Actual size of increase needs to be measuredBMJ 1996; 313 doi: https://doi.org/10.1136/bmj.313.7066.1209b (Published 09 November 1996) Cite this as: BMJ 1996;313:1209
- Sally Collins,
- Andrew Moore,
- Henry McQuay
- Research assistant Consultant biochemist Clinical reader in pain relief Pain Relief Unit and Nuffield Department of Anaesthetics, University of Oxford, Oxford Radcliffe Hospital, The Churchill, Oxford OX3 7LJ
EDITOR,—Underlying Anton J M de Craen and colleagues' conclusion that adding codeine 60 mg to paracetamol can result in a small but significant increase in analgesia is the crucial question “How big is small?”1 The benefit of adding 60 mg codeine to paracetamol in studies of single doses can be quantified.2 Thirteen of the 19 reports from de Craen and colleagues' search met our inclusion criteria of comparisons of single doses of oral paracetamol with single doses of paracetamol plus codeine 60 mg, double blind treatment with random allocation, baseline pain of moderate to severe intensity, and total pain relief as a derived outcome. Two studies used paracetamol doses of 1000 mg and the rest used 600 mg or 650 mg. We calculated the risk ratio and number needed to treat for paracetamol versus paracetamol with codeine 60 mg, using the criterion of pain relief of at least 50%.2
Only one of the 13 reports had a significantly increased risk ratio (the 95% confidence interval did not include 1 (fig 1)). The combined risk ratio (fixed effects model) for this homogeneous dataset was significant (1.25 (1.09 to 1.43)). Adding 60 mg codeine to paracetamol for single dose pain relief produced a number needed to treat of 9.1 (5.8 to 24). This means that one of every nine patients with pain of moderate to severe intensity will get at least 50% pain relief with paracetamol plus 60 mg codeine who would not have done had they been given the same dose of paracetamol alone.
We, like others, find the number needed to treat to be a clinically useful finding of quantitative systematic reviews of treatments. To get credible numbers, however, attention must be paid to the validity of the trials that are included in the review. We had to omit two of the trials in de Craen and colleagues' analysis because of their failure to ensure adequate sensitivity by requiring at least moderate or severe pain at baseline.3 They also used the mean descriptors from the trials. Means of asymmetrically distributed data (summed difference in pain intensity) can produce incorrect conclusions.4
Paracetamol-codeine combinations are often used in multiple doses. Assessing the additional benefit of the codeine in trials of single doses may underestimate the effect after multiple doses.