Lack Of Standardized Methods For Handling Multiples Sources For Model Parameters: A Systematic Review

• Published in: Value in health 2017-10, Vol.20 (9), p.A857
• Main Authors: Ribeiro, RA, Bertoldi, EG, Leal, LF, Stella, SF
• Format: Article
• Language: English
• Subjects: Analysis; Cardiology; Cost analysis; Economics; Health care expenditures; Health economics; Medical treatment; Meta-analysis; Oncology; Pharmacology; Sensitivity analysis; Systematic review
• ISSN: 1098-3015; 1524-4733
• DOI: 10.1016/j.jval.2017.08.2456

OBJECTIVES: Guidelines for cost-effectiveness models (CEM) usually recommend performing systematic review and meta-analysis for the treatment effectiveness estimate. However, CEM are populated with many other parameters, and guidelines do not usually provide guidance regarding how the review should be conducted.There is also no consensus for the optimal approach when multiple parameter sources (MPS) are encountered, particularly regarding pooling of results. We reviewed three health economics journals (MDM, Value in Health and Pharmacoeconomics) to evaluate MPS usage frequency and how these data were handled. METHODS: Systematic review of published CEM from 2013-2014 by pairs of two independent reviewers. We evaluated MPS frequency, summarization and use in sensitivity analysis. RESULTS: 69 articles were included. Most common areas were oncology (25%), infectology (12%) and cardiology (10%). In 61% of the articles, it was clear that MPS were used for model parameters aside the treatment effectiveness, in most cases for estimating probabilities (95%), and less often for utilities (40%) and costs (24%). In only 29% of the studies a formal systematic review was used. In studies with MPS, only 45% indicated some form of pooling; among such studies, methods for pooling was unclearly described in 53%. Only six articles used the term "meta-analysis", and only four clearly stated that a single-group meta-analysis was undertaken. Regarding sensitivity analysis, in 52% of the studies the MPS found were apparently not used. MPS were clearly used in 12 studies (either highest/lowest values from the literature, or simply a second analysis with a different source/estimate), and in an unclear fashion in five studies. In only three studies the 95% CI of single-group meta-analysis was used. CONCLUSIONS: We found that many CEM use MPS for populating their models, but proper methods for pooling values using meta-analysis are rarely used, and even more rarely properly reported, especially when single-group meta-analysis would be required.