Comparative Meta-Analysis of the Efficacy of Once-Daily Fluticasone Furoate 100 µG Versus Twice-Daily Fluticasone Propionate 250 µG in Adolescents and Adults with Persistent Asthma

Data from 433 patients included in the ITT populations of the FF100 and FP250 arms of the two clinical studies were combined for this meta-analysis (FFA109685: FF100 n = 105, FP250 n = 100; FFA112059: FF100 n = 114, FP250 n = 114). Baseline characteristics were comparable between treatment arms within and across the two studies, except for a higher rate of previous ICS + long-acting β₂ agonist treatment in study FFA109685.

Using a random-effects model, the mean difference between FF100 and FP250 in change from baseline in trough FEV₁ was approximately −1.7 mL (95% confidence interval [CI] −80.4, +77.0); this difference was not statistically significant (p = 0.9664) (Fig. 1). FF100 was non-inferior to FP250 for the primary outcome measure, as the lower boundary of the 95% CI was greater than the pre-defined non-inferiority margin of −200 mL. The Q test and I ² results indicated no statistically significant heterogeneity between the results from the two trials. A supporting meta-analysis using LS mean change from baseline in trough FEV₁ also demonstrated no statistically significant difference between FF100 and FP250 (−7.9 mL [95% CI –87.1, +71.3], p = 0.8450), and non-inferiority of FF100 to FP250. Similar findings were obtained using a fixed-effects meta-analytic model (data not shown).

Meta-analytic approaches allow for the combined analysis of statistical data to obtain a more robust measure of treatment effect. Only two company-sponsored studies from the GSK FF clinical trial programme met the criteria for inclusion in this meta-analysis, having evaluated both FF100 and FP250 treatment arms. Combining the findings from the individual trials, the results of this meta-analysis demonstrated no statistically significant difference between FF100 and FP250, and non-inferiority of FF100 to FP250, in mean change from baseline in trough FEV₁. With only two studies included in the meta-analysis, meta-regression to adjust for differences between patient populations was precluded; however, such adjustment was unlikely to have affected the results. Overall, the disparities between the two studies were minor and considered unlikely to have significantly impacted the trial outcomes. Furthermore, the Q test and I ² results indicated no significant statistical heterogeneity between the trial results.

However, limitations of the analysis should be considered when interpreting these results. Firstly, both included studies, FFA109685 and FFA112059, were placebo-controlled trials that included FP250 only as a reference arm. Neither study was originally designed for investigating the equivalence or non-inferiority of FF to FP. Secondly, FF and FP were delivered via different inhalers in the two studies (ELLIPTA and Diskus/Accuhaler, respectively [5, 6]). Thirdly, the timing of the primary endpoint assessment differed between the two studies (8 weeks in FFA109685; 24 weeks in FFA112059); however, the rationale for comparing data from the two time points using a LOCF approach is supported by the following: (i) from a clinical perspective, steady state (FEV₁ effect size) is already achieved by 8 weeks of treatment and (ii) trough FEV₁ did plateau after 8 weeks in study FFA112059 [6]. Lastly, caveats associated with the limited sample size in this study should be noted. Meta-analyses including few studies have lower statistical power to detect the effect of an intervention, and preclude from performing publication bias tests, meta-regressions (to test for the presence of effect modifiers), subgroup analyses, and sensitivity analyses, for example. The analysis was limited by the number of patients with available data for inclusion (N = 433 across the two studies) and it is acknowledged that the degree of random error contributing to imprecise estimates of treatment effect may be reduced with a larger sample size.