Comparing different definitions of prediabetes with subsequent risk of diabetes: An individual participant data meta-analysis involving 76 513 individuals and 8208 cases of incident diabetes
journal contributionposted on 01.01.2019, 00:00 authored by C M Y Lee, S Colagiuri, M Woodward, E W Gregg, R Adams, F Azizi, R Gabriel, T K Gill, C Gonzalez, A Hodge, D R Jacobs, J J Joseph, D Khalili, D J Magliano, K Mehlig, R Milne, G Mishra, M Mongraw-Chaffin, Julie PascoJulie Pasco, M Sakurai, P J Schreiner, E Selvin, J E Shaw, G Wittert, H Yatsuya, Rachel HuxleyRachel Huxley
© © Author(s) (or their employer(s)) 2019. Re-use permitted under CC BY. Published by BMJ. Objective There are currently five widely used definition of prediabetes. We compared the ability of these to predict 5-year conversion to diabetes and investigated whether there were other cut-points identifying risk of progression to diabetes that may be more useful. Research design and methods We conducted an individual participant meta-analysis using longitudinal data included in the Obesity, Diabetes and Cardiovascular Disease Collaboration. Cox regression models were used to obtain study-specific HRs for incident diabetes associated with each prediabetes definition. Harrell's C-statistics were used to estimate how well each prediabetes definition discriminated 5-year risk of diabetes. Spline and receiver operating characteristic curve (ROC) analyses were used to identify alternative cut-points. Results Sixteen studies, with 76 513 participants and 8208 incident diabetes cases, were available. Compared with normoglycemia, current prediabetes definitions were associated with four to eight times higher diabetes risk (HRs (95% CIs): 3.78 (3.11 to 4.60) to 8.36 (4.88 to 14.33)) and all definitions discriminated 5-year diabetes risk with good accuracy (C-statistics 0.79-0.81). Cut-points identified through spline analysis were fasting plasma glucose (FPG) 5.1 mmol/L and glycated hemoglobin (HbA1c) 5.0% (31 mmol/mol) and cut-points identified through ROC analysis were FPG 5.6 mmol/L, 2-hour postload glucose 7.0 mmol/L and HbA1c 5.6% (38 mmol/mol). Conclusions In terms of identifying individuals at greatest risk of developing diabetes within 5 years, using prediabetes definitions that have lower values produced non-significant gain. Therefore, deciding which definition to use will ultimately depend on the goal for identifying individuals at risk of diabetes.