Noninferiority testing in cost-minimization studies: Practical issues concerning power analysis

• Published in: International journal of technology assessment in health care Vol. 22; no. 2; pp. 261 - 266
• Main Authors: Span, Mark M., TenVergert, Elisabeth M., van der Hilst, Christian S., Stolk, Ronald P.
• Format: Journal Article
• Language: English
• Published: New York, USA Cambridge University Press 01.04.2006
• Subjects: Confidence intervals; Cost Control; Cost-minimization; Design; Hypotheses; Investigations; Medical economics; Medical research; Medical technology; Models, Econometric; Noninferiority; Outcome Assessment (Health Care) - economics; Outcome Assessment (Health Care) - statistics & numerical data; RESEARCH REPORTS; Resource allocation; Sensitivity analysis; Sensitivity and Specificity; Studies; Success; Sensitivity analysis; Medical economics; Cost-minimization; Noninferiority
• ISSN: 0266-4623; 1471-6348
• DOI: 10.1017/S0266462306051099

Objectives: In cost-minimization studies, it is important to establish noninferiority in the clinical effect of the treatments under investigation. The relationship between the proportion of patients reaching the end point in a study, equivalence limit (δ), and power is investigated in the context of cost-minimization studies with dichotomous clinical end points. Two formulations of the null-hypothesis, absolute and relative formulations of δ, will be explored. Methods: Sensitivity analysis was performed, in which the effect of the predicted proportions and δ on the power in a noninferiority setting was investigated. The patterns found are discussed in terms of the practical relevance within the cost-minimization framework. Results: Sensitivity analyses show different patterns of results for both null-hypotheses. The differences in these results originate from the way δ is expressed. By expressing δ as absolute difference, power grows quite fast when sample proportions are smaller than expected. In the case of a proportional δ at small sample proportions, the power to establish noninferiority remains low. Conclusions: To obtain valid results from a cost-minimization study, care has to be taken to adapt the correct methodology for noninferiority testing in clinical outcomes. Defining δ in terms of absolute differences between treatments can lead to obscured results. Although conservative, the expression of δ as a proportion of the effectiveness of the treatment as usual is found to be closer to clinical practice. The inflated δ, resulting from smaller clinical effects than expected when absolute formulation is applied, thus can be avoided.