External Validation of the International IgA Nephropathy Prediction Tool in a Colombian Hispanic Cohort: Insights...

To externally validate the International IgA Nephropathy Prediction Tool (IIgANN-PT) in a Colombian Hispanic cohort with biopsy-proven IgA nephropathy (IgAN).

We conducted a retrospective cohort study of 397 Colombian Hispanic adults with primary IgAN biopsied between 2004 and 2022 and followed through 1 April 2024. The race-unadjusted IIgANN-PT was applied at biopsy to predict the composite outcome of sustained 50% decline in estimated glomerular filtration rate (eGFR) or end-stage kidney disease (ESKD). Model performance was evaluated at the pre-specified 5-year horizon using time-dependent AUC and Harrell's C-statistic for discrimination, inverse-probability-of-censoring-weighted (IPCW) Brier score for overall prediction error and calibration-in-the-large (CITL) plus calibration slope (βLP) for calibration. Predicted and observed 5-year risks were also compared across the tool's four pre-specified linear predictor risk strata.

Over a median follow-up of 4.7 years, 132/397 (33.3%) patients reached the composite outcome. In the complete-case analysis (n = 234), discrimination at 5 years was excellent (AUC 0.94 [95% CI 0.89-0.97]; Harrell's C 0.88), with low overall prediction error (IPCW Brier score 0.103). Calibration was accurate on average (CITL 0.03), but the calibration slope was > 1 (βLP 1.40), indicating underprediction concentrated at the extremes; the largest underestimation occurred in the very-high-risk stratum (mean predicted ~67% vs. observed ~80%-85%).

In this Colombian Hispanic IgAN cohort, the IIgANN-PT retained excellent 5-year discrimination and good average calibration, but underestimated risk in the highest-risk patients, supporting its use for risk stratification while emphasising the need for calibration assessment (and potential local recalibration) before broader implementation.