The Relationship of Response on Time to Next Treatment Based on Evidence from Two RCTs in Newly Diagnosed Stem Cell Transplantation Ineligible Multiple Myeloma Patients

INTRODUCTION The outcome of patients with Multiple Myeloma (MM) improved considerably over the last years due to an increase in availability of novel agents. However, the optimal sequence is largely unknown. Besides efficacy, determination of the optimal sequence is also of importance in light of cost effectiveness. To this end the development of a health economic (HE) model would be of interest. For such a model time to next treatment (TTNT) is required. We previously showed that in stem cell transplant (SCT) eligible patients TTNT was not treatment but response (i.e. complete, partial or no response; CR, PR or NR) dependent. We here investigate whether response was predictive for TTNT independent of the treatment regimen in SCT ineligible patients (pts) with MM.

METHODS We analyzed patient level data of two phase 3 randomized controlled trials (RCTs) performed by the Dutch-Belgian cooperative HOVON group in newly diagnosed, transplant-ineligible patients. The HOVON-87 (N=668) compared Melphalan Prednisone (MP) with Thalidomide followed by Thalidomide maintenance (MPT-T) versus MP with Lenalidomide followed by Lenalidomide maintenance (MPR-R) (Zweegman et al Blood 2016) (EudraCT number 2007-004007-34). The HOVON-49 (N=333) compared MP versus MPT (including Thalidomide maintenance) (Wijermans et al J Clin Oncol 2010) (ISRCTN 90692740). The following data were included: treatment, best response, TTNT and survival status. Pts were censored after last date of contact. Pts who did not have information on TTNT and/or response were excluded from our analyses.

Kaplan-Meier curves and Cox proportional hazards models were fitted on this data to investigate whether the TTNT is mainly response dependent and whether the quality of response is predictive for TTNT.

RESULTS We included 519 HOVON-87 pts and 252 HOVON-49 pts. First, we analyzed the TTNT hazard ratios (HRs) of the active and comparative arm within each response category for both trials.

The TTNT HR of the HOVON-87 pts with a CR (N=75) was not significantly different between treatments (HR 1.340, 95% CI: 0.590-3.039, p-val = 0.484). We also found no significant difference in PR pts (HR 0.952, 95% CI: 0.753-1.204, p-val = 0.681, N = 375) and NR pts (HR: 0.898, 95% CI: 0.518-1.556, p-val = 0.701, N=69).

The number of CR pts in the HOVON-49 study was too low (N=8) to allow for a reliable analysis of this response category. Therefore, for this study we present very good PR (VGPR) as separate category. For both VGPR (N=44 pts) and PR pts (N=95 pts) we did see a significant difference in TTNT in the treatment arms (HR: 0.363, 95% CI: 0.174-0.757, p-val = 0.007 and HR:0.634, 95% CI: 0.417-0.965, p-val = 0.033 respectively). However, for NR pts the TTNT HR was similar (HR: 1.131, 95% CI: 0.746-1.714, p-val = 0.561, N=105 pts)

Second, we compared the TTNT between the different response groups within each trial.

For HOVON-87, the median TTNT for CR pts was 63.4 months (mos) (95% CI: 51.8 mos - not evaluable), for PR pts it was 24.6 mos (95% CI: 22.8 - 27.3 mos), and for NR pts it was 19.5 mos (95% CI: 9.7 - 25.3 mos). The TTNT of CR pts was significantly longer than of PR pts (HR: 3.923, 95% CI: 2.601-5.917, p-val = 0.000). And the TTNT of PR pts was significantly longer than of NR pts (HR: 1.411, 95% CI: 1.052-1.893, p-val - 0.000).

For HOVON-49, the median TTNT for VGPR pts is 23.4 mos (95% CI: 15.0 - 26.7 mos), PR pts was 19.0 mos (95% CI: 15.4 - 21.5 mos), and for NR pts was 3.7 mos (95% CI: 3.4 - 4.3 mos). Only the TTNT of PR pts was significantly longer than the TTNT of NR pts (HR: 3.978, 95% CI: 2.938 - 5.388).

CONCLUSIONS Data from the HOVON-87 showed a relationship between response and TTNT. Pts in this trial achieving a CR were observed to have a significantly longer TTNT compared to those achieving PR at best. Furthermore, TTNT was not significantly different for the treatment arms (e.g. patients with CR after MPR-R had similar TTNT than patients with CR after MPT-T).

Establishing a relationship between response and TTNT was challenging among pts from the HOVON-49 since i) too few CR pts were seen in this trial and ii) the experimental arm included maintenance while the comparator treatment did not.

We can conclude that the previous established relationship between response and TTNT was partly confirmed for newly diagnosed SCT ineligible MM pts. Further research is necessary to identify other relevant predictors for TTNT and to confirm the current findings before we can incorporate this into our HE model.