In 1991, the United Kingdom (UK) Department of Health together with the Foundation for the Study of Infant Deaths and Cot Death Research initiated campaigns to persuade mothers to stop placing their infants to sleep in the prone position and to start putting them on their sides or on their backs.^1,2 As time went on the emphasis became more and more to put the infant on its back and indeed the Department of Health campaign was called Back to Sleep. The reasons for the campaign were set out in detail by the Chief Medical Officers Expert Group,³ who found published evidence that the relative risk of cot death for infants sleeping prone was in the range 1.9 to 12.7. Decreases in the proportion of infants placed prone had been followed, in Avon and in New Zealand, with decreases in the incidence of cot death.^1,2,4 On the basis of these results, the Expert Group advised that mothers should not put infants to sleep prone. Whether the side was less safe than the back position was not clear, although there was some evidence to suggest that the back position was less likely to be associated with sudden infant death syndrome (SIDS) than the side position.⁵ Since then, further studies have provided evidence to show that the back position is indeed less likely to be associated with SIDS than the side position.⁶ From the temporal effect on SIDS rates within the UK, it is now clear that the change in sleeping position has been followed by a dramatic reduction in the incidence of SIDS.^2,6

From monitoring trends in mortality, it has been shown that the change in sleeping position has not resulted in increases in other causes of death in infancy.^1,2,5,8,9 Nevertheless, there remains the suspicion that other features of the child's health may be affected. The American Academy of Pediatrics Task Force listed the following reasons why prone sleeping had been thought to be better for infants¹⁰: a decreased likelihood of aspiration, reduced gastroesophageal reflux, less colic, less head molding and, in children with specific abnormalities such as the Pierre Robin syndrome, the risk of airway obstruction when supine. Advantages of prone position were described as improved pulmonary function, sleeping and psychomotor development, and the possible prevention of infant scoliosis.^10,11 Consequently, the American Academy of Pediatrics Task Force was worried that encouraging parents to put the infants to sleep supine might affect the health of the child adversely though the evidence upon which some of the concerns were based was poorly documented.

The Back to Sleep Campaign in Britain fortuitously occurred during the time in which pregnant mothers were being enrolled into the Avon Longitudinal Study of Pregnancy and Childhood (ALSPAC). The opportunity was therefore taken to assess the consequences of changing sleeping position of the infant population over a defined time period. In this article we report on the relationship between the sleeping position of the child and the health outcomes in the first 6 to 8 months of the child's life. As far back as September 1989, health care professionals in Avon had been made aware of the potential risks associated with prone sleeping, and had been encouraged to recommend supine or side position. By 1991 the prevalence of prone sleeping among infants in Avon had decreased considerably from that observed 2 years earlier, and that in other areas of the UK, though no public campaign to change infant sleeping position was mounted until the end of October 1991.^1,2,7,8

MATERIALS AND METHODS

In a questionnaire administered 4 weeks after the infant's birth, the mother was asked: "in what position is your baby (i) when (s)he goes down for the night?; (ii) when (s)he wakes up?" She was given the options "(a) back, (b) side, (c) front, and (d) varies" for each of these categories. It has been assumed that the position at 4 weeks will be similar to that for the first 6 months of the child's life. The validity of this was shown in a separate study of 152 randomly selected children in Avon^8,13 which showed a slight decrease in prevalence of side sleeping between 1 and 3 months of age, with a corresponding rise in both supine and prone sleeping, but little change between 3 and 5 months of age.

The health outcomes were assessed from questions asked in two questionnaires, administered at 4 weeks and at 6 months. The health outcomes considered were mainly based on signs and symptoms of the child rather than diagnoses. This was thought to provide a far more objective assessment, and be relatively independent of maternal experience or educational ability. It also rested less upon whether a doctor had made a diagnosis or not. The health items considered are listed in Table 1.

Table 1. Health Outcomes Considered in the Study [View Table]

In all there were 14 138 livebirths that survived the neonatal period. Of these, questionnaires administered at 4 weeks were completed for 12 348 children (an 87% response rate). Excluded from this analysis were all multiple births (n = 321), all whose mothers did not complete this questionnaire between June 1991 and February 1992 (n = 51), or between 4 and 6 weeks of the child's age (n = 2133). Thus for health outcomes at 4 to 6 weeks the study sample comprised 9777 infants. Questionnaires administered at 6 months were completed for 8852 of the 9777 children (91%). Of these, 328 who did not complete the second questionnaire between 6 and 8 months of the child's age were excluded.

There were slight biases in the proportion of the 9777 mothers who completed the 6-month questionnaire-they were more likely to be older, better educated, have breastfed, and lived in nonrented accommodation and were less likely to be smokers or to have had the infant admitted to the special care infant unit. All these variables have therefore been taken into account in the analyses.

Confounding factors likely to be related to the health outcomes were considered as follows:

1. The mother's highest educational level: (in five different groups: Certificate of Secondary Education or less; vocational qualification; O-level; A-level; degree);
2. Mother's age: (<25; 25 to 29; 30+);
3. Housing circumstances: (owned/mortgaged; housing association or private rented accommodation; council accommodation);
4. Parity: (0 or 1+);
5. Maternal postnatal smoking: (yes or no);
6. Breastfeeding: (no; yes but stopped by 4 to 6 weeks; yes still breastfeeding at 4 to 6 weeks);
7. Admission to special care infant unit: (yes vs no);
8. Sex of child;
9. Mother's ethnic origin: (white; African/Caribbean/other black; Indian/Pakistani/Bangladeshi; Chinese; other);
10. Month of completion of the questionnaire (to allow for seasonal effects in morbidity);
11. Age at completion of questionnaire (for the 4-week questionnaire this was classified as 4, 5, or 6 weeks of age; for the 6-month questionnaire it was classified as 6, 7, or 8 months);
12. Whether the child had been vaccinated against pertussis by the 6-month questionnaire completion date (yes vs no).

A variety of different logistic regression analyses were used to ensure that the data had been fully investigated. First, in the tables the unadjusted odds ratio (OR) for each sleeping position relative to the back is shown, and second, the adjusted OR allowing for the 12 items above and, where relevant, the month the questionnaire was completed and/or the age at completion. The adjustment used 85% to 89% of cases, (the differences being due to missing data). Not shown in the tables were the results of further logistic regression analyses using a stepwise process and allowing only those items that were statistically significantly associated with the particular outcome. These made little difference to the overall conclusions.

RESULTS

Sleeping Position on Being Put Down and Waking Up

Table 2. Stated Sleeping Position at 4-6 Weeks: "In What Position is Your Baby (i) When He Goes Down for the Night? (ii) When He Wakes Up?" [View Table]

To evaluate the effectiveness of the campaign we first subdivided the infants according to when the question about sleeping position was answered. There were three time periods considered: before the public campaign (ie, between June 1991 and October 1991); during the campaign (ie, between November 1991 and April 1992); after the campaign (ie, May 1992 to February 1993). The way in which sleeping position changed over these time periods is shown in Table 3. Note the strong and dramatic increase in back sleeping over time and the diminution in front sleeping. Infants were more likely to have slept on their side before the campaign and least likely to after the campaign. Table 3 also shows that little else changed that had been emphasized in the campaign, apart from a reduction in the use of a quilt particularly for the 4-week-old child. There were no changes in other items mentioned in the campaign(viz, mother's smoking habit, and that of the partner) or in breastfeeding rates.

Table 3. Factors Targeted or Mentioned in the Back to Sleep Campaign; Responses Grouped According to Relationship to Time of Campaign [View Table]

The change in sleeping position over time was found in subgroups of mother's education and mother's age, housing type, parity, smoking habit, breastfed and bottle-fed infants, and whether or not they were admitted to special care infant unit. Nevertheless there were significant variations between sleeping position and demographic factorsbefore the public campaign sleeping position was associated with housing type (council housing associated with more prone and less back sleepers), parity (more prone among multiparae), and breastfeeding (more back sleepers); after the campaign sleeping position was associated with mother's education level (the better educated had more back sleepers and more "varies"), housing (council more prone, owned/mortgaged more back), parity (more prone and side among multiparae), smoking (fewer back and more side sleeping among smokers), breastfeeding (those breastfeeding at 4 to 6 weeks were least likely to be prone and more likely to "vary") and the sex of the child (boys are more likely to "vary" and least likely to be on side). There were, however, no differences in the demographic variation in the mothers or children delivered before, during, or after the campaign; similar proportions of mothers were in each of the educational groups, maternal age groups, housing types, parity, admission to special care, and ethnic minority groups; similar proportions had had their child immunized against pertussis.

Health of the Child

Table 4. Questionnaire Responses Grouped by Sleeping Position on Going Down at Night: Odds Ratio (OR) With Adjustments for Confounding Variables [View Table]

A less subjective measure of ill health is whether or not the doctor was called to the home because of a problem with the infant. Although this was more likely to occur with the 4- to 6-week-old infant who was placed prone, this was not statistically significant; however, adjustment increased the OR to 1.47 [95% CI .97, 2.21], which just failed to reach statistical significance.

Assessment as to whether the doctor had been called to the home for the ill infant by 6 to 8 months of age indicated that the infant put to sleep on the back was least likely to have such a historyand unadjusted rates showed high ORs both for side sleeping and front sleeping. On adjustment, these ORs were reduced, but the side sleeping was significantly elevated in comparison with back sleeping with OR 1.17 [95% CI 1.01, 1.35] after adjustment for all factors. The OR for front sleeping after adjustment for all factors was 1.15 [95% CI .82, 1.60].

At the same age the mothers were asked whether the children had ever had a cough. Again there was a significant variation, this time however it was the children placed to sleep on their fronts who had the highest risk. After adjustment, the OR increased to 1.53 [95% CI 1.08, 2.17] statistically significantly greater than the back sleeping position. A similar question asked at 6 months of age also showed a greatly elevated risk for front sleepingeven after adjustment for all potential confounders, the OR for front sleeping was 1.52 [95% CI 1.07, 2.16]. However, when the data were confined to the response to the question at 6 months that the child had coughed for a period of at least 2 days, the statistical significance of the association with prone sleeping position disappeared.

Data regarding earache at 6 months showed a significant association with sleeping position when the data were unadjusted, with prone sleepers having the highest risk of earache, but after adjustment, the OR decreased from 1.71 to 1.33, and was no longer statistically significantly associated [95% CI .84, 2.09]. Other aspects of ear problems were addressed through a number of different questions such as "has anyone thought there might be a problem with hearing." This showed a significant unadjusted association, with children placed on their fronts being more likely to be reported to have a problem with hearinghowever on adjustment this was no longer statistically significant, and the children whose sleeping position was said to vary were those who had the greatest OR 1.53 [95% CI .90, 2.61]. In relation to whether the child had worse hearing during or after a cold, there was no significant association with sleeping position, nor was there any significant relationship with wheezing or breathlessness, episodes of stopping breathing or holding breath when asleep. Interestingly, in relation to the claims for prone sleeping in regard to respiratory function and apnea in preterm infants,¹⁴ after adjustment the prone sleepers were less likely to be reported as having had episodes of stopping breathing when asleep with an OR of .78 [95% CI .51, 1.20], but this was not statistically significant.

Children sleeping prone were reported to breath through their mouths more often than through their nose, but after adjustment this ceased to be statistically significant with an OR of 1.17 [95% CI .79, 1.72]. In regard to snoring, however, after adjustment there was a significant association with sleeping position, children who were put to sleep on their sides were significantly less likely to snore than either those put to sleep on their back or on their front, with an OR of .76 [95% CI .63, .93].

In regard to choking with feeds, there was slightly increased risk associated with side sleeping at 6 to 8 months, OR 1.17 [95% CI 1.01, 1.34], but this became nonsignificant after adjustment. Infants who were put prone were more likely after adjustment to have been reported as having a lot of wind (ie, burping) on feeding, although this was not statistically significant; OR 1.58 [95% CI .96, 2.60].

At 4 to 6 weeks the mother was asked whether she felt that the infant's crying was a problemalthough infants put prone were more likely to have such a description, this was not statistically significant on adjustment; OR 1.69 [95% CI .94, 3.04]. There were significant relationships with unhappiness, however (Table 4E). Infants who were put to sleep prone at 4 to 6 weeks and particularly those put in varying positions were significantly less likely to be reported as happy by their mothers. The adjusted OR for varied sleeping position was 1.53 [95% CI 1.13, 2.07].

Data for convulsions are also shown in Table 4F. Here it can be seen that although there were strikingly elevated ORs that changed little with adjustment, overall statistical significance was not reached, probably because the incidence is so small. However, because at this age at least half of all convulsions are related to febrile fits, it is feasible that prone sleeping could be associated with such convulsions.

DISCUSSION

There remain methodological questions concerning this study. First, it is feasible that taking the information on sleeping position when the children were 4 weeks of age is not necessarily an accurate reflection of sleeping position throughout the first 6 months of life. Unfortunately, details of sleeping position were not again requested until the child was 30 months of age. At this stage the usual sleeping position of the child was related to (but not synonymous with) that found when the child was 4 weeks of age in relation to going down to sleep. As noted above, in a separate study of child care practices in a group of 152 Avon families randomly selected from the birth register in 1990 to 1991, no significant changes in routine sleeping position for infants were identified between 1 and 5 months of age, suggesting that the sleeping position at 4 weeks was a good proxy for usual practice throughout the first 6 months.^8,13

There are also the questions that must always arise in any study concerning the representative nature of the sample under scrutiny. The ALSPAC study achieved a high response rate, and covered over 85% of the eligible population. Nevertheless, there were certain groups that were underrepresented in this studythese particularly involved mothers who were teenagers, those from the ethnic minority groups, and unsupported mothers. Nevertheless, the analysis showed similar time trends in sleeping position within strata of education, ethnic group, and education background. The message concerning sleeping position of the child appears to have reached all members of the community. It is unlikely that any systematic bias in response could therefore change our results.

A secondary hypothesis of this study was to see whether information on other conditions in these young infants would actually help inform the debate as to why front sleeping position should be associated with increased risk of SIDS. Whereas we only showed significant increases of two conditions among back sleepers, there were a large number of conditions where the front sleeper showed an elevated OR in comparison with the back sleeper. There were relatively small numbers of front sleepers, and consequently these elevated ORs rarely reached statistical significance. The front sleepers were significantly more likely to be reported as having had a cough (although not more likely to be reported as coughing for at least 2 days), both at 4 weeks and at 6 months of age, to have had episodes of pyrexia, and signs of eczema in the first weeks, compared with back sleepers. They were also significantly more likely to sleep for a prolonged length of time (in excess of 14 hours) at 6 months of age. Although not statistically significant at the 95% level, prone sleeping infants were also more likely to be reported as not very well, to have had the general practitioner called to the home, to have had earache and suspected hearing problems, colic, and crying problems.

These relationships may provide clues to the etiology of SIDS. Prolonged sleeping may be an indicator of an infant that was inappropriately difficult to arouse; increased risk of cough, pyrexia, earache and having to call the general practitioner to the home suggest that in the presence of an infection symptoms are worse for infants sleeping prone.

There remains the question as to whether the initial position of the infant on being put to sleep was actually a consequence of any of the symptoms that had occurred in the first 4 weeks rather than the symptom being initiated by the sleeping position. It may well be that children with wind were found to be more docile if put down on their fronts than if put in any other sleeping position. The relationship with colic may well be of similar origins-and the dramatic relationship shown with a varying sleep position was probably a direct consequence of trying to achieve a less unhappy infant by trying different sleeping positions.

In conclusion, this study has shown a variety of different relationships between sleeping position and morbidity-few are statistically strong, but the general tenor of the results indicate that infants put to sleep on their back were if anything at reduced risk of respiratory and gastrointestinal conditions, those put on their side were more likely to be visited at home by a doctor for health problem, to have colic, but also to sleep for a prolonged length of time and be at reduced risk of snoring. Front sleeping was associated with a wide range of different problems, mainly respiratory in origin.

Data from this study are reassuring in the light of the Back to Sleep Campaign, because they show no indications that this policy may be producing unwarranted adverse effects in the population of children born since the campaign. There are however a number of queries that should still be investigatedfor example is there any effect on the subsequent temperament, behavior, and development of these children according to the sleeping position in which they were put, and is health after 6 months related in any way to the sleeping position of the child?