Objective: To investigated the use of mixed treatment comparison (MTC) meta-analysis models to summarize results from randomized clinical trials (RCTs) on approved pharmacological treatments for neovascular age-related macular degeneration (AMD).
Methods: The number of patients with visual loss or visual gain of 3 or more lines of visual acuity (15 ETDRS letters) at 1 year was extracted from 10 RCTs including patients with neovascular AMD and comparing at least one of the following drugs to sham treatment (1080 control patients, 8 studies) or to each other: verteporfin photodynamic therapy (PDT, 1124 patients, 4 studies), pegaptanib (904 patients, 2 twin studies), ranibizumab (984 patients, 4 studies). Both frequentist and Bayesian methods were used to conduct MTCs.
Results: Direct and indirect evidence was available and found to be overall in good agreement for the comparisons: PDT vs. control, ranibizumab vs. control, ranibizumab vs. PDT. Bayesian model fit was better for a model including a covariate coding for the PIER study ranibizumab regimen, i.e. quarterly injections after three initial monthly doses. In the MTC model, monthly ranibizumab was superior to PDT and pegaptanib, and could not be shown to be better than PIER ranibizumab regarding visual loss, being estimates imprecise. Ranibizumab PIER retreatment regimen was better than PDT and pegaptanib regarding visual loss, whereas an advantage over them regarding visual gain was suggested by a frequentist MTC approach, but not by a Bayesian approach, which was more conservative. A limitation of our MTC model was that only two twin studies connected pegaptanib to the treatment network, and only one study was available for the PIER ranibizumab regimen.
Conclusion: The clinically heterogeneous and sparse typology of the evidence is a limitation to carry out MTC meta-analyses of approved pharmacological treatments for neovascular AMD. Ranibizumab was found to be the most effective treatment compared to PDT and pegaptanib, although this superiority cannot be demonstrated regarding visual gain for the PIER ranibizumab regimen in a Bayesian analytic setting. We did not find RCTs which investigated the current ranibizumab as needed retreatment regimen approved in Europe.