MORE ACCURATE MODEL TO ESTIMATE GFR FOR CARBOPLATIN DOSING B

A new model that improves estimation of glomerular filtration rate (GFR) could replace seven previously published models as the new standard of care for calculating the dose of platinum-based chemotherapy, such as carboplatin (Paraplatin, Bristol-Myers Squibb), say researchers.

GFR estimation is "a cornerstone of curative and palliative care" because it has a bearing on dosing, which in turn determines tumor response and risk for nephrotoxicity, they point out.

Importantly, the new model is based on standard biometric data and can be easily used by oncologists to improve carboplatin dosing accuracy and reduce risk of dosing errors in patients with lung, ovary, triple-negative, and germline BRCA1/2 mutation–positive breast cancer, as well as seminomas.

The new model may also prove useful for dose decision making in a broad range of clinical settings, including the care of patients without cancer, the study authors say.

More Accurate and Less Biased

Up to now, the most accurate creatinine-based published model for estimating GFR and for calculating carboplatin dose has been the body surface area (BSA)–adjusted chronic kidney disease epidemiology (CKD-EPI).

The new model, which is available online as a free tool, is more accurate and less biased, say lead author Tobias Janowitz, MB, BChir, PhD, of the University of Cambridge, United Kingdom, and colleagues.

An analysis of a large data set of patients with cancer showed that the new model reduced the absolute percentage error in carboplatin dose to 14.17% compared to 18.62% for the BSA-adjusted CKD-EPI and 25.51% for the Cockcroft-Gault formula. The findings were published online July 7 in the Journal of Clinical Oncology.

The findings were externally validated using a data set that showed that absolute error reduction ranged from 34.23% for the Cockcroft-Gault formula to 18.92% for the BSA-adjusted CKD-EPI model to 11.71% for the new model.

"Evidence from our internal and external validation work suggests that our new model is the best currently available model to predict GFR in patients with cancer," Dr Janowitz and colleagues say. "Our new model further improves the estimation accuracy for GFR and may present a new standard of care and should be investigated alongside BSA-adjusted CKD-EPI in clinical practice."

There is a lack of consensus as to the best way to estimate GFR using the current "gold standard" models, including chromium-51 EDTA excretion measurements (51Cr-EDTA GFR), the Cockcroft-Gault model, the Jelliffe model, or 24-hour urine creatinine level, Dr Janowitz and colleagues say.

Study Details

The study included data from 2471 white adult patients with histologically confirmed cancer. The data were obtained between August 2006 and January 2013. Serum creatinine was measured within 30 days of the 51Cr-EDTA GFR measurement. Most patients had near-normal kidney function, as demonstrated by a median 51Cr-EDTA GFR of 81 mL/min.

Data were divided randomly into two groups: 80% (representing 1977 patients) were used for model development, and 20% (representing 494 patients) were used for internal validation of the new model.

For external validation, the data set consisted of 111 male patients from another cancer center with stage I seminoma. The median 51Cr-EDTA GFR for these patients was 113 mL/min, and the median age was 39 years.

The full data set was used to compare 51Cr-EDTA GFR with eGFR from seven published models and BSA-adjusted models, including Martin, Wright, Mayo, Modification of Diet in Renal Disease (MDRD), and CKD-EPI. The Cockcroft-Gault and the Jelliffe models were also assessed and compared with the new model using the statistics root-mean-squared-error (RMSE) and median residual. Performance was compared in the internal and external validation data sets. An absolute percentage error of >20% was used to determine carboplatin dosing accuracy.

For the internal validation set, the BSA-adjusted CKD-EPI was the most accurate model for eGFR (RMSE, 16.30 mL/min; 95% confidence interval [CI], 15.34 - 17.38 mL/min), the study showed. However, eGFR accuracy was improved by the new model (RMSE, 15.00 mL/min; 95% CI, 14.12 - 16.00 mL/min) when compared to all published models, including the BSA-adjusted CKD-EPI.

For those interested in testing the model, the online tool provides the eGFR for any given set of input data, say Dr Janowitz and colleagues. It also provides "an estimated predictive confidence interval for the true GFR (default setting at 95%), an estimated probability of the true GFR being below or above an operator-chosen value (default setting at 50 mL/min), as well as the eGFR according to the BSA-adjusted CKD-EPI model."

More external validation is welcome, Dr Janowitz and coauthor Edward H. Williams told Medscape Medical News. "The model, coefficients, extended methods, and link to the shiny app for use of the model are also available in the manuscript that we have published with open access," they said in an email. "We invite colleagues to approach us if they have data sets available that may help with further external validation and refinement of the model. We also would be grateful if our colleagues were to test the model, evaluate the online tool, and provide feedback if indicated."

Clinicians are advised to be aware that "there is uncertainty attached to every estimation," they added. With this in mind, a prediction interval is provided "which can be thought of as a confidence interval for the prediction."

Limitations of the study include the fact that all the patients were white and were treated at a single center. Previous studies have shown that adjustment factors improve GFR prediction in black patients, and future studies should make this focus a priority, the authors say.

Funding for this study was provided by the Wellcome Trust Translational Medicine and Therapeutics Programme, the National Institute of Health Research Cambridge Biomedical Research Center, and the University of Cambridge. Dr Janowitz and Edward Williams have disclosed no relevant financial relationships. The original article lists disclosures of coauthors.

J Clin Oncol. Published online on July 7, 2017. Full text