We compared the GFR measured using Tc-99 m-DTPA or Cr-51-EDTA radiotracers as gold standard with the estimated GFR derived from four distinct formulas. Our results indicate that the CKD-EPI equation performed best among these four. For patients diagnosed with urothelial tract carcinoma, the meticulous assessment of renal function is imperative. This necessity arises from the elevated risk of treatment-induced nephrotoxicity as well as the intrinsic likelihood of tumor-mediated obstruction within the urinary system. Either condition poses a significant risk of irreversible renal failure.

While all four eGFR equations presented strong CCC with mGFR, there were variations in their biases, wide limits of agreement, and TDIs. All indicates substantial variability in the estimation of GFR, also evident by the TDI above 20%. Despite the variability and bias in the eGFR estimation methods when compared to mGFR, the CKD-EPI method displayed the strongest results compared with mGFR at baseline and for all measurements, the highest percentage of eGFR values within 10%, and 30% accuracy of mGFR, the highest CCC, the lowest TDI, and the slightest average error (bias) in eGFR estimation for baseline and all measurements combined.

The characteristics of the patients in our study (Table 2) align well with the patient profiles presented in the existing bladder cancer literature [26]. The patients underwent multiple GFR measurements throughout their treatment period, which enabled us to evaluate the consistency during treatment.

Our results mimic previous GFR studies, where the selected eGFR equations demonstrated a strong positive correlation with mGFR, although conclusions are disparate depending on the patient population [2, 8, 18, 27,28,29,30,31,32,33,34]. In a study examining testicular cancer patients receiving cisplatin, a high correlation was observed between eGFR (creatinine clearance) and mGFR both at baseline and 3-month post-treatment. However, during the treatment period, this correlation was not evident, a phenomenon that may be attributable to cisplatin nephrotoxicity, and also the significant weight loss observed during treatment, with an average loss of approximately 10 kg. [35] In our cohort, the weight remained stable during treatment and was comparable to the baseline weight (Figure S5). The less stringent and less effective administration of anti-emetic therapy during platinum-based treatments in 1988, as opposed to contemporary practices, may account for the observed absence of weight loss within our cohort. Nevertheless, there was a diminution in the accuracy of the estimation methods during the treatment phase for cisplatin-treated patients. This decrease of correlation was not as marked as in [35], and the methods retained reliability for clinical use during treatment. (Table 3, S4 and S5). Notably, this deterioration was not evident in the carboplatin-treated group, likely due to carboplatin being less nephrotoxic.

The measurement inaccuracy of EDTA/DPTA mGFR is shown to be within 5–15% compared to inulin, with a tendency to underestimate GFR [36,37,38], and a variation between mGFR and eGFR of 15% could be argued to be acceptable accuracy for eGFR equations [21].

CCC indicated a good but suboptimal concordance between measurements, with CKD-EPI showing the best but declining performance in later measurements for the cisplatin-treated part of the cohort. [21, 22] Later measurements revealed lower mGFR values, suggesting eGFR’s decreased accuracy in kidney failure due to various factors such as chemotherapy side effects. TDI values observed in this study exceeded those of the established gold standard, [21] although current KDIGO guidelines recognize the presence of within-person variation in measured GFR (mGFR) [38]. The literature has yet to reach a consensus on an acceptable level of agreement between mGFR and eGFR, as evidenced by the ongoing debate in recent publications [21, 39, 40].

Surface-normalized GFR standardizes for body size (1.73 m2), whereas formulas like Cockcroft–Gault or Wright inherently adjust for it [34, 41]. Given that chemotherapy dosage relies on non-normalized GFR, we propose that laboratory reports should adjust the eGFR with the patient’s BSA for those formulas not accounting for BSA, presenting both surface normalized and non-normalized eGFR, echoing previous suggestions regarding abandonment of BSA indexing [41,42,43]. This study employed enzymatically measured creatinine from accredited laboratories. Enhanced accuracy in such modern creatinine assays has improved eGFR equation reliability, even when these equations were initially based on older assays with greater variability [44]. Even with optimal assays, there is an inherent variability of creatinine based on factors such as muscle mass, protein intake, sarcopenia, and chronic illnesses (diabetes mellitus, hypertension, etc.) that complicates creatinine-based eGFR [38, 45]. In addition, the reliability of the eGFR equations varies among different ethnic groups. Studies have indicated that the equation tends to underestimate eGFR in some ethnic groups, such as African Americans, if specific ethnic coefficients are not applied [46]. Conversely, it may overestimate eGFR in other populations, including some Asian cohorts where modified equations are in use [47]. These discrepancies arise from variations in muscle mass, dietary protein intake, and other genetic and environmental factors that influence serum creatinine levels.

In the lack of a more stable and freely filtered serum marker than creatinine, knowledge of the inaccuracies of eGFR, and when to choose mGFR, has important clinical implications. The nephrotoxic effects of agents such as cisplatin may exacerbate the discrepancy between plasma creatinine levels and the GFR, as GFR may decline while plasma creatinine remains unchanged. This trend was evident in our cohort, with a decrease in the CCC (Table S3) in patients treated with cisplatin. In contrast, such a trend was not observed in patients who were administered carboplatin (Table S4).

The study’s finding suggesting CKD-EPI is the most precise method aligns with similar studies for urothelial carcinoma using different mGFR references (CrCl), although agreement is not consistent for all cancers [8, 18, 38, 48, 49]. We found no clinical characteristics, baseline creatinine values or treatment related characteristics that helps to guide the use of mGFR or eGFR for individual patients, except for the decline in accuracy during later cycles of cisplatin treatment, which we attribute to the nephrotoxic effect of cisplatin. Even for patients with a low GFR, the variability in LOA is as consistent as for the higher GFRs (Figures S2 and S3). This suggests that fluctuations in creatinine that do not reflect actual GFR changes affects the variation in agreement between eGFR and mGFR, partly this variation can be driven by cisplatin nephrotoxicity in the longitudinal measurements. On the other hand, the within-subject variation of mGFR of 15% reflects actual GFR changes and also affects the inconsistency in agreement [38]. Care should be taken when interpreting eGFR on patients with extreme values of BMI, and in the last cycles of cisplatin treatment [18, 38, 41].

As in other comparative studies, this study affirms the good performance of the CKD-EPI equation on enzymatically measured creatinine [8, 10, 38, 48, 49]. Clinical decisions on cisplatin eligibility and carboplatin dose calculation should be guided by standard protocols, with mGFR remaining accessible, particularly for patients receiving cisplatin. Presently, guidelines lack clarity and do not favor one method over another. We suggest an update in clinical guidelines for managing urothelial carcinoma, to recommend CKD-EPI as the preferred equation for eGFR echoing the recommendation of KDIGO, although we do acknowledge that the differences between the estimation methods are modest [50]. Exclusive reliance on eGFR to determine glomerular filtration rate may lead to an elevated risk of adverse events, especially in patients receiving cisplatin during subsequent cycles of treatment.

Strengths and limitations

A strength of this study is the large dataset with corresponding mGFR from a clinical gold standard reference and serum creatinine values from an accredited laboratory using enzymatic assays and well annotated patient cohort. The estimation methods do not meet optimal standards, but a larger cohort would not amend this and further comparative studies using the same methods for this patient population are unlikely to yield different results. Limitations of this study include its retrospective design with possible confounders unaccounted for such as actual changes in GFR between eGFR and mGFR if not measured in the same day. Inherent variabilities in reference measurements (mGFR) and external factors like day-to-day variations in performance of laboratory equipment and variations in patient conditions can influence results. Lacking an absolute eGFR gold standard achieving specific LOA, CCC, and TDI benchmarks, we focus on relative differences between available methods.