Introduction/Objective: To assess the differences in conclusions drawn from a traditional meta-analysis (MA) with that of individual patient data (IPD), using data from a review of the effectiveness of a single dose of nedocromil sodium (NCS) to prevent exercise-induced bronchoconstriction (EIB).
Methods: Randomised, double-blind, controlled trials that compared a single dose of nedocromil sodium to a placebo in prophylactic treatment of exercise-induced bronchoconstriction (EIB) were included. Using Cochrane methods, a systematic review (SR) of the literature was completed and submitted to the CDSR for dissemination. Authors of the twenty included crossover trials were contacted to provide individual patient data. Data collection and analysis: The main outcomes assessed were the maximum percent reductions in FEV1 and PEFR after standard exercise challenges compared to pre-challenge baselines, and the adverse effects caused by NCS. Results from the meta-analysis were reported as the weighted mean difference (WMD) with 95% confidence intervals between treatment and placebo effects using the random effects model. These overall pooled estimates and the subgroup estimates were compared to results obtained from analysis of IPD.
Results: From the 20 included crossover trials (280 participants) authors provided IPD from 9 trials (45%) representing 112 participants (40%). However, few investigators were able to provide data on the order of treatment sequence or results on the discrete exercise challenges, so period and carry-over effects could not be estimated. A total of 3 (15%) studies reported a test for these in the text. In the meta-analysis of these 9 studies, NCS had a significant effect on inhibiting EIB. The WMD for the mean maximum % fall FEV1 was 16.0% (95% Cl: 12.7, 19.2), similar results were obtained from the IPD analysis: the mean difference between treatment and placebo was 16.0 (95% Cl: 13.1, 19.9). Significant differences in the inhibitory effect of NCS were demonstrated in subgroup analyses based on severity of EIB (mild EIB: < 30% fall; moderate-severe EIB: >= 30% fall). The WMD for mean maximum % fall FEV1 in mild EIB was 12.8% (95% Cl: 8.6, 17.0) compared to 22.5% (95% Cl: 15.3, 25.8) in more severe EIB in the main meta-analysis. The IPD results (or mild EIB were 8.2% (95% Cl: 4.8, 11.5) and for more severe EIB were 25.6% (95% Cl: 22.1, 29.2). There were no other significant subgroup results and no important differences in the reported adverse effects using NCS.
Discussion: The results of the IPD analysis did not differ in important ways from the results of the traditional Cochrane meta-analysis. The effort required to complete IPD is impressive, and the yield in this review was low. Systematic reviews of crossover trials would benefit from improvements in reporting in original trials.