EstimatioN oF mORtality risk in type2 diabetiC patiEnts (ENFORCE): an inexpensive and parsimonious prediction model

CONTEXT: We previously developed and validated an inexpensive and parsimonious prediction model of 2-year all-cause mortality in real-life type 2 diabetic patients. OBJECTIVE: This model, now named ENFORCE, was now investigated in terms of i) prediction performance at 6 years, a more clinically useful time-horizon; ii) further validation in an independent sample; iii) performance comparison in real-life versus clinical trial setting. DESIGN: Observational prospective. Randomized clinical trial. SETTING: White patients with type 2 diabetes. PATIENTS: Gargano Mortality Study (GMS; n=1019), Foggia Mortality Study (FMS; n=1045), Pisa Mortality Study (PMS; n=972) as real-life samples and the standard glycemic arm of the ACCORD clinical trial (n=3150). MAIN OUTCOME MEASURE: The endpoint was all-cause mortality. Prediction accuracy and calibration were estimated to assess model's performances. RESULTS: ENFORCE yielded a 6-year mortality C-statistics of 0.79, 0.78 and 0.75 in GMS, FMS and PMS, respectively (P heterogeneity=0.71). Pooling the three cohorts, a 6-year mortality C-statistic of 0.80 was observed. In the ACCORD trial, ENFORCE achieved a C-statistic of 0.68, a value which is significantly lower than that obtained in the pooled real-life samples (P<0.0001). This difference resembles that observed with other models when comparing real-life vs. clinical trial settings, thus suggesting it is a true, replicable phenomenon. CONCLUSIONS: Time horizon of ENFORCE has been extended to 6 years and validated in three independent samples. ENFORCE is a free (http://www.operapadrepio.it/enforce/enforce.php) and user-friendly risk calculator of all-cause mortality in White type 2 diabetic patients from real-life setting.