First-line treatments for chronic lymphocytic leukaemia: interpreting efficacy data by network meta-analysis

When multiple treatments are available, network meta-analysis can synthesize evidence and rank relative effectiveness. We applied this approach to current treatments for previously untreated chronic lymphocytic leukaemia. Data search was conducted in PubMed and websites of regulatory agencies (year 2000 through present time). Our analysis included randomized controlled trials assessing treatments for previously untreated chronic lymphocytic leukaemia. The endpoint of the analysis was the rate of progression-free survival at 3 years. At least two reviewers abstracted study data and outcomes. Agents examined for their relative effectiveness included four monotherapies (chlorambucil, fludarabine, bendamustine, alemtuzumab) and four combination treatments (cyclophosphamide + fludarabine, cyclophosphamide + cladribine, cyclophosphamide + fludarabine + rituximab, cyclophosphamide + fludarabine + alemtuzumab). A Bayesian network meta-analysis was conducted to comparatively evaluate these treatments. Nine trials (3620 patients) were included in the analysis. Odds ratio (with 95 % credible intervals) was estimated for all direct and indirect comparisons. Combinations treatments were found to be significantly more effective than single-agent treatments. Ranking in effectiveness was as follows: (1) cyclophosphamide + fludarabine + rituximab, (2) alemtuzumab, (3) cyclophosphamide + fludarabine + alemtuzumab, (4) cyclophosphamide + fludarabine and (at same ranking) cyclophosphamide + cladribine, (6) fludarabine, (7) bendamustine and (8) chlorambucil. Bendamustine fared worse in our analysis than in its pivotal trial. Overall, the estimated rankings appeared to be robust according to probabilistic analysis. Numerous indirect comparisons were assessed in the absence of RCTs. In conclusion, we generated an updated synthesis of the effectiveness of these treatments and we ranked them according to a Bayesian probabilistic model. In our probabilistic analysis, cyclophosphamide + fludarabine + rituximab ranked first in the base case while the worst-case scenario of this analysis placed this treatment at a remarkable second place.