Session Title: Free Paper Session 11: Vascular Diseases & Diabetic Retinopathy III
Session Date/Time: Friday 08/09/2017 | 11:00-12:30
Paper Time: 11:48
Venue: Room 114
First Author: : J.Korobelnik FRANCE
Co Author(s): : N. Hawkins T. Reason I. Chatzitheofilou F. Ryan D. Muston P. Kaiser
Diabetic macular edema (DME) is the leading cause of vision loss in diabetes. Recommended treatments include laser and anti-vascular endothelial growth factor (VEGF) therapies, such as intravitreal aflibercept (IVT-AFL) and ranibizumab, which are both licensed treatments. Mean change in best-corrected visual acuity (BCVA) is a primary efficacy measure, but may be dependent on baseline BCVA due to a ceiling effect. Previous network meta-analyses (NMA) using meta-regression (Korobelnik et al, 2015) have considered this; however, meta-regression at trial level can be impacted by ecological bias. Incorporating individual patient-level data (IPD) provides more robust adjustment for differences in between-trial baseline BCVA.
The objective of this NMA was to perform an indirect comparison of anti-VEGF regimens and laser photocoagulation, incorporating IPD where available to allow for a robust adjustment for differences between trials in baseline BCVA.
A systematic review was conducted to identify relevant randomized controlled trials, and 13 trials were identified as eligible for inclusion. Trials were included that reported the randomized controlled results at 12 months for the outcomes of interest: mean change from baseline in BCVA, and patients achieving gains of ≥10 and ≥15 Early Treatment Diabetic Retinopathy Study (ETDRS) letters. These trials were used to form a network for a Bayesian NMA. The network included 12 treatment regimens; those of interest were IVT-AFL 2 mg every other month (2q8) after 5 initial doses, ranibizumab 0.5 mg pro re nata (PRN), ranibizumab 0.5 mg treat and extend (T&E), and laser photocoagulation. Baseline BCVA was adjusted for by incorporating IPD from 5 trials and aggregate data from 8 other trials. Analyses were performed using (i) a common covariate adjustment where baseline BCVA was assumed to modify treatment effects in the same way against a common reference comparator for all interventions, (ii) covariate adjustments specific to each treatment comparison, and (iii) no covariate adjustment. Binary outcomes are reported as odds ratios (OR), the continuous outcome is reported as the mean treatment difference (ETDRS letters). Uncertainty is reported using 95% credible intervals (CrI).
The common covariate model (with laser as the reference treatment) was best fitting. Results are presented for this model and no covariate adjustment for comparison. IVT-AFL, ranibizumab 0.5 mg PRN and ranibizumab 0.5 mg T&E were shown to be significantly superior to laser in all analyses. IVT-AFL showed statistically significant superiority versus ranibizumab 0.5 mg PRN for both the no covariate adjustment and common covariate adjustment analyses for gaining ≥10 letters (OR 1.72, 95% CrI 1.11-2.53 and OR 1.71, 95% CrI 1.09-2.56, respectively), gaining ≥15 letters (OR 2.07, 95% CrI 1.19-3.38 and OR 2.28, 95% CrI 1.25-3.85, respectively), and mean change in BCVA (mean difference 4.42, 95% CrI 2.07-6.76 and mean difference 4.95, 95% CrI 3.09-6.83 ETDRS letters, respectively). IVT-AFL also showed statistically significant superiority compared with ranibizumab 0.5 mg T&E in mean change in BCVA for both analyses (mean difference 5.06, 95% CrI 1.88-8.23 and mean difference 4.90, 95% CrI 2.06-7.72 ETDRS letters, respectively). Patients of a lower BCVA at baseline had a greater response to treatment, consistent with the direction of effect expected.
It is important for clinicians and policymakers to compare the relative efficacy of DME treatments using the most robust methods available. Head-to-head trials provide stronger evidence of comparative efficacy than indirect comparisons. This indirect comparison of IVT-AFL 2q8 versus ranibizumab 0.5 mg regimens provides a useful complement to the direct comparative evidence of IVT-AFL versus ranibizumab 0.3 mg given in Protocol T. To the author’s knowledge, this is the first NMA in DME that has adjusted for covariate imbalances using IPD. Previous analyses that used only aggregate data are prone to ecological bias. For all analyses, the covariate adjusted analysis was the best fitting model, with coefficient in the expected direction. The results show that these adjustments are important and methodologically appropriate. Availability and use of IPD can be important in ensuring NMAs are robust. This analysis has consistently shown superiority of IVT-AFL to laser and ranibizumab 0.5 mg PRN. IVT-AFL was superior to ranibizumab 0.5 mg T&E in the continuous outcome. These efficacy results were irrespective of adjustment for baseline BCVA; safety outcomes were not considered. Results are similar to the earlier published NMA that included fewer trials and no IPD.