Smoking During Pregnancy and Infantile Colic

DISCUSSION

The results indicate that maternal smoking during pregnancy increases the risk of IC. The major strengths of this study are its size and the prospective collection of information on smoking and potential confounders during pregnancy. It is widely known that smokers and nonsmokers differ also in respect to other lifestyle factors. It therefore is important to adjust for possible confounders. However, residual confounding or confounding by unknown confounders still may be present. We also performed the analyses on term children with a birth weight above 2500 g only, but the results remained essentially unchanged.

We strove to avoid misclassification by seeking information on IC through 2 different sources—parents and health visitors—neither of whom knew the purpose of the study. Both the mothers and the health visitors answered the questionnaires retrospectively, which may have introduced a recall bias. The interobserver agreement between the mothers' and the health visitors' reporting of IC was good with a Kappa coefficient of 0.8,²⁵ and our results remained essentially the same whether the analyses were performed with the maternal information on IC or with the health visitors' recordings. We do not have information about the time at which the health visitor filled in the questionnaires; if they consulted the mother first, then this may explain the high Kappa coefficient. The occurrence of IC among the children whose IC information was from the health visitors' questionnaires did not differ from that of the children whose IC information was from the parents' questionnaires, indicating no major misclassification between the parents' and the health visitors' questionnaires.

There is a remote risk of a certain recall bias as information on IC was collected up to 5 months after the symptoms disappeared, but it seems unlikely that recall was related to the mother's reporting of smoking during pregnancy. A differential misclassification would be possible if, for example, smokers reported more IC than the true value of IC compared with nonsmokers. This would overestimate the association between smoking and IC. A bias toward the null hypothesis would be possible if the smokers recorded less IC than the true value of IC compared with nonsmokers. We were not able to conclude on a direction of this possible bias, but an IC occurrence of 10.8% is in accordance with another study that uses the same definition of IC.²⁶

The primary limitation of our study is our inability to adjust for postnatal smoking, because only few pregnant women changed their smoking habits from pregnancy to the postnatal period. The association between prenatal and postnatal smoking and IC could be attributable to an association between postnatal smoking in the home and IC rather than to a pregnancy exposure effect or be a matter of undetected confounding. Furthermore, the temperament of the child might have been a confounder for which we were not able to adjust. An association between maternal smoking and child temperament has been found in a few studies,^27,,28 but conflicting results exit on IC and temperament.^29–31 Another limitation is the missing information on some of the participants on educational level, psychosocial status during pregnancy, and caffeine intake and smoking in the postpartum period. Excluding these mothers from the analyses reduced the bivariate estimate between maternal smoking during pregnancy and IC, indicating that this strategy led to selection bias or incomplete confounder control. Thus, these mothers were included in the final model as separate categories.

Previous studies focused on postnatal smoking, without taking prenatal smoking into consideration.^5,,13,15,16 Prenatal exposure to smoking is associated with impaired fetal growth³² and to diseases and development in later childhood.^33–35 It therefore is important to examine the effect of smoking during pregnancy and to take it into consideration when studying exposure to smoking postpartum and IC.

A Norwegian follow-up study of 885 women and their newborns found a higher frequency of IC among children of breastfeeding smokers (40%) compared with children of breastfeeding nonsmokers (26%). Information on postnatal smoking was collected 18 weeks postpartum, and it was noticed that smoking habits mostly remained the same during pregnancy and did not change until the postnatal period. They reported a higher frequency of IC compared with our study, which might be explained by differences in the definition of IC, if maternal smoking is associated with the reporting of IC.⁵ A Dutch cross-sectional study of 3345 infants found a twofold increased risk of IC when the mothers were smoking postpartum and a higher risk among infants who were formula fed and exposed to maternal smoking compared with infants who were breastfed and exposed to maternal smoking.¹⁶ A Swedish study of 376 mothers and their newborns found no statistically significant association between smoking in the breastfeeding period or smoking indoors at home and IC.¹⁷ However, only 12% were smokers, and the statistical power of this study may be limited. These studies did not adjust for prenatal smoking, which therefore cannot be used to elucidate which exposure period is most important to the development of IC.^5,,13,15,17 Only one study, which studied parental smoking and the risk of postprandial IC, controlled for socioeconomic status, breastfeeding (yes/no), and birth weight but did not control for smoking during pregnancy.¹⁵

Prenatal smoking is known to inhibit fetal growth, and a number of diseases have been associated with fetal growth retardation probably through an “organ programming effect.”¹⁸ IC may arise as a result of a delay in the development of the central nervous system or the gastrointestinal tract, caused by smoking or environmental tobacco smoke exposure postpartum. We cannot, with the data available, distinguish between these 2 mechanisms.

If smoking during pregnancy is a risk factor in IC, then we may expect children of mothers who smoked during their pregnancy to have an earlier onset of IC than children of mothers who did not smoke during their pregnancy. This prediction was corroborated by our observation that IC started in the first week in 21% of the children whose mothers were nonsmokers during their pregnancy but in 30% of the children of mothers who smoked during their pregnancy, although insignificant (results not shown). If postnatal smoking has an effect on the child's crying, then mothers who smoke during the postpartum period should expect their children to have IC for a longer period than mothers who do not smoke during the postpartum period. This was not found in our study, in which IC lasted for more than 12 weeks in 25% of the infants of mothers who smoked postpartum and in 35% of the infants of mothers who did not smoke postpartum, but is an insignificant difference (results not shown).