OBJECTIVES We have developed a model to evaluate type-2 diabetes prevention interventions. We aimed to validate this model against external data to test the accuracy of model predictions. METHODS An individual patient simulation was developed to predict longitudinal trajectories of HbA1c, 2-hr glucose, FPG, BMI, systolic blood pressure, total cholesterol and HDL cholesterol based on statistical analyses of the Whitehall II longitudinal cohort. Criteria for diabetes diagnosis were flexibly specified. Cardiovascular events were estimated from the QRISK2 algorithm. Microvascular complications of diabetes were estimated from the UKPDS outcomes model. Several validations were performed to compare model outcomes with reported data from external sources. We assessed the predicted diabetes incidence using data from the EPIC Norfolk cohort. Data from the Health Survey for England (HSE) 2003 cohort was simulated for eight years to compare predicted disease incidence and metabolic distributions with HSE 2011 data. We compared microvascular, cardiovascular and mortality outcomes in a diabetic population with those observed in the UKPDS. We assessed the performance of the model in predicting the results of the ADDITION trial for diabetes screening. RESULTS We found that the model overestimated three-year incidence of diabetes, particularly in high risk (HbA1c>6.0) individuals, but underestimated diabetes incidence in medium risk individuals (HbA1c 5.5-5.9) compared with the EPIC-Norfolk data. Predictions from HSE 2003 were fairly accurate. Predictions for microvascular events were similar to the UKPDS, but cardiovascular disease and mortality were slightly under-predicted. The model replicated the non-significant difference seen between control and intervention arms of the ADDITION trial, but overestimated total mortality and cardiovascular disease. CONCLUSIONS The SPHR Diabetes model appears to be fairly accurate at predicting external data, but has a tendency to overestimate mortality rates in a newly diagnosed diabetic cohort, and underestimate cardiovascular disease and mortality compared with the UKPDS.