Individual patient meta-analysis using information from clinically homogeneous acute pain trials with observations over 24 h was used to investigate different ways trials can be analysed and reported. There were 13 third-molar extraction trials, with 1,330 patients using rofecoxib 50 mg, 303 using ibuprofen 400 mg, and 570 using placebo. Pain relief scores were available at individual time points, plus time to remedication. Many more patients remedicated with placebo than ibuprofen 400 mg, and more with ibuprofen than rofecoxib 50 mg. Median time to remedication, the proportion remedicated at various times, or survival curves would be useful outcomes. In dealing with missing data points when patients remedicated, baseline observation carried forward was more conservative than last observation carried forward, resulting in higher (worse) NNTs and lower average pain scores after 12 and 24 h. Results based on both methods might be sensible for trials longer than eight hours. The distribution of pain relief was highly skewed, especially at later times, when almost no patient was average. Different cut points for pain relief (at least 25, 50 or 75% maxTOTPAR) and longer duration changed the NNT for ibuprofen compared with placebo, but less for rofecoxib, reflecting longer duration of action of rofecoxib. Reporting for each treatment group the percentage of patients with 25, 50 and 75% pain relief at various times after dose, and reporting the proportion of patients with good or complete pain relief, and inadequate pain relief, at each time point, would improve acute pain trial reporting.
Pain Research and Nuffield Department of Anaesthetics, University of Oxford, Oxford Radcliffe NHS Trust, The Churchill, Headington, Oxford OX3 7LJ, UK
*Corresponding author. Tel.: +44 1865 225775; fax: +44 1865 226978.
Received 29 July 2004; received in revised form 5 April 2005; accepted 3 May 2005.