Time to Next Treatment Is Response Dependent and Treatment Independent Based on Evidence from RCTs in Stem Cell Transplantation Eligible Multiple Myeloma Patients

INTRODUCTION Multiple Myeloma (MM) is an incurable disease and its treatment is characterized by treatment sequences (treatment lines). We questioned whether a health economic (HE) model could be developed that analyses treatment sequences (TS) rather than a particular line of treatment in transplant eligible patients with MM.

In this model we hypothesized that the time to next treatment (TTNT) is response dependent (i.e. complete, partial or no response; CR, PR or NR) rather than primarily dependent on the choice of specific regimens.

METHODS We have analyzed the patient level data of the following major phase III RCTs in newly diagnosed SCT eligible MM patients: HOVON-50 (N=536): TAD vs VAD, HDM/ASCT, Thalidomide vs Interferon maintenance (Lokhorst et al Blood 2010), HOVON-65/GMMG-HD4 (N=827): PAD vs VAD, HDM/ASCT, Bortezomib vs Thalidomide maintenance (Sonneveld et al J Clin Oncol 2012) and IFM 2005-01 (N=478): Bortezomib/Dexamethasone vs VAD induction (Harousseau et al J Clin Oncol 2010). The following data were included: treatment arm, best response, TTNT and survival status. Patients were censored after last date of contact. Patients who did not have any information on TTNT 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.

Data from PETHEMA/GEM05 (Rosiñol et al Blood 2012) and MRC Myeloma IX (Morgan et al Haematologica 2012) are currently being prepared to be included in our analyses.

RESULTS The hazard ratio (HR) of CR patients (pts) in the experimental treatment arm versus CR pts in the comparator arm is 0.664 (95% Confidence Interval: 0.418-1.055, p-value = 0.083) for HOVON-50 (N=134 pts), 0.922 (95% CI: 0.688-1.235, p-value = 0.586) for HOVON-65 (N=329 pts), and 2.512 (95% CI: 0.522-12.094, p-value = 0.251) for IFM 2005-01 (N=202 pts). The HR of PR pts in the active arm versus PR pts in the comparator arm is 0.777 (95% CI: 0.596-1.014, p-value = 0.063) for HOVON-50 (N=254 pts), 0.860 (95% CI: 0.674-1.098, p-value = 0.226) for HOVON-65 (N=337 pts), and 0.393 (95% CI: 0.128-1.207, p-value = 0.103) for IFM 2005-01 (N=175 pts). The HR of NR pts in the active arm versus NR pts in the comparator arm is 1.303 (95% CI: 0.754-2.250, p-value = 0.343) for HOVON-50 (N=62 pts) and 0.760 (95% CI: 0.402-1.436, p-value = 0.398) for HOVON-65 (N=67 pts). The number of NR patients in the IFM study was too low (N=15) to show any sensible results on the HR.

The HR of CR pts versus PR pts is 3.012 (95% CI: 2.300-3.945, p-value = 0.000) in HOVON-50, 1.935 (95% CI: 1.600-2.341, p-value = 0.000) in HOVON-65, and 2.269 (95% CI: 1.012-5.091, p-value = 0.047) in IFM 2005-01.

The comparator arms’ TTNT (i.e. VAD) of the HOVON trials were tested similar given the result of the log rank test (p-value = 0.5165 for CR pts, and p-value = 0.0579 for PR pts). For the median TTNT and corresponding confidence intervals data of the comparator arms from both HOVON studies were used. The median TTNT for CR pts is 58 months (95% CI: 51 – 70 months), for PR pts is 29 months (95% CI: 27 – 32 months), and for NR pts is 8 months (95% CI: 6 – 13 months). Even though the comparator arm of the IFM 2005-01 study was similar to the comparator arms of the HOVON studies, we did not include this data in the calculation of the median TTNT mainly because the median was not reached in the IFM trial.

The number of excluded pts was low (86, 94 and 86 pts in HOVON-50, HOVON-65 and IFM 2005-01 respectively) and scattered evenly across the response categories and treatment arms, implying no bias was introduced when excluding these patients from our data set.

CONCLUSIONS In our analyses we have discovered that in the pivotal phase III RCTs with Thalidomide and/or Bortezomib in SCT eligible MM patients, TTNT is response dependent and not treatment dependent. Furthermore, patients achieving CR were observed to have a significantly longer TTNT compared to those achieving PR at best.