Comparing Randomized Controlled Data with Grouped Data from the Literature to Inform Decision Making When Standard of Care Has Low Level of Evidence Available

Objectives: To apply indirect comparison methods to inform decision-making when standard of care has low level of evidence available.

Methods: Simulated treatment comparison is one of the methods available to generate indirect comparison of data in different formats and levels of evidence. It is particularly helpful when patient level data exists for one set of data and there is the need to compare it with single-arm studies. In theses cases evidence is deemed ''unanchored'' due to a lack of a common comparator. We use the case of second line treatment of chronic immune thrombocytopenic purpura (ITP). Splenectomy has been historically recommended in guidelines mainly due to the lack of evidence for other second-line therapies for ITP. We use individual patient data from eltrombopag clinical trials and one of the only studies published in the literature reporting long term outcomes for splenectomy to generate an indirect comparison of eltrombopag and splenectomy in second line treatment of ITP.

Results: A subset of patients (n=84) treated with eltrombopag but not subject to splenectomy in the randomised, phase 3 study RAISE (PMID:20739054) were indirectly compared with a large retrospective cohort of patients (n=233) who underwent splenectomy for ITP in the study by Vianelli et al (PMID:23144195). After controlling for gender, age, baseline platelet count and number of previous therapies, we estimate that the proportion of patients with ITP responding to second line therapy is higher with eltrombopag in comparison to splenectomy: complete response (platelet count >100x10⁹/L) 68.6% vs 54.0%; partial response 18.6% vs 24.6%; no response (platelet count <30x10⁹/L)) 12.8% vs 21.4%.

Conclusions: Indirect comparison with simulated treatment comparison is useful for decision making when no head to head data exists between new treatment options and standard of care and also when standard of care has low level of evidence available.