Comparison of age-related changes in glomerular filtration rate: creatinine-based formulae vs. radioisotopic method

Introduction

Chronic kidney disease (CKD) is a significant health issue worldwide. The asymptomatic nature of CKD in its early stages can result in the loss of up to 90 % of kidney function by the time the disease becomes clinically noticeable. Without effective management, CKD significantly elevates the risk of cardiovascular morbidity, mortality, and progression to end-stage renal failure [1], 2]. Clinically, CKD is defined by the presence of positive markers of kidney damage or a glomerular filtration rate (GFR) persistently below 60 mL/min/1.73 m² present for a minimum of three months [3]. The low level of public awareness regarding CKD, in addition to its early asymptomatic phase, underscores the importance of GFR as a critical parameter for clinicians. GFR is widely recognized as the best overall indicator of renal function and plays a pivotal role in the screening, diagnosis, and staging of CKD, as well as the adjustment of drug dosages [2], [3], [4].

GFR is defined as the plasma volume of blood filtered from the glomerular capillaries into Bowman’s capsule per unit of time and reflects the kidney’s filtration capacity [5]. As the direct measurement of GFR is not feasible, it is typically measured indirectly based on the clearance of endogenous and exogenous substances. While urinary inulin clearance remains the gold standard for GFR quantification, it is impractical for routine clinical use due to its limited availability and the analytical challenges related to measuring inulin levels [6]. Given the challenges associated with clearance calculation using exogenous markers, the use of endogenous markers has advantages in terms of clinical applicability and cost-effectiveness. The main endogenous marker available for routine use is creatinine, and the Kidney Disease: Improving Global Outcomes (KDIGO), Clinical Practice Guideline for the Evaluation and Management of Chronic Kidney Disease recommends using serum creatinine and an estimating equation for initial GFR assessments. While creatinine-based estimated GFR (eGFR) is generally valid for most clinical scenarios, the use of eGFR formula incorporating both creatinine and cystatin C, or the measurement of GFR using exogenous markers, is recommended when greater accuracy is required and where GFR plays a critical role in clinical decision-making [3].

Owing to its cost-effectiveness and widespread availability, eGFR calculations based on creatinine have been rapidly integrated into clinical laboratories. Among the most commonly used equations are the Cockcroft-Gault formula (CG, 1976) [7], the Modification of Diet in Renal Disease (MDRD, 1999) [8], and the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI, 2009) [9] equations. Over time, these formulae have been modified to align with standardized, isotope dilution mass spectrometry (IDMS)-traceable creatinine methods, improving accuracy [1], 10]. In 2021, a new CKD-EPI 2021 equation was introduced, omitting race as a variable recognizing that it is a social rather than a biological construct [11]. However, controversy has arisen regarding the potential overestimation of GFR in non-Black populations using this updated equation [10], 12]. Despite these concerns, the CKD-EPI equation has gained widespread adoption in clinical laboratories due to its overall performance, particularly in populations with near-normal renal function and is recommended by KDIGO guidelines [3].

In elderly individuals, serum creatinine levels may remain within normal ranges despite a significant decline in GFR. Thus, serum creatinine levels and creatinine-based estimations may not reliably assess renal function in this population [10], 13]. The MDRD equation is considered to overestimate GFR, especially in elderly people with reduced muscle mass [14]. In contrast, cystatin C-based equations provide a more accurate alternative for GFR estimation in older adults, as cystatin C levels are less affected by age and sex [15]. However, cystatin C has the disadvantages of limited availability and high cost compared to creatinine. These limitations highlight the critical need for more reliable and accessible methods of GFR assessment, especially for people over 60 years of age.

Exogenous markers are increasingly recognized as valuable tools in cases where creatinine-based equations are limited and when precise GFR assessment is required for clinical decision-making. Among these, GFR measurement using technetium 99m-diethylene triamine penta-acetic acid (^99mTc-DTPA) with dual plasma sampling is a relatively inexpensive and practical method. ^99mTc-DTPA is exclusively cleared from plasma by glomerular filtration, without tubular secretion or reabsorption, making it an ideal marker for accurate GFR estimation [16]. GFR measurement using ^99mTc-DTPA is a widely accessible method in nuclear medicine departments and is considered a more reliable method for evaluating and monitoring renal function [17]. This clearance method can be performed using one or two plasma samples, and its accuracy is considered to be comparable to that of inulin clearance [18].

In this study, we aimed to assess the performance of three different creatinine-based GFR estimation methods by comparing them to measured GFR using ^99mTc-DTPA with dual plasma sampling, a more reliable method for GFR assessment. Additionally, we aimed to determine the effects of serum creatinine concentration, age, and body mass index (BMI) on the discrepancies between measured GFR (mGFR) and eGFR methods.

Materials and methods

Results

The demographic and clinical data of 905 adult patients included in the study are presented in Table 1. The median value of GFR measured by ^99mTc-DTPA using dual plasma sampling was 25 mL/min/1.73 m² (interquartile range (IQR)= Q1-Q3: 15–43), while the creatinine-based formulae overestimated GFR. The results of the comparisons of the formulae to measured GFR values are given in Table 2. Pearson’s correlation analysis showed that the CKD-EPI 2021 (r=0.854), MDRD (r=0.824), and CG (r=0.856) equations were highly correlated with ^99mTc-DTPA. Passing-Bablok regression analysis revealed that all formulae had positive proportional systematic errors (Figure 1). Bland-Altman plots and the wide limits of agreement demonstrated that the eGFR formulae exhibited a positive bias and low agreement compared to ^99mTc-DTPA (Figure 2). The agreement between the eGFR and mGFR methods, stratified by CKD stages, was evaluated using weighted kappa statistics (Table 3). The MDRD equation demonstrated the highest concordance rate (47.5 %) and weighted kappa value (0.527), indicating better agreement with ^99mTc-DTPA compared to the CKD-EPI 2021 and CG formulae. However, the overall moderate kappa values suggested variability in the performance of these formulae, particularly in moderate CKD stages.

Figure 1:

Comparison of measured GFR and creatinine-based eGFR formulae: Passing-Bablok regression analysis. The Passing-Bablok regression equation and scatter plot indicate the presence of a positive proportional systematic bias in the eGFR formulae in comparison to ^99mTc-DTPA. The black lines (rectangles) highlight the stages of chronic kidney disease. The normal serum creatinine levels were defined as<0.9 mg/dL for women and <1.2 mg/dL for men.

Figure 2:

Comparison of GFR methods using Bland-Altman plots. The Bland-Altman analyses and the wide limits of agreement demonstrate that the eGFR formulae exhibit a positive bias compared to ^99mTc-DTPA. Furthermore, at normal creatinine concentrations, reduced agreement, increased variability, and a negative correlation trend are particularly observed for the CKD-EPI 2021 and MDRD equations.

We assessed the correlations between bias and variables to identify error-causing conditions (Table 4). We found mild to moderate correlations for creatinine. Creatinine values were categorized as normal (female<0.9 and male<1.2 mg/dL) and high according to reference ranges. Accordingly, at normal creatinine values, a high correlation was found between mGFR and the bias of the CKD-EPI 2021 (r= −0.807) and MDRD (r= −0.637) formulae. This correlation trend at normal creatinine concentrations, accompanied by decreased agreement and increased variability, was also evident in the Bland-Altman plots. Thus, the bias observed in these formulae can be partially explained by normal creatinine values. The mean differences (bias) between calculated and mGFR values in groups categorized according to age, creatinine, and BMI values are given in Table 5. Accordingly, all formulae had the highest bias in patients with normal creatinine values.

Discussion

In this study, we evaluated the performance of three creatinine-based eGFR formulae (CKD-EPI 2021, MDRD, and CG) by comparing them to GFR values measured using ^99mTc-DTPA dual plasma sampling, a reference method for GFR assessment. Our findings revealed the systematic overestimation of GFR by all formulae despite strong correlations (r>0.82) between eGFR and mGFR values. The median mGFR of 25 mL/min/1.73 m² (Q1-Q3: 15–43) highlighted a cohort predominantly within advanced CKD stages, yet even in this context, creatinine-based equations exhibited positive proportional biases in the Passing-Bablok regression analysis and wide limits of agreement in the Bland-Altman analyses, underscoring moderate concordance with the reference method. While the MDRD equation demonstrated marginally superior agreement (weighted κ=0.527) in CKD staging compared to CKD-EPI 2021 and CG, the overall moderate kappa statistics and variability across CKD stages revealed persistent limitations. Notably, the highest discrepancies were observed in patients with normal creatinine values, where strong inverse correlations (e.g., CKD-EPI 2021: r=−0.807) linked normal serum creatinine to increased bias and reduced agreement. This conflict, wherein creatinine levels within reference ranges mask declining renal function, underscored inherent flaws in creatinine-based models, such as their dependence on muscle mass and metabolic factors, which disproportionately affect accuracy in older adults and individuals with preserved creatinine despite reduced GFR. Collectively, these findings emphasized the critical need for refining eGFR formulae or incorporating supplemental biomarkers like cystatin C to improve accuracy, especially in populations where normal creatinine values may erroneously imply renal health despite significant functional decline.

In a study involving 85 elderly patients (51 males and 34 females), the MDRD equation demonstrated the highest consistency with GFR measured using dual plasma sampling ^99mTc-DTPA. In contrast, the CG and CKD-EPI 2009 equations underestimated GFR [21]. A comparable study involving 76 patients demonstrated no significant superiority among the creatinine-based eGFR formulae in individuals under 70 years of age. However, in patients aged 70 years and older, the CG formula showed greater consistency with dual plasma sampling ^99mTc-DTPA, while the MDRD and CKD-EPI 2009 formulae were deemed less reliable and appeared to overestimate GFR. These findings suggested that the CG formula may be more appropriate for GFR estimation in older adults [22]. In particular, the better performance of the CG formula in elderly patients may be attributed to the possibility that the decrease in muscle mass and creatinine production in this group is partially compensated by the weight variable of the formula. Conflicting findings in the literature may be due to heterogeneity in patient populations, methodological variations, and differences in formula calibration ranges. In a study involving 551 patients, Huang et al. reported that the MDRD equation underestimated GFR, while the CKD-EPI 2009 equation overestimated GFR in comparison to the reference method of dual plasma sampling ^99mTc-DTPA [23]. Xie et al. conducted a study in the Chinese population with 154 CKD patients aged 18–83 years and found that the CKD-EPI 2009 equation provided the closest agreement with GFR measured by dual plasma sampling ^99mTc-DTPA [24]. On the other hand, our results showed that these three equations tended to overestimate GFR. These discrepancies highlighted the variability in the performance of creatinine-based GFR estimation formulae that may arise from demographic characteristics (e.g., differences in body composition in Chinese populations), methodological variations, or different CKD stage distributions.

The CKD-EPI 2021 equation, which omits race as a variable, has been widely adopted due to its improved performance in diverse populations. However, concerns remain regarding its potential misclassification and overestimation CKD diagnosis in black populations and delayed CKD diagnosis in non-black populations [12], 25]. The results of a study in cancer patients highlighted that excluding race from eGFR equations led to inadequate chemotherapy dosing and poorer outcomes in black patients [26]. In a validation study in kidney transplant recipients, the authors reported that the race-free CKD-EPI 2021 equation showed the poorest agreement with mGFR, which was consistent with our findings [27].

Between two distinct studies involving healthy kidney donors, one revealed that the CKD-EPI 2009 equation provided GFR estimates closest to those obtained with the ^99mTc-DTPA method [28]. In contrast, the other study concluded that the CG formula demonstrated superior performance, which was in agreement with our results [29]. In a cross-sectional study of 129 patients, the CKD-EPI 2009 equation demonstrated the best overall agreement with GFR measured by dual plasma sampling ^99mTc-DTPA, while the CG formula exhibited the poorest agreement, particularly in patients younger than 65 years old [30]. In our study, the MDRD formula had the best overall agreement, while the CKD-EPI 2021 formula had the worst.

With aging, GFR decreases physiologically due to structural changes and nephron loss in the kidneys. After the age of 30, GFR decreases by 0.8/mL/min per year [31]. Furthermore, in older adults, serum creatinine levels often remain within normal reference ranges despite a significant reduction in GFR, as age-related declines in muscle mass lead to reduced creatinine production. This discrepancy may lead to misleadingly normal creatinine values, thereby limiting the reliability of creatinine-based formulae in the elderly population. Our findings aligned with those of Schaeffner et al., who reported that serum creatinine levels often remained within normal ranges in elderly individuals despite significant declines in GFR [13]. In our study, the CG formula provided the closest agreement with mGFR in patients over 60 years of age, while the MDRD formula performed better in younger individuals. These findings also suggested that age-specific formulae or an alternative method based on cystatin C, which is less affected by age and sex, may be necessary for accurate GFR estimation in older adults.

It is important to consider the limitations of the MDRD equation, particularly when eGFR falls below 60 mL/min/1.73 m². The MDRD equation, derived primarily in populations with CKD, systematically underestimates measured GFR at higher values (>60 mL/min/1.73 m²), leading to the overdiagnosis of CKD in individuals with mild renal impairment or normal kidney function [32]. This bias is particularly pronounced in younger populations, women, and non-Black individuals, as demonstrated in large cohort studies [33]. In elderly populations (>70 years), the MDRD and CKD-EPI 2009 equations show minimal differences, but CKD-EPI 2009 may even underestimate GFR in the very elderly (>90 years), complicating CKD staging in this demographic [33]. Furthermore, older formulae do not account for standardized creatinine assays, which became widespread after 2006, potentially introducing variability in historical comparisons [32]. In obese individuals, reliance on serum creatinine poses challenges, as altered body composition may skew results, necessitating alternative equations or cystatin C-based estimates [34]. The CKD-EPI 2009 equation, while more accurate overall, still exhibits racial bias due to its inclusion of race coefficients, prompting recent revisions to race-free equations that integrate cystatin C for improved precision across diverse populations [11]. These advancements highlight the need for context-specific formula selection, emphasizing cystatin C or combined creatinine-cystatin C equations in obese populations, elderly populations, and non-Black subgroups to mitigate systematic errors and ensure equitable clinical decision-making [11], 34].

The discrepancies observed in our study have important clinical implications. The overestimation of GFR by creatinine-based formulae can lead to delayed diagnosis and challenges in CKD management, particularly in older adults and individuals with normal creatinine values. The use of cystatin C-based equations or exogenous markers may improve the accuracy of GFR estimations and enhance clinical decision-making in these populations.

Our study had some limitations. The retrospective design of the study limited our access to comprehensive clinical data, such as additional laboratory results and detailed information on comorbid chronic conditions. The single-center nature of the study may limit the generalizability of our findings. Additionally, there was no information on whether all patients had been evaluated under optimal conditions, such as adequate hydration status, which may have influenced GFR measurements. Future multicenter studies with standardized protocols are needed to validate our results and enhance their applicability to diverse clinical settings.

Conclusions

The findings of this study suggest that in patients older than 60 years, the CG formula may serve as a suitable alternative when GFR cannot be determined by cystatin C or ^99mTc-DTPA. These results, which differ from some findings in the literature, may be particularly significant given the large patient cohort included in this study or the variation in creatinine measurements. In older adults, serum creatinine levels and creatinine-based GFR estimation may not reliably reflect renal function due to age-related physiological changes, such as reduced muscle mass and creatinine production. A formula such as the weight-corrected Cockroft-Gault equation may overcome these limitations when GFR cannot be determined by cystatin C or ^99mTc-DTPA.