Rapid Responses to:

PRIMARY CARE: Peter Bower, Sarah Byford, Julie Barber, Jennifer Beecham, Sharon Simpson, Karin Friedli, Roslyn Corney, Michael King, and Ian Harvey Meta-analysis of data on costs from trials of counselling in primary care: using individual patient data to overcome sample size limitations in economic analyses BMJ 2003; 326: 1247-1250 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Validity of post-hoc meta-analysis of secondary cost data from medical research studies

M Jawad Hashim   (13 June 2003)

Validity of post-hoc meta-analysis of secondary cost data from medical research studies 13 June 2003
M Jawad Hashim, Attending Physician Port Orchard WA 98366 Send response to journal: Re: Validity of post-hoc meta-analysis of secondary cost data from medical research studies	Dear Editor, Bower and colleagues(1) in their meta-analysis show that pooling patient data from several trials may yield additional information regarding cost-effectiveness that may not be detected in individual studies. I am concerned about the validity of this approach due to the multiple assumptions involved and the heterogeneity of the data pooled together. Although the authors allude to this limitation, the soundness of this approach has been questioned before in medical research.(2, 3) The problems include variability in cost measurement, missing data, patient population differences, and time differences. A review(2) of published randomized controlled trials showed that only 0.2% of the studies had cost analysis included. This may indicate the lack of cost as primary outcome consideration in most of these studies. The risk of inferring from data that were not primary outcome measure prior to data collection is well known. Even those trials that did include economic review suffered from shortcomings including lack of sensitivity analysis and inappropriate portioning of overhead costs. Precision may be lacking in summary measures obtained from cost meta-analysis due to the heterogeneity of individual studies included.(4) While the type 2 errors may occur in small randomized controlled trials, the reporting of a difference when none exists (type 1 error) is worse in meta-analyses especially when multiple assumptions can be manipulated to yield a desired conclusion. Sincerely, M Jawad Hashim MBBS Port Orchard Washington 1. Bower P, Byford S, Barber J, Beecham J, Simpson S, Friedli K, et al. Meta-analysis of data on costs from trials of counselling in primary care: using individual patient data to overcome sample size limitations in economic analyses. Bmj 2003;326(7401):1247-50. 2. Adams ME, McCall NT, Gray DT, Orza MJ, Chalmers TC. Economic analysis in randomized control trials. Med Care 1992;30(3):231-43. 3. Lyman GH. Economic analysis of randomized, controlled trials. Curr Oncol Rep 2001;3(5):396-403. 4. Saint S, Veenstra DL, Sullivan SD. The use of meta-analysis in cost-effectiveness analysis. Issues and recommendations. Pharmacoeconomics 1999;15(1):1-8. Competing interests:   None declared