Relationship between modeling technique and reported outcomes: case studies in models for the treatment of schizophrenia

Abstract

Complex modeling techniques such as discrete-event simulation and simpler Markov or decision-tree models have been used to estimate the cost-effectiveness of treatment for chronic diseases such as schizophrenia. A systematic literature review of MEDLINE, EconLit, Embase, and the Cochrane Library identified schizophrenia modeling studies presenting incremental cost-effectiveness ratios. The relationship between modeling technique used and reported outcomes was examined. Fifty-four studies reporting results of 69 pairs of drug comparisons were identified. Of the paired-drug comparisons, 27 were conducted in at least two studies; in 14 of the 27, the results agreed (i.e., drug A cost-effective compared with drug B) despite differences in modeling techniques. Thirteen of the 27 paired-drug comparisons had contradictory study results even when the same modeling technique was used. Different modeling techniques did not appear to explain different findings about cost-effectiveness.

Keywords::

• antipsychotic agents
• economic
• models
• review
• schizophrenia

Financial & competing interests disclosure

This study was funded by F Hoffmann-La Roche. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. Third-party medical editing support was provided by ApotheCom.

Key issues

• The majority (14 of 27 studies) of cost–effectiveness results in schizophrenia comparing the same pair of drugs gave consistent results despite using different modeling techniques (Markov or decision-tree model or discrete-event or Monte Carlo simulation).

• In 7 of the 13 paired-drug comparisons where inconsistent results were reported, inconsistent results were obtained using the same modeling technique.

• In the two most commonly modeled paired-drug comparisons with contradictory results, the following factors differed:

  • – The comprehensiveness of the model in its inclusion of the elements of efficacy (response, relapse), adherence and discontinuation rates and adverse events;

  • – The clinical efficacy and safety parameter values and the strength of the sources providing these data, for example, meta-analysis and randomized clinical trial data versus observational data and expert opinion;

  • – The utility weights associated with the different health states in the model.

• Differences in the disease and treatment events included in the model, as well as differences in the input parameter values, may be more important than modeling technique in determining the results of a cost–effectiveness analysis.

Notes