Results

ODDS RATIO AND DOSE–RESPONSE RELATION

Crude odds ratios (95% confidence intervals) for maternal smoking before, during, and after pregnancy were 3.0 (2.2 to 4.1), 4.0 (2.9 to 5.6), and 3.8 (2.8 to 5.3) respectively (table 1). Odds ratios for paternal smoking were 1.6 (1.2 to 2.1), 1.7 (1.3 to 2.3), and 2.0 (1.5 to 2.8), and for other household members smoking postnatally 2.0 (1.3 to 3.3). Odds ratios were adjusted for age of infant as well as age and educational level of mother. Additional adjusting for birth weight reduced the effects marginally. Paternal smoking and other smoking were adjusted for maternal smoking in pregnancy. After adjustment, there was no effect of paternal smoking or other smoking (table 1). This was confirmed in a stratified analysis in which there was no effect of paternal smoking if the mother did not smoke (table2).

To evaluate dose–response relations, crude odds ratios for both maternal and paternal smoking before, in the first trimester, in the third trimester, and after pregnancy were calculated in different dose strata (nil, 1–9, 10–19, and > 20 cigarettes/day). We found an obvious dose–response relation for maternal smoking (fig1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Crude odds ratios (log scale) for maternal and paternal smoking doses in different dose strata by exposed period in Scandinavia 1992 to 1995 (the Nordic epidemiological SIDS study); cig = cigarettes.

SMOKING CESSATION

To assess possible effects of maternal smoking cessation, mothers who stopped smoking were divided into three groups. First, those who smoked before pregnancy but stopped permanently when they realised that they were pregnant (early cessation); second, those who stopped when the baby was born (late cessation); and third, those who smoked before pregnancy, stopped during it, but resumed smoking postpartum (partial cessation, table 3). Relatively few mothers changed their smoking behaviour (23 of 151 cases (15%) and 150 of 334 controls (45%)), but in neither group did we find any significant risk increase compared with mothers who never smoked. The odds ratio for early cessation was 0.7 (0.3 to 1.4), late cessation 1.1 (0.5 to 2.4), and partial cessation 1.1 (0.4 to 3.2) (table 3). The mothers who continued to smoke throughout pregnancy and thereafter had an odds ratio of 4.5 (3.1 to 6.5).

To determine why late cessation was not associated with an increased risk, even though the mothers of these infants had smoked during pregnancy, we compared the number of cigarettes consumed by the late ceasers and the non-ceasers. We found that the late ceasers smoked less than the non-ceasers during pregnancy, both in the first and the third trimester. In the first trimester, late ceasers consumed 5.1 cigarettes/day (95% confidence interval 2.1 to 8.2) against 9.3 cigarettes/day (8.7 to 9.9) for non-ceasers. The numbers in the third trimester were 5.3 (2.9 to 7.6) and 8.3 (7.7 to 8.9) respectively. We also tried to adjust for the amount smoked, but the results did not clarify the issue.

PERINATAL AND POSTNATAL PROBLEMS

The differences between SIDS cases and controls exposed and not exposed to maternal smoking in utero and postnatally are shown in tables 4 and 5. The variables initially investigated were: age, birth weight, age at weaning, prematurity, induction of labour, vacuum extraction, caesarean section, resuscitation, admission to neonatal ward, problems during the first week (floppiness, irritability, weak sucking, and vomiting), problems after the first week (floppiness, activity, and irritability), feeding habits, infantile colic, symptoms of infection, antibiotic treatment, admission to hospital after discharge from maternity ward, use of a dummy (pacifier), regurgitation/vomiting, and depth of sleep as estimated by the parents. We chose variables which had effects that might be modified by maternal smoking. Of these variables, only the ones given in tables 4 and 5showed significant differences. There was no age difference between the groups in this study: mean age at death was 16.1 weeks in the non-exposed group and 15.9 in the exposed group (p = 0.90). Cases had lower birth weight than controls, regardless of exposure, and smoke exposed infants had lower birth weight than non-exposed infants. Exposed cases stopped breast feeding earlier (8.0 weeks v11.9 weeks, p = 0.036) (table 4).

Among the controls, infants of smoking mothers more often had problems related to breast feeding (49.2% v 29.9%), as well as infections (51.8% v 48.8%), and were more often admitted to hospital (24.2% v 19.0%) and treated with antibiotics (24.2% v 17.1%) (table 5). The only exception was poor sucking, which was more prevalent among infants of non-smoking mothers. The effect of poor sucking was much stronger in infants of smoking mothers (p = 0.01 in a multiplicative model). The same tendency (not significant) was observed for breast feeding. Otherwise, there was a tendency (not significant) towards smaller effects among infants of smoking mothers. We assessed possible effect modifications of smoking on the following variables: low birth weight, low maternal education, prone sleeping position, high parity, and low maternal age. No significant interactions were observed, except for an additive interaction with position placed to sleep the last time (p < 0.001), which is described elsewhere.14

DIFFERENCES BETWEEN THE NORDIC COUNTRIES

Smoking habits differed between the Scandinavian countries. In the control group Sweden had the lowest prevalence of smoking mothers during pregnancy, 23.5% against 34.7% in Norway and 35.8% in Denmark (table 6). The odds ratios for smoking were somewhat lower in Denmark and Norway, but the differences were not statistically significant (table 6). Adjusted for age of infant, maternal age, and education, the odds ratios were somewhat lower, but the pattern was the same, with the exception that the statistical significance for smoking in pregnancy disappeared in the Danish data.

The attributable fraction (AF) was 45% in the whole material. Due to the high prevalence of smoking, Norway had the highest AF (54%), followed by Sweden (44%) and Denmark (41%).

Discussion

Nicotine is a powerful stimulant of cholinergic neurones, and has been postulated to act upon nicotinic acetylcholine receptors either centrally19 or peripherally.20 Given prenatally in animal experiments, it alters ventilatory and awakening responses to hypoxia.21-23 Nicotine also acts by vasoconstriction and diminution of placental blood flow, thereby reducing the amount of oxygen and nutrients available to the fetus. Prenatal smoking has been shown to have significant negative effects on neonatal airway development and lung elasticity.24 Carbon monoxide might theoretically reduce oxygen transport, and tar and smoke particles are considered responsible for an increased susceptibility to airway infections, a factor associated with SIDS.25

Our major findings were an independent effect of smoking on the risk of SIDS with a dose–response association, no effect of passive smoking, and a strong effect of smoking cessation.

It is well known that smoking reduces birth weight, and it has been argued that adjusting for birth weight carries a risk of over adjustment.26 We therefore calculated odds ratios with and without this adjustment and found slightly lower values with the adjustment, but the effects were still significant. This suggests an effect in addition to that mediated by the reduction in birth weight.

As in any case-control study, the result is probably affected by recall bias. Smoking might be underreported by control mothers as well as by case mothers.

To evaluate a possible non-response bias, marital status, maternal age, and birth order in the Norwegian cases and controls were compared with data from the Medical Birth Registry of Norway (MBRN) for the same period, including SIDS deaths. The proportions of single mothers, young mothers, and first births were lower in the case-control data than in the MBRN, but the reduction was similar in cases and controls.

The results from the Nordic study confirm that tobacco smoking is an independent risk factor for SIDS, and after the implementation of the supine sleeping position it has now emerged as the most important risk factor. Most infants exposed to smoking were exposed both in utero and postpartum, so the effects of prenatal and postnatal smoking are difficult to separate.12 27 We showed that paternal smoking had no significant effect as such; it tended to augment the effect of maternal prenatal smoking, but the confidence intervals overlapped. This was also found by Blair et al.27 In our study, the mean cigarette consumption of the mother in the first trimester, when the father did not smoke, was 8.1 (7.2 to 8.9) cigarettes/day, and when the father did smoke, 9.2 (8.4 to 9.9) cigarettes/day. The figures for the third trimester are similar: 6.3 (5.5 to 7.1) and 7.9 (7.2 to 8.6) cigarettes/day, respectively. An association between maternal and paternal smoking could account for the findings in both studies.

Table 3 contains the components recently suggested by Golding12—that is, mothers who smoked only postnatally (partial cessation) as well as mothers who smoked both prenatally and postnatally—and from this it is probable that it is maternal smoking during pregnancy that is important. We consider that paternal and other passive smoking is of relatively minor importance. This is supported by the dose–response figures (fig 1) but we cannot rule out a small effect of maternal smoking postpartum.

The question of the influence of maternal smoking on perinatal and postnatal problems has been raised on the grounds that some children die without having been exposed to tobacco smoke, either prenatally or postnatally. In our study, 30.7% of the case mothers had never smoked. The hypothesis in this section was that it might be shown that there are differences between cases exposed and not exposed to tobacco smoke. Of the 21 variables initially investigated, only those related to feeding (poor sucking, breast feeding) and to infection (infectious symptoms, antibiotics, and hospital admission) remained.

Our interpretation of these findings is that the early feeding difficulties seen in smoke exposed cases are in some way related to the biological effects of tobacco smoke on the fetus, and that the increased risk of SIDS due to smoking is not caused by an increased risk of infections. Infection might be an additional factor that contributes to the fatal event. It is conceivable that infants who die of SIDS come from a population of infants with compromised health, and the ill effects of this are enhanced by maternal smoking.

In contrast to the findings of Haglund and Cnattingius in 1990,28 we could not show a difference in age at death between the smoke exposed and non-exposed SIDS cases. We have no explanation for this.

Comparing the percentage of smoking mothers in the Scandinavian countries with the incidence of SIDS at the end of the 1980s, we note that both Norway and Denmark had very high SIDS rates, and Sweden a much lower rate.29 During the study period, Norway and Denmark both had a significantly higher proportion of smoking mothers than Sweden. These differences in smoking habits may partly reflect the varying incidences, but none of these analyses provides a convincing explanation of why Denmark and Norway differed as much as they did in the peak years 1985 to 1989. In spite of different smoking habits, all Scandinavian countries now have virtually the same SIDS incidence. This indirectly highlights the major influence on SIDS incidence exerted by the switch to the non-prone sleeping position.

We believe we have shown a remarkable effect of maternal smoking cessation. This conveys an important message—that is, that smoking cessation is potentially beneficial. We find it curious that the infants of mothers who stopped smoking postpartum were not at increased risk, even though they had been exposed to nicotine in utero. One explanation could be that these mothers had already begun to reduce smoking during pregnancy, as was shown in the first and third trimesters. This is consistent with our demonstration of a dose–response relation between smoking dose in pregnancy and SIDS. The fraction of mothers who changed their smoking habits was low, only 23 of 151; therefore the interpretation remains uncertain.

We conclude that smoking is an independent risk factor for SIDS, and that it is mainly mediated through maternal smoking during pregnancy. There are differences in smoking habits among women in Scandinavia, which are probably related to the different odds ratios observed and may have contributed to the varying incidences in the Scandinavian countries at the end of the 1980s. Analysis of smoking careers indicates that smoking cessation, or even a reduction in smoking, may be beneficial in reducing the risk of SIDS.