OBJECTIVES: Central limit theorem states that with a sufficiently large sample size, it can be assumed that the sample mean is normally distributed regardless of the distribution of the initial population. Despite this, in cohort-level economic models, costs are often varied in probabilistic sensitivity analysis using the generalised gamma distribution. A targeted literature review was conducted to see whether this was being considered in National Institute for Health and Care Excellence (NICE) submissions.

METHODS: All the NICE health technology appraisal (HTA) committee papers published in the three months prior to June 2019 were reviewed, with a focus on the distribution reported to test the uncertainty costs in the respective economic model.

RESULTS: Within this time period 14 HTA committee papers were published, of these, 8 were for cohort-level economic models that reported the use of the generalised gamma distribution to test the uncertainty of costs in the conducted probabilistic sensitivity analysis. This approach was not something that was questioned by any of the evidence review groups associated with each submission.

CONCLUSIONS: The assumption of normality is regularly overlooked when producing probabilistic sensitivity analysis for cohort-level health economic models, with the generalised gamma distribution used to test the uncertainty around mean cohort costs with no clear justification. Although this may be appropriate in some circumstances, in cases where the sample size of values informing the mean cost is large enough to allow the use of the central limit theorem, it may be more appropriate to assume costs are normally distributed for probabilistic sensitivity analysis.