The Kidney Failure Risk Equation may help predict the risk of end-stage renal disease in children with chronic kidney disease
EBM Focus - Volume 13, Issue 4
- Models that estimate the risk of end-stage renal disease (ESRD) in patients with chronic kidney disease can help guide management, but such models have not been evaluated in children.
- A retrospective cohort study evaluated the Kidney Failure Risk Equation (KFRE, derived and validated in adults) in 603 children with eGFR < 60 mL/minute/1.73 m2.
- The median KFRE risk estimates for low-, moderate-, and high-risk tertiles closely matched the observed rates of ESRD within 2 years: 1% estimated vs. 0.5% observed for low-risk, 5% vs. 3.5% for moderate-risk, and 33% vs. 29% for high-risk.
Management of chronic kidney disease (CKD) in adults can be informed by models that estimate the risk of end-stage renal disease (ESRD) based on demographic and laboratory-derived factors associated with disease progression, but the applicability of these models to children is not known. A recent retrospective cohort study evaluated the Kidney Failure Risk Equation (KFRE, derived and validated in adults, JAMA 2011) in 603 children and adolescents (median 12 years old) from the Chronic Kidney Disease in Children (CKiD) study. The children included in the current study had an estimated glomerular filtration rate (eGFR) < 60 mL/minute/1.73 m2 and did not have hyperoxaluria. The KFRE calculates an estimated risk (0% to 100%) for ESRD within a specified time frame based on age, sex, eGFR, albumin to creatinine ratio (ACR), and serum levels of calcium, phosphate, bicarbonate, and albumin. The ACR was measured directly in 24% of children and calculated from the protein to creatinine ratio in 76%. The children were followed for median 3.8 years.
ESRD, defined as receipt of long-term dialysis or kidney transplant, developed in 27 children (4.5%) within 1 year, 62 (10%) within 2 years, and 144 (24%) within 5 years. When the cohort was divided into tertiles, the median KFRE 2-year risk estimates closely matched the observed rates of ESRD: 1% (range 0%-2.5%) estimated vs. 0.5% observed for the low-risk group, 5% (2.6%-13.1%) estimated vs. 3.5% observed for moderate-risk, and 33% (13.2%-100%) estimated vs. 29% observed for high-risk (the median values were estimated from a figure; statistical comparison not reported). The KFRE also had very good discrimination, as quantified by the c-statistic, which represents the probability that a child who developed ESRD would have a higher estimated risk than a child who did not; a c-statistic of 1 has perfect performance and 0.5 indicates no predictive value. The c-statistic was 0.91 for predicting ESRD within 1 year, 0.87 for 2 years, and 0.82 for 5 years, with similar values reported for a version of the KFRE that uses only the first 4 variables described above.
The retrospective study design and reliance of ACR values calculated from the protein to creatinine ratio in most children reduces confidence in these results to a small degree. Also, applicability may be limited as children enrolled in the CKiD study were closely followed and may have adhered to recommended management more carefully than children in the general population. It should also be emphasized that while median ESRD risk estimates were similar to observed rates, each tertile—particularly the high-risk group—contained a range of a risk estimates. Thus, it is unclear to what level of precision the KFRE can estimate ESRD risk. Nonetheless, risk models for children with CKD are lacking, and the results of this study suggest that the KFRE can help identify children with CKD who are at an increased risk of ESRD and thus help with management decisions.