Objective:
Preventive services including screening for diabetes and its potential risk factorsare available to more Americans under Obamacare Preventive Care. Stratifying individuals by the predicted risk of developing type 2 diabetes mellitus (T2DM) would be useful for improving public health with efficient interventions. Although a number of T2DM prediction models have been reported, there is little evidence in East Asians, especially that from long-term follow-up studies. They are reported to have lower ability of innate insulin secretion and develop diabetes at much lower body mass index (BMI) than Caucasians and African Americans. Thus, this study aims to develop a point-based prediction model for 10-year risk of developing T2DM incidence in middle-aged Japanese men.
Method:
We followed 3,540 males in a worksite in Japan who were aged 35-64 years and free of diabetes in 2002 until March 31, 2015. Relationships of baseline age (continuous), BMI (<23, 23- <25 [reference category (Ref)], 25- <27.5, ≥27.5 kg/m
2
), current smoking status (yes, no [Ref]), alcohol consumption(0 [Ref], <23, 23- <46, ≥46 g/day), regular exercise of a moderate or higher intensity, an interval of ≥3 days per week, and a duration of ≥30 minutes per time (yes [Ref], no), medication use for dyslipidemia (yes, no [Ref]), family history of diabetes (having the first degree’s relatives with diabetes, not having [Ref]), serum triglycerides (<150 [Ref], ≥150 mg/dl), high density lipoprotein cholesterol(≥40 [Ref], <40 mg/dl), and fasting blood glucose (<100 [Ref], 100- <110, 110- <126 mg/dl) with incidence of T2DM were examined by Cox proportional hazard model. Variables significantly associated with T2DM (p<0.10) in the univariate models were simultaneously entered into a multivariate model, and backward variable selection procedure was done to determine the final multivariate model. Points were assigned for each predictor according to the method used in the Framingham Study.
Result:
During the median follow-up of 12.2 years, 342 males developed T2DM. The point-based model employing BMI, current smoking status, family history of diabetes, and blood levels of triglycerides and fasting blood glucose showed reasonable discrimination (c-statistics: 0.73) and goodness of fit (Hosmer-Lemeshow p=0.22).
Conclusion:
Our point-based prediction model showed applicability in terms of identifying middle-aged Japanese men at high risk of developing T2DM. The present findings warrants further investigations to determine whether using the point-based prediction models is effective to reduce T2DM incidence.