The stakes behind these numbers are high. GFR, the rate at which the glomeruli of the kidneys filter plasma, determines how quickly many drugs leave the body. Nephrotoxic chemotherapy agents and renally cleared antimicrobials, including aminoglycosides, sit squarely in this category, and both underestimation and overestimation of kidney function carry real clinical consequences. Overestimating GFR in a child receiving chemotherapy can lead to overdosing and inappropriate agent selection, raising the risk of adverse effects including kidney injury. Underestimating GFR can result in underdosing and the exclusion of effective drugs, compromising treatment efficacy. Similar logic applies to antibiotics such as aminoglycosides, which carry a substantial global burden of drug-related adverse effects. Yet the reference standard for measuring GFR, a 24-hour urinary clearance of an exogenous marker such as inulin, iohexol or 51Cr-EDTA, is rarely performed in paediatric practice because it demands a continuous intravenous infusion, repetitive blood and urine collections and often bladder catheterisation.

Nuclear medicine GFR scans, which inject a radioactive filtration marker such as 99mTc-DTPA and collect a handful of blood samples, are a more practical surrogate, but they remain time-consuming, expensive and invasive, and guidelines recommend reserving them for specific circumstances such as malnutrition, heart failure or cancer, where precise measurements are crucial. Notably, these indications do not extend to dosing guidance for widely used renally cleared drugs. That leaves creatinine-based estimation equations as the workhorse of routine assessment. In adults, tools such as CKD-EPI and Cockcroft-Gault are embedded in clinical practice and dosing software. In children, however, their accuracy has been repeatedly questioned, with studies showing substantial inconsistency, both under- and overestimation, across a broad range of GFR values. The root cause is physiology: GFR does not reach adult levels until approximately the second year of life, and both muscle mass and creatinine production undergo rapid changes throughout childhood, undermining any formula that treats creatinine as a stable marker of filtration.

The most commonly used paediatric tool, the Modified Schwartz equation, has itself shown troubling bias. Studies of children with and without kidney disease have reported under- and overestimation trends that risk inappropriate drug dosing, and a previous study of 96 children aged 0.25 to 17.5 years found a p30 accuracy of just 53% with a median bias of 15.5 mL/min/1.73 m2. Newer equations, including the Chronic Kidney Disease in Children Under 25 years (CKiD-U25) equation and the Full Age Spectrum (FAS) equations, have demonstrated improved accuracy, but they have only been validated in children over one and two years of age, respectively, leaving a critical gap in infancy. The GROW-GFR equation was designed to close that gap by taking a fundamentally different route: rather than being fitted directly to measured GFR values, it was derived from a population pharmacokinetic (popPK) model of tobramycin, an aminoglycoside antibiotic that is almost exclusively cleared by the kidneys, with approximately 90% to 95% eliminated unchanged by glomerular filtration.

The underlying tobramycin model was built from a multicentre prospective study conducted over two years, from 2020 to 2022, at The Royal Children’s Hospital in Melbourne and The Children’s Hospital at Westmead in Sydney. It included 63 children with cystic fibrosis, with a median age of 13.3 years and a range of 0.2 to 19.3 years, who received intravenous tobramycin and had a median of four tobramycin concentrations determined per patient, with a range of one to 34. In the present study, the published popPK model was not refitted. Instead, individual estimates of tobramycin clearance derived from the model were used as a surrogate measure of GFR, an assumption justified by the drug’s near-total renal clearance. The resulting equation considers a patient’s age in years, weight in kilograms, height in metres and serum creatinine in mg/dL, and incorporates a published renal maturation model that accounts for changes in GFR with increasing post-menstrual age, a feature the authors believe underpins its improved performance in the youngest age groups.

Validation came from a retrospective audit of nuclear medicine GFR scans performed at The Royal Children’s Hospital Melbourne between July 2016 and November 2023 in children aged 0 to 19 years. Of 1,673 extracted results from 754 children, 1,560 results from 711 children met the inclusion criteria, which required an NM-GFR of 150 mL/min or less, along with height, weight and serum creatinine recorded within age-specific time windows of the scan. The median age was 8.5 years, with a range of 0.1 to 19 years, and the median NM-GFR was 55 mL/min, ranging from 0.5 to 150. Nearly all children, 710 of 711, had comorbidities, with malignancy the most common at 72%, followed by prior solid organ or haematopoietic stem cell transplant at 39%, cardiac disease at 13% and kidney disease at 10%. None of the patients had cystic fibrosis. All scans used 99mTc-DTPA with two blood samples at 90 and 180 minutes post-injection, with a transition in October 2021 from weight-based dosing analysed by gamma counter to an activity-based approach following the Fleming guidelines.

The headline result was stark. Across all 1,560 scans, GROW-GFR achieved a p30 accuracy, the percentage of estimates within 30% of the measured NM-GFR, of 76%, with a bias of minus 1.9 mL/min, and this performance was remarkably consistent across all age groups, at 80%, 74%, 75% and 78% for children under 3 months, 3 months to under 2 years, 2 to 12 years and over 12 years, respectively. The Modified Schwartz equation, by contrast, managed a p30 accuracy of only 46% with a bias of 17.3 mL/min, systematically overestimating kidney function. In children aged 2 years and over, GROW-GFR’s 76% p30 accuracy comfortably beat the Modified Schwartz (49%), CKiD-U25 (58%), FAS (51%) and FAS-height (53%) equations, and the equation also yielded a considerably lower root mean square error of 18% compared with 27% to 30% for the alternatives. In children over 12 years, the other equations overestimated GFR by 20.3 to 27.8 mL/min, while GROW-GFR showed minimal bias of 0.2 mL/min.

The advantages were most dramatic in the youngest children, where competing equations are least reliable or simply not applicable. In infants under 3 months, GROW-GFR reached a p30 accuracy of 80% while the Modified Schwartz equation achieved just 10%. In children aged 3 months to under 2 years, the figures were 74% versus 35%, and GROW-GFR outperformed the CKiD-U25 equation in children between 1 and 2 years old by 75% to 40%. When the team fitted generalised linear mixed-effects models accounting for repeated measures nested within individuals, GROW-GFR’s overall p30 accuracy rose to 80%, with a mean bias of minus 2.5 mL/min, and pairwise contrasts showed the odds of achieving p30 accuracy were significantly lower for every competing equation, with odds ratios of 0.21 for Modified Schwartz and 0.08 to 0.16 for the FAS and CKiD-U25 equations, all at p less than 0.001. The authors also note that GROW-GFR’s accuracy appears comparable to, and in some cases better than, established cystatin C-based equations, despite relying only on widely available serum creatinine.

A further technical distinction sets the equation apart: it estimates absolute GFR rather than a value standardised to a body surface area of 1.73 m2, a reference that reflects the mean BSA of young American adults in 1925 and is inappropriate for children, most of whom fall well below 1.60 m2. Studies comparing standardised and absolute GFR have shown greater discrepancies at BSA extremes, and guidelines including KDIGO recommend that drug dosing in children and adults with extreme body sizes be based on absolute rather than standardised GFR. The study does carry limitations. The equation was originally developed from a relatively small cohort of 63 children with cystic fibrosis, raising questions about generalisability, although its strong performance in this large, diverse, non-CF validation cohort suggests any CF-specific effects are minimal. The cohort predominantly had normal renal function, NM-GFR values were capped at 150 mL/min so renal hyperfiltration was not assessed, and 99mTc-DTPA’s 10% to 13% protein binding may contribute to slight underestimation of GFR. Hydration status could not be evaluated, and a methodology change during the study period was addressed with sensitivity analyses showing similar results across both eras. The authors conclude that GROW-GFR should replace existing tools for estimating GFR in children to guide dosing of renally excreted drugs, though performance in populations with significantly reduced kidney function warrants further investigation.

Subject of Research: Refining absolute GFR estimation in children: development and validation of the GROW-GFR equation

Article Title: Refining absolute GFR estimation in children: development and validation of the GROW-GFR equation

Article References: Harwood, K. H., Duffull, S., Davey, C., Quinlan, C., Forbes, T., Francis, P., Lai, T., Ranganathan, S., & Gwee, A. (2026). Refining absolute GFR estimation in children: development and validation of the GROW-GFR equation. Frontiers in Pharmacology, 17, Article 1855031. https://doi.org/10.3389/fphar.2026.1855031

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1855031

Keywords: Refining, absolute, estimation, children, development, validation, GROW-GFR, equation, scientific research