To evaluate the performance, methodological quality, and translational readiness of prediction models for progression from diabetic kidney disease (DKD) to end-stage renal disease (ESRD), emphasizing their clinical applicability.
Key Findings:
Fifteen studies met inclusion criteria, all judged at high risk of bias, which raises concerns about the reliability of the findings.
Pooled AUCs were 0.896 for training models and 0.863 for validation models, indicating promising discrimination.
Sensitivity analyses yielded similar pooled AUCs, but results remained exploratory due to high heterogeneity, limiting their applicability.
Interpretation:
While DKD to ESRD models show promising discrimination, methodological limitations and lack of external validation hinder their clinical applicability, necessitating further research.
Limitations:
High risk of bias in all included studies, which may affect the validity of the results.
Limited applicability due to reliance on biopsy-proven DKD, which may not reflect the broader patient population.
Inconsistent reporting and assessment of predictors, complicating comparisons and generalizability.
Conclusion:
Future prospective multicenter studies with rigorous external validation and calibration are urgently needed to enhance the clinical utility of DKD to ESRD prediction models, addressing the identified methodological limitations.
