Whats revealed is often half the truth

An interesting review: what’s revealed is half the truth (almost)

This morning came across a nice review of a cochrane review on the efficacy of ibuprofen alone or in combination with anti emetics for the management of migraine. It turned out that ibuprofen at 400 mg dose is really a useful and quite effective medicine. Indeed with overall estimates of the meta analysis, as the authors wrote,

‘ Nine studies (4373 participants, 5223 attacks) compared ibuprofen with placebo or other active comparators; none combined ibuprofen with a self-administered antiemetic. All studies treated attacks with single doses of medication. For ibuprofen 400 mg versus placebo, NNTs for 2-hour pain-free (26% versus 12% with placebo), 2-hour headache relief (57% versus 25%) and 24-hour sustained headache relief (45% versus 19%) were 7.2, 3.2 and 4.0, respectively. For ibuprofen 200 mg versus placebo, NNTs for 2-hour pain-free (20% versus 10%) and 2-hour headache relief (52% versus 37%) were 9.7 and 6.3, respectively. The higher dose was significantly better for 2-hour headache relief than the lower dose. Soluble formulations of ibuprofen 400 mg were better than standard tablets for 1-hour, but not 2-hour headache relief.’

from the abstract of the study

Now, if you review the l`abbe plots of the study and the figures in the actual review, you start to see a different story. Take the l`abbe plots for instance (check Figures1-3 in the original report for instance here), and a different story emerges. One gets to see that about 25% of the studies had any form of concealment of allocation, and most of the studies for the two hour relief of pain were concentrated around the centre line of the l abbe plot, and that the largest study in the meta analysis was equivocal and had the results plotted near the centre line.

I think meta analysis like this presents a very interesting perspective on the choice of treatment. One needs to look at the fine line.