Evaluation of novel candidate filtration markers from a global metabolomic discovery for glomerular filtration rate estimation

A comparison of neural networks and regression-based approaches for estimating kidney function in pediatric chronic kidney disease: Practical predictive epidemiology for clinical management of a progressive disease

PURPOSE

Clinical management of pediatric chronic kidney disease requires estimation of glomerular filtration rate (eGFR). Currently, eGFR is determined by two endogenous markers measured in blood: serum creatine (SCr) and cystatin C (CysC). Machine learning methods show promise to potentially improve eGFR, but it is unclear if they can outperform regression-based approaches under clinical constraining requiring real time measurement and only two predictors. We constructed a neural network for eGFR (NNeGFR) and compared it to the clinical standard Under 25 (U25eGFR) equations using the same data for training and validation.

METHODS

The U25eGFR data comprised 1683 training and 843 validation observations that included iohexol measured GFR (mGFR), SCr and CysC. Sex-stratified feed forward NNs included the same predictors as U25eGFR (i.e., age, height/SCr, CysC) with additional nonlinear transformations. Performance was evaluated by bias (for calibration), proportions within 10 % and 30 % of mGFR (P10 and P30, for accuracy), root mean square error (RMSE, for precision) and R2 (for discrimination).

RESULTS

NNeGFR performed comparably to the U25eGFR equations on all metrics. Biases were minimal, slightly favoring U25eGFR. NNeGFR and U25eGFR had similar P10 (>37 %), P30 (>86 %) and RMSE.

CONCLUSIONS

NNeGFR performed as well as established equations to estimate GFR. Without additional biomarkers related to kidney function, which are not currently clinically available in real time, NN methods are unlikely to substantially outperform regression derived GFR estimating equations. Implications for translation of these advanced epidemiologic methods to clinical practice are discussed.

Panel estimated Glomerular Filtration Rate (GFR): Statistical considerations for maximizing accuracy in diverse clinical populations

Assessing glomerular filtration rate (GFR) is critical for diagnosis, staging, and management of kidney disease. However, accuracy of estimated GFR (eGFR) is limited by large errors (>30% error present in >10-50% of patients), adversely impacting patient care. Errors often result from variation across populations of non-GFR determinants affecting the filtration markers used to estimate GFR. We hypothesized that combining multiple filtration markers with non-overlapping non-GFR determinants into a panel GFR could improve eGFR accuracy, extending current recognition that adding cystatin C to serum creatinine improves accuracy. Non-GFR determinants of markers can affect the accuracy of eGFR in two ways: first, increased variability in the non-GFR determinants of some filtration markers among application populations compared to the development population may result in outlying values for those markers. Second, systematic differences in the non-GFR determinants of some markers between application and development populations can lead to biased estimates in the application populations. Here, we propose and evaluate methods for estimating GFR based on multiple markers in applications with potentially higher rates of outlying predictors than in development data. We apply transfer learning to address systematic differences between application and development populations. We evaluated a panel of 8 markers (5 metabolites and 3 low molecular weight proteins) in 3,554 participants from 9 studies. Results show that contamination in two strongly predictive markers can increase imprecision by more than two-fold, but outlier identification with robust estimation can restore precision nearly fully to uncontaminated data. Furthermore, transfer learning can yield similar results with even modest training set sample size. Combining both approaches addresses both sources of error in GFR estimates. Once the laboratory challenge of developing a validated targeted assay for additional metabolites is overcome, these methods can inform the use of a panel eGFR across diverse clinical settings, ensuring accuracy despite differing non-GFR determinants.

Refining GFR estimation: a quest for the unobservable truth?

Assessing glomerular filtration rate (GFR), which is central to evaluating kidney health, remains challenging. Measured GFR is not widely available and lacks standardization. Estimated GFR can be highly inaccurate for some patients and has limited applicability to many patient populations, such as those who are acutely ill. Recent metabolomic advances show promise for identifying new filtration markers that might enhance GFR estimation. Improving GFR assessment will require refinement in both GFR measurement and estimation methods.