ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005


The Latest on X

Kidney Week

Please note that you are viewing an archived section from 2023 and some content may be unavailable. To unlock all content for 2023, please visit the archives.

Abstract: FR-PO935

On the Accuracy of the Kidney Failure Risk Equation

Session Information

Category: CKD (Non-Dialysis)

  • 2301 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention


  • Ålund, Oskar, AstraZeneca PLC, Gothenburg, Sweden
  • Unwin, Robert J., AstraZeneca PLC, Cambridge, United Kingdom
  • Kalra, Philip A., Salford Royal Hospital, Salford, United Kingdom
  • Taal, Maarten W., University of Nottingham, Nottingham, United Kingdom
  • Soderberg, Magnus, AstraZeneca PLC, Gothenburg, Sweden

Group or Team Name

  • On Behalf of the NURTuRE Investigators.

The accuracy of Kidney Failure Risk Equation (KFRE) is based on the Receiver Operating Characteristic (ROC) curve, Area Under the ROC Curve (AUC), and Harrell's C-statistic. We examined the predictive power of the 4-variable KFRE described in [1] and considered the limitations of ROC curves and C-statistics.


We used data from the UK's NURTuRE CKD cohort with outcomes for 2426 CKD patients. 149 patients had reached ESKD at follow up. Model performance was assessed by:
* Harrell’s C-statistic
* Average precision
* Average negative predictive value


The C-statistic was 0.89 and AUC for 2-year discrimination 0.91, both corroborating [1]. The C-statistic on the subset of patients with known time to ESKD was 0.64. Average negative predictive value was 0.99. Average precision was 0.32. Excluding patients above age 40, average precision was 0.42.


The C-statistic is computed by counting patient pairs that the model orders correctly for time to ESKD. Most pairs include a patient who reached ESKD before follow-up and one who did not. Therefore, the C-statistic overestimates the KFREs ability to order patients for time to ESKD. For the subset of patients whose time to ESKD was known, the C-statistic was only 0.64. In other words, the KFRE cannot be used reliably to sort NURTuRE patients by time to ESKD.

The AUC is the probability that a randomly chosen positive patient would be assigned a larger risk score than a randomly chosen negative patient. Since most such pairs can be easily sorted for time to ESKD, this metric turns out well. In contrast, precision measures the probability that a patient will reach ESKD given that (s)he is predicted to by the KFRE. A precision of 0.32 in this case means that only 32% of patients predicted to reach ESKD within 2 years will do so. Interestingly we see an increase in precision in younger patients, which we believe is due to a higher correlation between eGFR and time to kidney failure in lower age brackets. In summary: With a negative predictive value of 0.99, the KFRE can rule out near term kidney failure with high probability but cannot confirm it. Therefore, while its application for referrals is appropriate, the KFRE should not be used as a basis for costly or risky medical intervention.

1. Tangri, Navdeep, et al. JAMA 305.15 (2011): 1553-1559.


  • Commercial Support – AstraZeneca