Disufenton sodium may continue to play a role in treating certain CLL patients due to lower toxicity profiles, convenience as an oral agent, and lower costs. The clinical value of these therapies could be evaluated in additional network meta analyses among different CLL patient populations with varying prognostic factors. Limitations This study has several limitations worth noting. First, a network meta analysis entails extrapolation from available data to indirectly make comparisons that have not actually been conducted. As a result, the validity of indirect estimates generated by a network meta analysis is uncertain. Although taurine requiring additional research, there is evidence to suggest that adjusted indirect comparisons usually agree with results of head to head randomized trials and the validity of the indirect estimates is dependent on the internal validity and homogeneity of the included trials.25 Second, the network meta analysis method is based on model simulation and as with all modeling methods, the results are dependent on data quality and the accuracy of model inputs.
Overall, the studies included in our treatment comparison HSP network received relatively lower Jadad quality scores because they were not double blinded and most did not report the method for randomization. Bias contained in the trials may inflate the estimates of relative and absolute treatment effects. Third, our comparison network was small with only five RCTs and did not assess all available therapeutic options such as bendamustine, pentostatin or cladribine based regimens due to limited published evidence. We conducted a literature search in the Medline database and The Cochrane Library, and also performed manual searches of reference citations. Though this methodology allowed us to conduct a broad search, this review may not have captured all relevant studies. In general, doxorubicin larger comparison networks would allow each study to,borrow more strength, from other studies and could improve the precision and certainty of indirect estimates.
Our comparison network included a three arm trial and our random effects model did not adjust for correlation induced by having more than two arms. Because our network only included one multi arm trial, any potential correlation was likely minimal. If a comparison network contains several multi arm trials, it is recommended that models are developed to account for the potential correlation within each trial.8 Finally, our study was limited to only one endpoint and does not provide information about adverse effects from therapy or therapy administration. Conclusion Among the five therapy options evaluated in this network meta analysis, our results suggest that FCR should be consideredan optimal initial treatment strategy for younger, healthier, treatment naïve CLL patients with low to intermediate risk disease. However, due to limitations of using model simulations, additional evaluations of FCR are necessary in order to clinically validate its therapeutic potential. Conflict of interest David Veenstra has served as a consultant to Genentech, Inc. No other authors declare potential conflicts of interest. Funding Mindy Cheng was supported by a Pharmaceutical Research and Manufacturers of America Foundation pre doctoral fellowship.