Systematic assessment of decision-analytic models for chronic myeloid leukemia

Abstract

Background: Several tyrosine kinase inhibitors (TKIs) are approved for the treatment of chronic myeloid leukemia (CML). Decision-analytic modeling can help to extrapolate data from short-term clinical trials and also consider quality of life when evaluating different treatment strategies.

Objective: Our goal was to describe and analyze the structural and methodological approaches of published decision-analytic models for various treatment strategies in CML and to derive recommendations for the development of future CML models.

Data sources: We performed a systematic literature search in electronic databases (MEDLINE/PreMEDLINE, EconLit, EMBASE, NHS EED, and Tuft's CEA Registry) to identify published studies evaluating CML treatment strategies using mathematical models. The search was updated in August 2013.

Study selection: The models were required to compare different treatment strategies in relation to relevant clinical and patient-relevant health outcomes [e.g., life-years gained, quality-adjusted life-years] over a defined time horizon and population.

Study appraisal and synthesis methods: We used standardized forms for data extraction, description of study design, methodological framework, and data sources for each model.

Results: We identified 18 different decision-analytic modeling studies. Of these, 17 included economic evaluations. Modeling approaches included decision trees, Markov cohort models, state-transition models with individual (Monte Carlo) simulations, and mathematical equations. Analytic time horizons ranged from 2 years to a lifetime. Treatment strategies compared included bone marrow or stem cell transplantation, conventional chemotherapy, interferon-α, and TKIs. Only one model evaluated a second-generation TKI. Most models did not report a model validation. All models conducted deterministic sensitivity analyses and four reported a probabilistic sensitivity analysis.

Limitations: Articles that were not published in English or German were not included in this review. Our literature search was restricted to published full-text articles in certain databases. Therefore, publications that met our inclusion criteria but were published in different databases, different languages, or as abstracts only may have been missed.

Conclusions: While several well-designed models of CML treatment strategies exist, there remains a need for the assessment of the long-term efficacy and cost effectiveness of novel treatment options such as second-generation TKIs. Additionally, these models should be validated using independent data.