (age at delivery, marital status, height, weight at booking, ethnicity and whether they were working or not), gestational (gestational age at delivery, in weeks, and parity) and clinical parameters, such as hypertension and bleeding during pregnancy, and also delivery via Caesarean section (CS), were collected. Gestational age was mainly based on an early ultrasound examination, and this practice did not change significantly between 2003 and 2005. Lifestyle factors before pregnancy, such as alcohol intake and smoking, were also ascertained. Smokers were classified as never, former or current smokers. Smoking status across different periods of gestation and any information on maternal SHS exposure during pregnancy, either in the workplace or at home, were not available.

Analyses

Univariate and multivariable logistic regression analyses were performed, using sas 9.1 (SAS Institute Inc., Cary, NC, USA). The mean birthweight in grams (95% CI) of babies in 2005 was compared wth the value for 2003 across different categories of smoking status using Cochrane–Armitage trend tests to assess significance. Subgroup analyses (separately for non-CS and non-PIH mothers) were conducted to estimate adjusted mean LBW and preterm birth rates/1000 live births before and after the smoking adjustment. Multivariable logistic regression analyses were performed to estimate the adjusted odds ratios (ORs) for the two main study outcomes: LBW and preterm birth risks in 2005 relative to 2003, which was the reference period. Separate models were estimated. The goodness of fit of the models were assessed using the Hosmer–Lemeshow test.¹⁸ All models showed good fits (P > 0.05).

Results

Table 1 shows descriptive analyses across all the covariates studied for the years 2003 and 2005. Table 1 shows that the mothers in general were significantly more likely to be taller, older, single and mostly working in 2005 versus 2003. A significant proportion of pregnant mothers (around 20%) had alcohol intake before their current pregnancies. A significant increase in gestational hypertension in 2005 versus 2003 was also observed (Table 1). Most importantly, there was a significant increase in CS in 2005 versus 2003 (19.5 versus 15.4%, respectively), a relative increase of almost 27% (Table 1).

Table 1.   Socio-demographic, gestational and clinical characteristics of mothers with singleton live-births in Dublin (2005 versus 2003)

Variables	2003 (n = 7593) %	2005 (n = 7648) %
*Not statistically significant.
Gestational age (n = 15 241)
<37 weeks	5.5	5.5
37–40 weeks	40.0	42.0
>40 weeks	54.5	52.5
Maternal age (n = 15 241)
<20 years	5.2	4.5
20–34 years	73.2	70.4
>34 years	21.6	25.1
Marital status (n = 15 241)
Single	30.7	33.7
Height (n = 13 173)
<1.58 m	20.9	19.7
≥1.58 m	64.8	67.6
Employment (n = 14 942)
Working	59.4	66.5
Not working	40.6	33.5
*Sex of the baby (n = 15 241)
Boy	50.4	50.1
Girl	49.6	49.9
Ethnicity (n = 15 241)
White	84.6	85.8
Non-white	15.4	14.2
*Parity (n = 15 241)
Primiparous	33.6	32.3
≥1 order	66.4	67.7
Alcohol intake (n = 15 241)
No	21.8	19.2
Occasional	68.2	72.6
Regular	10.0	8.2
Gestational hypertension (n = 15 241)
Yes	8.1	14.4
No	91.9	85.6
*Bleeding in pregnancy (n = 15 241)
Yes	19.8	20.2
Caesarean section (n = 15 241)
Yes	15.4	19.5

Table 2 shows gestational-age-adjusted mean birthweights and maternal smoking rates in 2005 vis-à-vis 2003, and the testing for a significant difference in mean birthweights and in maternal smoking rates. In general, there was a decrease in mean birthweights in 2005, but the observations were statistically significant in mothers who currently smoked. Second, there was a significant 12% relative decline in maternal smoking rates in 2005 compared with 2003 rates (20.6 versus 23.4%). A significant increase in smoking cessation was also observed in mothers who delivered in 2005 compared with those who delivered in 2003 (P = 0.047). Overall, currently smoking mothers had the lowest mean birthweights, whereas the former smoking mothers had the highest mean birthweights.

Table 2.   Gestational-age-adjusted mean birthweights (in grams) with 95% CIs of singleton live-births to smoking mothers in Dublin (2005 versus 2003)

Variables	2003 (95% CI) n = 7593	2005 (95% CI) n = 7648	Mean difference
*Statistically significant.
Overall
Mean birthweight (n = 15 207)	3467 (3414; 3520) (n = 7574)	3456 (3403; 3509) (n = 7633)	P = 0.45
LBW (mean weight) (n = 659)	1909 (1838; 1980) (n = 305)	1922 (1851; 1993) (n = 354)	P = 0.16
Never smokers (n = 8108)	52.6% (51.5%; 53.7%)	54.0% (52.9%; 55.1%)	P = 0.08
Mean birthweight (n = 8108)	3527 (3450; 3604) (n = 3985)	3503 (3426; 3580) (n = 4123)	P = 0.41
LBW (mean weight) (n = 280)	1972 (1860; 2084) (n = 129)	1921 (1809; 2033) (n = 151)	P = 0.58
Former smokers (n = 3755)	23.9% (23.0%; 24.9%)	25.3% (24.4%; 26.3%)	P = 0.047*
Mean birthweight (n = 3755)	3549 (3435; 3663) (n = 1818)	3547 (3433; 3661) (n = 1937)	P = 0.41
LBW (mean weight) (n = 115)	1867 (1688; 2046) (n = 50)	1828 (1649; 2007) (n = 65)	P = 0.51
Current smokers (n = 3344)	23.4% (22.5%; 24.4%)	20.6% (19.7%; 21.6%)	P < 0.001*
Mean birthweight (n = 3344)	3250 (3157; 3343) (n = 1771)	3220 (3127; 3313) (n = 1573)	P = 0.006*
LBW (mean weight) (n = 264)	1862 (1753; 1971) (n = 126)	1968 (1859; 2077) (n = 138)	P = 0.25

Subgroup analyses are shown in Tables 3 and 4. Overall, a significant increase in adjusted mean LBW rates/1000 live births was observed in both categories of mothers, who either had no PIH or had no CS interventions. Such patterns were strengthened when the smoking status of mothers was accounted for. In contrast, adjusted mean preterm birth rates/1000 live births showed a statistically significant decline only in mothers who had no CS interventions, but such patterns weakened when the smoking status of mothers was accounted for.

Table 3.   Adjusted mean low birthweight (LBW) and preterm birth rates/1000 singleton live births (before and after smoking behaviour adjustments) in 13 518 pregnant mothers without pregnancy-induced hypertension (PIH) in Dublin (2005 versus 2003)

	*Rates/1000	**Rates/1000
*Adjusted for maternal age, parity, sex of the child, maternal height, working status of mother, marital status, ethnicity, bleeding during pregnancy, alcohol intake, caesarean section, gestational age (for LBW births only) and weight of child (for preterm births only) **Adjusted for maternal age, parity, sex of the child, maternal height, working status of the mother, marital status, ethnicity, bleeding during pregnancy, alcohol intake, caesarean section, gestational age (for LBW births only), weight of child (for preterm births only) and smoking status of mother.
LBW
2003 (n = 6975)	35.1	35.1
2005 (n = 6543)	41.1	41.1
P values	0.02	0.004
Preterm births
2003 (n = 6975)	48.2	48.2
2005 (n = 6543)	51.1	51.1
P values	0.21	0.51

Table 4.   Adjusted mean low birthweight (LBW) and preterm birth rates/1000 singleton live births (before and after smoking behaviour adjustments) in 12 577 pregnant mothers without Caesarean section in Dublin (2005 versus 2003)

	*Rates/1000	**Rates/1000
*Adjusted for maternal age, parity, sex of the child, maternal height, working status of the mother, marital status, ethnicity, bleeding during pregnancy, alcohol intake, pregnancy-induced hypertension (PIH), gestational age (for LBW births only) and weight of child (for preterm births only). **Adjusted for maternal age, parity, sex of the child, maternal height, working status of the mother, marital status, ethnicity, bleeding during pregnancy, alcohol intake, PIH, gestational age (for LBW births only), weight of child (for preterm births only) and smoking status of mother.
LBW
2003 (n = 6422)	39.8	39.8
2005 (n = 6155)	44.8	44.8
P values	0.01	0.003
Preterm births
2003 (n = 6422)	55.9	55.9
2005 (n = 6155)	55.1	55.1
P values	0.009	0.06

Table 5 shows the risks of LBW and preterm births in 2005 relative to 2003. There was a 43% significantly increased LBW risk in 2005 when all the potential confounders were accounted for (OR, 1.43; 95% CI, 1.10–1.85). In contrast, there was a 25% significantly decreased preterm birth risk in 2005 when all the potential confounders were accounted for (OR, 0.75; 95% CI, 0.59–0.96).

Table 5.   Odds ratios (ORs) for the risk of preterm/low birthweight (LBW) for two time periods: unadjusted and multivariate models

	2003 (Reference)	2005 (95% CI)
*Statistically significant **Adjusted for maternal age and sex of the child. Model 1: Adjusted for maternal age, sex of the child, smoking status of mother Model 2: Model 1 + Caesarean section. Model 3: Model 2 + pregnancy-induced hypertension (PIH) Model 4: Model 3 + parity, maternal height, working status of mother, marital status, ethnicity, bleeding during pregnancy, alcohol intake, gestational age (for LBW births only) and weight of child (for preterm births only).
LBW
**Unadjusted	1.00	1.14 (0.98, 1.33)
Model 1	1.00	1.17 (1.01, 1.37)*
Model 2	1.00	1.15 (0.99, 1.34)
Model 3	1.00	1.05 (0.90, 1.23)
Model 4	1.00	1.43 (1.10, 1.85)*
Preterm births
**Unadjusted	1.00	1.00 (0.87, 1.15)
Model 1	1.00	1.02 (0.89, 1.18)
Model 2	1.00	1.02 (0.89, 1.18)
Model 3	1.00	0.96 (0.84, 1.11)
Model 4	1.00	0.75 (0.59, 0.96)*

Discussion

This cross-sectional observational study has three significant findings 1 year after the successful implementation of the comprehensive workplace smoking ban in Ireland in March 2004: a 12% (from 23.4 to 20.6%) relative decline in maternal smoking rates; a 25% decreased preterm birth risk; and, finally, a 43% increased LBW risk. The current maternal smoking rate (20.6%) is lower than the overall smoking rate of 27% among the general Irish population,¹⁹ but the maternal smoking rates are substantially higher when compared with a 7% maternal smoking rate in Massachusetts, a state with a comprehensive tobacco control program and a workplace smoking ban.²⁰ Nonetheless, smoking cessation before pregnancy has marginally increased after the smoking ban (P = 0.047).

To our knowledge, this is the first epidemiological study to have looked at the potential impact of a comprehensive workplace smoking ban on LBW and preterm birth rates, but with no historical control groups. Earlier, the 2006 US Surgeon General report had concluded that the evidence is ‘sufficient’ to infer a causal relationship between maternal SHS exposure and LBW, but that the evidence is only ‘suggestive’ and not sufficient to infer a causal relationship for preterm births.¹⁵ Such conclusions were apparently not reflected in our study. Maternal SHS exposure, especially in domestic settings, might have diluted the overall effects. In 2008 it was also reported that declining maternal smoking prevalence during pregnancy did not reduce LBW prevalence in Massachusetts.²¹ However, the observed mean birthweight difference (∼300 g) between currently and not currently smoking mothers is in agreement with earlier evidence of the impact of active smoking on mean birthweight.^3,4

Our study findings have limitations. Because of the cross-sectional nature of the study, a causal link could not be established. Despite adjusting for potential known confounders through robust multivariable logistic analyses, residual confounding is a possibility. Although minimal, exposure misclassification bias cannot be ruled out. Recent evidence suggests a specificity of almost 95% between self-reported smoking behaviour among pregnant women and an objective measurement, but under-reporting among current smokers is a possibility.²² It should also be emphasised that the workplace smoking ban was not targeted to achieve a reduction in smoking rates. Nevertheless, we observed a marginal but significant decline in maternal smoking rates. The Coombe Hospital is a national referral centre in Dublin, and therefore the study findings cannot be generalisable. Nonetheless, a single source of data centre, using a uniform data collection system, might have minimised any underlying potential biases.

Both PIH and CS have been shown to influence adverse birth outcomes,^9,10 but our subgroup analyses accounted for PIH and CS, and yet showed significant patterns in both LBW and preterm birth rates (Tables 3 and 4). From a public health policy context, the observed significant increase in CS should be explored further. A significant increase in PIH observed in our study is interesting, especially in the context of the apparent ‘protective’ effect of maternal smoking on PIH. A recent prospective study reported a significant inverse exposure–response relationship of maternal smoking and eclampsia only, and not for PIH.⁹ In our study, we had no information on pre-eclampsia and, therefore, such effects could not be ascertained. A recent New York City study suggested that smoking is only protective against pre-eclampsia without pregestational hypertension, and even then principally among young women.²³ A recent review also concluded that a reduced risk of pre-eclampsia among smoking mothers is unlikely because of residual confounding.¹⁰ Nevertheless, the absence of deprivation measures in our study might have influenced the study findings, because a social gradient does exist in preterm birth trends that cannot be fully explained through adjustment for maternal smoking.²⁴

Furthermore, the long-term health implications of infants born to mothers who actively smoked during pregnancy are well-established.^2,25 Therefore, it is imperative that factors potentially contributing to an increased LBW risk should be identified. Such a rise in LBW rates is not only observed in Ireland but also in many industrialised nations.²⁶ Reductions in mean birthweights are also observed elsewhere.²⁷ An increase in CS interventions might have influenced such a pattern, as observed in Brazil,²⁸ but an overall reduction in mean birthweights was maintained, despite accounting for gestational age (Table 2). Finally, LBW and preterm births are modifiable and preventable risk factors. Therefore, additional perinatal, infant, as well as adult morbidity and mortality could be prevented, or at least reduced, if comprehensive health promotional programs surrounding prenatal maternal smoking and SHS exposure are actively pursued and integrated into prenatal care, both in the community and in clinical settings.

In conclusion, our study demonstrated a significant fall in maternal smoking rates during pregnancy after the workplace smoking ban, with a marginal increase in smoking cessation among mothers 1 year after the ban. A significant reduction in preterm birth rates was also observed 1 year after the workplace smoking ban, which is a welcome sign. The observed significant increase in LBW risk is intriguing, and needs further exploration, especially in the light of the 2006 US Surgeon General report, which had concluded that the evidence is ‘sufficient’ to infer a causal relationship between maternal SHS exposure and LBW. Nevertheless, the findings of this cross-sectional study are to be interpreted with caution, and a causal inference cannot be drawn.

Disclosure of interest

None.

Contribution to authorship

LC, VC and SD conceived the study and secured the grant. VC and EM contributed to data management. RC performed the initial statistical analyses. ZK conducted the literature review, performed the final analyses and drafted the first original manuscript. LC, SD, RC, VC and ZK contributed to the interpretation of the results. All the co-authors contributed to the final manuscript. LC is the guarantor of the paper.

Details of ethics approval

No ethical approval was sought because of the secondary analyses of anonymised data without any patient identifiers.

Funding

The Health Research Board of Ireland funded this project.

Acknowledgements

We thank Ms Helen Finnerty for her earlier contribution to this project, the Health Research Board for the financial support and Ms Sheila Keogan for her contribution to the initial grant proposal.