Refining KFRE With Mortality Risk Boosts Prediction Accuracy in Multimorbidity, CKD

New research suggests adjusting the kidney failure risk equation (KFRE) to account for competing mortality risks in patients with chronic kidney disease (CKD) can improve 5-year kidney failure prediction outcomes, particularly for those with multimorbidity.1

Findings from a 2-cohort external validation and recalibration study of the 4-variable KFRE highlight the clinical usefulness of an updated competing mortality risk model to increase individualized referral rates and improve subsequent treatment burden in patients with CKD, with and without multimorbidity.

“This is the first study that we are aware of validating KFRE in those with multiple long-term conditions, and that also considers the competing risk of mortality,” wrote study investigator Heather Walker, MBChB, a PhD candidate of cardiovascular and metabolic health at the University of Glasgow, and colleagues.

KFRE’s variables are age, sex, estimated glomerular filtration rate (eGFR), and urine albumin-to-creatinine ratio. Typically, eGFR is calculated using creatinine, which can be influenced by reduced muscle mass related to aging or multimorbidity. In contrast, as a biomarker, cystatin C is less affected by muscle mass. Prior investigation has questioned whether using a cystatin C-based eGRF can improve prediction in this patient population, but no updated KFRE models have been developed.2

To address this gap, investigators developed and validated an updated model using the same KFRE variables while accounting for competing mortality risks. They also compared performance using cystatin C-based eGFR (eGFRcys) or combined creatinine-cystatin-based eGFR (eGFRcr-cys) versus creatinine-based eGFR (eGFRcr), assessing discrimination, calibration, and overall fit at 2 and 5 years in both patient populations.

The observational cohort study included 24,489 individuals from the research-based UK Biobank, and 30,147 from the population-based Stockholm Creatinine Measurements project (SCREAM), identifying multimorbidity as ≥2 long-term conditions with CKD and kidney failure as long-term dialysis or kidney transplantation. In competing risk analyses, all-cause mortality was identified through linkage to national death registers.

In the UK Biobank and SCREAM cohorts, respectively, 61.2% and 70.3% of individuals had multimorbidity, the mean age was 62.8 and 70.1 years, while 54% and 66% of patients were female.

Upon analysis, investigators observed consistently strong discrimination for kidney failure across eGFRcr, eGFRcys, and eGFRcr–cys, with Area Under the Curve (AUC) and concordance index (C-index) ≥0.86 in individuals with and without multimorbidity. Adding the biomarker cystatin C did not improve discrimination or calibration. When calibration was assessed without accounting for the competing risk of mortality, investigators noted an underestimated kidney failure risk(Observed-to-expected ratio [O/E] ratio range, 1.01-2.18), and cystatin C did not enhance accuracy.

In competing risk analysis, investigators saw the 5-year cumulative incidence of kidney failure in multimorbidity cohorts was 3.0% (95% CI, 2.7-3.2) for SCREAM and 1.6% (95% CI, 1.4-1.8) for UK Biobank. In no multimorbidity cohorts, incidence was 3.6% (95% CI, 3.2-3.9) for SCREAM and 0.7% (95% CI, 0.6-0.9) for UK Biobank.

Mortality risk was higher in multimorbidity groups, with a 5-year cumulative incidence of death of 34.0% (95% CI, 33.4-34.6) in SCREAM and 7.2% (95% CI, 6.7-7.6) in UK Biobank. In comparison, no multimorbidity groups had lower mortality: SCREAM 11.1% (95% CI, 10.5-11.7) and UK Biobank 3.3% (95% CI, 3.0-3.7).

A model accounting for the competing risk of death improved performance compared to the UK-calibrated 5-year KFRE, with an O/E ratio of 0.98 at 5 years in multimorbidity groups. Calibration improvements were most notable in these groups, and overall model fit was enhanced.

Accounting for competing risk also influenced clinical decision-making. Applying a kidney failure risk threshold >5% increased total referrals in both cohorts compared to KFRE. Referred individuals had decreased multimorbidity, fewer deaths within 5 years, and were younger in SCREAM. Overall, the competing risk model captured more kidney failure events within 5 years than KFRE.

“KFRE has good discrimination accuracy to predict kidney failure in individuals with multimorbidity, but has less robust calibration and crucially does not account for the competing risk of death,1” investigators concluded. “A model accounting for competing mortality risk improves model performance, particularly calibration, in individuals with multimorbidity.

References

1. Walker H, Carrero JJ, Sullivan MK, et al. The kidney failure risk equation in people with CKD and multimorbidity: the effect of competing mortality risks. Nephrology Dialysis Transplantation. Published online November 24, 2025. doi:https://doi.org/10.1093/ndt/gfaf252

2. Benoit SW, Ciccia EA, Devarajan P. Cystatin C as a biomarker of chronic kidney disease: latest developments. Expert Review of Molecular Diagnostics. 2020;20(10):1019-1026. doi:https://doi.org/10.1080/14737159.2020.1768849

‌