Maternal Smoking during Pregnancy and Necrotizing Enterocolitis-associated Infant Mortality in Preterm Babies

Few studies have examined the possible pregnancy-related risk factors for necrotizing enterocolitis (NEC)-associated deaths during infancy. Infant death due to NEC in preterm babies was identified from the US Linked Livebirth and Infant Death records between 2000 and 2004. The average number of cigarettes per day reported by the mothers who were smoking during pregnancy was classified in three categories: non-smoking, light smoking (<10 cigarettes/day) and heavy smoking (≥10 cigarettes/day). Logistic regression analyses examined the association between prenatal smoking and NEC-associated infant mortality rates with adjustment for potential confounders. Compared with non-smoking mothers, light and heavy smoking mothers have a higher risk of NEC-associated infant mortality [light smoking: adjusted odds ratio (aOR) = 1.21, 95% confidence interval (CI), 1.03–1.43; heavy smoking: aOR = 1.30, 95% CI, 1.12–1.52], respectively. Moreover, the association was stronger among white race (light smoking: aOR = 1.69, 95% CI, 1.34–2.13; heavy smoking: aOR = 1.44, 95% CI, 1.18–1.75) and female babies (light smoking: aOR = 1.31, 95% CI, 1.02–1.69; heavy smoking: aOR = 1.62, 95% CI, 1.29–2.02). Maternal smoking during pregnancy is associated with increased risks of infant mortality due to NEC in preterm babies, especially in white race and female babies.

a pathologically hypoxic environment 14 . Despite the risks, a considerable number of women continue to smoke when they become pregnant. As much as one-fourth to two-thirds women who smoke prior to pregnancy will continue to smoke while pregnant 15,16 .
To date, only few epidemiological studies investigated the possible associations between maternal smoking during pregnancy and NEC with inconsistent results, some reporting a negative relationship 6 , while others showing no association 17 .
We hypothesized that maternal smoking during pregnancy increases the risk of death in NEC-affected premature babies. Thus, this population-based cohort study was designed to examine the potential relationship between maternal smoking throughout pregnancy and NEC-associated infant death in preterm babies.

Methods
In this study, we used the data from the US Linked Livebirth and Infant Death files from 2000 to 2004, which included all births in the USA and linked infant (< 1 year of age) deaths, recorded by death certificates 18 . The dataset was published by the National Center for Health Statistics (NCHS), US Centers for Disease Control and Prevention (CDC). A detailed description of this data is provided elsewhere 18 . The available information in this data repository included demographic characteristics of mothers, obstetric history, maternal smoking, birth outcomes, infant death, and underlying cause of death.
Maternal smoking identified on the birth certificate included a "yes" or "no" response and the average number of cigarettes smoked per day during pregnancy. We coded maternal smoking during pregnancy as a trichotomous variable: (1) nonsmokers, (2) light smokers, women who reported smoking 1 to 10 cigarettes per day during pregnancy, and (3) heavy smokers, women who reported smoking more than 10 cigarettes per day during pregnancy. Maternal smoking was not reported in California, Indiana, South Dakota, and New York State (except New York City) during the study period. These subjects were excluded from the analysis.
Women were classified in three groups based on maternal age at delivery: 19 years or less, 20-34 years, and 35 years or older. Maternal race/ethnicity was classified as non-Hispanic white, non-Hispanic black, Hispanic, and Asian. Maternal educational attainment was collapsed into four categories: less than high school (< 12 years), completed high school ( = 12 years), college (13-16 years), and graduate school (≥ 17 years). Marital status was recorded as a dichotomos variable (married/unmarried). Infant birthweight was expressed in grams and infant gender was recorded as male/female. Gestational age at delivery was estimated based on the date of the last menstrual period (LMP); if the LMP was unreliable or if there was a substantial discordance between the clinical estimate of gestational age and LMP (> 2 weeks), the first clinical estimate of gestational age was used 19 . Based on the month when prenatal care services were first accessed, women were classified into four groups: received prenatal care from first trimester of pregnancy (months 1-3), from the second trimester (months 4-6), from the third trimester (months 7-9), and no prenatal care.
NEC-associated infant deaths (< 1 year of age) were defined as deaths for which International Classification of Diseases ICD-10 code P77 was recorded on the death certificate 20 . NEC-associated infant deaths were further divided into three categories: early neonatal death (0-6 days), late neonatal death (7-27 days), and postneonatal death (28-364 days).
We restricted our analysis to singleton live births at gestational ages of 24-36 weeks and with birthweights of 500 g or more. There were 20,365,992 live births in the linked 2000-2004 birth and infant death data set ( Fig. 1). We excluded 668,171 multiple gestations, 17,964,345 newborns delivered before 24 or after 36 weeks of gestation, and 4,919 births of less than 500 g. Subjects with missing information on important variables including maternal smoking (335,300), maternal race/ethnicity (15,388), education (18,736), month prenatal care (37,638), method of delivery (5,861), and infant birthweight (141) were further excluded, leaving 1,315,493 subjects for the final analysis.
NEC infant mortality rates, calculated as the number of deaths per 10,000 live births, were determined overall and by several variables such as birthweight, gestational age, infant sex, and maternal race/ethnicity. Unconditional logistic regression models were constructed to examine the association between maternal smoking during pregnancy and risk of NEC-associated infant mortality. Odds ratios (ORs) and the corresponding 95% confidence intervals (CIs) were computed after adjusting for potential confounders. The confounders were chosen for inclusion in the adjusted models based on information from previous literature: maternal race/ethnicity, age, education, marital status, month that prenatal care started, infant sex, gestational age, and birth year [9][10][11][12] . Infants born to non-smoking mothers were the reference group. Cox proportional hazard model was used to estimate risk ratio of infant mortality among the nonsmokers, light smokers, and heavy smokers, adjusting for the above confounders. All analyses were performed with Statistical Analysis System (version 9.2, SAS Institute, Cary, N.C.).
Since these files contain anonymized public data, an ethics approval was not required by an Institutional Review Board. Table 1 presents the characteristics of preterm pregnancies (24-36 gestational weeks) resulted in live births and NEC-related infant mortality in the USA from 2000 to 2004. From a total of 1,315,493 eligible singletons, live preterm babies, 172,096 (13.1%) were exposed to prenatal smoking. Within the first year of life, 1,122 NEC-associated deaths were recorded leading to a mortality rate of 8.5 per 10,000 live births due to this cause. The majority (82.7%) of infant deaths due to NEC occurred in the late neonatal period, 7 to 27 days after birth.

Results
Maternal race/ethnicity comprised of non-Hispanic White (55.9%), non-Hispanic Black (23.4%), Hispanic (17.1%) and Asian (3.7%). Nearly three-quarters (72.3%) of the women were between 20 and 34 years old, and two-thirds (59.2%) were married. The majority of women have attained education at the level of high school or higher (76.5%), and attended prenatal care from the first trimester of pregnancy (80.8%). Approximately 83% of preterm deliveries were late preterm (33 to 36 weeks), little over half (53.9%) of the preterm babies were male.
SCiEnTifiC REPORtS | 7:45784 | DOI: 10.1038/srep45784 Supplementary Table 1 compares characteristics between smokers and non-smokers. Several characteristics including maternal age, race/ethnicity, education, marital status, prenatal care, infant sex, gestational age, and birthweight were significantly different between the two groups. Compared with non-smokers, smokers were less likely to attain high educational level and attend prenatal care, and more likely to be white, single, and young childbearing age.
Using the Cox proportional hazard models, we analyzed the survival curves among the nonsmokers, light smokers, and heavy smokers stratified by (white) race and (female) gender of the babies. The three survival curves were similar, showing a slight difference of NEC-associated infant mortality among the three maternal groups (S upplementary Figures 1 and 2).

Discussion
In the present study, we examined the potential relationship of maternal smoking during pregnancy with NEC-associated infant mortality in preterm babies. Our results suggest that increased risk of infant mortality due to NEC is associated with smoking during pregnancy, both light and heavy prenatal smoking. This association appears to be stronger among infants of female gender and white race. The Cox proportional hazard models showed a slight difference of NEC-associated infant mortality among the three maternal groups.
A previous case-control study from the US identified maternal smoking during pregnancy as the only risk factor that was associated with the development of NEC in infants 6 . In contrast, a recent Canadian retrospective cohort study, examining the adverse impacts of maternal cigarette smoking on preterm infants, found no association between maternal smoking and NEC-associated mortality 17 . Our results support the relationship of prenatal smoking with NEC-associated mortality in preterm infants. Although studies on maternal smoking during pregnancy and infant mortality are not new, the mechanism through which tobacco causes death during infancy is still poorly understood. The proposed biological mechanisms include fetal hypoxia due to increased levels of carboxyhemoglobin and resulting decreased levels of blood oxygen, nicotine induced placental vasoconstriction, and placental vascular diseases leading to short-and long-term adverse effects on their offspring health 21 .
The effect of maternal smoking during pregnancy on the developing placenta has long been of clinical interest. Nicotine exposure impairs the initial stage of placental vascularization, the invasion of maternal blood vessels into the cytotrophoblasts 6,14 . Nicotine from the mother can also readily cross the placenta and transfer to the fetus, rendering the fetus susceptible to direct poisoning 13 . Maternal cigarette smoking has been related with lower gestational age at birth, decreased birth weight, and shorter crown-heel length in many studies 22,23 . Despite the aforementioned studies linking smoking to abnormal vascular development, few studies actually show a causality for the mechanism for this association.
We further found that the association between maternal smoking during pregnancy and NEC-associated infant mortality was stronger among white race and female babies. Although it is not clear what biological mechanisms could explain these findings, gender-and racial-specific effects of prenatal smoking on infant deaths due to NEC may exist. In previous animal studies, nicotine was showed to have sex-dependent effects on offspring health. Perters and Tang (1982) found that prenatal nicotine treatment of adult female rats reduced both the number of male born and male birth weight, whereas female offsprings tended to be unaffected 24 . Subsequently, several human studies suggested that the adverse effect of prenatal smoking on offspring growth and development  was more pronounced among males than females [25][26][27] . However, a recent large study (n = 1,815,318), using the 1995 -1997 German birth statistics, revealed that the negative effect of maternal smoking during pregnancy on the birth weight and risk of small-for-gestational-age (SGA) was significantly greater in newborn female than in newborn male 28 . Thus, our findings are consistent with the latter study.
Regarding the racial and ethnic differences in the association between maternal smoking during pregnancy and offspring health, there have been inconsistent findings. While some studies reported that the negative effect was greater for black children than for white children 29,30 , others reported stronger associations among white than black children 31,32 . In the present study, the association between maternal smoking and NEC-associated infant mortality was stronger among white babies as compared with the babies from other ethnic/racial groups. The following two factors may explain the differences found across race/ethnicity categories. Firstly, different genetic polymorphisms affecting nicotine metabolism probably exists among different racial/ethnic groups 33 . Variations in enzyme activity could alter the toxic effects on the offspring. Secondly, dose of nicotine per cigarette may vary, for instance due to racial/ethnic group preferences for certain types of cigarettes 34 . Thus, the same quantity of cigarettes may not mean the same dose of toxins across various groups.
We also observed that 82.7% of infant deaths occurred in the late neonatal period (i.e., in infants who survived early causes of mortality). This pattern is consistent with the expected timing of NEC onset reported in the literature 5 . Despite the decline in the overall infant mortality observed during the last several decades 5 , improvements in survival rates of very early preterm/low birth weight babies from early causes of death, may increase the rates of NEC-associated mortality in these immature infants, and explain the persistent high mortality rate (8.5 per 10,000 live preterm singleton births) reported here and elsewhere 5 . Intervention strategies to decrease the incidence of prematurity may have a potential greater impact on reducing the number of NEC-associated morbidity and mortality.
The large population-based database linking birth and infant death data is a major strength of our study. However, some limitations are worth mentioning. Firstly, smoking on the birth certificates was self-reported, which may have introduced some reporting bias, and potentially underestimated the percentage of smokers in pregnancy, and, therefore, likely affected the reliability of smoking prevalence. With the highly advertised adverse perinatal effects of smoking in pregnancy and the mounting social and medical pressure on women to quit smoking during pregnancy, the accuracy of the information on smoking during pregnancy may be questionable 35 . Secondly, uncontrolled genetic and environmental confounding effects may exist. For example, human milk is the preferred diet for preterm babies as it protects against NEC, but we are unable to control for the type of feeding or breastfeeding rate. Evidence from observational and trial data suggests that human milk diet reduces the incidence and severity of NEC 5,36 . A dose-related association was reported between human milk feeding and a reduction of risk of NEC or death after the first 2 weeks of life among extremely low birth weight infants 37 . In our study, we did not control for the type of feeding, as no data regarding feeding and nutrition was available; thus, we acknowledge here this limitation of our study. Smoking may be a proxy for other exposures such as alcohol and stress, for which there are no reliable measures. Finally, a diagnosis of NEC could include Bell's stage I, which encompassed a significant proportion of non-NEC entities. However, by focusing on NEC mortality, this criticism is largely obviated. NEC as a cause of mortality is a clearly defined entity. While the coincidence with other conditions may have occurred (i.e., sepsis first, NEC second), it would likely be infrequent. In summary, our study demonstrates that maternal smoking during pregnancy is associated with NEC-specific infant mortality in premature babies. These findings may have implications in basic and epidemiologic research on the mechanisms that underlie NEC-associated development and death.