CKD Prediction Models Are “In Their Infancy”

Although much more work is needed in development and clinical application, risk models to predict chronic kidney disease (CKD) and its progression show “acceptable” discrimination, concludes a systematic review in PLOS Medicine.

The investigators performed a critical assessment of CKD risk models. A literature search identified 26 publications reporting on the development, validation, or impact assessment of models to predict the risk of CKD occurrence or progression. Discrimination, recalibration, and reclassification performance were assessed, along with validation and impact assessment.

In derivation samples, most of the CKD risk models showed acceptable to good discriminatory performance, with area under the receiver operating characteristics curve values greater than 0.70. Calibration was generally acceptable, although less frequently evaluated. External validation was performed for only eight out of 30 occurrence models and five out of 17 progression models, with modest-to-acceptable discrimination.

The studies provided little information on the predictive value of newer circulatory or genetic CKD biomarkers, or on the clinical impact of the prediction models. In addition to a lack of validation studies, the derivation samples were limited by a lack of ethnic diversity. Limitations of the review included the lack of a consensus approach to rating prediction models and the difficulty of assessing publication bias.

Risk assessment of CKD has important implications for prevention and early detection. Although risk factors for CKD development and progression have been identified, their value in CKD risk stratification through clinical prediction models has yet to be established.

“These findings suggest that the development and clinical application of CKD risk models is still in its infancy,” the investigators conclude. Although published models show acceptable discriminatory performance, their value in clinical practice remains to be demonstrated. More work on calibration and external validation is needed before the models are incorporated into clinical guidelines [Echouffo-Tcheugui JB, Kengne AP. Risk models to predict chronic kidney disease and its progression: a systematic review. PLOS Med 2012; 9(11):e1001344; doi:10.1371/journal.pmed.1001344].

February 2013 (Vol. 5, Number 2)