Cost‐effectiveness analysis based on the number‐needed‐to‐treat: common sense or non‐sense?

Abstract

This paper explores and critically discusses some of the methodological limitations of using the number-needed-to-treat (NNT) in economic evaluation. We argue that NNT may be a straightforward measure of benefit when the effect of an intervention is immediate, but that serious problems arise when the effect is delay rather than avoidance of an adverse event. In this case, NNT is not a robust or accurate measure of effect, but will vary considerably and inconsistently over time. This weakness will naturally spill over onto any CEA based on NNT. A literature review demonstrated that CEAs based on NNT were all published within the last five years, and that all studies suffered from important limitations. A major weakness of using NNT is the imposed restrictions on the outcome measure, which can only be strictly uni-dimensional and non-generic. Using NNT in economic evaluations is obtained at a cost in terms of both methodological shortcomings, and a reduced ability for such evaluations to serve as a useful tool in decision making processes. The use of NNT in economic evaluations might be better avoided. To every complicated question, there is a simple, straightforward, easy--and probably wrong answer (Occam's Sledgehammer).

Cost‐effectiveness of long‐acting insulin analogues vs intermediate/long‐acting human insulin for type 1 diabetes: A population‐based cohort followed over 10 years

“…However, due to a lack of representative utility data for Taiwanese patients with T1D, a cost‐utility analysis was not conducted. Fifth, including NNT in CEA studies may increase the understanding and relevance of CEA findings to clinical decision‐makers, but we acknowledge its limitations . For instance, our study used NNT to quantify the treatment effectiveness as a function of the difference in the probability of developing an outcome event between two treatment groups, which can only measure one type of benefit (eg, any diabetes‐related complications in the present study) at one time.…”

From Clinical Trial Efficacy to Real-Life Effectiveness: Why Conventional Metrics do not Work

“…Several authors have warned against the sensitivity of NNT to factors that change baseline risk, e.g. patients’ characteristics, secular trends in incidence and case fatality and delay to event [38, 39]. The value of NNT is not the same if the treatment effect is immediate or if the effect is to delay an outcome rather than prevent it [40].…”

On the clinical meaningfulness of a treatment's effect on a time‐to‐event variable

“…Some authors have criticized the use of NNT specifically in the setting of a time‐to‐event variable (e.g. Kristiansen and Gyrd‐Hansen 9), but it is not clear why an absolute measure of benefit would be of little value for a time‐to‐event variable, but great value for a binary variable. This and other questions are left for future evaluations.…”