ditorial 

Divergent GFR Estimates and Clinical Implications

Menaka Sarav, Robert J. Alpert

JAMA Published Online: November 7, 2025

doi: 10.1001/jama.2025.14043

Chronic kidney disease (CKD) affects more than 800 million people worldwide, significantly increasing the risk for kidney failure, cardiovascular disease, and all-cause mortality.¹ Accurate assessment of glomerular filtration rate (GFR) is essential for identifying and managing CKD. Although GFR is most accurately measured by analyzing the rate of kidney excretion of infused exogenous substances, such as inulin or iohexol, this is not practical in most cases. Therefore, estimated glomerular filtration rates (eGFRs) have become important for CKD diagnosis and management.

Creatinine-based eGFR (eGFRcr) is currently the standard method for estimating GFR because it is widely available, cost-effective, and familiar to clinicians. Serum creatinine levels depend in part on muscle mass, creating variability across age, sex, and nutritional status. Additionally, creatinine undergoes variable tubular secretion, which is influenced by medications such as cimetidine and trimethoprim, declining kidney function, and individual physiology, which can compromise eGFRcr accuracy.²

This has led to the use of plasma cystatin C as a potentially improved approach for estimating GFR (eGFRcys). Cystatin C is a member of the cystatin family of proteins that bind and inhibit lysosomal cysteine proteases.³ It has a low molecular weight, is composed of 120 amino acid residues, and is synthesized continuously by all nucleated cells. It is not protein bound and is positively charged, allowing it to be freely filtered by the glomerulus. Filtered cystatin C is primarily reabsorbed in the kidney proximal tubule where it is metabolized. Because cystatin C is produced at a constant rate and freely filtered at the glomerulus, its plasma levels are a measure of GFR. Unlike creatinine, plasma levels of cystatin C are not affected by muscle mass.

However, cystatin C levels can increase in inflammation, hyperthyroidism, adiposity, corticosteroid administration, and malignancy,⁴ which may affect its accuracy as a marker for GFR. Additionally, the measurement of cystatin C is significantly more expensive than creatinine (cost of reagents per test: $4 vs $0.20), which may limit its routine use.⁵

Standard Chronic Kidney Disease Epidemiology Collaboration formulas, which utilize plasma creatinine (eGFRcr), plasma cystatin C (eGFRcys), and the combination of both (eGFRcr-cys), together with age, sex, and empirically derived parameters, have been developed.⁶ Utilizing these formulas, several studies have found that while these estimates are sometimes consistent, often the values display significant discrepancies. Fu et al compared measurements of eGFRcr, eGFRcys, and eGFRcr-cys with GFR measured as plasma clearance of iohexol in patients with concurrent measurements.⁷ In 49% of the samples, eGFRcr was significantly higher than eGFRcys and associated with an overestimation of GFR by eGFRcr and an underestimation by eGFRcys, compared with the iohexol-measured GFR. In 11% of samples, eGFRcys exceeded eGFRcr, associated with an overestimation of GFR by eGFRcys and an underestimation by eGFRcr. In both scenarios, the combined eGFRcr-cys provided estimates closer to the actual iohexol-measured GFR. These results were consistent with findings from Wang et al, which reviewed 12 studies with similar comparisons.⁸ These findings highlight the variability in eGFR estimates and the potential benefit of combining both markers for greater accuracy in the GFR estimating equations.

In the current issue of JAMA, Estrella et al present a meta-analysis involving 821 327 individuals from 23 ambulatory cohorts and 39 639 individuals from 2 inpatient cohorts, assessing the prevalence and implications of discordant eGFRcys and eGFRcr.⁹ They found that 11% of ambulatory patients and 35% of inpatients exhibited substantial negative eGFR differences, defined as an eGFRcys that was at least 30% lower than eGFRcr. A total of 3.8% of ambulatory patients and 14.5% of inpatients exhibited substantial positive eGFR difference, defined as an eGFRcys at least 30% higher than eGFRcr. Importantly, Estrella et al found that because of these discrepancies, use of eGFRcys compared with eGFRcr frequently led to reclassification to a more advanced stage of CKD. Consistent with prior literature, the proportion of measurements in which eGFRcys was more than 30% lower than eGFRcr varied widely among patients and across cohorts, ranging from 3% to 50%. Older age, smoking, and comorbid conditions, including heart failure, peripheral artery disease, chronic obstructive pulmonary disease, liver disease, and obesity, were associated with greater negative discordance. When interpreting the results of this study, it is important to acknowledge that GFR was not directly measured. Thus, it remains unclear whether the observed greater negative discordance was primarily due to inaccuracies in eGFRcr or eGFRcys. A large negative discordance between eGFRcys and eGFRcr (ie, eGFRcys ≥30% lower than eGFRcr) was associated with increased risk of all-cause (28.4 vs 16.8 per 1000 person-years; hazard ratio [HR], 1.69) and cardiovascular (6.1 vs 3.8 per 1000 person-years; HR, 1.61) mortality, atherosclerotic cardiovascular disease (13.3 vs 9.8 per 1000 person-years; HR, 1.35), heart failure (13.2 vs 8.6 per 1000 person-years; HR, 1.54), and kidney failure requiring replacement therapy (2.7 vs 2.1 per 1000 person-years; HR, 1.29).⁹

The Estrella et al study highlights the frequency of large discordances between eGFRcr and eGFRcys and extends previous research by quantifying the association of eGFR discordance with long-term health outcomes. The numerical difference between eGFR, calculated from eGFRcys and eGFRcr, is not a biological entity or clinical marker. Rather, when eGFRcys and eGFRcr differ significantly, the discrepancy typically arises from factors that preferentially affect one marker over the other or affect them in opposite directions. For instance, inflammation, adiposity, thyroid dysfunction, and steroid use primarily influence cystatin C levels, altering eGFRcys.⁴ In contrast, muscle mass, medications such as cimetidine and trimethoprim, and dietary protein intake predominantly affect creatinine levels, altering eGFRcr.² Understanding the underlying components associated with the discordance between eGFRcys and eGFRcr is essential for accurate clinical interpretation, as discrepancies may reflect non–kidney-related influences rather than true differences in kidney function.

Large negative discordance between eGFRcys and eGFRcr, when eGFRcys is at least 30% lower than eGFRcr, and the association with older age, cigarette smoking, and comorbid conditions, could be related to effects of decreased muscle mass on creatinine production, leading to a calculated eGFRcr greater than measured GFR. However, the study by Fu et al also suggests that these patients could have an eGFRcys lower than measured GFR, implying an increased rate of cystatin C production⁷ possibly related to effects of inflammation or stress. Similarly, the association of a large negative discordance between eGFRcys and eGFRcr with all-cause mortality and cardiovascular and kidney disease may be attributable to low muscle mass, reflected in the higher eGFRcr, and/or to the presence of chronic inflammation and stress, reflected in the low eGFRcys.

Furthermore, Chen et al have shown that longitudinal analysis of patients whose eGFRcys declined more quickly than eGFRcr (widening the negative discordance) revealed an 8.2-fold higher risk of mortality regardless of baseline discordance, suggesting that annual measurements of both creatinine and cystatin C provided more comprehensive insights into health status changes than a single cystatin C measurement, as current guidelines suggest.⁵While these findings are promising, more data are needed to support integrating sequential measurements of the difference between eGFRcys and eGFRcr into daily clinical use.

Considering the frequency and importance of large discrepancies between eGFRcys and eGFRcr when eGFRcys is lower than eGFRcr by 30% or more, the question arises as to when measuring cystatin C levels is most appropriate. Given its higher cost, cystatin C cannot fully replace creatinine measurement. However, it is valuable at specific times to improve diagnostic and prognostic accuracy. The KDIGO (Kidney Disease: Improving Global Outcomes) 2024 guidelines emphasize the importance of integrating cystatin C into routine practice, representing a significant advancement in CKD management.¹⁰ Measuring plasma cystatin C allows for the calculation of eGFRcys and eGFRcr-cys, providing greater confidence in assessing a patient’s kidney function and classifying the severity of CKD. Additionally, a high negative discordance between eGFRcys and eGFRcr offers important information regarding the patient’s risk for death, cardiovascular events, or kidney disease progression.

The primary limitation is not determining the clinical utility of cystatin C testing, as this study and others reinforce the valuable insights from eGFRcys, but rather the issue is addressing the implementation barriers, including cost, availability, and clinical awareness. Health care systems that have successfully implemented in-house cystatin C testing with rapid turnaround times have observed marked increases in use and improved clinical decision-making and operational workflow.¹¹

Future research should focus on standardizing eGFR estimation techniques, exploring the cost-effectiveness of routine cystatin C testing, and evaluating the effects of measurement on long-term patient outcomes. Addressing implementation barriers through clinician education, workflow improvements, and integration into electronic health records is important to fully realize the benefits of cystatin C testing. By embracing new methods for improving accuracy of kidney function assessment and monitoring, the future of CKD management appears promising, offering enhanced diagnostic precision and the potential for improved patient outcomes.