Photo Credit: Marco Marca

The following is a summary of “Enhancing individual glomerular filtration rate assessment: can we trust the equation? Development and validation of machine learning models to assess the trustworthiness of estimated GFR compared to measured GFR,” published in the January 2025 issue of BMC Nephrology by Lanot et al. 

Creatinine-based estimated glomerular filtration rate (eGFR) equations are commonly used in clinical practice but have limitations. Measuring GFR is time-consuming and not routinely available. 

Researchers conducted a retrospective study to develop machine learning models assessing the trustworthiness of the European Kidney Function Consortium (EKFC) eGFR equation in estimating measured GFR within 10%, 20%, or 30% at the individual level. 

They used data from 22,343 participants across European and US cohorts. About 4 machine learning and 2 logistic regression models were trained on 9,202 participants to predict EKFC eGFR accuracy within 30% (p30), 20% (p20) or 10% (p10) of mGFR, validated internally on 3,034 participants and externally on 10,107. Predictors included creatinine, age, sex, height, weight, and EKFC. 

The results showed the random forest model was the most robust, achieving an area under the curve of 0.675 (95% CI 0.660–0.690) and accuracy of 0.716 (95% CI 0.707–0.725) for the P30 criterion in the external validation cohort. Sensitivity was 0.756 (95%CI 0.747–0.765) and specificity was 0.485 (95% CI 0.460–0.511) at the 80% probability level. The model’s PPV was 89.5%, higher than the EKFC P30 of 85.2%. A free web application was developed to assess EKFC trustworthiness at the individual level. 

Investigators found that the machine learning model marginally improved GFR estimation trustworthiness at the population level and provided valuable individual-level assessments. 

Source: bmcnephrol.biomedcentral.com/articles/10.1186/s12882-025-03972-0