Development of a Novel Risk Prediction Model for Cisplatin Nephrotoxicity

Abstract

Cisplatin is a chemotherapy drug that will be used to treat approximately 850,000 new cancer patients in 2015. Despite decades of its use, toxicity to the kidney remains a problem in 25-30% of the patients. Given the high frequency of kidney toxicity, we see as Nephrologists, a need to readdress methods for evaluating candidacy for cisplatin use in Oncology. Through this project, we may change the paradigm of selecting candidates and evaluating risk of cisplatin chemotherapy, which is an essential step prior to offering this drug as an option to patients. We propose to challenge the current method used for estimating kidney function, which is using serum creatinine or estimated creatinine clearance by the Cockcroft-Gault equation. Our first aim is to compare the estimated glomerular filtration rate (eGFR) equation by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) for determining eligibility to receive cisplatin, since it is the most accurate equation available for estimating kidney function at the present time. Our second aim is to derive a risk prediction model using readily available clinical and laboratory variables that would enable clinicians to instantly compute the risk of kidney toxicity during the office visit and arrive at a graded score. This score could then be presented to patients as a low-, medium- or high-risk measure to further inform their decision-making regarding the treatment and to jointly (with their physicians) make an informed choice in the context of other options available. We propose to achieve the above aims by analyzing a large group of over 6000 patients that we have assembled using the Research Patient Data Repository (RPDR) and Oncology Data Retrieval System (OncDRS), which are large data repositories of patients treated at Partners affiliated hospitals and Dana Farber Cancer Institute, respectively. Detailed information collected from these patients who have been treated with cisplatin over the past 15 years includes demographics, chemotherapy, other medications, medical history and laboratory values. By performing sophisticated statistical analysis on these data, we will be able to understand the strength of these associations and thus identify major risk factors causing kidney toxicity. The model will be developed from the larger of the two group of patients treated at one hospital (derivation cohort) and testing our findings on the second group (validation cohort). We expect the results from this study to influence clinical practice and have lasting, positive short- and long-term impact on the overall health of patients receiving this drug; as well as identifying and redirecting patients with heightened risk of toxicity. If successful, ours will be the largest study on this topic and offer a novel method to address an old problem.