Abstract
Purpose:
Produce meaningful comparisons of clinical trials of anti-vascular endothelial growth factor (VEGF) therapies in the treatment of visual impairment (VI) due to diabetic macular edema (DME).
Methods:
The analysis consisted of five steps: (1) identify the baseline characteristics impacting the increase in best-corrected visual acuity (BCVA) with anti-VEGF therapies (2) create a model with RESPOND/RESTORE patient-level data to estimate the impact of key baseline characteristics on BCVA gains from baseline to month 12 (3) predict the average BCVA gain with ranibizumab 0.5 mg pro re nata (PRN) (monotherapy and ranibizumab+laser combination therapy) and laser monotherapy, assuming the same baseline characteristics as in VIVID/VISTA (4) compare the predicted gains with aflibercept bi-monthly (2q8) gains in VIVID/VISTA using a Bayesian network meta-analysis with fixed treatment effect (5) run sensitivity analyses to assess the robustness of the results.
Results:
The baseline characteristics frequently associated with BCVA gains in the literature are baseline age, central retinal thickness (CRT) and BCVA. Without adjusting for those baseline variables (i.e. using raw clinical data), RESTORE/RESPOND patients on ranibizumab 0.5 mg PRN gained, on average, 6.6 letters (95% credible interval [95% CrI]: 4.5-8.7) more than those on laser monotherapy. If RESPOND/RESTORE patients were similar to those in VIVID/VISTA (in terms of age, CRT and BCVA at baseline), the model predicted 9.3 more letters (95% CrI: 6.8-11.8) for patients on ranibizumab 0.5 mg PRN than for those receiving laser monotherapy and 9.4 more letters (95% CrI: 6.8-11.9) for combination therapy patients. Those results were non-statistically different (2-sided p-value: 0.65 for ranibizumab monotherapy and 0.68 for combination therapy) compared to the BCVA gain observed for aflibercept 2q8 in VIVID/VISTA (10.0 more letters than laser, 95% CrI: 8.3-11.7). Adding baseline characteristics to the model (gender, baseline duration of diabetes, baseline HbA1c) also produced a predicted gain of 9.3 letters (95% CrI: 6.5-12.1) for ranibizumab patients over laser.
Conclusions:
It is important to adjust for baseline characteristics influencing outcomes when comparing clinical trials. After doing so, the difference in letters gained between patients receiving ranibizumab and aflibercept is non-significant.