An accurate assessment of the Glomerular Filtration Rate (GFR) is essential in oncology patients to ensure safe prescribing of chemotherapy drugs, detecting kidney injury and assessing prognosis. Inulin clearance is considered as gold standard of the GFR measurement. Nonetheless, it cannot be applied in routine clinical practice. Radioisotopic methods have a high correlation with inulin clearance and thus remains the most widespread method for accurate calculation of the GFR. The most common radioisotopes used include chromium 51 EDTA(51Cr-EDTA)and technetium-99m diethyl triamine penta-acetic acid (99mTc-DTPA) [1, 2]. However, these methods are relatively expensive, invasive and time consuming. Consequently, oncologists often rely on formulae to estimate the GFR on the basis of serum creatinine and other parameters.
Oncologists should pay attention to the limitation of eGFR formulae. Oncology patients may have low muscle mass and reduced dietary protein intake, which would influence the concentration of serum creatinine and thus the performance of eGFR formulae [14]. In our study, the fraction of patients with eGFR absolute percentage error >30% is more than one-third in all formulae, which is consistent with previous studies [7, 15]. This means that a great number of oncology patients received wrong diagnosis and drug dosage. Therefore, it is inappropriate to apply eGFR formulae to all oncology patients. Oncologists should identify populations where eGFR formulae based on serum creatinine is not likely to provide an accurate estimate and thus alternative measurements of the GFR should be considered. Some studies have demonstrated that the performance of eGFR formulae may be affected by age, weight and GFR [10-12, 16]. However, a limitation among prior studies was that only univariate analysis was performed and thus confounding factors were not controlled. To address this limitation, we used multivariate logistic regression analysis to identify the independent factors associated with the poor accuracy of eGFR formulae.
Oncologists need to be vigilant when using formulae to assess the kidney function in overweight or obese populations. It has been reported that the Cockcroft–Gault formula showed a tendency to overestimate GFR in oncology and other populations where overweight or obesity were considered [10, 17]. We confirmed this finding in our study. It seems that lean oncology patients suffer more from cachexia and malnutrition in clinical practice and would exhibit worse estimates of the GFR. However, in fact, more attention is needed in overweight patients instead of lean ones when estimating the GFR. Of great interest is that Cockcroft and Gault clearly stated it was probably not applicable to use their formula in obese populations [3]. This is understandable because the Cockcroft–Gault formula is only one that incorporates body size as an important index. Besides, some studies suggested that application of the alternate weight descriptors can improve the accuracy of Cockcroft–Gault formula, such as ideal body weight [18] and adjusted body weight [19], but these strategies is not yet validated in large population.
BUN/Scr ratio may be an indicator of the accuracy of eGFR formulae. BUN/SCr ratio greater than 20 is known as a marker of pre-renal renal dysfunction [20]. Poggio et al. found that the Cockcroft–Gault and MDRD formulae were not reliable in sick hospitalized patients, especially those with high BUN/SCr ratio [21]. However, little is known about the association of elevated BUN/SCr ratio and accuracy of eGFR formulae in oncology patients. We found that the Cockcroft–Gault formula was likely to be inaccurate in oncology patients with high BUN/SCr ratio. Possible explanation for our observations is that BUN/SCr ratio may rise in oncology patients with low rate of creatinine generation, and creatinine-based eGFR formulae would perform poorly accordingly.
Normal estimates of the GFR might not be actually that normal. In subgroup patients of higher eGFR, Cockcroft-Gault formula showed low accuracy and great degree of overestimation, which is consistent with reports from studies consisting of population with normal renal function, such as kidney donors [22, 23]. This means that kidney injury will be wrongly labeled as normal kidney function and the degree of renal damage will be underestimated, which encourages oncologists to make wrong decisions regarding the administration of iodinated contrast medium, employment of nephrotoxic drugs and the time to initiate renal replacement therapy. Besides, narrow therapeutic index is a pharmacokinetic characteristic of most chemotherapy agents. A well-known example of such agents is carboplatin, whose dose is adjusted by Calvert formula incorporating the GFR as an important variable [24]. As a consequence, overestimated GFR may result in overdosage of chemotherapy agents, particularly those agents which are entirely eliminated by the kidneys in unchanged active form. Ultimately, inaccurate assessment of the GFR means severe side effects, as well as increasing incidence of renal insufficiency, or even death.
An important facet of this study is that a nomogram was developed to predict the reliability of eGFR as calculated by the Cockcroft-Gault formula in oncology patients. The nomogram is a simple graphical prediction tool. By assigning points to the four variables, oncologists can assess the predictive risk of individuals. This provides clinically useful information and guide personalized clinical decision-making regarding whether to use accurate GFR measurements for oncology patients. Furthermore, our nomogram is constructed on the basis of readily available clinical data making it easy for clinicians to use. Internal validation indicated good performance with area under ROC of 0.743 and accurate calibration.
We acknowledge several limitations in this study. Firstly, due to the retrospective nature of our study, there might have been selection bias and unknown confounders in the analysis. Secondly, even though we collected many characteristics in our patient’s population, there still exists some factors that were not analyzed in our study, including cancer type, dietary intake, prior treatments and other medications. Finally, although the nomogram was validated internally by bootstrap resampling, external validation using an independent data set was required before routine use.