Statistics from Altmetric.com
Editor,—In the paper by Mitchellet al the use of duvets was associated with sudden infant death syndrome (SIDS) on univariate (odds ratio 1.65 (95% confidence interval 1.31, 2.08)) but not multivariate analysis, and the SIDS risk associated with duvet use was not modified by sleeping position.1 These findings contrast with our finding of a strong association (adjusted odds ratio 6.16 (2.01, 18.87)) between quilt (duvet) use and SIDS among non-prone infants.2
Mitchell et al postulated that one reason for the discrepancy may be that our Tasmanian study had not controlled for socioeconomic factors. We did, in fact, adjust for a large number of additional potential confounders that could not be listed in the short report due to space limitations. We found that adjustment for unemployment, maternal education, or maternal parity did not alter the association between duvet use and SIDS among non-prone infants. We agree that the conflicting findings may relate to local differences in duvet characteristics, although alternative explanations may also contribute.
One possibility is non-differential misclassification in that the New Zealand study asked a single question to determine exposure whereas the Tasmanian study employed visual verification of the actual bedding items for cases and controls by interviewer where possible. From 1991 to 1995 this also included a sample of different types of duvets, which the nurse took to the home interview to assist with classification. As non-differential misclassification will bias an association based on dichotomous exposures towards or beyond the null,3 this may explain the weaker strength of association between duvet use and SIDS in the New Zealand study compared with ours.
It is also critically important not to adjust for any factor that may be on the causal pathway between exposure and disease, as this will lead to an underestimate of the true association.4 A classic example is adjustment for birth weight when examining any association between maternal smoking and infant mortality. This is clearly inappropriate as the adverse effect of smoking is partially mediated through birth weight.4
The causal pathways between duvet use and SIDS are less clear but our data indicated that part of the adverse effect resulted from facial obstruction.2 Thus, it is not surprising that little adverse effect remained for duvet use on SIDS in the UK study after adjustment for a large number of factors, including head covered during last sleep, and the authors correctly pointed out that duvet use appeared to increased the risk of SIDS partially through a propensity for total covering.5
Mitchell et al report that duvet use was inversely associated with being tucked in firmly, a protective factor for SIDS. They included “firm tucking in” as a confounder in their analyses, and thus report adjusted odds ratios for duvet use that reflect only the residual effect of duvet use on SIDS, excluding any adverse effect that is actually meditated through looser bedclothes. To rely on these adjusted results may underestimate the true association between duvet use and SIDS.
Dr Mitchell and colleagues comment:
Our results have similarities and differences with those from the Tasmanian study. We have both shown that the risk of SIDS from thermal stress is only among infants sleeping prone.1-1 1-2 Given that there is a positive relation between duvet use and excess thermal insulation, we were surprised that the Tasmanian study subsequently found that duvet use increased the risk of SIDS only among infants sleeping supine.1-3
We suggest several explanations for the difference between the studies:
- the characteristics of the duvets differed between New Zealand and Tasmania
- the Tasmanian study did not adjust for confounders; their letter indicates that they did, but this was not reported in their paper.
Ponsonby and colleagues suggest two additional explanations:
- non-differential misclassification
- adjustment for “firm tucking in” is inappropriate as it may be on the causal pathway.
Misclassification is unlikely; although we used a simple question to determine whether a duvet was used it is unlikely that parents would mistake its use. Furthermore fewer than 2% of cases and controls did not answer this question.
To exclude the possibility that we inappropriately included a factor in the causal pathway we have rerun the multivariate analysis without “firm tucking in”. The risk of SIDS with duvet use remains insignificant (adjusted odds ratio 1.18 (95% confidence intervals 0.87, 1.60) compared with “variable included” adjusted odds ratio 1.04 (0.77, 1.58)). Our impression is that duvets in New Zealand are larger than those in Australia and are tucked in at the foot of the cot. This may reduce the possibility of the duvet covering infants' heads. In our study there was no increased risk of being found with head covered when using a duvet compared with those who were not using a duvet.
We conclude that these explanations do not account for our findings and that our study does not support the recommendation to avoid duvets.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.