Predicting the development of diabetes in older adults: the derivation and validation of a prediction rule.

Abstract

To create a simple prediction rule that could perform as well as the 2-h postchallenge plasma glucose (PCPG) test to predict those at risk for diabetes. We created a prediction rule in one sample and prospectively validated it for incident diabetes in a separate cohort. A cross-sectional analysis with data from the Rancho Bernardo Study (age 67 +/- 11 years) to derive a rule predicting abnormal PCPG >/=140 mg/dl, using demographic, clinical, and laboratory data of nondiabetic participants with fasting plasma glucose (FPG) <126 mg/dl. Data from the Health, Aging and Body Composition study (age 74 +/- 3 years) were used to prospectively validate this rule for incident diabetes and compare it with the predictive ability of the PCPG test. Of 1,549 RBS participants, 514 (33%) had PCPG >/=140 mg/dl. Female sex, age, triglycerides, and FPG were most significantly associated with abnormal PCPG. Based on standardized beta-coefficients, we allotted 1 point for female sex, triglycerides >/=150 mg/dl, or FPG 95-104 mg/dl. Age >/=70 years or FPG 105-115 mg/dl were given 2 points, and FPG 116-125 mg/dl received 3 points. In the validation cohort, this simple prediction rule was as good as the 2-h PCPG test for predicting incident diabetes (C-statistic: 0.71 for both). Advanced age, female sex, FPG, and triglycerides were able to predict adults at risk for diabetes equally well as the 2-h PCPG test. Using this rule, clinicians may better identify older persons who should receive intensive lifestyle intervention to prevent type 2 diabetes.