Sarcopenia, adiposity and large discordance between cystatin C and creatinine‐based estimated glomerular filtration rate in patients with cancer

Abstract Background Creatinine‐based estimated glomerular filtration rate (eGFRCRE) may overestimate kidney function in patients with sarcopenia. While cystatin C‐based eGFR (eGFRCYS) is less affected by muscle mass, it may underestimate kidney function in patients with obesity. We sought to evaluate the relationship between body composition defined by computed tomography (CT) scans and discordance between creatinine, eGFRCRE and eGFRCYS in adult patients with cancer. Methods This study is a cross‐sectional study of consecutive adults with cancer with an abdominal CT scan performed within 90 days of simultaneous eGFRCRE and eGFRCYS measurements between May 2010 and January 2022. Muscle and adipose tissue cross‐sectional areas were measured at the level of the third lumbar vertebral body using a validated deep‐learning pipeline. CT‐defined sarcopenia was defined using independent sex‐specific cut‐offs for skeletal muscle index (<39 cm2/m2 for women and <55 cm2/m2 for men). High adiposity was defined as the highest sex‐specific quartile of the total (visceral plus subcutaneous) adiposity index in the cohort. The primary outcome was eGFR discordance, defined by eGFRCYS > 30% lower than eGFRCRE; the secondary outcome was eGFRCYS > 50% lower than eGFRCRE. The odds of eGFR discordance were estimated using multivariable logistic regression modelling. Unadjusted spline regression was used to evaluate the relationship between skeletal muscle index and the difference between eGFRCYS and eGFRCRE. Results Of the 545 included patients (mean age 63 ± 14 years, 300 [55%] females, 440 [80.7%] non‐Hispanic white), 320 (58.7%) met the criteria for CT‐defined sarcopenia, and 136 (25%) had high adiposity. A total of 259 patients (48%) had >30% eGFR discordance, and 122 (22.4%) had >50% eGFR discordance. After adjustment for potential confounders, CT‐defined sarcopenia and high adiposity were both associated with >30% eGFR discordance (adjusted odds ratio [aOR] 1.90, 95% confidence interval [CI] 1.12–3.24; aOR 2.01, 95% CI 1.15–3.52, respectively) and >50% eGFR discordance (aOR 2.34, 95% CI 1.21–4.51; aOR 2.23, 95% CI 1.19–4.17, respectively). A spline model demonstrated that as skeletal muscle index decreases, the predicted difference between eGFRCRE and eGFRCYS widens considerably. Conclusions CT‐defined sarcopenia and high adiposity are both independently associated with large eGFR discordance. Incorporating valuable information from body composition analysis derived from CT scans performed as a part of routine cancer care can impact the interpretation of GFR estimates.


Introduction
Accurate assessment of the estimated glomerular filtration rate (eGFR) is critical, especially in patients with cancer who commonly receive renally cleared medications with narrow therapeutic indices.Estimating GFR using serum creatinine (SCr) remains the most widely used method in both clinical practice and research. 1,2Yet, because creatinine is a byproduct of muscle metabolism, creatinine production is lower in patients with sarcopenia, a common condition affecting patients with cancer. 2,3As a result, creatinine levels may be falsely low in patients with sarcopenia and lead to overestimation of GFR.Cystatin C is a low-molecular-weight (13 K Dalton) protein produced by all nucleated cells that is filtered by the glomerulus and is not reabsorbed or secreted. 4nlike SCr, cystatin C is not readily affected by age, sex, muscle mass or diet and has been increasingly used as an alternative to SCr to estimate GFR, with some caveats (e.g., cystatin C is elevated in obesity, uncontrolled thyroid disease and acute inflammatory states).S1 Cancer cachexia is a multifactorial syndrome characterized by the loss of skeletal muscle leading to progressive functional impairment, which is strongly associated with increased mortality. 5However, recent studies have shown that visibly wasted patients are increasingly rare in clinical practice; a clinical phenotype of 'sarcopenic obesity' has emerged to describe patients with a high body mass index (BMI) but with underlying depleted muscle mass.S2 Therefore, clinical impressions of cachexia or BMI are inadequate markers for sarcopenia in patients with cancer, and more precise evaluation of body composition using imaging techniques has been increasingly used in clinical practice over the past 15 years. 6,7omputed tomography (CT) scans are widely used for body composition assessment 8,9 ; the cross-sectional areas of tissues on single images at the level of the third lumbar vertebra (L3) strongly correlate with whole body adipose tissue, muscle and lean tissue mass. 10,11Using different diagnostic thresholds, studies have shown that 40-60% of patients with locally advanced or metastatic cancer had imaging-defined sarcopenia.][14] Given that patients with active cancer undergo frequent CT scans for diagnosis and monitoring of disease progression, CT is a convenient tool to evaluate body composition and quantify skeletal muscle and adipose tissue and, thus, may aid in the accurate interpretation of SCr or cystatin C-based eGFR (eGFR CRE and eGFR CYS ).Because patients with sarcopenia may have a falsely low SCr and obesity is associated with elevated cystatin C, we hypothesized that CT-defined sarcopenia and high adiposity will each be associated with a large discordance between eGFR CRE and eGFR CYS .Our objective was to evaluate the association between body composition (skeletal muscle index [SMI] and adiposity index) assessed by abdominal CT scan and discordance between eGFR CRE and eGFR CYS in patients with cancer.

Patient population and study design
Using the centralized data warehouse of our health system, the Research Patient Data Registry (RPDR), we designed a retrospective cohort study of consecutive adult patients with a pre-existing diagnosis of malignancy who had simultaneous measurements of SCr and cystatin C as a part of routine care between May 2010 and January 2022.We defined the baseline as the date of the simultaneous eGFR CRE and eGFR CYS measurements.In patients with multiple simultaneous assessments, the first instance was used.We calculated eGFR CRE using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 2021 race-free equation 2 and eGFR CYS using the CKD-EPI 2012 race-free equation. 15,16We included all patients who had an abdominal CT scan performed within 90 days before or after the baseline date.We excluded patients with eGFR CRE < 15 mL/min/1.73m 2 and those whose CT scans failed quality control for the determination of SMI (Table S1).

Data collection
Comorbidities were defined based on diagnosis codes appearing any time prior to the baseline date.Corticosteroid use was defined by an active prescription within 30 days of the baseline date.The cancer type was determined by the most frequently coded cancer-related diagnosis prior to the baseline.Cancer stage was determined by chart review and defined as early stage versus locally advanced or advanced (stage 3 or 4) based on oncologists' documentation.Patients who had previously completed antineoplastic treatment or underwent surgery without evidence of disease as documented by their oncologist were considered to be in remission.Baseline chronic kidney disease was defined by the CKD-EPI 2021 race-free equation that incorporates both SCr and cystatin C (eGFR CRE-CYS < 60 mL/min/1.73m 2 ). 17

Body composition analysis on computed tomography scans
9][20][21] Each segmentation label map was independently reviewed by a trained analyst with 2 years' experience performing body composition analysis blinded to clinical outcomes, including eGFR CRE and eGFR CYS (IT).We calculated SMI (cm 2 /m 2 ) by dividing the skeletal muscle cross-sectional area by the patient's height squared.Visceral and subcutaneous fat areas were normalized for height (divided by height squared) to calculate the visceral adiposity index (VAI) and the subcutaneous adiposity index (SAI) in cm 2 /m 2 (Figure 1).The total adiposity index was calculated by adding VAI and SAI. 6,21When possible, we selected outpatient CT scans; if no outpatient CT scan was available within 90 days, we included the inpatient CT scan closest to the baseline date.
Primary exposure CT-defined sarcopenia was defined using previously published SMI cut-offs of <39 cm 2 /m 2 for women and <55 cm 2 /m 2 for men. 6High adiposity was defined as the highest sex-specific quartile of total adiposity index (VAI plus SAI), as no cut-offs for low or high adiposity index exist.

Primary outcome
The primary outcome was eGFR discordance, defined as eGFR CYS more than 30% lower than the eGFR CRE .The reference group consisted of all patients who did not meet the criteria for eGFR discordance.The 30% cut-off was chosen as it is commonly used in clinical studies to define the accuracy of eGFR compared with measured GFR. 22,23

Secondary outcomes
To evaluate a more severe eGFR discordance, we additionally identified a subset of patients with eGFR CYS that was more than 50% lower than eGFR CRE .Finally, to assess the absolute eGFR difference (eGFR DIFF ), we subtracted eGFR CRE from eGFR CYS ; negative values indicated that the eGFR CRE was higher than the eGFR CYS .We evaluated the continuous relationship between eGFR DIFF and SMI.We defined a large eGFR DIFF as having an eGFR CYS more than 15 mL/min/ 1.73 m 2 lower than eGFR CRE . 24gure 1 Body composition analysis at the level of the third lumbar vertebral body.Representative axial computed tomography images at the level of the third lumbar vertebral body (left column) and matching segmentation label maps (right column).Patient A was a 26-to 30-year-old male with a skeletal muscle (red) index (SMI) of 65.9 (4th quartile), a subcutaneous adiposity (green) index (SAI) of 38.6 (2nd quartile) and a visceral adiposity (yellow) index (VAI) of 25.8 (1st quartile).Patient B is a 46-to 50-year-old female with an SMI (red) of 18.9 (1st quartile), an SAI (green) of 77.3 (2nd quartile) and a VAI (yellow) of 36.92 (3rd quartile).

Statistical analysis
We reported baseline characteristics for the entire cohort by SMI quartile using counts and percentages for categorical variables, means with standard deviations (±SD) for normally distributed continuous variables and median and interquartile range (IQR) for skewed variables.Variables with missing values were imputated using the multiple imputation method; five imputed datasets were obtained through chained equations, and parameter estimates from each imputed dataset were pooled using Rubin's rules.S8 We examined the unadjusted associations between baseline demographics, comorbidities, medications, laboratory studies and body composition (CT-defined sarcopenia and high adiposity) with eGFR discordance.We fit a multivariable logistic regression model to determine the association between eGFR discordance (>30% and >50%) and body composition, accounting for potential confounders.For the sensitivity analyses, we fit a multivariable logistic regression model to determine the association between the lowest sex-specific SMI quartile and >30% eGFR discordance and fit a multivariable logistic regression model to determine the association between CT-defined sarcopenia and large eGFR DIFF (eGFR CYS more than 15 mL/min/1.73m 2 lower than eGFR CRE ), accounting for potential confounders.Haemoglobin and serum albumin were evaluated in clinically relevant categories.Unadjusted spline regression was performed to evaluate the relationship between SMI and eGFR DIFF (eGFR CYS minus eGFR CRE ) using the B-spline basis for cubic polynomial splines with three degrees of freedom.
To address bias, we also performed additional sensitivity analyses excluding patients with inpatient CT scans, liquid tumours (lymphoma, leukaemia or myeloma) and acute kidney injury (AKI) at the time of eGFR CRE and eGFR CYS assessments.AKI was defined as an SCr that was 50% higher than the lowest SCr in the 365 days prior to baseline.
All comparisons were two-sided, with P < 0.05 considered significant.All analyses were performed using R 4.1.1(R Foundation), SAS 9.4 (SAS Institute) and GraphPad Prism V.9.1.0(GraphPad Software).The design and reporting of this cohort study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for reporting observational studies.

Patient and computed tomography scan characteristics
Of the 1988 patients with cancer originally included in this study, 662 had one or more abdominal CT scans within 90 days of the baseline date.We excluded 46 patients with eGFR CRE < 15 mL/min/1.73m 2 and 48 patients whose scans failed to accurately segment skeletal muscle or failed quality control (Figure S1 and Tables S1 and S2).The included 545 patients had a median age of 64 years (IQR 55-72), 45% were male and 80.7% were non-Hispanic white (Table 1).Patients had a wide array of cancer types (Table S3); the most common cancer types were breast (12.5%), gynaecological (11.4%) and gastrointestinal (10.8%).Most patients (N = 338; 62%) had advanced disease.The mean BMI was 27.6 kg/m 2 (SD ± 7.2).Most CT scans (N = 375; 69%) were performed in the outpatient setting, and N = 334 (61%) were performed with intravenous contrast.
We evaluated the relationship between CT-defined sarcopenia and the eGFR differences (eGFR DIFF ) defined by subtracting eGFR CRE from eGFR CYS (negative values mean that eGFR CYS is lower than eGFR CRE ).The mean eGFR DIFF was À20.9 ± 25.1 mL/min/1.73m 2 in patients with CT-defined sarcopenia compared with À8.6 ± 22.4 mL/min/1.73m 2 in patients without CT-defined sarcopenia (P < 0.001), and 247 (45%) of patients met the definition of large eGFR DIFF (eGFR CYS more than 15 mL/min/1.73m 2 lower than eGFR CRE ).A sensitivity analysis also demonstrated an independent association between CT-defined sarcopenia and a large eGFR difference (aOR 1.97, 95% CI 1.13-3.43,P = 0.017; Table S6).A spline model showing the continuous relationship between SMI and eGFR DIFF demonstrated that as SMI decreased, the predicted difference between eGFR CRE and eGFR CYS widened considerably (Figure 6).

Sensitivity analyses
Exclusion of patients with AKI, inpatient CT scans or liquid tumours did not meaningfully alter the relationship between SMI quartiles and eGFR discordance (Figure S4A-C, respectively).

Discussion
In a cohort of patients with cancer, we found that CT-defined sarcopenia and high adiposity, defined by CT scan, are independent risk factors for eGFR discordance.Furthermore, the lower a patient's SMI, the wider the discordance between eGFR CRE and eGFR CYS .Accurate assessment of eGFR is important but challenging in patients with cancer due to the high prevalence of sarcopenia and sarcopenic obesity.Patients with cancer are commonly exposed to antineoplastic therapies or supportive medications (antibiotics, anticonvulsants and anticoagulants) that require dose adjustments based on kidney function; therefore, a large discordance between eGFR CRE and eGFR CYS poses significant challenges in clinical management and may be associated with adverse medication events and lower survival. 257][28][29] We hypothesize that medication overdose due to inaccurate estimation of GFR when relying on eGFR CRE in patients with sarcopenia may cause dose-related toxicity in renally excreted chemotherapies 26 and may thereby contribute to the decreased survival in these patients.Body composition analysis using CT scans is a more sensitive and specific way of assessing the quantity and distribution of skeletal muscle and adipose tissue than BMI. 6,10,11rior studies have suggested that obesity may increase cystatin C levels independent of kidney function, which leads to an underestimation of GFR using the cystatin C equation.However, this observation is largely based on obesity defined by BMI rather than a detailed analysis of body composition. 30rior studies utilized clinical and anthropomorphic measurements (BMI, muscle strength, skin fold thickness, etc.) as an assessment of sarcopenia and obesity, which may not as accurately reflect the body composition of muscle and adipose tissue.S9-S11 As cystatin C is secreted by nucleated cells and constitutively expressed in almost all organs, 31 higher body cell mass, which correlates with a higher BMI, could lead to a higher cystatin C level independent of kidney function.A study in children showed that adjustment with body cell mass, estimated by bioimpedance, increased the accuracy of cystatin C-based GFR estimation. 32Other studies have suggested that increased cystatin C mRNA expression in adipose tissue may be contributing to higher cystatin C levels in obese individuals. 33To our knowledge, this is the first study that demonstrated an association between CT-defined total adiposity index and a higher odds of discordance between eGFR CRE and eGFR CYS .
Additionally, we found that patients with liquid tumours have significantly higher rates of eGFR discordance than patients with solid tumours.This possibly reflects a higher cancer cell burden in liquid tumours that may lead to higher cystatin C levels, 34 suggesting that there is a complex relationship between serum biomarkers and individual disease states, and future studies are needed to best individualize GFR estimation in patients with cancer.
Other recent studies demonstrate an association between large differences in eGFR CRE and eGFR CYS and adverse clinical outcomes in non-cancer populations, including frailty, falls, hospitalizations and death. 35,36Of note, the rate of large eGFR discordance in our cohort was significantly higher than prior studies conducted in populations without cancer (Figure S3B), suggesting that the high prevalence of sarcopenia in patients with cancer is an important risk factor for discordance between eGFR CRE and eGFR CYS . 35,36A recent study of 327 patients in China with resectable gastric cancer showed an association between a lower serum creatine/ cystatin C ratio and the presence of imaging-defined sarcopenia. 37Our cohort included patients with various tumour types and stages and found an independent relationship between both CT-defined sarcopenia and high adiposity with eGFR discordance.A study of 22 316 US veterans who had a simultaneous SCr and cystatin C showed that patients whose creatinine to cystatin C ratio was ≤0.75 had a higher mortality risk (3.03 [2.81-3.27])compared with those with a creatinine to cystatin C ratio of 1.00 to <1.25; this affirms that the relationship between creatinine and cystatin C (or eGFR CRE and eGFR CYS ) may be an important indicator of health status independent of kidney function. 38ur study has several limitations.First, cystatin C was ordered as part of clinical care, which likely selected a population in whom clinicians questioned the accuracy of creatine-based eGFR; therefore, the rate of eGFR discordance in our study is likely to be higher than in the general cancer population.Second, our study included a significant number of patients who were hospitalized or had unstable kidney function at the time of their CT scan, suggesting that a significant fraction of patients were experiencing acute illness.However, sensitivity analyses excluding patients with inpatient CT scans and AKI did not meaningfully change the relationship between CT-defined sarcopenia and eGFR discordance.Third, we used a single measurement of SCr and cystatin C, which may not reflect a steady state at the time of measurement.Although these may limit the generalizability of our findings, these variables were adjusted for in the multivariable model.Fourth, a subset of patients either did not have a CT scan or were excluded during quality assurance.Of note, automated segmentation was not possible in patients with soft tissue oedema, which particularly affects the subcutaneous adipose tissue metric.Fifth, we evaluated adiposity quartiles as previously used by others 39 ; however, this may affect the generalizability of our definition of 'high adiposity' across other populations.Sixth, CT-defined sarcopenia does not quantify muscle quality or strength, important aspects of the clinical definition of sarcopenia that more accurately predict adverse outcomes. 40Finally, our study lacks the gold standard (measured GFR) given the retrospective design; prospective studies are needed to validate the accuracy of GFR estimating equations in consecutive patients with sarcopenia and high adiposity against measured GFR.
In conclusion, we demonstrate that CT-defined sarcopenia is associated with a higher risk of eGFR discordance; as SMI decreases, the discordance between eGFR CRE and eGFR CYS widens considerably.High adiposity is also associated with a higher risk of large eGFR discordances.Deep learning models facilitate fully automated body composition analysis on abdominal CT scans done as a part of routine cancer care, allowing for the potential for widespread clinical use in the future.Patients with altered body composition may require measured GFR to determine an accurate GFR for drug dosing. 6,10,11,21Future studies that account for body composition are needed to improve and personalize the approach to GFR estimation in patients with cancer who rely on accurate eGFR to guide chemotherapy dosing.

Figure 2 ( 6 Figure 3
Figure 2 (A, B) Histograms showing the distribution of skeletal muscle index (SMI) by sex.The red shaded area denotes the sex-specific cut-offs for sarcopenia as described by Prado et al.6

Figure 6
Figure 6Spline model predicting the relationship between skeletal muscle index (SMI) and estimated glomerular filtration rate (eGFR) difference (eGFR CYS minus eGFR CRE ).Unadjusted spline regression was performed to evaluate the relationship between SMI and eGFR DIFF (eGFR CYS minus eGFR-CRE ) using the B-spline basis for cubic polynomial splines with three degrees of freedom using the 10th to 90th SMI percentile for the cohort.Lower SMI values indicate a greater degree of sarcopenia.

Table 1
Baseline patients' characteristics Note: Counts and percentages or median and interquartile ranges are shown.Missing data: Serum albumin is missing for 4 patients (0.7%), the visceral adipose index is missing for 9 patients (1.7%) and the subcutaneous adipose index is missing for 77 patients (14.1%).Abbreviations: eGFR, estimated glomerular filtration rate; SCr, serum creatinine; SMI, skeletal muscle index.

Table 2
Predictors of estimated glomerular filtration rate discrepancy Note: eGFR discrepancy is defined by eGFR CYS > 30% lower than eGFR CRE .Baseline eGFR CRE-CYS was defined using the Chronic Kidney Disease Epidemiology Collaboration 2021 race-free combined cystatin C and creatinine equation.Abbreviations: CI, confidence interval; CT, computed tomography; eGFR, estimated glomerular filtration rate; HIV, human immunodeficiency virus; OR, odds ratio; REF, reference.aCorticosteroidusewasdefinedwithin30 days of baseline.Figure 4Forest plot illustrating the strength of associations with 30% and 50% estimated glomerular filtration rate (eGFR) discrepancies.The final model for both >30% and >50% eGFR discrepancies is adjusted for the variables shown in Table2.The unadjusted and multivariable logistic regression model for >50% eGFR discrepancy is shown in TableS4.CI, confidence interval; REF, reference group.Journal of Cachexia, Sarcopenia and Muscle 2024; 15: 1187-1198 DOI: 10.1002/jcsm.13469