GFR Estimating Equations and Their Association with Adverse Clinical Outcomes: A Population-Based Retrospective Cohort Study

Abstract

Chronic kidney disease (CKD) is a progressive condition characterised by abnormalities in kidney structure or function, and is classified by cause of injury, eGFR and albuminuria.1 It carries an enormous global burden, affecting approximately 788 million adults worldwide, accounting for nearly 1.5 million deaths annually, and currently ranking as the ninth leading cause of death globally.1,2 In Canada alone, there are an estimated 4 million people who have CKD, costing Canadians approximately $40 billion annually in both direct and indirect costs.1,2 Altogether, the irreversible nature of CKD, the absence of curative options, increasing risk beyond kidney outcomes, and the financial burden of CKD underscore the critical importance of accurate kidney function assessment to facilitate early recognition and intervention. Glomerular filtration rate (GFR) is widely considered the best method in determining the kidney's health and function in both research and clinical settings.1 It is also used in estimating the risk for kidney failure using tools such as the Kidney Failure Risk Equation.3 Multiple eGFR equations have been developed; however, serum creatinine (SCr) based equations are more utilised given that it is inexpensive, widely standardised, and routinely measured.1,4 In 2021, two new equations were developed, the 2021 Chronic Kidney Disease Epidemiology Collaboration (CK21) and the European Kidney Function Consortium Equation (EKFC), both of which were race-free equations. In North America, the CK21 has been widely adopted; however, the EKFC has been shown to have potential benefits in other populations. Current evidence shows that EKFC shifts CKD prevalence but does not meaningfully improve prediction of outcome risk, with discrimination remaining the same across equations.5,6 Important methodological gaps remain, however. Prior analyses report only relative measures such as hazard ratios and reclassification metrics; none model the cumulative incidence function directly to estimate and assess the accuracy of a patient's absolute outcome risk. Calibration has also not been formally assessed, a clinically relevant metric that provides insight as to whether predicted probabilities match reality. Furthermore, current studies are limited to two European populations or focus on a cardiovascular high-risk cohort, with no general North American population assessed. Lastly, there have been limited studies comparing the usage of eGFR in kidney failure risk tools using the EKFC. This thesis addresses these gaps by comparing CK21 and EKFC in a population-based Ontario cohort, evaluating discrimination, calibration, and reclassification for kidney and cardiovascular outcomes, and assessing their performance as inputs to the KFRE.