|
|
| |||||||||||||||||||||||||||||||||||||||
Background In several studies the sudden infant death syndrome (SIDS) has been significantly associated with sleeping in the prone position. It is not known how the prone position increases the risk of SIDS.
Methods We analyzed data from a case-control study (58 infants with SIDS and 120 control infants) and a prospective cohort study (22 infants with SIDS and 213 control infants) in Tasmania. Interactions were examined in matched analyses with a multiplicative model of interaction.
Results In the case-control study, SIDS was significantly associated with sleeping in the prone position, as compared with other positions (unadjusted odds ratio, 4.5; 95 percent confidence interval, 2.1 to 9.6). The strength of this association was increased among infants who slept on natural-fiber mattresses (P = 0.05), infants who were swaddled (P = 0.09), infants who slept in heated rooms (P = 0.006), and infants who had had a recent illness (P = 0.02). These variables had no significant effect on infants who did not sleep in the prone position. A history of recent illness was significantly associated with SIDS among infants who slept prone (odds ratio, 5.7; 95 percent confidence interval, 1.8 to 19) but not among infants who slept in other positions (odds ratio, 0.83). In the cohort study, the risk of SIDS was greater among infants who slept prone on natural-fiber mattresses (odds ratio, 6.6; 95 percent confidence interval, 1.3 to 33) than among infants who slept prone on other types of mattresses (odds ratio, 1.8).
Conclusions When infants sleep prone, the elevated risk of SIDS is increased by each of four factors: the use of natural-fiber mattresses, swaddling, recent illness, and the use of heating in bedrooms.
Epidemiologic research identifying an effect modification or interaction can assist in our understanding of the causal process of disease. Effect modification occurs when the estimate of the effect of one factor depends on another factor in the study base9. The term "statistical interaction" denotes a similar phenomenon10. The search for an effect modification of the relation between the prone position and SIDS is likely to be fruitful. Parents tend to place their infants in the same sleeping positions regardless of the season, yet the incidence of SIDS is highest in the winter in most countries11. This could be explained by an independent effect of the season or by a seasonal factor potentiating the effect of the prone position. The effect of the prone sleeping position on the risk of SIDS may also depend on nonseasonal conditions or intrinsic abnormalities, such as a developmental deficit or injury. This investigation, from the Tasmanian prospective cohort and case-control studies, attempted to identify effect modifiers of the relation between the prone sleeping position and SIDS.
Methods
The Case-Control Study
The source population consisted of infants who were born and resided in Tasmania, Australia. Case infants were those classified after a postmortem examination as having died of SIDS. As reported previously,12 two control infants were selected for each case infant. Control 1 was matched for age, and control 2 was matched for age and birth weight (<1500 g, 1500 to 2499 g, or 
2500 g).
Two sets of data were obtained and are described in detail elsewhere12. First, information on the thermal environment at the time of a case infant's death was collected. For a subgroup of case infants, this information was collected by ambulance officers investigating the death scene. For other case infants, this information (excluding room temperature) was obtained from the parents at an interview conducted approximately six weeks after the event. For the control infants, this information was gathered by a research assistant at the infants' homes. Additional information was obtained with a comprehensive verbal questionnaire administered to the case parents approximately six weeks after their child had died and to control parents directly after the measurement of the thermal environment or as soon as possible thereafter. Meteorologic information for the day on which the thermal environment was measured was obtained from regional weather stations.
The Cohort Study
A concurrent prospective cohort study to investigate the cause of SIDS is being conducted and involves the approximately 20 percent of live-born infants in Tasmania who are classified according to a perinatal scoring system as being at an increased risk of SIDS. The methods of this study have been described in detail elsewhere6. Data on infants in the cohort born from January 1988 to the end of June 1991 were analyzed to confirm an effect modification of the prone position related to the type of mattress used.
Statistical Analysis
For the case-control analysis, the infant's usual sleeping position was examined, rather than the position in which the infant was found after death or after awakening from sleep. The level of agreement between these two terms was 74 percent for the case infants (kappa = 0.40) and 96 percent for the control infants (kappa = 0.90). Though the latter measure better reflects the period of risk, it involves artifacts caused by agonal processes, such as cervical hyperextension among case infants or position changes made after waking among the controls. The term "prone position" is used to refer to the usual prone sleeping position throughout this paper.
To assess the effect of the variables used for control matching, subgroups were formed according to age, birth weight, region, and minimal 24-hour external temperature in a region. Because of the closeness of matching, the exclusion of discordant case-control sets resulted in minimal loss of information. Case infants who died during a particular season were classified together, and their controls were included in the same category. These procedures provided strata of the matched variables; hence, the difference in risk estimates between strata could be examined13. The common odds-ratio test was used to test for interactions14.
For variables not used in matching, conditional logistic regression was used to evaluate risk and interaction effects. Univariate odds ratios and 95 percent confidence intervals were logit-based and calculated by multiple logistic regression with Epidemiological Graphics, Estimation and Testing (EGRET) 0.25.1 software. The 2-by-2 tables are derived from a basic model that used the usual prone position, the putative effect modifier, and an interaction term for the usual prone position and the effect modifier. Hence, the risk estimates for each factor and the interaction term were adjusted for the confounding effect of the other components in the model. This regression procedure is based on the multiplicative model. It implies that the odds ratios for each exposure present can be multiplied to provide an overall estimate of risk when no interaction is present13,15. The hierarchy principle was always observed16. Tests for an interaction often use a higher significance level than P = 0.0516,17. Here, interaction terms that improved the model fit (P<0.10) were considered to indicate significance.
For the cohort study, a nested case-control analysis was conducted. Case infants for whom prospective home-visit information was available were matched to controls (2 to 10 per case infant) according to birth weight (<1500 g, 1500 to 2499 g, or 2500 g), date of birth, post-conceptional age, and the availability of home-interview data. Conditional logistic regression was used to assess the interaction between the usual sleeping position and the type of mattress used. This was not undertaken for other possible effect modifiers examined in the case-control data because the available measures in the prospective study were not regarded as directly comparable to the retrospective measures.
Results
Characteristics of the Infants in the Case-Control Study
From October 1, 1988, to October 1, 1991, 62 cases of SIDS occurred. Fifty-eight families participated in the retrospective interview (94 percent). The initial response rate of the control families for the 62 case families was 83 percent (101 of 121 families). When a control family did not respond, in all but one instance another control family was selected. A small number of subjects had missing or uncodable responses for some variables. Such data were excluded from the analysis, so there is some variation in the numbers of infants included in the analyses in Table 1. The protocol followed by the ambulance crew at the death scene was obtained for 33 of the 62 cases of SIDS. An earlier analysis of this study data (for which the death-scene protocol was available for 28 of 41 cases) reported an association between overheating and SIDS12. Death-scene information was obtained by parental interview in 29 cases of SIDS, partly because data collection by ambulance crews was phased out in 1991.
|
Risk-Factor Analysis
In the case-control study, infants who usually slept in the prone position had a significantly higher risk of SIDS (unadjusted odds ratio, 4.5; 95 percent confidence interval, 2.1 to 9.6) than those who slept on their backs or sides. The use of a natural-fiber mattress (odds ratio, 1.3; 95 percent confidence interval, 0.6 to 2.9), swaddling at the time of death (odds ratio, 1.3; 95 percent confidence interval, 0.6 to 2.6), recent illness (odds ratio, 1.6; 95 percent confidence interval, 0.8 to 3.5) and the use of heat in the bedroom at the time of death (odds ratio, 1.1; 95 percent confidence interval, 0.5 to 2.2) were not significant risk factors for SIDS in a univariate analysis. The proportion of control infants who usually slept prone did not appear to vary according to the season (P = 0.66).
The interaction of several factors with the prone sleeping position on the risk of SIDS was examined first for the variables used to select the matched controls. The effect of the prone position on the risk of SIDS did not vary significantly according to birth weight (P = 0.84), minimal external temperature (P = 0.46), or region (P = 0.62). The odds ratio for the prone position and SIDS was not significantly higher in young infants (age, 9 days to 11 weeks) (odds ratio, 6.7; 95 percent confidence interval, 1.9 to 24) than in older infants (age, 12 to 33 weeks) (odds ratio, 3.3; 95 percent confidence interval, 1.2 to 8.7). The odds ratio for the prone position and SIDS according to the season was 2.8 for autumn (95 percent confidence interval, 0.53 to 15), 4.0 for winter (95 percent confidence interval, 1.7 to 35), and 3.2 for spring (95 percent confidence interval, 0.80 to 13). The odds ratio for summer could not be calculated, since all case infants who died in summer slept prone.
Second, data on factors not used to match controls with case infants were examined. Particular emphasis was placed on the sleeping environment because of the postulated mechanisms linking sleeping position and SIDS. No significant multiplicative interaction was found between the prone sleeping position and maternal age, parity, maternal level of education (>10th grade vs. 
10th grade), paternal employment (full-time vs. other), infant sex, history of breathing problems in the infant, type of feeding (full or partial breast-feeding vs. only bottle-feeding), maternal postnatal smoking status, type of room ventilation, type of home heating, the placing of a quilt (doona) over the infant, the use of heavy thermal insulation (the degree of thermal insulation of bedding or clothing relative to room temperature) over the infant, the level of excess thermal insulation, the use of a pillow, the use of a plastic or rubber mattress lining, the use of any underlying sheepskin, or the use of an uncovered sheepskin. (Thirty-eight percent of the control population in the case-control study slept on sheepskins. Of these infants, 53 percent slept on sheepskins that were not covered by a sheet or other material.) The strength of the association between sleeping in the prone position and SIDS did vary significantly according to the type of mattress used, the use of swaddling, the occurrence of recent illness, and the use of heating in the room at the time of the case infant's death or the control infant's last sleep.
Risk Potentiation by a Natural-Fiber Mattress
The types of mattresses used in the control population were as follows: foam for 80 infants (67 percent), kapok for 2 (2 percent), ti-tree bark for 25 (21 percent), other materials for 12 (10 percent), and unknown for 1 (1 percent). The term "natural-fiber mattress" refers to mattresses filled with flakes of ti-tree bark or kapok fibers and generally enclosed in a permeable cotton cover. Manufacturers state that mattresses made of ti-tree bark "allow the free passage of air." They are also described as "soft and fluffy" and able to be "shaped for comfort." It is recommended that these mattresses be shaken well after washing to even out the filling. On testing, we found that placement of a 1-kg sugar bag on ti-tree and kapok underbedding left an indentation that was still present five minutes after the bag had been removed. For infants sleeping on natural-fiber mattresses, the risk of SIDS increased nearly 20-fold if the infants usually slept prone. For infants sleeping on mattresses that were not filled with natural fibers, the prone position was associated with only a threefold increase in the risk of SIDS (Table 1). This difference in effect was significant (P = 0.05).
The relation between the sleeping position and the type of mattress used was examined for 22 sets of infants selected from the data base for cohort infants (22 case infants and 213 controls). The usual use of the prone sleeping position at one month of age was associated with a nonsignificant higher risk (odds ratio, 6.6; 95 percent confidence interval, 1.3 to 33) among infants who slept on natural-fiber mattresses than among infants who slept on other types of mattresses (odds ratio, 1.8; 95 percent confidence interval, 0.5 to 6.2) (P = 0.20).
Risk Potentiation by Swaddling
Swaddling was said to have been used if an infant was wrapped in any item of bedding (for example, a sheet or light blanket) while asleep. The strength of the association for the prone position and SIDS was higher (P = 0.09) among swaddled infants (12-fold increase in risk) than among non-swaddled infants (3-fold increase in risk).
Risk Potentiation by Recent Illness
Illness is a composite variable and refers to the presence of one or more of the following: nasal congestion, cough, chest noises, fever, and episodes of vomiting or diarrhea on the day of death or control interview or the day before. The prone position increased the risk of SIDS more than 10-fold among ill infants, but it was associated with only a slight increase in risk among well infants. This difference in risk was significant (P = 0.02). No significant positive interaction was observed for any subcategory of illness. Illness was not associated with a higher risk of SIDS among infants who did not usually sleep prone (odds ratio, 0.83; 95 percent confidence interval, 0.27 to 2.6), but it was associated with a significant sixfold increase in risk among infants who usually slept in the prone position (odds ratio, 5.7; 95 percent confidence interval, 1.8 to 19) (Figure 1).
|
Table 1 shows that the prone position was a stronger risk factor (P = 0.006) among infants who slept in a room heated by any heat source than among infants who slept in an unheated room. Only five case-control sets matched a case infant who slept in a heated room in the prone position with at least one control who slept in an unheated room in some other position; thus, the confidence interval for the risk estimate for SIDS among infants who slept prone in heated rooms was wide. To assess this issue further, the relation between the prone position, room temperature, and SIDS was examined in the subgroup of infants (32 case infants and 61 controls) for whom there were complete thermal data. The effect of the prone position on the risk of SIDS was greater (P = 0.05) in warm rooms (15 to 29 °C) than in cooler rooms (6 to 14 °C).
Adjustment for Confounding Factors
After adjustment for maternal age, birth weight, the use of sheepskins, the use of a plastic or rubber mattress liner, and the use of a quilt underneath the infant, a significant interaction effect was still observed between the prone sleeping position and each of the following: the type of mattress used, the use of swaddling, the occurrence of recent illness, and the use of heating in the bedroom. Adjustment for the use of a pillow gave the following P values for each effect modifier with prone position: P = 0.03 for the use of a natural-fiber mattress, P = 0.16 for the use of swaddling at the time of the infant's death, P = 0.01 for the occurrence of recent illness, and P = 0.03 for the use of heating in the bedroom at the time of the infant's death.
Discussion
The association between the prone position during sleep and SIDS in these two studies has been reported previously. This analysis identified conditions that potentiate the risk associated with the prone position. The adverse effect of the prone position was significantly increased by several factors: the use of a natural-fiber mattress, swaddling, recent illness, and heating of the infant's room.
The multiplicative model was used to screen for interactions. Any positive interaction would also be present in an additive model, but negative or absent findings cannot be taken to mean that additive interactions are not present18. In addition, since the sample used in this study was not large, the statistical power may have been insufficient to detect some interaction effects. Recall bias is unlikely to explain the observed potentiation of the risk associated with the prone position. Prospective and retrospective reports of the association of SIDS and sleeping position show good agreement6,19. The prospective study confirmed the increased risk of SIDS among infants who sleep in the prone position on natural-fiber mattresses. The observational data on room temperature confirmed the finding that the effect of the prone position was stronger in heated rooms.
The adverse effect of the prone sleeping position was six times stronger among infants sleeping on ti-tree or kapok mattresses than among infants sleeping on other types of mattresses. The potential for rebreathing and asphyxia on these types of natural-fiber mattresses requires further investigation. Kemp and Thach20 have previously described 25 infants with SIDS discovered prone on polystyrene cushions. Studies in animals suggested that accidental suffocation by rebreathing was the most likely cause of death in most of the infants studied20. Respiratory obstruction in relation to the position of an infant on a pillow was studied by Emery and Thornton,21 who concluded that the air passage of an infant is impaired if the baby is placed face down on any type of mattress or pillow; the degree of obstruction generally increased with the increasing softness of the material. Alternatively, the natural fibers could conceivably induce an abnormal immunologic response.
Swaddling significantly increased the risk associated with the prone position but was not a risk factor for SIDS among infants who did not sleep in the prone position. Beal and Porter have observed that in countries such as China, where infants are swaddled but sleep supine, SIDS is uncommon22. Conditions that provide greater stability, such as swaddling, may be advantageous for infants who sleep on their sides or backs but disadvantageous to infants who are prone and attempt to change position.
Two seasonal factors, recent illness and the use of heating in rooms, potentiated the association between the prone position and SIDS. This may explain why attempts in our region to decrease the use of the prone position since 1991 have been accompanied by a decline in the incidence of SIDS,23 particularly during the winter months.
The finding that the prone position was a stronger risk factor in heated than unheated rooms supports the theory that hyperthermia and the prone position may interact to predispose an infant to SIDS24. Swaddling and the use of natural-fiber mattresses may decrease heat loss in infants. Illness may increase heat production. However, this study, in accord with previous work,2,12 failed to demonstrate that the prone position was a stronger risk factor among infants with heavy or excess clothing and bedding over them.
Several other mechanisms could also explain our results. In particular, the theory of rebreathing of carbon dioxide20,25,26 deserves further investigation. Conditions that increase an infant's respiratory dead space, such as soft, porous bedding,25 or reduce air movement would increase the likelihood of substantial rebreathing of carbon dioxide by an infant sleeping in the prone position25. Other postulated mechanisms for the prone position as a risk factor for SIDS, such as asphyxia27 and arousal deficit,28 remain possibilities, and it may well be that more than one of these mechanisms is responsible.
The strength of the association between the prone position and SIDS differs among studies conducted in different communities. For example, in 1989 Nelson et al. commented that the rate of SIDS in Sweden was relatively low, even though the prone sleeping position was advised for infants24. These results suggest that differences in the prevalence of effect modifiers between communities may explain part of this variation.
Healthy infants should not be placed prone to sleep. These results suggest that if infants have to sleep in the prone position for a specific medical reason, they should be placed on a firm mattress and not swaddled. Future physiologic work should be directed at understanding how the type of mattress used, swaddling, illness, and the use of heating in the bedroom operate to potentiate the adverse effect of the prone position on SIDS early in life.
Supported by the Tasmanian state government, Australian Rotary Health Research Fund, National Health and Medical Research Council, National Sudden Infant Death Syndrome Council, Wyeth Pharmaceuticals, Zonta International, and Tasmanian Sanatorian After-Care Association.
We are indebted to the parents and infants who participated in this study; to H. Bain and other research assistants for data collection; to the professional officers of the Tasmanian Ambulance Service for the collection of death-scene data; to Ego Pharmaceuticals for the provision of their products; to the departments of Hospital Pathology, Police, and Justice for their cooperation; to the hospitals participating in the cohort study; to J. Kaldor, S. Kasl, and D. Peterson for epidemiologic advice; to J. Williams for computing assistance; and to D. Lees for photography and graphics.
Source Information
From the Menzies Centre for Population Health Research, University of Tasmania, 43 Collins St., Hobart, Tasmania, Australia 7000, where reprint requests should be addressed to Dr. Ponsonby.
References
| |||||||||||||||||||||||||||||||||||||||
Related Letters:
The Risk of Sudden Infant Death
Cohle S. D., Kemp J. S., Thach B. T., Li D.-K., Shannon D. C., Ponsonby A.-L., Dwyer T., Gibbons L. E.
Extract |
Full Text
N Engl J Med 1994;
330:63-64, Jan 6, 1994.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2010 Massachusetts Medical Society. All rights reserved. |
