Reliability of glomerular filtration rate estimating formulas compared to iohexol plasma clearance in critically ill children.

Accurate renal function assessment is crucial to guide intensive care decision-making and drug dosing. Estimates of glomerular filtration rate (eGFR) are routinely used in critically ill children; however, these formulas were never evaluated against measured GFR (mGFR) in this population. We aimed to assess the reliability of common eGFR formulas compared to iohexol plasma clearance (CLiohexol) in a pediatric intensive care (PICU) population. Secondary outcomes were the prevalence of acute kidney injury (AKI) (by pRIFLE criteria) and augmented renal clearance (ARC) (defined as standard GFR for age + 2 standard deviations (SD)) within 48 h after admission based on mGFR and eGFR by the revised Schwartz formula and the difference between these two methods to diagnose AKI and ARC. In children, between 0 and 15 years of age, without chronic renal disease, GFR was measured by CLiohexol and estimated using 26 formulas based on creatinine (Scr), cystatine C (CysC), and betatrace protein (BTP), early after PICU admission. eGFR and mGFR results were compared for the entire study population and in subgroups according to age, using Bland-Altman analysis with calculation of bias, precision, and accuracy expressed as percentage of eGFR results within 30% (P30) and 10% (P10) of mGFR. CLiohexol was measured in 98 patients. Mean CLiohexol (± SD) was 115 ± 54 ml/min/1.73m2. Most eGFR formulas showed overestimation of mGFR with large bias and poor precision reflected by wide limits of agreement (LoA). Bias was larger with CysC- and BTP-based formulas compared to Scr-based formulas. In the entire study population, none of the eGFR formulas showed the minimal desired P30 > 75%. The widely used revised Schwartz formula overestimated mGFR with a high percentage bias of - 18 ± 51% (95% confidence interval (CI) - 29; - 9), poor precision with 95% LoA from - 120 to 84% and insufficient accuracy reflected by P30 of only 51% (95% CI 41; 61), and P10 of 21% (95% CI 13; 66) in the overall population. Although performance of Scr-based formulas was worst in children below 1 month of age, exclusion of neonates and younger children did not result in improved agreement and accuracy. Based on mGFR, prevalence of AKI and ARC within 48 h was 17% and 45% of patients, respectively. There was poor agreement between revised Schwartz formula and mGFR to diagnose AKI (kappa value of 0.342, p < 0.001; sensitivity of 30%, 95% CI 5; 20%) and ARC (kappa value of 0.342, p < 0.001; sensitivity of 70%, 95% CI 33; 58). CONCLUSION: In this proof-of-concept study, eGFR formulas were found to be largely inaccurate in the PICU population. Clinicians should therefore use these formulas with caution to guide drug dosing and therapeutic interventions in critically ill children. More research in subgroup populations is warranted to conclude on generalizability of these study findings. CLINICALTRIALS: gov NCT05179564, registered retrospectively on January 5, 2022. WHAT IS KNOWN: • Both acute kidney injury and augmented renal clearance may be present in PICU patients and warrant adaptation of therapy, including drug dosing. • Biomarker-based eGFR formulas are widely used for GFR assessment in critically ill children, although endogenous filtration biomarkers have important limitations in PICU patients and eGFR formulas have never been validated against measured GFR in this population. WHAT IS NEW: • eGFR formulas were found to be largely inaccurate in the PICU population when compared to measured GFR by iohexol clearance. Clinicians should therefore use these formulas with caution to guide drug dosing and therapeutic interventions in critically ill children. • Iohexol plasma clearance could be considered an alternative for accurate GFR assessment in PICU patients.