Mattresses, microenvironments, and multivariate analyses

Sir,

Does parity confound the association between used infant mattress and
Sudden Infant Death Syndrome (SIDS) in Scotland (1)? That is to say, is
the observed association (2,3) in fact a result of the confounding effect
of parity and not due to used infant mattress at all. Clarification of
this issue is important as 33% of babies in Scotland sleep on a used
infant mattress as do many infants in New Zealand (4). We explore this
issue using case-control data from two studies in Scotland from January
1992 to May 2000 (2,3).

Questionnaires were completed on 278/317 (88%) cases notified in time
for an interview within 28 days of death and 554/634 (87%) controls.
Parity was determined by asking How many live births has mother had,
including this one? Exposure to a used infant mattress was assessed by
asking parents about routine sleeping place for both cases and controls
and ascertaining the mattress status (new for this baby or not). Routine
bed sharers were excluded (35 cases and 33 controls). Multivariate
analysis, to control for possible confounding factors including parity,
confirmed a valid statistical association between used infant mattresses
and SIDS, OR 2.51 (95%CI 1.39, 4.52) (2) and OR 3.07 (1.51, 6.22) (3).
However further criticism advised: ‘A careful analysis of mattress re-use,
stratified by families of different sizes to clarify the possible
importance of parity, which is obscured by the multivariate model used’
(1). Table 1 shows the stratified analysis of the pooled data. The
observed association between used infant mattress and SIDS remains, OR
2.26 (1.40, 3.63).

Of more interest is the alternative hypothesis that the observed
association between parity and SIDS is in fact due to the confounding
effect of used infant mattress and not due to parity at all. The
multivariate fixed effects analysis (3) inferred this conclusion as parity
became insignificant only when used infant mattress was added to the
model. Using stratified analysis we confirm that the increase in the risk
of SIDS associated with each one unit increase in parity, OR 1.49 (1.28,
1.73), almost disappears, OR 1.20 (0.95, 1.50), when controlled for used
infant mattress and the matched nature of the data.

In summary, parity does not confound the association between used
infant mattress and SIDS; on the contrary, used infant mattress is in fact
the probable reason for the observed association between parity and SIDS
in Scotland.

References

1. Fleming PJ, Blair PS, Mitchell EA. Mattresses, microenvironments, and
multivariate analyses. BMJ 2002; 325: 981-982.

2. Brooke H, Gibson A, Tappin D, Brown H. Case-control study of Sudden
Infant Death Syndrome in Scotland 1992-5. BMJ 1997; 314: 1516-20.

3. Tappin D, Brooke H, Ecob R, Gibson A. Used Infant Mattresses and Sudden
Infant Death Syndrome (SIDS) in Scotland. A Case-control study. BMJ 2002;
325:1007-9.

4. Gregory G. Expert rebuffs cot death study. New Zealand Herald; Section
1, page 3, 4th November 2002.

Table 1: Stratified unmatched analysis of the effect of
Used Infant Mattress on Sudden Infant Death Syndrome
controlling for Parity

Used infant mattress
Parity	   New Used	 OR(95%CI)	       MHOR(95%CI)
3+	SIDS	    8	 28	3.75(1.14,12.69)	2.26(1.40, 3.63)

controls 15	 14	1.00	
3	SIDS	   17	 37	2.46(1.13,15.41)	

controls 43	 38	1.00	
2	SIDS	   24	 50	1.88(1.02,3.46)	

controls 83	 92	1.00	
1	SIDS	   51	 28	3.92(2.06,7.48)	

controls207	 29	1.00	
Total	SIDS	  100	 143	2.88(2.08, 3.99)	

controls348	 173	1.00	
MHOR is the Mantel Haenszel Odds Ratio showing
the association between used infant mattress
and SIDS after controlling for parity and the
matched nature of the data.

Competing interests:  
None declared