When to act on evidence?

BMJ 2002; 325 doi: (Published 02 November 2002) Cite this as: BMJ 2002;325:h

Assembling good evidence is hard. Knowing how and when to act on the evidence is even harder.

A group from Glasgow find that routine use of an infant mattress previously used by another child is associated with a threefold increase in risk of sudden infant death (p 1007). The same authors found such an association in 1997 and set out to test it again. That they have found the association again makes it more likely to be “real.” Another study from England found a similar but weaker association, but it disappeared after adjustment for known risk factors. So should parents throw out old mattresses? A cautious editorial advises “no.” Make sure, however, that you put your baby on his or her back and provide a smoke free environment.

Should twins be delivered by caesarean section? Observational studies from the 60s suggested that second twins had a higher risk of perinatal death than first twins. Larger studies failed to confirm the risk. Now Gordon Smith and others have studied a large sample retrospectively and examined in detail the cause of death (p 1004). They find that second twins are at higher risk of “delivery related perinatal death.” There were no deaths among 454 second twins delivered by caesarean section. Is this good enough evidence to recommend routine caesarean section? The authors calculate that a randomised trial to answer the question would require 6500 women with twin pregnancies. “We propose,” conclude the authors, “that women with twins should be counselled about the risk to the second twin and the theoretical possibility of a protective effect of planned caesarean section when considering mode of delivery at term.” Perhaps too they should be counselled on the difficulties of interpreting evidence. Not so easy to do with 50 pregnant women in the waiting room.

What about giving free smoke alarms to deprived populations? A before and after study conducted in Oklahoma showed an 80% drop in admissions to hospital and deaths related to fire. Now a cluster randomised trial (a superior design) finds no reduction in injuries related to fire (p 995). The researchers also found that half of the alarms were not working 15 months after being installed (p 998). This is the difference between an efficacy trial (what works in perfect condition) and an effectiveness trial (what works in the real world) (p 979). It wouldn't seem sensible after this trial simply to distribute smoke alarms, but it might be worth ensuring they are maintained. But that needs another trial and an economic evaluation—it might work but be unaffordable.

Sometimes the problem is acting on the evidence gathering rather than acting on the results. In July the Women's Health Initiative trial in the United States of hormone replacement therapy was stopped because of poor outcomes in the women taking combined treatment. (Editorials in the BMJ and JAMA, incidentally, gave conflicting advice to practitioners (p 1036)). The question arose on whether a similar trial in Britain should be stopped. The answer is “yes” (p 987).


View Abstract