Validation of the IHE type 2 diabetes cohort model in the Japanese clinical setting.

AIMS

Economic simulation models, such as the IHE Type 2 Diabetes Cohort Model (IHE-DCM-T2), are used widely to inform resource allocation for Type 2 Diabetes (T2D) treatments. Recently, IHE-DCM-T2 was augmented with Japanese-specific risk equations to align with the Japanese healthcare context. This study extends prior model validation of IHE-DCM-T2 to cover the Japanese risk equations for applications in Japan's clinical setting and healthcare system.

MATERIALS AND METHODS

Face validity was assessed through expert review of model assumptions and structure. Model programming was verified by code review and 728 stress tests. Predictive accuracy was tested by comparing model predictions to real-world outcomes from 28 Japanese studies, assessing concordance visually, with regression lines, and with mean absolute percentage error (MAPE), root mean square percentage error (RMSPE), mean squared logarithmic error (MSLE), and mean squared log-accuracy ratio (MSLAR). Subgroup analyses examined dependent and independent endpoints, along with mortality, microvascular, and macrovascular outcomes. Sensitivity analyses assessed robustness to variations in scale and sample size.

RESULTS

IHE-DCM-T2 demonstrated face validity and correct implementation. External validation against 120 endpoints showed good alignment between predicted and observed events, with regression line slope=0.96 and R=0.98. Overall, prediction errors were: MAPE=0.83, RMSPE=1.21, MSLE=0.61, and MSLAR=0.53. Predictions were more accurate for dependent than independent endpoints. Among endpoint categories, macrovascular events had the lowest average errors, whereas mortality endpoints had the highest MAPE and RMSPE, and microvascular endpoints had highest MSLE and MSLAR. Predictive accuracy was consistent across alternative test specifications.

LIMITATIONS

Limitations included gaps in validation data, and the requirement for long-term follow-up that inherently reflects past treatment patterns. Only studies with at least 1,000 patients were included, which may introduce selection bias.

CONCLUSIONS

This comprehensive validation of the IHE-DCM-T2, augmented with Japanese-specific risk equations, demonstrated its suitability for health technology assessments and resource allocation decisions for T2D in the Japanese clinical setting and healthcare system.