ABSTRACT
Background Among heart transplantation (HT) recipients, the accuracy of serum creatinine (sCr)-based estimated glomerular filtration rate (eGFR) may be limited by fluctuations in muscle mass. Cystatin C (cysC) is less influenced by muscle mass; however, its levels may increase with obesity and steroid use. Herein, we: i) longitudinally compared changes in eGFRcysC and eGFRsCr among HT recipients; ii) investigated the association of body mass index (BMI), steroid use and muscle mass with discrepancies between eGFRs; and iii) explored the implications of eGFRcysC use on valganciclovir (VGC) dosing.
Methods cysC and sCr were measured in 294 blood samples obtained from 80 HT recipients. Intra-individual differences between eGFRs (eGFRdiffcysC-sCr) were calculated. Negative eGFRdiffcysC-sCr values correspond to eGFRsCr > eGFRcysC and positive values to eGFRcysC > eGFRsCr. In a patient subset (n=21), pectoralis muscle measures were derived from computed tomography scans.
Results Marked differences between eGFRcysC and eGFRsCr were observed, particularly early post-HT (1-week post-HT, median eGFRdiffcysC-sCr −28 ml/min/1.73 m2). eGFRcysC demonstrated stability following a transient post-operative decline, while eGFRsCr decreased in the first year post-HT. Lower BMI and higher prednisone dose displayed a modest association with more negative eGFRdiffcysC-sCr values. Conversely, pectoralis muscle measures indicative of greater muscle mass and better tissue quality exhibited a stronger association with more positive eGFRdiffcysC-sCr values. eGFRcysC use would have resulted in VGC dose adjustment in 46% of samples, predominantly towards lower doses.
Conclusions Among HT recipients, eGFRcysC and eGFRsCr markedly differ with implications for VGC dosing. The observed discrepancies may reflect changes in body composition and steroid use.
INTRODUCTION
Kidney dysfunction is a frequent complication following heart transplantation (HT) and is associated with increased morbidity and mortality1–3. Several factors are implicated in the pathogenesis of kidney dysfunction, including patients’ comorbidities (e.g., diabetes, hypertension) and HT-related factors (e.g., post-operative kidney injury, prolonged calcineurin inhibitor use)4,5. An accurate assessment of kidney function is essential both pre- and post-HT for identification of patients who may benefit from concomitant kidney transplantation6 and to guide medical therapy (e.g., renally-dosed medications, calcineurin inhibitor sparing regimens). However, serum creatinine (sCr)-based equations utilized to calculate estimated glomerular filtration rate (eGFR) may have limited accuracy among HT recipients7.
In advanced heart failure (HF), high rates of sarcopenia and malnutrition lead to decreased sCr generation, potentially resulting in overestimation of kidney function8. Following HT, the interpretation of changes in sCr levels is further complicated by several GFR-independent factors, such as post-operative immobility causing further muscle mass loss and changes in nutritional status.
Cystatin C (cysC) is an alternative eGFR marker which is less influenced by muscle mass, nutritional status, and demographics than sCr9. Cystatin C-based equations have shown superior accuracy in the general population10,11, and across various patient subgroups, including those with HF12,13. However, prior research investigating cysC-based assessment of kidney function in a HT population is limited to a single report in a small cohort of predominantly White males14,15. Notably, the potential impact of cysC confounders, such as changes in body mass index (BMI)16 and steroid use17,18, which are relevant in this patient population19, has not been previously explored. Furthermore, the implications of cysC-based estimates for post-HT medication dosing remain to be defined.
Valganciclovir (VGC), an antiviral medication commonly used for prophylaxis against cytomegalovirus (CMV) infection in HT, requires dose adjustments based on kidney function. Inappropriately high VGC dose in relation to kidney function may contribute to the high rate of leukopenia observed among HT recepients20–22. The management of this complication often requires dose reduction or discontinuation of vital medications, including immunosuppressants, potentially increasing morbidity and mortality22. Therefore, an accurate estimation of GFR is of paramount importance to prevent VGC toxicities and optimize outcomes.
In a cohort of HT recipients, we aimed to: i) longitudinally compare cysC- and sCr-based kidney function estimates at multiple time points; ii) investigate the potential contributions of BMI, steroid use and muscle mass to the discrepancies observed between cysC- and sCr-based estimates; and iii) explore the implications of cysC-based estimates for VGC dosing.
MATERIALS AND METHODS
Study design and data collection
We identified all patients who received HT at Columbia University Irving Medical Center between June 2016 and April 2019. The multi-organ and retransplant recipients were excluded. Cystatin C and sCr were measured simultaneously at the following time points: pre-HT, and serially post-HT: 1 week, 1 month, 2 to 3 months, 5 to 6 months, 1 year. Clinical data were obtained from electronic medical records. All participants provided informed consent, and the study was approved by the institutional review board.
Estimated glomerular filtration rate
In the primary analysis, eGFR was calculated using the 2012 CysC-based (eGFRcysC) and the 2021 sCr-based (eGFRsCr) Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations10,11. Additional analyses included eGFR calculated with the 2021 CKD-EPI equation combining both markers (eGFRsCr+cysC)11. For each concurrent cysC and sCr measurement, the difference between the two estimates (eGFRdiffcysC-sCr) was calculated as the difference between log-transformed eGFRcysC and eGFRsCr, which is equivalent to the log-transformed ratio eGFRcysC/eGFRsCr23. Negative values of eGFRdiffcysC-sCr correspond to eGFRsCr > eGFRcysC and positive values to eGFRcysC > eGFRsCr.
Pectoralis muscle measures
Pectoralis muscle measures were performed in a patient subset with computed tomography (CT) scan obtained for clinical purposes within 14 days of cysC and sCr sampling. Methods were adapted from previous studies24,25. In brief, the analysis was performed using Aquarius iNtuition Viewer (TeraRecon Inc., San Mateo, CA) by a skilled radiologist blinded to cysC and sCr values. Unilateral cross-sectional pectoralis muscle area (cm2) and attenuation (PHUm; pectoralis muscle mean Hounsfield units) were measured on the first axial slice above the superior aspect of the aortic arch. To account for body size, the pectoralis muscle index (PMI; cm2/m2) was obtained by dividing unilateral cross-sectional pectoralis muscle area by the height in square meters. Muscle attenuation (PHUm) represents a proxy for the tissue quality, where lower values indicate intramuscular fatty infiltration associated with impaired muscle function26.
Valganciclovir dosing and leukopenia events
In accordance with the manufacturer’s recommendation, VGC dosing is based on creatinine clearance (CrCl) calculated with the Cockcroft-Gault equation27. Among patients receiving VGC at 1, 2 to 3 and 5 to 6 months post-HT, the implications of different kidney function estimates on VGC dosing were examined. To allow for comparison with CrCl, eGFRcysC, eGFRsCr and eGFRsCr+cysC values were re-expressed in mL/min by multiplying by body surface area derived from the DuBois formula divided by 1.73 m2 28,29. In addition, the association of different GFR estimates with leukopenia events was explored. Leukopenia was defined as a white blood cell count <4 × 109/l within 1-month of sampling and after excluding other identifiable causes (e.g., CMV infection, bacterial sepsis).
Only samples collected at 1, 2 to 3 and 5 to 6 months post-HT contributed to this analysis, due to the following: i) the rate of leukopenia in the first month post-HT is low22; and ii) our institutional protocol recommends discontinuing VGC after 6 months of prophylaxis.
Statistical analysis
Kidney function estimates were log-transformed to ensure normality. Linear mixed models regressed cysC- and sCr-based estimates as well as eGFRdiffcysC-sCr longitudinally at different time points. Patients were modelled as random effects to account for within-person correlation of repeated measurements. All models had an exchangeable correlation structure and included age and sex. To more rigorously control for time-invariant confounders, we performed sensitivity analyses restricted to patients who had samples available at both: pre-HT and 1-week post-HT; pre-HT and 1-month post-HT; 1-week and 5 to 6 months post-HT; and 1-month and 5 to 6 months post-HT. Comparisons between kidney function estimates were performed using the Wilcoxon signed-rank test. Using linear mixed models with the same correlation structure and covariates as described above, we analyzed: i) in the entire study population, longitudinal changes in BMI and prednisone dose, and their association with cysC- and sCr-based estimates as well as eGFRdiffsCr-cysC; ii) among patients with available CT scan, the association of PMI and PHUm with cysC- and sCr-based estimates as well as eGFRdiffcysC-sCr.
Generalized linear models with an exchangeable correlation structure were used to investigate the association of different kidney function estimates with leukopenia. Each observable period began at the time of sampling and continued up to the occurrence of leukopenia. Censoring took place at the time of: i) leukopenia attributable to an identifiable cause (e.g., CMV infection, bacterial sepsis); ii) any change in VGC dosing or VGC discontinuation; and iii) 1 month from sampling. All models were adjusted for white blood cell count at the time of sampling, VGC dose, mycophenolate dose, and sulfamethoxazole-trimethoprim use. A two-sided p-value <0.05 was considered statistically significant. All analyses were performed using SAS Version 9.4 (SAS Institute, Cary, NC).
RESULTS
Cystatin C-versus serum creatinine-based estimated glomerular filtration rate
Among 80 HT recipients, 294 blood samples were collected. Baseline characteristics are reported in Table 1.
We found significant discrepancies between eGFRcysC and eGFRsCr over the study follow-up (Figure 1, Table S1). Estimated GFRcysC markedly decreased from pre-HT to 1-week post-HT and subsequently increased with time post-HT. In contrast, eGFRsCr remained stable early post-HT, but showed a decline at 5 to 6 months that was sustained at 1-year post-HT (Table S1). The difference between the two estimates (eGFRdiffcysC-sCr) widened early after HT, with cysC yielding markedly lower estimates than sCr at 1-week and 1-month post-HT (median eGFRdiffcysC-sCr −28 [interquartile range {IQR}: −40 to −18] and −22 [IQR: −31 to −6] ml/min/1.73 m2, respectively). This early discrepancy between eGFRcysC and eGFRsCr attenuated over time, with no statistically significant differences observed at 5 to 6 months and at 1-year post-HT.
Sensitivity analyses restricted to patients with paired samples were consistent with the main analysis (Figure S1). Estimated GFR calculated with an equation combining both glomerular filtration markers (eGFRsCr+cysC) yielded values which were intermediate between eGFRcysC and eGFRsCr (Table S1, Figure S2). Following a transient decline at 1-week, eGFRsCr+cysC remained stable post-HT.
Compared with eGFRsCr, eGFRcysC reclassified majority of samples to a different eGFR category, predominantly towards more advanced kidney dysfunction (Figure 2). Specifically, reclassification to worse eGFR category occurred in 10 (42%) patients pre-HT, 48 (87%) patients at 1-week post-HT and 48 (72%) patients at 1-months post-HT (Table S2). Overall, 38 (13%) samples exhibited an eGFR <30 mL/ min/1.73 m2 based on cysC compared with only 4 (1%) based on sCr (Figure 2). The reclassification resulting from eGFRsCr+cysC is shown in Table S3.
Body mass index, prednisone dose and pectoralis muscle measures: association with kidney function estimates and eGFRdiffcysC-sCr
Post-HT changes in BMI and prednisone dose are depicted in Figure 3. After an initial decline at 1 month, BMI increased and was significantly higher than pre-operative values at 5 to 6 months and 1-year post-HT. Prednisone was administered at a median dose of 40 mg/day at 1-week post-HT and subsequently tapered. Higher BMI was associated with lower eGFRsCr and eGFRsCr+cysC but had no significant association with eGFRcysC (Table S4). Prednisone dose was not associated with any cysC or sCr-based estimates.
We found a modest but statistically significant association of BMI and prednisone dose with eGFRdiffcysC-sCr. A 10% increase in BMI was associated with a 2.1% change in eGFRdiffcysC-sCr towards more positive values (95% confidence interval [CI]: 0.2 to 4.1%, p=0.029). For each 10% increase in prednisone dose, there was a 1.5% change in eGFRdiffcysC-sCr towards more negative values (95%CI: 0.8 to 2.2%, p<0.001). Similar results were obtained in models including both BMI and prednisone dose (Table S4).
Baseline characteristics of the subset of study participants with available pectoralis muscle measures are shown in Table S5. Among 21 patients with a total of 40 CT scans obtained at a median of 3 (IQR: 1 to 8) days from sampling, average PMI was 4.9 ± 2.1 cm2/m2 and average PHUm was 27 ± 14 mean Hounsfield units. Higher PHUm was associated with higher eGFRcysC (β=0.1154, 95%CI: 0.0017 to 0.0214, p= 0.023). No other association between pectoralis muscle measures and individual cysC- or sCr-based eGFR reached statistical significance. Higher PMI and PHUm were associated with more positive eGFRdiffcysC-sCr values (Figure 4). In adjusted models, for each 1-standard deviation (SD) increase in PMI and PHUm, eGFRdiffcysC-sCr changed towards more positive values by 12.9% (95%CI: 1.7 to 25.2%, p=0.023) and 17.8% (95%CI: 6.6 to 30.2%, p=0.002), respectively. Further adjustment for BMI and prednisone dose did not significantly alter the association between pectoralis muscle measures and eGFRdiffcysC-sCr (data not shown).
Implications of different kidney function estimates for valganciclovir dosing and association with leukopenia
The baseline characteristics of the 57 study participants receiving VGC at 1, 2 to 3, and 5 to 6 months post-HT are presented in Table S6. Over 113 total samples, the use of eGFRsCr for VGC dosing would have resulted in dose adjustment in 30 (26%) samples; of those, 16 (14%) would have been underdosed and 14 (12%) overdosed (Figure 5).
The use of eGFRcysC for VGC dosing would have resulted in dose adjustment in 52 (46%) samples; of those, 11 (10%) would have been underdosed and 41 (36%) overdosed. Of note, only 74 (65%) of samples had appropriate VGC dosing per the manufacturer’s recommended approach based on CrCl. Due to the marked discrepancies between cysC- and sCr-based estimates observed early post-HT, VGC dose adjustment resulting from eGFRcysC use would have been observed mainly within the first 3 months post-HT (Figure S3).
A total of 30 leukopenia events occurred among 27 individual patients. In adjusted models including VGC dosing, for each 1-SD decrease in eGFRcysC, eGFRsCr+cysC and CrCl, the risk for leukopenia approximately doubled (Table 2). The association of eGFRsCr with leukopenia did not reach statistical significance (p=0.083).
DISCUSSION
In a cohort of 80 HT recipients, we found marked discrepancies between cysC- and sCr-based estimates, particularly in the early post-HT period. While eGFRsCr remained stable in the post-operative period and declined afterwards, eGFRcysC worsened early post-HT but improved long-term. Body mass index and prednisone dose were modestly associated with the difference between the two estimates (eGFRdiffcysC-sCr). However, in a patient subset with available CT scan, a stronger association was present between pectoralis muscle measures indicative of greater muscle mass and better tissue quality and more positive eGFRdiffcysC-sCr values. The use of cysC-based estimates would have led to frequent dose adjustments for VGC, primarily favoring dose reductions.
The measurement of clearance of exogenous markers (e.g., iohexol) represents the gold standard for kidney function assessment. However, its utility in clinical practice is limited by complexity of administration and cost, making estimates based on endogenous markers such as cysC and sCr more common. Despite the traditional reliance on sCr for kidney function assessment, two major nephrology societies have advocated for the routine use of cysC, citing its potential for greater accuracy and its independence from race30. Yancy et al. have similarly endorsed wider implementation of cysC-based estimates in the care of patients with HF31. In a recent consensus statement on heart-kidney transplantation, Kobashigawa et al. suggested the use of the CKD-EPI equation combining sCr and cysC6. Additionally, the most recent guidelines from the International Society for Heart and Lung Transplantation present cysC as an alternative for kidney function monitoring in HT recipients32. Our findings indicate that cysC- and sCr-based estimates differ significantly and cannot be used interchangeably.
A marked difference between cysC and sCr-based estimates occurs when factors unrelated to kidney function differentially affect the two markers. As previously reported by our group23,33, the discrepancy between eGFRs widens in the setting of worsening HF, potentially due to variations in sCr confounders such as muscle mass and nutritional status. The present study supports the hypothesis that muscle mass represents a major determinant of eGFRdiffcysC-sCr among HT recipients. This is in line with prior evidence in patients implanted with left ventricular assist device (LVAD), where early post-operative changes in sCr correlated with changes in muscle mass as assessed by PMI24. Notably, in the same report, sCr-based eGFR showed a transient improvement after LVAD implantation, whereas cysC-based eGFR remained stable. In the early post-HT phase, characterized by muscle wasting and reduced protein intake, sCr may underestimate the degree of kidney dysfunction resulting from surgical insult and acute calcineurin inhibitor nephrotoxicity. Subsequent improvements in the quality and quantity of muscle mass as well as in nutritional status could account for the divergent trajectories of eGFRsCr and eGFRcysC. Factors influencing cysC levels independently from kidney function are also likely to explain variations in eGFRdiffcysC- sCr. The association of higher BMI with more positive eGFRdiffcysC-sCr values may reflect the influence of adipose tissue on cysC generation34 or the effect of muscle mass on sCr levels. The modest association of prednisone dose with eGFRdiffcysC-sCr in our population is consistent with prior studies demonstrating a steroid-induced, dose-dependent cysC increase not related to kidney function17,18.Overall, our findings indicate a large impact of non-GFR determinants of cysC and sCr in the early post-operative period, particularly in light of improved convergence of the estimates at 1-year post-HT.
Establishing an evidence-based approach to assess kidney function pre- and post-HT is critical for current clinical practice. Since the 2018 implementation of the new United Network for Organ Sharing allocation system, patients demonstrate higher acuity at the time of HT, as evidenced by increased rates of mechanical ventilation, intensive care unit stay and temporary mechanical circulatory support, and are more likely to require post-operative kidney replacement therapy35–37. The current paradigm for kidney function assessment, which relies heavily on sCr, might have its limitations further accentuated in a patient cohort with a rising incidence of critical illness. Additionally, a new Organ Procurement and Transplantation Network policy restricted simultaneous heart-kidney transplant to patients with an GFR <30 ml/min/1.73 m2 and introduced a “safety net” in the presence of persistent kidney dysfunction (eGFR <20 ml/ min/1.73 m2, or need for kidney replacement therapy) within the first year following HT15,38. Notably, the policy provided no guidance on which marker or equation should be used to calculate eGFR. The significant variations observed in kidney function estimates in the present study may have relevant implications for clinical practice if cysC were to become routinely used. This makes our findings timely and pertinent to the ongoing discussion regarding kidney transplantation eligibility among HT candidates and recipients6,15,39.
Valganciclovir dosing recommendations using the Cockcroft-Gault equation are based on the VICTOR trial40, a study conducted nearly two decades ago. There are numerous limitations to the Cockcroft-Gault equation with respect to accurately estimating drug clearance. Developed in 1973 using data from 249 men, this equation was not adjusted for body surface area. Also, since the publication of VICTOR trial, sCr assays have been standardized to reference methods. Contemporary sCr-based equations, but not Cockcroft-Gault equation, have been expressed for these reference methods. Given that the assay used in the derivation of Cockcroft-Gault equation yielded higher sCr levels than current methods41, its application may lead to higher drug dosage recommendations than originally intended in pharmacokinetic studies. Using cysC, we identified a significant number of HT recipients who were potentially overdosed on VGC. Future studies could leverage this pilot data to directly compare VGC dosing using cysC-based equations against other kidney function estimates and evaluate differences in rates of leukopenia and CMV infection.
Our study has several limitations. First, the absence of a gold-standard measure of GFR prevents us from determining which marker offers the most precise kidney function estimate. The single-center design represents another limitation, as our practices and transplantation eligibility criteria might not be representative of other institutions. The number of missing samples at certain time points, particularly pre-HT, raises concerns about potential unmeasured confounders affecting our findings on longitudinal eGFR changes. Nevertheless, our sensitivity analyses on paired samples corroborated the conclusions of the main analysis. We must also acknowledge that the eGFR equations used in this study were initially derived from patients with stable kidney function, limiting their reliability when applied to acutely ill patients. Lastly, although all the findings of the present study are hypothesis-generating, the analysis on leukopenia events is exploratory in nature and should be interpreted with particular caution.
In conclusion, among HT patients, we found significant discrepancies between cysC- and sCr-based kidney function estimates, which could be attributed to changes in body composition and steroid use. These findings may have relevant implications for VGC dosing and other aspects of clinical management. Further research is warranted to define a new standard for kidney function assessment in this patient population.
AUTHORS’ CONTRIBUTIONS
Conception and Design: AP, DLJ, PCC, MY
Statistical Analysis: AP, RTD, BB
Analysis and Interpretation of the Data: AP, AL, PD, AOS, IM, BB, RTD, KT, GTS, NU, JSL, SAH, SM
Supervision or Mentorship: PCC, MY
FINANCIAL DISCLOSURE STATEMENT
GTS has received consulting fees from Abbott. NU has received consulting fees from Leviticus and LiveMetric. PCC serves as consultant for Abbott and Roche. MY has received research grant support from Abbott. The remaining authors have no conflict of interest to declare.
Data Availability
All data produced in the present study are available upon reasonable request to the authors
ACKNOWLEDGMENTS
None