Optimal Treatment Sequencing In Multiple Myeloma: An Exploratory Modeling Approach

The introduction of the novel agents thalidomide, lenalidomide and bortezomib has created a perspective of chronic, intermittent treatment of patients with multiple myeloma (MM). The important question today is how and in which sequence these treatments should be administered in order to achieve the best outcome. This question cannot be addressed by means of a randomized clinical trial (RCT), due to the number of trial arms, patients and years needed. There is therefore a need for an evidence-based methodology to address this question. Finding a robust conclusion is also relevant to balance costs and efficacy of treatments by taking a longer perspective.

The goal of this study was to establish an analytic framework aimed at comparing the outcomes of treatment sequences for MM by incorporating the results of published studies into a single coherent unifying Markov model. We focused on treatment strategies for patients not eligible for transplantation. The ultimate aim was to predict the overall survival (OS) of each treatment sequence, by expressing the additive effect on OS from single treatments at presentation or at relapse.

The basic structure of the model contains three lines of treatment and one phase for “later lines of treatment”. Within each of the three treatment lines, patients could either enter, stay in or move to complete response, partial response or no response states. In order to populate the model, this meta-modeling study consisted of the following steps. Firstly, we performed a systematic literature review (SLR), extracting outcome measures that could serve as inputs for the model (cut-off date November 2009) and analyzing them by traditional meta-analysis (fixed effect modeling (FEM)). We then performed a first meta-regression identifying the response for each treatment in each treatment line. The next meta-regression established a model framework based on the following mechanism: response to treatment from meta-regression 1 is predictive for time to progression (TTP). TTP in its turn could be translated into time to next treatment (TTNT). A relationship was established between complete response (CR), partial response (PR) and no response (NR) and OS. Finally, the synthesized data were incorporated in the Markov sequencing model that estimates overall survival for the considered treatment sequences. In total the model can predict results for 245 different treatment sequences. Outcomes are expressed as mean OS, the mean response rates of the individual treatment combinations per line of treatment and corresponding TTNT estimates. Uncertainty in outcomes was addressed in sensitivity analyses.

The SLR provided 57 relevant clinical studies with 84 treatment arms. The meta-analysis on response showed that in first line MPV (melphalan/prednisone + bortezomib) showed the highest CR (33%), followed by MP + lenalidomide (17%), MP + thalidomide (10%) and MP (3%). Internal validation showed consistency with the results from the meta-analysis. External validation showed consistency with results presented after the SLR cut-off date, like the MM-015 trial, e.g. 16% complete response (CR) predicted compared to 18% observed CR for MP/lenalidomide and 3% CR predicted vs. 5% CR observed for the MP arm. In the second meta-analysis, a linear pattern established for the relationship between overall response and TTP. Following extrapolation from TTP to TTNT, external validation versus data from the VISTA trial showed predicted TTNT for MPV to 29.29 month compared to the observed 28.1 month and predicted 19.14 month for MP compared to the observed 19.2. Finally, the exploratory model showed that mean survival results for the sequences starting with MP, MPT, MPR and MPV ranged from 3.86–4.50, 4.72–5.09, 5.07 to 5.05–5.65 years respectively. The survival of sequences starting with one of the novel agents in combination with MP was consistently and significant better than sequences starting with MP alone as the confidence intervals did not overlap.

The numerical result of this exploratory analysis indicates that starting with one of the novel agents in combination with MP increases survival compared to starting with MP alone. We were able to develop a methodological framework in which we can evaluate the additive effect of single treatments on the overall OS as a result of the treatment sequence as well as the TTNT per treatment line.

Heeg:Pharmerit: Consultancy. van Agthoven:Janssen-Cilag BV: Employment, Equity Ownership. Liwing:Janssen-Cilag AB: Employment, Equity Ownership. Off Label Use: All drugs are approved for the treatment of multiple myeloma. However, non-approved combinations of drugs or use of drugs in non-approved lines of therapy are included in the sequencing model. The data used for the model regarding such combinations and lines of therapy have been taken from published clinical trials. Mellqvist:Jansen-Cilag: Honoraria; Celgene: Honoraria. Logman:Pharmerit: Employment. Donatz:Janssen-Cilag GmbH: Employment. Aschan:Janssen-Cilag AB: Employment. Kropff:ORTHO BIOTECH: Honoraria; Celgene: Honoraria. Treur:Pharmerit: Consultancy. Barendse:Janssen-Cilag AB: Employment. Harousseau:Janssen-Cilag: Advisory Board, Honoraria; Cellgene: Advisory Board, Honoraria. Palumbo:CELGENE: Honoraria, Membership on an entity's Board of Directors or advisory committees; JANSSEN-CILAG: Honoraria, Membership on an entity's Board of Directors or advisory committees. Sonneveld:Ortho-Biotech: Membership on an entity's Board of Directors or advisory committees; Millennium: Membership on an entity's Board of Directors or advisory committees; Celgene: Membership on an entity's Board of Directors or advisory committees.