Low response rates generate considerable uncertainty
Editor – This study found that in infants who died from SIDS, the
reported use of a dummy in the index sleep was 4.1%, in the 50% of
eligible mothers who agreed to take part. The use of a dummy in the other
50% of eligible infants with SIDS is unknown. In the control infants the
reported use of a dummy in the last sleep before interview was 23.6%, in
the 40% of eligible mothers who agreed to take part. The use of a dummy in
the other 60% of control infants is also unknown.
This means that potentially the overall dummy use in all the cases
was between 2% and 54% and in all the controls between 9% and 69%. None of
the confidence intervals reported in the paper reflects uncertainty
generated by those who declined to take part. Due to the low response
rates, unadjusted odds ratios of anywhere between 0.01 and 7.7 are
compatible with the study findings, when the total eligible population is
The authors conclude in their abstract that “use of dummy seems to
reduce the risk of SIDS”, but it could just be that reported dummy use in
infants who died from SIDS was low because more mothers who gave their
infants dummies declined to take part in the study.
The abstract also concludes that “use of a dummy possibly reduces the
influence of known risk factors in the sleep environment” but this is
based on a spurious use of confidence intervals for the subgroups involved
(1). The more correct test of interaction reported in the results section
showed that “the difference in odds ratios between infants who did or did
not use a dummy was not significant.”
There is much more uncertainty in this paper than the authors would
have us believe from the abstract. The reporting contrasts sharply with
the careful reporting of Fleming et al (2), and in my view Fleming’s more
cautious conclusion still stands: “Further epidemiological evidence and
physiological studies are needed before pacifier use can be recommended as
a measure to reduce the use of SIDS”.
1. Altman DG, Bland JM. Statistics Notes: Interaction revisited: the
difference between two estimates. BMJ 2003;326:219
2. Fleming PJ, Blair PS, Pollard K, Platt MW, Leach C, Smith I, et al.
Pacifier use and sudden infant death syndrome: results from the CESDI/SUDI
case control study. Arch Dis Child 1999;81(2):112-116.
Competing interests: No competing interests