Case-control study of sudden infant death syndrome in Scotland, 1992-5BMJ 1997; 314 doi: http://dx.doi.org/10.1136/bmj.314.7093.1516 (Published 24 May 1997) Cite this as: BMJ 1997;314:1516
- Hazel Brookea, executive director,
- Angus Gibson, chairmana,
- David Tappin, senior lecturer in community paediatrics, University of Glasgowa,
- Helen Brown, research associateb
- a Scottish Cot Death Trust, Royal Hospital for Sick Children, Glasgow G3 8SJ
- b Medical Statistics Unit, University of Edinburgh Medical School, Edinburgh EH8 9AG
- Correspondence and requests for reprints to: Mrs Brooke
- Accepted 17 February 1997
Objective: To investigate the relation between routine infant care practices and the sudden infant death syndrome in Scotland.
Methods: National study of 201 infants dying of the sudden infant death syndrome (cases) and 276 controls by means of home interviews comparing methods of infant care and socioeconomic factors.
Results: Sleeping prone (odds ratio 6.96 (95% confidence interval 1.51 to 31.97)) and drug treatment in the previous week (odds ratio 2.33 (1.10 to 4.94)) were more common in the cases than controls on multivariate analysis. Smoking was confirmed as a significant risk factor (odds ratio for mother and father both smoking 5.19 (2.26 to 11.91)). The risk increased with the number of parents smoking (P<0.0001), with the number of cigarettes smoked by mother or father (P=0.0001), and with bed sharing (P<0.005). A new finding was an increased risk of dying of the syndrome for infants who slept at night on a mattress previously used by another infant or adult (odds ratio 2.51 (1.39 to 4.52)). However, this increased risk was not established for mattresses totally covered by polyvinyl chloride.
Conclusions: Sleeping prone and parental smoking are confirmed as modifiable risk factors for the sudden infant death syndrome. Sleeping on an old mattress may be important but needs confirmation before recommendations can be made.
Parental smoking is currently the most important modifiable risk factor in the sudden infant death syndrome
In this study sleeping prone and, to a lesser extent, sleeping on the side increased the risk of the syndrome, so babies should be put down to sleep only on their back
Bed sharing with an infant should be discouraged if the mother smokes
Sleeping on an old mattress may carry an increased risk
There has been considerable interest in recent years about the role of infant care practices and environment in the sudden infant death syndrome. Previous studies showed the risk associated with sleeping prone,1 2 and modification of this practice has been associated with a major reduction in the syndrome worldwide.3 This improvement prompted our search for other modifiable risk factors. We report a four year case-control investigation of risk factors for the sudden infant death syndrome in Scotland from 1992 to 1995, when the rate of deaths from the syndrome fell from 1.1 to 0.7 per 1000 live births.
The registrar general for Scotland reported to us all infant deaths occurring after the seventh day of life to the end of the first year. The computerised maternity record for each infant was provided by the information and statistics division of the Common Services Agency. In the case of deaths from the sudden infant death syndrome our office was also notified directly by the pathologist responsible for the necropsy. We defined the sudden infant death syndrome according to Beckwith as the sudden death of any infant or young child which is unexpected from the history and in which a thorough postmortem examination fails to show an adequate cause for death.4
Consistency in classification was sought by the use of a standard necropsy protocol with agreed diagnostic criteria.5 In addition, all death certificates of infants aged 1 week to 1 year were scrutinised for possible misclassification of explained deaths. Overall, 201 out of 798 postperinatal infant deaths were diagnosed as the sudden infant death syndrome. Six other sudden deaths may have been misdiagnosed as bronchopneumonia and were not included in the study.
We identified two controls for each case of the syndrome–the births immediately before and after the index case in the same maternity unit. In this way controls were matched for age, season, and maternity unit. After permission had been obtained from the general practitioner a fieldworker employed by the study contacted the mother by letter, asking for her cooperation in completing a questionnaire during a home visit. All home visits were made within 21 days of the index case's death to minimise differences in age related circumstances between cases and controls. Questionnaires were completed on 147 cases out of a total of 201 reported and on 276 controls. The failure to acquire data on the remaining 54 cases was largely due to delay in notification by the pathologist or parents not being at home on at least two occasions, making a visit within 21 days of the death impossible. For 108 cases there were two controls, for 27 cases there was one control, and for 12 cases there were no controls. For 29 controls there was no case interview. The characteristics of the cases without an interview were compared with the characteristics of those with an interview and were similar in terms of maternal age, social class, and deprivation category.6 There was a small difference in age at death of cases (the mean age of cases whose parent was interviewed was 15 weeks and that of those who were not was 18 weeks; P=0.04).
The questionnaire provided core medical and social data about the infant, as well as details of infant care practices in the home. Data were collected on routine childcare practices for cases and controls and on practice on the night of death for cases only.
The questionnaire was divided into six main categories: social and prenatal factors, feeding regimen, sleeping habits, sleeping environment, exposure to smoking, and illnesses.
Socioeconomic status was assessed by two methods. The first was assessment of deprivation on the basis of postal code in seven categories in ascending order of deprivation.6 These categories take cognisance of overcrowding, male unemployment, low socioeconomic status, and a lack of a car. The second was the registrar general's social class from standard occupational classification.
The tog value (warmth rating) of both clothing and bedding was calculated with the scoring system supplied by the Shirley Institute, Manchester.7 We used the average of day and night tog values in analyses. We took special note of the extra thermal implications of swaddling and how much of the body was swaddled.
Exposure to smoking was assessed in two ways. The first used the following ordinal scale: neither parent smoked, father only smoked, mother only smoked, mother and father both smoked. The second calculated a dose response by determining the total number of cigarettes (0, 1-9, 10-19, ≥20) smoked daily by mother, father, or other household member.
Exposure to old mattresses was assessed by asking parents if their infant routinely slept at night on a new mattress or on one previously used by another infant or an adult.
The primary analyses focused on routine childcare practices (see table 1), but routine practice and practice on the night of death were compared for cases (see table 2). The baseline comparison group always had the opposite definition–for example, maternal age <27 was compared with maternal age ≥27–unless otherwise stated in the footnotes to table 1).
Binary and categorical factors were analysed by conditional logistic regression analyses, which were fitted using the proc phreg program in SAS. This method allows for the matched nature of the data but is unable to use information on the unmatched cases and controls. Quantitative factors were analysed in random effects models fitted by the proc mixed program in SAS. Matched set was fitted as a random effect to allow for any correlations in the data due to matching. This method has the advantage that it uses information from unmatched cases and controls.
Over 100 factors were analysed, and about half of these were significant. However, many of the significant factors became non-significant when adjusted for obvious confounding factors or for socioeconomic factors. Results for the factors that remained significant after these adjustments are presented in table 1). Quantitative factors were categorised in this table so that all factors could be directly compared by odds ratios. Cut off points correspond to values expected to be most relevant on the basis of previous knowledge or as defined within other studies. All results in table 1) are based on comparisons between routine sleeping practices of cases and controls.
A multivariate analysis was carried out to determine which factors were independently significant when adjusted for all other factors found to be important in the study. This was carried out using a conditional logistic regression model that included all of the factors listed in table 1).
Interactions with factors of particular interest were tested in conditional logistic regression models.
Frequencies, odds ratios, and P values for factors that remained significant after adjusting for obvious confounding factors and for socioeconomic factors are presented in table 1).
Nearly all of the people who smoked did so during and after pregnancy. Overall, 79% of mothers of cases (115/146) smoked compared with 34% of control mothers (93/275; univariate odds ratio 5.91 (95% confidence interval 3.61 to 9.68)). There was a dose relation, with the risk of the sudden infant death syndrome increasing with the number of cigarettes smoked by the mother (P=0.0001), father (P=0.0001),and other people in the household (P=0.001); each factor was analysed separately. The risk also increased significantly with exposure (table 1). The risk caused by maternal smoking increased when the infant shared a bed (P<0.005). Given that smoking is causal, the population attributable risk of smoking during and after pregnancy was 62%.
Sleeping prone remained a significant risk factor, although few infants in the control population were routinely placed prone (2%(5/275)); 9% of the index mothers (13/146) opted for this position routinely, resulting in an increased risk for their infants (table 1). At death 50 of the 147 infants were found prone (34%), though only 19 (13%) had been placed prone (table 2).
Sleeping on the side was also a significant risk factor on univariate analysis (odds ratio 1.58 (1.01 to 2.46)). 51% of the index mothers (75/146) placed their babies on their sides, compared with 38% of controls (104/275). Sleeping on the side was the most labile sleeping position, but cases and controls routinely tended to move from sleeping on their side to sleeping supine if they moved at all rather than to sleeping prone. We noted that routinely only 44% (33/75) of index babies placed on their sides were found in a different position on wakening, compared with 68% (70/104) of controls (odds ratio 0.37 (0.20 to 0.68). Indeed, regardless of the position in which they were put down or found, controls were more likely than cases to change position regularly during sleep (odds ratio 0.53 (0.40 to 0.99)).
Mattresses used previously by at least one other infant or an adult seemed to place an infant at increased risk of the sudden infant death syndrome. The risk from routinely being on an old mattress at night also increased for infants who had undercovers with a lower tog value (P=0.05), who had colds (P=0.02), and who were off their feeds (P=0.02). No significant interactions occurred with smoking, sleeping place–for example, cot or pram–sleeping position, and birth order, so we could not draw any firm conclusions on the influence of these factors. There was no detectable increase in risk with old mattresses completely covered by polyvinyl chloride (table 3). The increase in risk was associated with both the so called combination mattresses, in which the bottom two thirds of the mattress is covered by polyvinyl chloride and the top third consists of ventilated foam covered by netting, and with cloth covered mattresses. We had insufficient data to investigate interaction of old mattresses with routine bed sharing. However, 34% of the index cases (48/142; table 2) were sleeping with parents at death–30% (42/142) in an adult bed and therefore on mattresses used by others.
We asked parents whether they had ever found that their baby had moved under the bedclothes. Overall, 24% of parents of cases (35/146) said that they had compared with 13% of controls (35/274), and this difference was significant on multivariate analysis (odds ratio 2.18 (1.03 to 4.64)) (table 1). On the day or night of death 13% (19/146) of the index cases were found under bedcovers.
Receiving any drug treatment in the week before death emerged as a strong risk factor in the multivariate analysis (table 1). After adjustment for symptoms in the previous week, consultation with a general practitioner, prematurity, and low birth weight, no individual drug was significant. We noted that the index cases were more likely to have had one or more of a range of symptoms and to have been seen by their general practitioner because of illness during the previous week. The symptoms with the highest odds ratios on univariate analysis were unusual sleepiness (odds ratio 2.61 (1.26 to 5.51)), snuffles (1.61 (1.07 to 2.44)), and sickness (1.69 (0.92 to 3.10)). The only symptom more common in controls was increased irritability.
Poverty was confirmed as a significant risk factor for the syndrome, the rate increasing with deprivation score, as shown in table 4). Low socioeconomic status (classes IV and V) was also significant even when adjusted for deprivation score in the multivariate model.
Factors significant on univariate but not on multivariate analysis were being male, sleeping on the side, non-routine use of cot bumper, routine sleeping with parent(s), any symptoms in previous week, gestation ≤36 weeks, usually being swaddled in previous week, previous infant death in family (sibling, half sibling, or first cousin), usually being sweaty on wakening, tog value of bedding and clothing ≥10, mother leaving school at ≤16, bottle feeding at time of death, two or more previous live births, birth weight <2500 g. However, their significance on univariate analysis makes them noteworthy. In particular, the finding of a high thermal score of bedding plus clothing is consistent with other published data.8 A high thermal score seemed to be more risky for boys than for girls (P=0.03).
Examples of factors that were not significant on univariate analysis included the time between the last two pregnancies, twin birth, non-European mother, complementary feeding (combined breast and bottle feeding), age at introduction of solids, sleeping room, sleeping place (other than parents' bed), use of pillow, use of duvet, type of mattress covering, use of sheepskin, swaddling, heated sleeping room, and admission to hospital in week before death. Significantly more babies died on a Saturday or Sunday than would be expected by chance (42% (84/201), P<0.01); 11% (17/158) of deaths occurred when the infants were away from their usual place of residence.
The decrease in the rate of the sudden infant death syndrome over the four years was 0.4 per 1000 live births. We assessed whether attributable risks were likely to have changed during the study by testing whether factors had changed during the study in the control group. The following factors changed significantly: the number of parents who smoked was reduced (P=0.02), more infants were placed supine (P=0.01), and mothers were older (P=0.008). Thus, assuming constant relative risks, population attributable risks had decreased for parents' smoking, use of prone and side sleep positions, and younger mothers.
Parental smoking during and after pregnancy is a major, potentially modifiable, risk factor found in many other studies.9 10 If only the father smoked the risk was almost significant. The finding of a dose response with the number of people smoking in the household adds weight to the possibility of smoking being causally related to the sudden infant death syndrome, as does the increased risk related to the number of cigarettes smoked. If smoking is causal two thirds of the cases of the syndrome might be avoided if mothers did not smoke during and after pregnancy. Health promotion initiatives to discourage young girls from starting to smoke and to help smokers reduce their habit are urgently required.11 12
Sleeping position and place
Although sleeping prone remains a strong risk factor, its low prevalence in the infant population in Scotland indicates that only a small percentage of deaths from the syndrome can now be attributed to this. Sleeping on the side was more risky than sleeping supine and since 38% of control infants were routinely placed this way, a significant number of deaths may be attributed to this sleep position. Sleeping on the back is the safest, so parents should be advised to use this position wherever possible.
The significance of failure to change position during sleep (table 1) supports observations by Schechtman et al that infants considered at increased risk of the syndrome show fewer spontaneous arousals from sleep and fewer movements during sleep than do control infants.13
The New Zealand cot death study found an increased risk for infants sleeping with a parent only if the mother smoked.14 A subsequent report from California failed to confirm this risk,15 but our data are consistent with the New Zealand findings. In addition, the greatly increased incidence of bed sharing in cases at death concerns us. We accept that routine bed sharing may be underreported, but it is difficult to believe that it could account for an increase from 8% routinely to 34% at death.
In Scotland most infants routinely share a room with their parent(s) at night–78% of index cases and 75% of controls. As nearly all the mothers in the study were European, the high prevalence of room sharing was not associated with an ethnic minority group, as noted in some studies.16 17 In our study it was not a significant factor (odds ratio 1.20 (0.69 to 2.09) on univariate analysis. This differs from the data of Scragg et al, who found it to be protective.18 We recognise that our data, in contrast to those of the New Zealand study,14 were collected when the incidence of the syndrome and the rate of prone sleeping (2%) were low and the rate of room sharing was high (75%) in controls, making comparison difficult. On the basis of our results, however, we believe that advice on room sharing is not at present indicated in Scotland.
Mattresses and other factors
The results from the mattress analysis were unexpected. There is an increased risk of some kind, regardless of parity and social deprivation, for infants sleeping on mattresses previously used by others, although the risk was not established for mattresses completely covered by polyvinyl chloride. Our findings therefore lend no support to the hypothesis that household fungi interact with fire retardant chemicals in the plastic covering of cot mattresses and release toxic gases, which in turn cause sudden infant death.19 The failure to establish risk with mattresses completely covered with polyvinyl chloride may be because they can be kept clean, regardless of age, while others cannot.
Although drug treatment was an independent risk factor for the sudden infant death syndrome no one drug was significant after adjustment. However, these tests lacked statistical power owing to the small numbers of infants taking each drug. Drug treatment may be a surrogate for some other risk factor(s) that we have not identified.
Young maternal age has historically been associated with an increased risk of the sudden infant death syndrome,3 and our study confirmed this. There was a shift in the mean maternal age nationally at delivery for the four years of our study compared with the mean age during a previous Scottish study.20 Mothers were, on average, older during the later study (72% ≥25 in 1992-5 v 30% in 1981-2.) We estimate that this swing accounts for about 9% of the decrease in the rate of the syndrome between the two studies.
Pillows and duvets were more commonly used on the night of death than routinely. On further examination the use of a duvet proved to be a surrogate for sharing the parental bed, whereas use of a pillow was not; therefore pillows may pose an increased risk.
The higher incidence of cases at weekends confirms previous findings.21
We have considered the possibility that some of the associations are the result of recall bias. The design of our study, requiring the home interview to take place within 21 days of death for cases, was aimed at limiting recall bias. Any drugs given may have been more easily remembered by parents of cases. For other findings–for example, smoking–recall bias would be more likely to lead to an underestimation by parents of cases compared with parents of controls as a self protective mechanism to allay guilt.
Our study confirms that smoking and sleeping prone are significant modifiable risk factors for the sudden infant death syndrome. It also supports suggestions that bed sharing when the mother smokes carries increased risk. Sleeping on an old mattress may be a new modifiable risk factor but merits further investigation.
We thank Sheila Bartholomew for her help in planning and designing the study, the pathologists throughout Scotland who deal with sudden infant deaths, the crown agent and procurators fiscal, the registrar general for Scotland, and the Common Services Agency for their cooperation. We also thank our fieldworkers and clerical staff for their valuable contribution.
Funding: Scottish Cot Death Trust.
Conflict of interest: None.