A Markov Model of Phase I Clinical Trials before or after Autologous Stem Cell Transplantation for Relapsed Multiple Myeloma

Background: High-dose chemotherapy followed by autologous stem cell transplantation (ASCT), performed either at the time of initial diagnosis or at relapse, is considered the standard of care for younger patients (<70 years of age) with multiple myeloma (MM). Currently, the optimal timing of ASCT, i.e., early transplantation versus transplant upon relapse, is under investigation (clinical trial NCT01208662). In addition, there has been an unprecedented pace of anti-myeloma compound discovery tested through phase I trials (i.e., 500% increase) during the last decade, and its timing in relation to ASCT is not clear. Further, the current perception of low therapeutic benefit from participation in phase I trials is the main obstacle for patient recruitment and makes the phase I trial a “last resort” in the overall therapeutic plan (Meropol et al. 2003). Here, we present a landscape of therapeutic benefit and toxicity of all MM phase I trials over the past decade.  Also, in order to determine the optimal timing of ASCT and phase I trial we constructed a Markov model to examine two different approaches: early ASCT and subsequent phase I trial recruitment (early ASCT) vs. phase I trial recruitment before ASCT (late ASCT).

Methods: The primary decision examined in this study is the timing of ASCT in relation to a phase I trial. We systematically reviewed the outcome of all MM phase I trials between 2004-2014 to build a comparable cohort for decision analysis. Response rate, adverse effects and mortality from single agents, as well as the combination of the experimental agents with immunomodulators (IMiDs) and proteasome inhibitors (PIs) are reported. Two strategies were tested through Markov modeling: early ASCT vs. late ASCT (Figure-1). The model was built using the commercially available DecisionMaker software. An annual discount rate of 3% was used for all clinical outcomes. Quality of life (QOL) information for post-ASCT and phase I trials were extracted from similar research papers.

Results: Phase I systematic review: A total of 43 phase I clinical trials with 946 patients (530 males and 416 females) were included. The precipitants' median age was 60 years old. 21 trials tested single agents and the remainder were done using combination therapies (i.e., with an IMiD or PI). Median overall response rate (ORR) was 34% for all trials, 16% with single agents and 42% with combination therapy (Figure-2). 89% of trials were completed in less than a year. The therapeutic benefit of single vs. combination therapy phase I clinical trials does not show an increasing trend during the last decade (Figure-3). Patients who participated in combination therapy phase I trials had more grade III-IV toxicity than single agents (HR: 1.35, p-value: 0.04). There were only 4 patients (less than 1% of all participants) who experienced therapy-related death. Although, time from diagnosis was a predictor of response in univariate analysis, but it was not statistically correlated with response by multivariate analysis. Also, ORR was not affected by the number of prior therapies.

Markov model: The discounted life expectancies for two strategies were calculated; fixed-time intervals of 6, 12, 18 and 24 months were chosen for analysis. The gains in life expectancy were the same in patients undergoing combination therapy phase I trials in early or late ASCT strategies (6.76 and 6.64, respectively; p-value: 0.21). Importantly, for patients undergoing single agent phase I therapy, early ASCT was associated with a higher life expectancy than was the strategy of delayed ASCT (7.96 vs. 6.86, respectively; p-value: 0.036).

Conclusions: Taken together, our study indicates that phase I trials demonstrate a higher ORR than reported response rates from traditional chemotherapy phase I trials, i.e. 5%, Horstmann et al. 2005, even with single agents. We conclude that phase I participation should not be viewed as the “last resort”. Our decision-making analysis shows that combination therapy phase I trials can be offered irrespective of ASCT. However, single agent phase I trials should be offered after ASCT for transplant-eligible patients.

 

Figure-1: Markov decision model. A relapsed MM base case transitions after each one-month cycle to other health states. Patients could have remained in an alive state for any number of cycles without transitioning to another health state. The ASCT and Phase I states are transitory states.

 

Figure-2:

Figure-3: