Estimating Long-Term Life Expectancy in Philadelphia Chromosome Positive (Ph+) Chronic Phase Chronic Myeloid Leukemia (CML-CP): Results of A Microsimulation Model

In ENESTnd, newly-diagnosed adult patients with Ph+ CML-CP were stratified by Sokal score (low [LS], intermediate [IS], and high [HS]) and randomized to frontline (FL) tyrosine kinase inhibitor (TKI) therapy with nilotinib (NI) 300 mg BID, NI 400 mg BID, or imatinib (IM) 400 mg QD. NI 300 mg BID demonstrated superior rates of molecular response and significantly fewer progressions to accelerated phase (AP)/blast crisis (BC) vs. IM. NI 300mg BID is the approved FL dose.

To use microsimulation to estimate the life expectancy of individual Ph+ CML-CP patients treated with FL IM or NI 300 mg BID based on ENESTnd.

Life expectancy of individual patients was predicted over a 60-year period beginning from FL therapy initiation. Simulated patients were tracked through the following (groups of) health states: i) FL in CP, ii) 2nd-line TKI therapy in CP or AP, iii) off TKI in CP, AP, or BC, iv) stem-cell transplant and post-transplant in CP, BC, or AP. Time spent in each state was dictated by transition probabilities that depended on each patient's FL treatment (IM or NI), Sokal score, and baseline age. Sokal score and baseline age are related (LS: 41.21 yrs; IS: 50.73 yrs, HS: 49.81 yrs), and inform non-CML mortality. FL treatment discontinuation probabilities by Sokal score and therapy were estimated by fitting parametric survival functions (e.g., Weibull) to ENESTnd (36 month minimum follow-up). Transition probabilities among the health states occurring after FL therapy discontinuation were estimated from published sources and population-based life tables. Modeled outcomes are reported by FL treatment, Sokal score, and in aggregate. In the base case, outcomes reflect a population matching the ENESTnd mean age at trial entry (47 years) and Sokal score distribution (37%, 36% and 28%, for LS, IS, and HS). Sensitivity analyses were conducted using the IRIS trial mean age at study entry (46 years) and Sokal score distribution (52%, 29%, and 19%) to compare results with previously reported survival estimates based on IRIS and to determine the impact of the higher proportion of high Sokal score patients in ENESTnd. Two additional scenarios were examined for comparison to other models: a) the absence of a 2^nd-line TKI treatment, and b) adjusting baseline age to 57.

Modeled life expectancy estimates are presented in the Table. Life expectancy in the base case scenario, with patient Sokal score distribution and baseline age from ENESTnd and use of 2^nd line TKI treatment, is presented in aggregate and by Sokal score. Alternate scenario life expectancy is presented in aggregate.

Patients treated with FL NI are predicted to live 3–8 years longer than patients treated with FL IM. Sokal score, baseline age, and availability of 2^nd-line TKI therapy are factors influencing long-term CML survival estimates. Presented life expectancy estimates are comparable to previous estimates derived with similar methodologies. Results observed in IRIS long-term follow up (86% overall survival at 8 years) and findings from an alternative model that relied on cytogenetic and molecular response to predict FL discontinuation both independently suggest actual life expectancy may be somewhat higher than reported herein. This microsimulation approach is unique in that it combines individual patient characteristics with clinical trial findings to estimate life expectancy within a flexible framework. This framework can be easily adapted to accommodate additional patient characteristics and treatment strategies and provide insight into the impact of treatment decisions on patient survival.

Snedecor:Novartis: Consultancy. Ji:Novartis: Consultancy. Magestro:Novartis: Employment, Equity Ownership. Stein:Novartis Pharmaceuticals: Employment. Botteman:Novartis: Consultancy.