Utility of long-term systolic blood pressure variability for predicting the development of type 2 diabetes mellitus.

Better identification of individuals at high risk for type 2 diabetes mellitus (T2DM) requires risk-prediction models incorporating novel predictors. Accordingly, this study aimed to evaluate the merits of including long-term systolic blood pressure variability (SBPV) in predicting T2DM incidence in a Japanese cohort of 3017 participants (2446 men, 571 women; age, 36-65 years) in 2007, who were followed up until March 2019. Consecutive SBP values, recorded between 2003 and 2007, were regressed annually for each participant. The slope and root-mean-square error of the regression line were calculated for each individual to represent SBPV. The significance of SBPV was examined by adding it to a multivariate Cox model incorporating age, sex, smoking status, regular exercise, family history of diabetes, body mass index, blood levels of triglycerides, high-density lipoprotein cholesterol, and fasting blood glucose. The c-index, net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were used to compare the performance of the prediction models without (Model 1) and with (Model 2) SBPV. During the 9.8-year follow-up period, 135 participants developed T2DM. Although a statistically significant difference in c-index between Model 1 (0.785) and Model 2 (0.786) was not found, the NRI (8.312% [p < 0.001]) and IDI (0.700% [p = 0.012]) demonstrated that the performance of Model 2 improved compared with Model 1. In conclusion, results suggested that long-term SBPV slightly improved predictive utility for T2DM when added to a conventional prediction model. The study was registered at University Hospital Medical Information Network Clinical Trial registry (UMIN000052544, https://www.umin.ac.jp/).