Underpowered or negative? A crucial distinction

Abstract

To the Editor: Schrauwen and colleagues have recently reviewed the emerging field of NAD metabolism in Diabetologia [1]. Their manuscript includes a description of a clinical trial of dietary nicotinamide riboside supplementation effects on whole body insulin sensitivity in obese men, conducted by us, together with our colleagues at Aarhus and Copenhagen Universities, Denmark [2]. As posted on clinicaltrials.gov (NCT02303483), this placebo-controlled, randomised clinical trial used M value determined by the hyperinsulinaemic–euglycaemic clamp (HEC) technique, the gold-standard approach for directly quantifying insulin sensitivity, as the primary endpoint. We report a negative outcome of the trial, and while this may be disappointing in terms of the possible future use of nicotinamide riboside supplementation for the prevention of type 2 diabetes, we believe it is important to publish such outcomes and to cite these correctly. We were therefore disconcerted to find the following description of our manuscript in the review by Schrauwen and colleagues: ‘The authors concluded that this study was underpowered and future studies should be larger and focus on other variables of metabolic health’. We find this to be a severe misrepresentation of a central conclusion of our work. As outlined in our manuscript, power calculations were conducted prior to enrolment into our study to determine an appropriate sample size for detecting clinically meaningful differences in the primary endpoint and to try to evade type I and type II errors resulting from insufficient sample size. We powered the study to detect a difference in the HEC-derived M value of 1.5 mg kg min, which has previously been achieved with insulin-sensitising drugs or exercise [3, 4]. We used a conservative estimate of the SD of 1.6 mg kg min in the power calculation, based on in-house experience from previous clamp examinations [2]. The primary endpoint, the M value, was not affected by nicotinamide riboside supplementation (p for interaction = 0.68). However, the actual SD for the difference between the baseline and the end-of-study M value within the placebo group was only 0.48 mg kg min. Therefore, in retrospect, the assumed SDwas probably too conservative.With an SD of 0.5 and n of 40, we would be able to detect a significant difference in theM value of 0.45 mg kg min with a power of 0.80. Hence, we do not suspect that the lack of effects of nicotinamide riboside can be attributed to the study being underpowered (i.e. a type II error). In addition to the M value, our study included several predefined secondary endpoints, and the combined outcomes of these were also negative. In line with standard study design, the trial was powered to detect significant differences in the primary endpoint and, therefore, where the risk of a type II error was discussed in our paper, this was in terms of the secondary endpoints [2]. When citing a clinical trial, it is important to respect the distinction between endpoints. When quoting the authors as * Niels Jessen [email protected]