In may of last year, our original article “Glucose Variability Assessed with Continuous Glucose Monitoring: Reliability, Reference Values, and Correlations with Established Glycemic Indices―The Maastricht Study” by Y.D. Foreman, M.C.G.J. Brouwers, C.J.H. van der Kallen, D.M.E. Pagen, M.M.J. van Greevenbroek, R.M.A. Henry, A. Koster, A. Wesselius, N.C. Schaper, and C.D.A. Stehouwer was published in Diabetes Technology & Therapeutics 2020;22(5):395 − 403. In this article, we assessed the minimum number of days needed to reliably measure daily glucose variability (GV) with continuous glucose monitoring (CGM), assessed GV reference values, and studied the correlation of GV with established glycemic indices. Unfortunately, we have recently discovered that the reliability analyses were incorrectly performed. We want to sincerely apologize for this mistake and wish to formally rectify it. Below, we explain our error and its impact on our original conclusions.
To determine the minimum number of days required for a reliable assessment of several CGM-derived indices, we calculated reliability (R)―per glucose metabolism status category―using the Spearman-Brown prophecy formula. This formula uses the variance between individuals and the variance within an individual. We employed variance component analysis to quantify these variances. However, we made a mistake when applying them to the Spearman-Brown prophecy formula. In the article, we described the following procedure in the Methods section: “R = σB2/(σB2 + [σW2/n]), where σB is the interindividual variance, σW is the intraindividual variance, and n is the number of consecutive monitoring days. A R of >0.80 was considered reliable”. The correct formula should be (differences in bold): R = σ2B / (σ2B + [σ2W / n]), where σ2B is the interindividual variance, σ2W is the intraindividual variance, and n is the number of consecutive monitoring days. Thus, we made the mistake of regarding σ as the ‘quantity’ of variance, and, as a consequence, erroneously entered the squared variances into the formula.
We have now correctly applied the Spearman-Brown prophecy formula, and, as a consequence, had to somewhat alter our conclusions on reliability (see Addendum 1 for comparison): for mean sensor glucose, individuals with normal glucose metabolism (NGM) now require 3 days instead of 2; for standard deviation, individuals with prediabetes and type 2 diabetes now require 2 days instead of 1, and individuals with NGM now require 3 days instead of 2; and for coefficient of variation, participants with NGM, prediabetes, and type 2 diabetes now all require 3 days of CGM.
Accordingly, we have rewritten our manuscript to incorporate our corrections, in the Abstract-Results section, Statistical Analysis, Reliability of continuous glucose monitoring, Discussion and conclusion. Apart from the main text, we have corrected Figure 1, Supplementary Figures S2, S3 and S6, and Supplementary Table S11.
The corrections have been made to the online version.
Yours sincerely, on behalf of all authors,
Yuri Foreman
Maastricht, 24 February 2021
Addendum 1: Comparison of the original Figure 1 to the corrected Figure 1
Original
Corrected
Reliability for mean sensor glucose (A), standard deviation (B), and coefficient of variation (C) per number of consecutive continuous glucose monitoring days in individuals with normal glucose metabolism, prediabetes, or type 2 diabetes. The horizontal line represents the cut-off for sufficient reliability (R > 0.80). NGM, normal glucose metabolism; PreD, prediabetes; T2D, type 2 diabetes.