Lean Mass Index Adjustment of eGFR Strengthens Associations with Cardiometabolic Risk Factors

1. Introduction

It is now well recognized that the clinical significance of CKD extends beyond drug dosing and renal replacement therapy and that it is a potent, independent driver of cardiometabolic risk, including hypertension, dyslipidemia, and cardiovascular mortality.[1,2,3,4]

The accurate assessment of glomerular filtration rate (GFR) is fundamental to the diagnosis, staging, and management of chronic kidney disease (CKD). Traditionally, estimated GFR (eGFR) derived from serum creatinine (SCr) serves as the primary tool in clinical practice due to its cost-effectiveness and accessibility. However, the reliability of SCr as a filtration marker is intrinsically limited, most notably but not exclusively by individual variations in skeletal muscle mass, which serves as the primary reservoir for creatine and its byproduct, creatinine.[5,6]

Standard eGFR equations, such as the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) 2009 and 2021 formulae, attempt to account for muscle mass variations by using surrogate demographic markers like age and sex.[7,8] Despite these adjustments, significant intra-individual variability persists. In patients with abnormally high or low muscle mass—common in aging populations, wasting diseases, or young athletes—creatinine-based eGFR can lead to significant misclassification.[9,10,11] While these limitations can be ameliorated by the inclusion of Cystatin C, a now widely available more modern filtration marker, misclassification of GFR from SCr-based estimates likely limit our understanding of historic CKD trajectories, health risks associated with CKD as well as of differences in current and past trends in prevalence.[12]

Several methods are available for the quantification of an individuals’ muscle mass and Dual-energy X-ray absorptiometry (DEXA) is currently one of the most widely used high-accuracy technique for quantifying body composition.[13] By calculating the Lean Mass Index (LMI)—lean mass divided by height squared— a standardized, validated metric of muscularity can be obtained.[14,15] Integrating LMI into the interpretation of serum creatinine may offer a physiological correction that demographic surrogates cannot reasonably achieve.

While previous studies have explored the impact of muscle mass on creatinine levels and the usefulness of creatinine levels as markers of muscle mass in the proven or assumed absence of CKD, it is unknown whether a LMI-adjusted serum-creatinine based eGFR (eGFR_LMISCr) improves the stratification of cardiometabolic risk compared to standard, unadjusted serum-creatinine based eGFR (eGFR_Scr).[9,16,17]

Based on these considerations, we hypothesize that LMI-adjusted eGFR will provide a more accurate reflection of a patient's renal-metabolic status and will demonstrate a stronger association with markers of cardiometabolic risk than traditional creatinine-based eGFR in a well-characterized Austrian general population cohort.

2. Materials and Methods

Study Design

The study was conducted within the framework of the Austrian LEAD (Lung, hEart, sociAl, boDy) study (ClinicalTrials.gov identifier: NCT07127518) which was approved by the Ethics Committee of the City of Vienna (Ethikkomission der Stadt Wien; protocol number EK-11-117-0711). All study procedures were performed in accordance with the Declaration of Helsinki, and written informed consent was obtained from all participants and/or their legal guardians. LEAD is an ongoing longitudinal, observational, population -based cohort study with repeated follow-up assessments conducted from 2011 onward (mean interval between visits 1 and 2: 4.25 ± 0.33 years). All participants aged ≥18 years with valid and no missing creatinine values were included in the present study; participants on dialysis or taking related medications were excluded.

Measurements

Lean mass-adjusted creatinine was derived to account for interindividual differences in muscle mass. Sex- and age-specific reference values for LMI were calculated using the median DEXA-derived LMI within strata defined by sex and single-year age. Individual serum creatinine concentrations were then normalised to the corresponding reference LMI using the following formula:

$$LMISCr=SerumCreatinine*{\displaystyle \frac{LMIof50thageandsexspecificLMIpercentile}{LMIofindividualparticipant}}$$

eGFR was calculated using both raw and LMI-adjusted serum creatinine values. The CKD-EPI 2009 equation was applied using sex- and age-specific coefficients.[7] For categorical analyses, eGFR values were classified according to Kidney Disease: Improving Global Outcomes (KDIGO) GFR categories, with a collapsed five-category classification, combining the two lowest GFR strata due to small sample sizes. Five-category groupings were defined as: <10, 30-<45, 45-<60, 60-<90 and ≥90 ml/min/1.73m2.[5] For regression modelling, KDIGO categories were encoded with ≥90 ml/min/1.73m2 as the reference group.

Height, weight and waist circumference were measured using a calibrated stadiometer and precision scale and were used to calculate body mass index (BMI, kg/m2) and waist-to-height ratio. Fat mass index (FMI) was obtained using DEXA scans. Participant characteristics including age, smoking status and cumulative smoking exposures were collected using standardised, interviewer-administered questionnaires.

Metabolic syndrome was defined according to International Diabetes Federation (IDF) criteria, based on central obesity, triglyceride levels, HDL cholesterol, blood pressure and fasting glucose.[18] Dyslipidemia was defined as at least one of the following: LDL-C ≥100mg/dL, triglycerides ≥150mg/dL, total cholesterol ≥200mg/dL, sex-specific low HDL-C or statin use. Obesity was defined based on the Lancet Diabetes and Endocrinology Commisssion (Obesity_LDEC) or WHO (Obesity_WHO; BMI ≥30kg/m2) definitions: participants were classified as having obesity_LDEC if they had BMI ≥40kg/m2; BMI ≥30kg/m2 in combination with either a waist-to-height ratio ≥0.50 or sex-specific waist circumference thresholds (≥102cm in men or ≥88cm in women); or BMI <30kg/m2 in combination with both elevated waist-height ratio and sex-specific waist circumference thresholds.[19,20] Prediabetes and diabetes were classified using fasting plasma glucose and HbA1c thresholds, and hypertension was defined as systolic blood pressure ≥140mmHg, diastolic blood pressure ≥90mmHg, or current use of anti-hypertensive medication. Cardiovascular disease was defined as self-reported physician-diagnosed myocardial infarction, coronary artery disease or heart failure.

Statistical Analysis

All statistical analyses were performed in R (version 4.3.0; R foundation for Statistical Computing, Vienna, Austria), with the level of significance set at <5%. Data are presented as mean ± standard deviation or relative frequency (percentage of group). Group comparisons between eGFR-CKD stages were performed using one-way ANOVA or t-tests for continuous variables and χ² tests for categorical variables. Age and sex effects on eGFR or eGFR-CKD stages were assessed using linear and logistic regression models and p-values were obtained by Wald’s test. Reclassification of eGFR-CKD stages were visualised as an alluvial plot. To compare the strength of association between outcomes and eGFR definitions, we estimated the difference in log-odds ratios between the two predictors using bootstrap resampling. For each outcome, separate age- and sex-adjusted logistic regression models and the difference in log(OR) between standardised raw and adjusted eGFR was evaluated using 1000 bootstrap samples to obtain confidence intervals and two-sided p-values.

Individual eGFR slopes were calculated for participants with complete and valid serum creatinine measurements at both study visits (n=7487). Annualised eGFR change was derived as the difference in eGFR between visits divided by the time interval and was computed separately for eGFR based on raw and LMI-adjusted creatinine. To compare longitudinal decline between raw and LMI-adjusted eGFR, slopes were assessed in a mixed-effects linear model including eGFR definitions, age, and sex, with a random intercept for participant to account for within-person pairing.

Data Availability Statement

Due to institutional agreements and ongoing data collection activity, unrestricted access cannot be granted to our primary data. However, data specifically pertaining to this manuscript may be made available upon reasonable request made to the corresponding author.

Use of Generative AI Statement

During the preparation of this work the authors used Google Gemini 3 to improve sentence structure in some instances. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the content of the published article.

3. Results

The study cohort comprised 11,143 adults with a mean age of 46.0 ± 17.2 years, of whom 47.3% were male. A high percentage of included subjects exhibited cardiometabolic risk factors, including dyslipidemia (84.2%), metabolic syndrome (26.5%), pre-diabetes or diabetes (20.8%), and hypertension (17.0%). A summary of baseline characteristics of the total cohort with and without stratification for CKD stages is provided in Table 1. A side-by-side comparison for group characteristics between eGFR_SCr and eGFR_LMISCr is provided in Supplementary Table S1.

Standardized, age- and sex-adjusted logistic regression analyses demonstrated that eGFR_LMISCr was significantly more strongly associated with dyslipidemia, arterial hypertension, pre-diabetes/diabetes, and metabolic syndrome. Manifest cardiovascular disease also showed a stronger association with eGFR_LMISCr compared to eGFR_SCr , although the 95% confidence intervals for these associations overlapped. Additionally, eGFR_LMISCr showed stronger associations with anthropometric parameters for obesity according to both WHO and LDEC Commission criteria[19,20] (Obesity_WHO and Obesity_LDEC). (Figure 1 and Table 2)

Comparison between standard eGFR_SCr and eGFR_LMISCr showed a high correlation (R^2 = 0.897). Error analysis revealed a negligible fixed bias of -0.26 mL/min/1.73m² and a precision (SD) of 8.41. The 95% limits of agreement (LoA) ranged from -16.75 to 16.23 mL/min/1.73m². Accuracy metrics indicated that 75.2% of estimates were within 10% (P10) and 99.2% were within 30% (P30) of the reference raw eGFR. A small proportional bias was observed (slope: -0.057, p < 0.001), indicating that individual differences were more pronounced at certain levels of kidney function. Overall, eGFR_SCr and eGFR_LMISCr showed good agreement: Weighted kappa= 0.76 (0.77, 0.78), z=80.5, p<0.001.

The distribution of eGFR in the total population remained largely unchanged and importantly remained in close approximation of a normal distribution. (Supplementary Figure S1) Similarly, eGFR decline with increasing age remained largely unchanged between eGFR_SCr and eGFR_LMISCr. (Supplementary Figure S2) There were overlapping 95% confidence intervals when comparing age- and sex effects on eGFR_SCr and eGFR_LMISCr. (Table 3a and Table 3b)

The application of lean mass correction resulted in shifts in the classification of chronic kidney disease (CKD) stages, with the highest relative number of reclassifications occurring between stage G1 and G2 as well as G3a and G2. (Table 4 and Figure 2)

Longitudinally, eGFR_LMISCr showed significantly greater yearly decline than eGFR_SCr , (Supplementary Table S2 and Supplementary Figure S3). This was largely driven by a faster decline of eGFR_LMISCr particularly in young women, with a significant effect of sex and age. (Table 5 and Figure 3)

4. Discussion

In line with our hypothesis, the present study demonstrates that adjusting serum creatinine for DEXA-derived LMI significantly enhances the strength of association between estimated glomerular filtration rate (eGFR) and cardiometabolic risk factors. By accounting for the biological dependence of creatinine on muscle mass, eGFR_LMISCr revealed substantially stronger associations with dyslipidemia, hypertension, pre-diabetes, diabetes, and metabolic syndrome compared to standard eGFR_SCr. These findings suggest that conventional eGFR_Scr may systematically underestimate the true extent of the relationship between kidney health and metabolic disease in individuals with relatively lower muscle mass, while also mischaracterizing subjects with favourable metabolic profile and high muscle mass.

The prevalence of subjects with decreased glomerular filtration rate impairment remained largely unchanged, however, we saw a relatively large number with mild impairment reclassified between CKD G stages. This was particularly pronounced between CKD G1 and G2 as well as G2 and G3a. Importantly, the diagnostic accuracy for eGFR_SCr is known to be limited in mildly decreased filtration function, making improved association through more accurate reflection of true filtration rate through eGFR_LMISCr more plausible.[21]

eGFR trajectories with increasing age remained largely unchanged with adjustment, which is in line with similar age- and sex effects shown between eGFR_SCr and eGFR_LMISCr, suggesting that effects of inter-individual differences in muscle mass on filtration function remain of equal relative importance across age spectrums. Longitudinally, our data suggest slower yearly GFR decline after LMI adjustment in men but faster decline in women. We also saw a stabilizing effect LMI adjustment on yearly eGFR change in young men, which appears biologically plausible as this demographic typically exhibits a gain of muscle mass over time.

Overall, our findings align with recent literature suggesting that alternative markers less dependent on muscle mass, such as Cystatin C, provide superior risk stratification for cardiovascular mortality compared to eGFR_SCr.[12]

Furthermore, our data showed that eGFR_LMISCr classified more subjects as exhibiting CKD G2 was slightly lower than raw eGFR_SCr on average, suggesting that standard, solely SCr-based formulas may be providing an overly optimistic view of renal health in the general population where more accurate estimation tools of GFR are not available and/or in historic cohorts.

Previous attempts were made to investigate the link between creatinine homeostasis and muscle mass, however, with different goals and endpoints. A paper by Kim et al. reported that muscle mass could accurately be estimated based on serum creatinine after adjusting for Cystatin C-based eGFR.[22] These findings, functionally representing an approach inverse from our own, are directly relevant to our own as this provides proof-of-concept for our own considerations. To the best of our knowledge, Nankivell et al. performed a study with a rationale closest to our own. Nankivell et al. compared the diagnostic accuracy of eGFR_SCr with Tc^99m DTPA plasma clearance in 137 kidney transplant recipients and demonstrated a strong inverse correlation of estimation error with muscle mass measured via DEXA scan, highlighting the potential for improved accuracy with correction.[9] Overall, investigative approaches with the goal of linking creatinine to muscle mass date back to as early as 1983.[23] Notably, adding muscle assessments from bioimpedance analysis (BIA) did not improve GFR assessments compared to measured GFR in CRIC study, albeit these findings need to be viewed in light of lower sample size compared to ours and of known limitations regarding measurement accuracy of BIA compared to DEXA.[24,25]

A major strength of this study is the use of the LEAD cohort, a large-scale, well characterized general population observational study in Austria, utilizing highly accurate DEXA measurement for body composition assessment. However, several limitations must be acknowledged. While eGFR_LMISCr showed stronger statistical associations, we cannot directly prove that the proposed LMI-adjustment results in a better reflection of true GFR rather than just improving its association with metabolic comorbidities. This requires validation against gold standard measured GFR (e.g., iohexol or inulin clearance) or, albeit with higher uncertainty, against formulas incorporating Cystatin C which is currently unavailable for our cohort.[5,26] Additionally, our study is limited by the current lack of data on albuminuria, which may represent an important confounder. However, the LEAD data are obtained from a general population cohort and therefore a relatively low prevalence of albuminuria can safely be assumed.[27] A recent study from a neighbouring European country reported a 5-year prevalence of chronic albuminuria of 6.8% in all subjects, however, age of inclusion into this study was 35 years vs 18 years in LEAD. Nevertheless, our data need to be interpreted with a degree of caution in light of this unknown factor and we encourage external validation.[28]

Adjusting serum creatinine for lean body mass significantly improves the epidemiological utility of GFR estimation by strengthening its associations with key cardiometabolic risk factors. The implementation of muscle-mass-corrected eGFR could refine risk stratification in general populations, particularly in cohorts with high variability in body composition. Our approach offers a novel method to "salvage" the utility of creatinine by using DEXA-measured muscle mass, which may be particularly useful in epidemiological settings where DEXA is already utilized for bone density or body composition analysis.