Multiparametric quantitative magnetic resonance imaging of skeletal muscle in CKD

Abstract

Skeletal muscle dysfunction causes functional decline and disability in patients with chronic kidney disease (CKD). Identification of muscle pathology prior to significant loss of physical function would be a major advance. Multiparametric, quantitative magnetic resonance imaging (qMRI) of 7 leg muscle groups (3 thigh, 4 calf) was conducted in patients with CKD stages 4-5 (n=6), end-stage kidney disease (ESKD, n=3), and healthy controls (n=10) using a 3 Tesla MRI scanner. Measurements included T1 relaxation time in the rotating frame (T1ρ) and transverse relaxation time (T2) mapping, Dixon imaging of intramuscular fat content, diffusion tensor imaging (DTI) for muscle structure, and ¹H-MR spectroscopy for intra- and extra-myocellular lipid (IMCL and EMCL, respectively) and physiologically relevant muscle metabolites. T1ρ and T2 times were prolonged and fat fraction (FF) was higher in CKD patients compared with controls (differences of 4.99 ms (95% CI 1.71-8.27), 6.72 ms (95% CI 3.78-9.66), and 6.67% (95% CI 0.65-12.68), respectively). T1ρ and FF were similarly elevated across muscle groups, whereas T2 differences may have been greater in calf muscles. T1ρ and T2 were lower in ESKD patients compared with CKD and similar to controls, consistent with prior histologic assessment of muscle fibrosis. No significant differences by CKD status were observed for DTI parameters. Compared with controls, IMCL was higher in CKD patients and trimethylamine and creatine concentrations were lower. In sum, multiparametric qMRI of skeletal muscle in CKD patients non-invasively identified differences in metrics associated with fibrosis, fat infiltration, and metabolic dysregulation.

Graphical Abstract

New & Noteworthy

In this study, we demonstrate that multiparametric, quantitative magnetic resonance imaging (qMRI) can quantify multiple distinct anatomic and pathologic features of skeletal muscle pathology in patients with CKD prior to significant functional decline. qMRI metrics of fibrosis and fat infiltration were elevated, and muscle metabolite concentrations were reduced, in patients with CKD compared with controls. This non-invasive approach offers a valuable alternative to traditional muscle biopsies for evaluating muscle health in CKD patients.

INTRODUCTION

Disability is an important contributor to poor quality of life, excess hospitalizations, and early death among adults with CKD (1). Prevention of disability in the CKD population is a major unmet need. Disability in patients with CKD is often preceded by skeletal muscle dysfunction, which causes weakness, fatigue, and functional impairment (1). Therefore, a critical gap in our knowledge is the ability to identify muscle pathology before significant functional decline occurs. Early detection of these abnormalities may allow for early interventions to prevent or slow muscle deterioration, ultimately reducing disability and improving quality of life for CKD patients.

Magnetic resonance imaging (MRI) shows promise as a non-invasive biomarker of muscle pathology in CKD (2). Recent technological advances enable quantitative mapping of skeletal muscle structure, composition, and metabolic health (3). Quantitative MRI (qMRI) protocols have been employed in genetic muscle diseases, which are characterized by changes in muscle structure including fiber atrophy, fibrosis, and fatty replacement (3). Similar changes are observed in patients with CKD: pathologic descriptions include disorganized sarcomeres, myofibril degeneration, glycogen and lipid accumulation, and fibrosis with expanded, densely packed collagen altering the extracellular matrix (4-8). A key strength of the qMRI approach is the ability to assess distinct anatomic and pathologic features of muscle pathology within a single imaging session. This non-invasive approach offers a valuable alternative to traditional muscle biopsies for evaluating muscle health in CKD patients.

Prior work suggests that qMRI mapping is sensitive to early pathology in skeletal muscle disorders. Several methods within qMRI could be informative in CKD. A novel method for non-invasive detection of fibrosis is the T1 relaxation time in the rotating frame (T1ρ). T1ρ has been shown to correlate with collagen, glycosaminoglycan, and hyaluronan content in fibrotic muscle disorders (9, 10). Another hallmark of fibrotic transformation of skeletal muscle is fatty replacement. In genetic muscle diseases, quantification of intramuscular fat using chemical shift-based imaging correlates with histological assessment and changes over time with disease progression (11-14). Intramuscular fat accumulation may be preceded by muscle water changes that are detectable by transverse relaxation time (T2) mapping (15). Furthermore, muscle lipid content quantified by ¹H magnetic resonance spectroscopy (MRS) correlates with strength and meaningful functional outcomes (16-18). ¹H MRS also quantifies muscle metabolites and has been used in genetic dystrophies, demonstrating altered metabolite concentrations that associate with poor outcomes (19, 20). These pathologic changes are common to a variety of dystrophic and fibrotic muscle diseases, and are useful biomarkers independent of the underlying etiology (15). Thus, qMRI mapping has the potential to provide a wealth of information relevant to muscle health in patients with CKD.

Here, we report the results of a proof-of-concept multiparametric qMRI protocol evaluating skeletal muscle in patients with CKD. The protocol included T1ρ and T2 relaxation time mapping, chemical shift-based Dixon imaging, and ¹H MRS. In addition, we investigated diffusion tensor imaging (DTI), which uses the directionality of water diffusion to probe muscle structure.

MATERIALS AND METHODS

Study Population

This study was nested within a previously reported prospective cohort study (21, 22). Patients were eligible for participation if they had an eGFR <30 ml/min/1.73 m² or had ESKD and were receiving dialysis therapy, were ≥21 years of age, able to provide informed consent, and were ambulatory. Exclusion criteria included bilateral lower extremity amputations, use of immunosuppressive medications in the prior 3 months, and an active cancer diagnosis or receiving treatment for cancer. Generally healthy, sedentary individuals without evidence of kidney disease (eGFR>60 mL/min/1.73m² and urine albumin:creatinine ratio <30 mg/g) were recruited as control participants. Sedentary was defined as not being engaged in strenuous work, regular brisk leisure physical activity, or a formal exercise session more than once per week for at least the previous 3 months (23, 24). Individuals of both sexes were eligible for participation. No formal power analysis was conducted to determine sample size.

Study Design

Health-related quality-of-life was assessed with the 36-Item Short Form Health Survey (SF-36) and the Patient-Reported Outcomes Measurement Information System (PROMIS) Physical Function 4a and Fatigue 7a short forms (25-27). Unilateral knee extensor strength was measured using isometric dynamometry with a handheld dynamometer (Manual Muscle Test System, Lafayette Instrument, Lafayette, IN). Two trials were conducted and subjects were instructed to perform a maximal exertion, with the highest result for the dominant leg used for analysis. Endurance capacity was measured by the 2-minute walk test conducted using a 50-foot course (28); the distance covered is highly correlated with 6-minute walk distance (29). The Short Physical Performance Battery (SPPB) is an established measure of mobility in older adults, including a 4-meter walk test, standing balance test, and 5-repetition sit-to-stand test (range 0-12, higher better) (30). Handgrip strength was measured twice in each hand using a handheld dynamometer (North Coast Medical, Morgan Hill, CA). The maximum value attained using the dominant hand was used for analysis. All laboratory tests were conducted in the clinical laboratory of Montefiore Medical Center. eGFR was calculated by the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation (31). Study data were collected and managed using REDCap (Research Electronic Data Capture; RRID:SCR_003445) electronic data capture tools hosted at the Albert Einstein College of Medicine (32). The study was approved by the Institutional Review Board of the Albert Einstein College of Medicine and conducted in adherence to the Declaration of Helsinki and all participants provided written informed consent.

Quantitative MRI Protocol

All imaging was performed on a 3T MR scanner (Ingenia Elition X, Philips Healthcare, Best, the Netherlands) with a 16-channel transmit-receive knee coil. Each patient was first centered at the thigh and then at the calf, and the same image sequences were performed within the same study. In addition to standard anatomical imaging, such as T1-weighted imaging, all subjects underwent scanning using the following quantitative MRI sequences. T1ρ and T2 mapping were not performed in one control participant, and DTI of the calf muscles was not performed in one dialysis patient due to technical limitations.

mDixon.

A multi-echo Dixon-based MRI protocol (mDixon-Quant) was used to assess the degree and distribution of fat infiltration in muscle compartments of the thigh and calf. Imaging parameters included: repetition time (TR) = 8.5 ms, six echoes with first echo time (TE) = 1.31 ms, and ΔTE = 1.2 ms, field of view (FOV) = 200 × 200 × 201 mm³, flip angle = 3°; matrix size = 168 × 121, acquisition resolution = 1.19 × 1.65 × 6 mm³, number of slices = 67, and scan duration ~ 35 sec. Water-only and fat-only images, along with quantitative fat-fraction and T2* maps, were generated using the vendor’s default reconstruction algorithm. This algorithm, while approved for liver fat fraction measurements, does not separate intramuscular from extramuscular fat. However, prior studies have validated this sequence in skeletal muscle (33, 34).

T1ρ and T2 mapping.

Quantitative T1ρ and T2 mapping sequences were performed to study muscle fibrosis and different water compartments and their fractions. The 3D T1ρ magnetization-prepared angle-modulated partitioned k-space spoiled gradient echo snapshots (MAPSS) sequence parameters included FOV = 200 × 200 × 201 mm³, matrix size = 200 × 154 × 64, TR = 4.4 ms, TE = 2.3 ms, resolution = 1 × 1.3 × 6 mm², number of slices = 67, and compressed SENSE factor = 3, spin-lock frequency = 500 Hz, 4 spin-lock time (TSL) durations = 0, 20, 40, 60 ms. acquisition duration ~ 3 min 28 sec (35-39). The 3D T2 mapping sequence had the same imaging parameters and scan duration except TE preparation time = 0, 20, 40, 60 ms. T1ρ and T2 maps were computed on a pixel-by-pixel basis using a mono-exponential decay model using the following equations: M(TSL) = M₀ exp(−TSL/T1ρ) for T1ρ, and M(TE) = M₀ exp(−TE/T2) for T2 mapping, where M₀ and M(TSL) denote the equilibrium magnetization and T1ρ-prepared magnetization with the spin-lock time of TSL, respectively. M(TE) denotes the magnetization acquired with the echo time TE. These two mono-exponential equations were linearized by logarithm. Data were not obtained for one control participant due to technical difficulties.

DTI:

The DTI sequences had the following parameters: FOV = 200 × 200 mm, acquisition matrix size = 80 × 78, slice thickness = 3mm, gap = 0 mm, TR/TE = 2259/66 ms, resolution = 2.50 × 2.56 mm², number of slices = 50, SENSE factor = 2, b values = 0, 800 s/mm² with 15 diffusion directions, half scan (partial Fourier factor) = 0.8, number of averages = 2, and acquisition duration = ~4 min 45 sec. Both spectral presaturation with inversion recovery (SPIR) and slice-selection gradient reversal (SSGR) were employed in the DTI sequences to minimize the impact of fat contamination on diffusion quantification. Apparent diffusion coefficient (ADC) and fractional anisotropy (FA) maps were generated automatically on the MR scanner console after image co-registration. To assess the potential impact of SNR-related bias on our diffusion metrics (40, 41), we estimated the mean SNR in two representative patient subjects. The measured values (ranging from 70 to 121) suggest that SNR-related bias had a limited impact on the quantitative diffusion MRI obtained with this DTI technique.

Muscle segmentation.

Three thigh muscles (vastus lateralis, vastus medialis, biceps femoris longus) and 4 calf muscles (tibialis anterior, soleus, gastrocnemius medialis, gastrocnemius lateralis) were manually segmented to define regions of interest (ROIs) by a blinded examiner. For all metrics except fat fraction, ROIs were drawn excluding fat, fascia, and large blood vessels using vendor-provided software on the MRI console. For fat fraction, ROIs did not exclude fat within the border of the muscle belly as doing so would have underestimated fatty replacement of skeletal muscle. Measurements for each qMRI metric were extracted from 5 consecutive slices and averaged.

Cross-sectional area.

Muscle contractile cross-sectional area (CSA) size was evaluated on the water-only images from the mDixon Quant sequence using open-source software ImageJ (National Institutes of Health, Bethesda, MD; RRID: SCR_003070). The center slice from the 5 consecutive slices was selected for CSA analysis. Images were first automatically thresholded using the isodata algorithm to generate binary images, with all muscle pixel values assigned to a value of one and background/fat pixels a value of zero. Manual ROIs were then drawn to encompass each muscle group, and the total number of muscle pixels within the ROI was quantified to calculate contractile CSA. Quadriceps CSA was calculated from an ROI encompassing the vastus lateralis, vastus medialis, and vastus intermedius. In addition, total thigh muscle CSA was measured on all 5 slices and the mean used for analysis.

¹H MRS.

A water-suppressed point-resolved spectroscopy (PRESS) sequence was used to localize a single voxel with size of 15 × 10 × 20 mm³ in the vastus lateralis muscle of the thigh, as well as the soleus and gastrocnemius medialis muscles in the lower leg, respectively. Care was taken to avoid muscle boundaries, large vessels, and visible intermuscular adipose tissue and intramuscular fat. Water suppression involved an excitation pulse followed by a frequency-modulated pulse with a 140 Hz bandwidth and a 300° flip angle to minimize the water signal at RF excitation. Imaging parameters included: TR = 2000 ms, TE = 30 ms, readout duration = 1024 ms, number of signal averages (NSA) = 32, spectral bandwidth = 2000 Hz, number of samples = 2048, and scan duration = 1 min 40 sec. Water-unsuppressed reference scans were acquired using the same parameters, except with NSA = 4. High-order “pencil beam” shimming was used to optimize the B₀ field homogeneity. The MRS data were exported offline for spectral peak fitting and the calculation of the nominal concentration of each metabolite component. The nominal concentration is defined as the area under the fitted peak. A custom software tool in MATLAB (The MathWorks, Natick, MA; RRID:SCR_001622) was developed using the open-source MRS library (MRS_MRI_libs, https://github.com/chenkonturek/MRS_MRI_libs) to fit the MRS data with a linear combination of symmetric Gaussian functions for the following six peaks: water, Cr (creatine)-CH₂, trimethylamine (TMA), Cr-CH₃, extramyocellular lipid (EMCL), and intramyocellular lipid (IMCL). TMA was quantified by the area of the peak at 3.2 ppm. Total creatine included the peak areas at 3.0 and 3.9 ppm. All measures were normalized by the water peak area from the water-unsuppressed reference spectrum. More details about the processing tool are provided in the Supplemental File.

Statistical Analysis

Comparisons of participant characteristics among the 3 participants groups were performed using analysis-of-variance (ANOVA) or Kruskal-Wallis tests for continuous variables and χ² tests or Fisher’s exact test for categorical variables. Differences between control participants and CKD patients were tested using 2-tailed t-tests or Wilcoxon rank-sum tests. Spearman correlation coefficients were used to test pairwise correlations of qMRI metrics. To test associations of CKD or ESKD status with qMRI metrics across muscle groups, which include multiple measures per person, mixed effects models including random intercepts were fit to account for clustering within an individual. Pairwise comparisons of CKD and control participants within muscle groups were performed using tests of contrast. Due to their skewed distributions, all ¹H MRS measures were log-transformed prior to analysis. All analyses were performed using Stata 17.0 (StataCorp, College Station, TX; RRID:SCR_012763). A p-value <0.05 was considered statistically significant.

RESULTS

Participant Characteristics

Of the 19 participants, 6 had non-dialysis dependent CKD, 3 had ESKD and were receiving hemodialysis, and 10 were controls (Figure 1). The mean age among the CKD, ESKD, and control participants was 62±16, 51±13, and 55±6 years, respectively (Table 1). Mean eGFR was 21.3 ± 5.0 and 87.9 ± 9.7 mL/min/1.73m² in CKD and control participants, respectively. All participants were free from disability and fully independent in activities of daily living. Self-reported physical function and fatigue and objective measures of performance including the SPPB, sit-to-stand time, and handgrip strength were poorer among CKD and ESKD patients than controls.

Figure 1. Flow diagram of study participation.

Table 1. Participant Characteristics.

	Control	CKD	ESKD	p
Number	10	6	3
Age (years)	55 ± 6	62 ± 16	51 ±13	0.36
Sex – n (%)				0.22
Female	4 (40%)	4 (67%)	3 (100%)
Male	6 (60%)	2 (33%)	0 (0%)
Race/ethnicity – n (%)				0.57
Non-Hispanic White	4 (40%)	2 (33%)	0 (0%)
Non-Hispanic Black	5 (50%)	2 (33%)	2 (67%)
Hispanic	1 (10%)	2 (33%)	1 (33%)
Hypertension – n (%)	1 (10%)	6 (100%)	3 (100%)	<0.001
Diabetes – n (%)	1 (10%)	4 (67%)	1 (33%)	0.05
eGFR (mL/min/1.73 m2)	87.9 ± 10.1	21.3 ± 5.4	….	<0.001
SF-36 Physical Component Score	50.6 ± 7.0	41.7 ± 7.4	42.0 ± 15.5	0.04
SF-36 Physical Functioning Domain	54.7 ± 5.9	47.2 ± 7.6	43.0 ± 20.7	0.02
PROMIS Physical Function	54.7 ± 6.3	47.5 ± 5.0	51.4 ± 9.5	0.14
PROMIS Fatigue	39.9 ± 4.9	49.4 ± 7.6	47.9 ± 8.7	0.02
SPPB	11.4± 0.7	10.2 ± 1.6	10.7 ± 2.3	0.40
5-repetition sit-to-stand time (s)	10.4 ± 2.5	12.4 ± 2.9	12.8 ± 6.6	0.60
Gait speed (m/s)	1.2 ± 0.2	1.1 ± 0.3	1.0 ± 0.2	0.56
2-minute walk distance (feet)	559 ± 167	546 ± 124	….	0.88
Handgrip strength (kg)	39.8 ± 8.3	29.3 ± 6.4	30.3 ± 7.8	0.09
Hip flexion strength (kg)	17.9 ± 4.5	15.1 ± 6.9	15.7 ± 4.5	0.35
Quadriceps strength (kg)	23.4 ± 8.9	22.7 ± 5.7	26.5 ± 11.0	0.91

Abbreviations: CKD, chronic kidney disease; ESKD, end stage kidney disease; eGFR, estimated glomerular filtration rate; SF-36, 36-Item Short Form Health Survey; PROMIS, Patient-Reported Outcomes Measurement Information System; SPPB, Short Physical Performance Battery.

SF-36, PROMIS, SPPB and its components, 2-minute walk, and strength testing were not available for 2 control participants.

Data are presented as mean ± standard deviation or median (interquartile range) for continuous variables. Comparisons of participant characteristics among the 3 participants groups were performed using analysis-of-variance or Kruskal-Wallis tests for continuous variables and χ² tests or Fisher’s exact test for categorical variables. Differences between control participants and CKD patients were tested using 2-tailed t-tests or Wilcoxon rank-sum tests.

Quantitative MRI

Representative T1-weighted images and quantitative MRI maps for control, CKD, and ESKD are presented in Figure 2 (T1ρ, fat fraction, T2) and Supplemental Figure S2 (T2*, FA, ADC). To determine the extent to which qMRI metrics clustered within an individual across muscle groups, we compared measures within each muscle with those of the vastus lateralis (VL). VL was selected as the comparator because it has been extensively characterized in humans, is the muscle within which we previously demonstrated biopsy-proven fibrosis accompanying loss of kidney function (8, 21), and prior genetic dystrophy studies demonstrated strong correlations of VL qMRI measures with functional endpoints (16, 18). Visual inspection indicated that values for T1ρ, fat fraction, T2, T2*, and ADC tended to cluster within an individual across muscle groups (Figure 3). In contrast, FA showed large within-person variability across muscles. Generally strong correlations were present with the VL for T1ρ and fat fraction, and within thigh muscles in particular for T2 and T2* (Table 2).

Figure 2. Representative quantitative magnetic resonance imaging maps.

T1-weighted images and T1ρ, fat fraction, and T2 maps of thigh muscles for control, CKD, and ESKD participants. Abbreviations: FF, fat fraction.

Figure 3. Quantitative magnetic resonance imaging metrics by person and muscle group.

Values for each muscle group for an individual participant are plotted relative to that individual’s values for the vastus lateralis muscle. For each metric, the range of Spearman correlation coefficients across muscle groups is displayed in the upper left corner. n=19 for all metrics except T1ρ and T2 (n=18). Abbreviations: VM, vastus medialis; BFL, biceps femoris longus; TA, tibialis anterior; S, soleus; GM, gastrocnemius medialis; GL, gastrocnemius lateralis; ADC, apparent diffusion coefficient.

Table 2. Muscle-Specific Correlations of Muscle Quantitative Magnetic Resonance Imaging Metrics with Vastus Lateralis Values.

Metric	VM	BFL	GM	GL	S	TA
T1ρ	0.93 p<0.001	0.78 p<0.001	0.76 p<0.001	0.72 p<0.001	0.58 p=0.01	0.75 p<0.001
Fat Fraction	0.88 p<0.001	0.76 p<0.001	0.61 p=0.006	0.63 p=0.005	0.81 p<0.001	0.84 p<0.001
T2	0.73 p<0.001	0.60 p=0.008	0.45 p=0.06	0.37 p=0.13	0.60 p=0.008	0. 83 p<0.001
T2*	0.84 p<0.001	0.74 p<0.001	0.43 p=0.07	0.35 p=0.14	0.51 p=0.03	0.42 p=0.07
FA	0.41 p=0.09	−0.07 p=0.76	0.17 p=0.50	0.52 p=0.03	−0.07 p=0.78	0.29 p=0.24
ADC	0.15 p=0.54	0.26 p=0.27	0.37 p=0.13	0.44 p=0.07	−0.19 p=0.44	0.02 p=0.94

Abbreviations: VM, vastus medialis; BFL, biceps femoris longus; TA, tibialis anterior; S, soleus; GM, gastrocnemius medialis; GL, gastrocnemius lateralis; ADC, apparent diffusion coefficient; FA, fractional anisotropy.

Pairwise correlations are presented as Spearman correlation coefficients and associated p-values. n=19 for Fat Fraction and T2*, n=19 for FA and ADC for VM and BFL, n=18 for other comparisons.

Given the clustering we observed, rather than focusing on differences within an individual muscle group, we tested global differences across muscles by CKD status for each qMRI parameter. These analyses revealed marked elevations in T1ρ and T2 relaxation times and in muscle fat fraction among patients with CKD compared with controls which were independent of differences in age (Table 3). In contrast, no differences were observed between CKD patients and controls for T2*, ADC, and FA. Age-adjusted estimates showed that T1ρ and fat fraction were similarly elevated across muscle groups, whereas differences in T2 may have been driven largely by changes in calf muscles (Figure 4). FA was numerically higher in CKD patients in all muscle groups but these differences were not statistically significant. Fat fraction correlated with T1ρ, T2, and ADC across muscles and participant groups (r=0.46, p<0.001; r=0.28, p=0.002; r=−0.18, p=0.04, respectively); associations with T1ρ and T2 may have been of greater magnitude among participants with CKD (Figure 5). T1ρ strongly correlated with T2 (r=0.69, p<0.001) without meaningful differences observed across groups (Figure 5).

Table 3. Associations of CKD Status with Muscle Quantitative Magnetic Resonance Imaging Metrics.

	Control vs. CKD		ESKD vs. CKD
Parameter	Estimate (95%CI)	P-value	Estimate (95%CI)	P-value
Unadjusted
T1ρ	−5.46 (−8.71 to −2.21)	0.001	−5.48 (−9.84 to −1.12)	0.01
Fat Fraction	−6.77(−12.55 to −0.99)	0.02	−5.82 (−13.73 to 2.09)	0.15
T2	−4.63 (−7.52 to −1.73)	0.002	−2.55 (−6.43 to 1.33)	0.20
T2*	0.40 (−2.73 to 3.52)	0.81	1.38 (−2.91 to 5.66)	0.53
ADC (10−3 mm2/s)	−0.05 (−0.12 to 0.01)	0.12	−0.06 (−0.15 to 0.04)	0.24
FA	−0.01 (−0.03 to 0.004)	0.14	0.005 (−0.02 to 0.03)	0.67
Age Adjusted
T1ρ	−4.99 (−8.27 to −1.71)	0.003	−4.68 (−9.17 to −0.17)	0.04
Fat Fraction	−6.67(−12.68 to −0.65)	0.03	−5.64 (−14.04 to 2.77)	0.19
T2	4.25 (−7.19 to −1.31)	0.005	−1.90 (−5.93 to 2.13)	0.36
T2*	1.07 (−2.00 to 4.14)	0.49	2.52 (−1.77 to 6.81)	0.25
ADC (10−3 mm2/s)	−0.04 (−0.10 to 0.02)	0.23	−0.04 (−0.13 to 0.06)	0.20
FA	−0.01 (−0.03 to 0.003)	0.11	0.01 (−0.01 to 0.03)	0.37

Abbreviations: CKD, chronic kidney disease; ESKD, end stage kidney disease; ADC, apparent diffusion coefficient; FA, fractional anisotropy.

Estimates and associated p-values determined using mixed effects models including random intercepts to account for clustering within an individual across muscle groups. n=19 for all analyses except T1ρ and T2 (n=18).

Figure 4. Quantitative magnetic resonance imaging metrics by muscle group and CKD status.

Data are presented as mean ± SE. Mixed effects models including random intercepts were fit to account for clustering within an individual. Pairwise comparisons of CKD and control participants within muscle groups were performed using tests of contrast. n=19 for all metrics except T1ρ and T2 (n=18). Abbreviations: VL, vastus lateralis; VM, vastus medialis; BFL, biceps femoris longus; TA, tibialis anterior; S, soleus; GM, gastrocnemius medialis; GL, gastrocnemius lateralis. *p<0.05, **p<0.01, ***p<0.001.

Figure 5. Correlations of quantitative magnetic resonance imaging metrics.

Scatterplots denoting values for selected metrics within each muscle group. Spearman correlation coefficients and associated p-values are presented separately for CKD, ESKD, and control participants. n=19 for all metrics except T1ρ and T2 (n=18).

We previously reported regression of muscle fibrosis in CKD patients following dialysis initiation (21). Therefore, we explored differences in qMRI parameters between patients with non-dialysis dependent CKD and 3 patients with ESKD receiving hemodialysis (Table 3). Of note, these 3 ESKD patients were included in our prior muscle biopsy study and had levels of muscle collagen that were similar to healthy control participants (21). Consistent with our biopsy findings, T1ρ relaxation times – which were elevated in CKD patients compared with controls –were significantly lower among dialysis patients and similar to control levels. In addition, although not statistically significant in this small sample, fat fraction was numerically lower in the ESKD patients compared with CKD.

Muscle contractile CSA of the 7 muscle groups examined was lower in CKD patients and Dialysis patients than Controls (Controls: 81.3 ± 13.3 cm², CKD: 64.9 ± 20.4 cm², Dialysis: 56.3 ± 9.1 cm²; p=0.05 across groups). Total thigh muscle CSA measured across all 5 slices showed a similar pattern (Controls: 88.9 ± 12.5 cm², CKD: 77.0 ± 23.5 cm², Dialysis: 63.0 ± 6.4 cm²; p=0.05 across groups). Quadriceps CSA on the center slice was highly correlated with total thigh muscle CSA measured across all 5 slices (r=0.93, p<0.001). Associations of CKD status with qMRI metrics in age-adjusted models were unchanged after adjustment for CSA (T1ρ: −5.51 ms (95% CI −8.55 to −2.48, p<0.001); FF: −8.21% (95% CI −12.89 to −3.53, p=0.001); T2: −4.57 ms (95% CI −7.43 to −1.70, p=0.002)). In contrast, associations with Dialysis status were attenuated (T1ρ: −3.02 ms (95% CI −7.45 to 1.40, p=0.18); FF: −0.71% (95% CI −7.67 to 6.25, p=0.84); T2: −0.89 ms (95% CI −5.07 to 3.29, p=0.68)).

¹H MRS was used to further investigate differences in intramuscular fat and to quantify muscle metabolites (Figure 6). IMCL was numerically higher in the VL, soleus, and gastrocnemius and was overall significantly higher in CKD patients compared with controls (0.71% (95% CI 0.03 to 1.39, p=0.04). EMCL was numerically higher in CKD patients, but this difference was not statistically significant (0.47 (95% CI −0.04 to 0.99, p=0.07). Both TMA and creatinine concentrations were lower in skeletal muscle of CKD patients (2.68% (95% CI 0.18 to 5.18, p=0.04) and 1.54% (95% CI 0.37 to 2.71, p=0.01), respectively).

Figure 6. Skeletal muscle lipid content and metabolite concentrations measured by ¹H-magnetic resonance spectroscopy.

Mixed effects models including random intercepts were fit to account for clustering within an individual. Pairwise comparisons of CKD and control participants within muscle groups were performed using tests of contrast. Concentrations are presented referenced to the water peak. n=9 control, n=6 CKD. Abbreviations: IMCL, intramyocellular lipid content; EMCL, extramyocellular lipid content; TMA, trimethylamine. *p<0.05, **p<0.01, ***p<0.001.

DISCUSSION

In this pilot study of multiparametric qMRI, we identified elevations in T1ρ and T2 relaxation times and higher fat fraction in the skeletal muscle of patients with advanced non-dialysis dependent CKD. These changes were largely absent in a small group of patients receiving hemodialysis, in whom imaging parameters were similar to controls. Differences in T1ρ, T2, and fat fraction were highly correlated across the muscles we studied, suggesting that structural muscle pathology in CKD affects appendicular muscle broadly rather than being limited to specific muscle groups. ¹H-MRS yielded deeper insight into the finding of elevated fat fraction in CKD patients by demonstrating significantly higher IMCL in skeletal muscle. Furthermore, ¹H-MRS showed that concentrations of metabolites important to muscle function were reduced in CKD patients. Importantly, although physical function was poorer in the patients with CKD than in healthy volunteers, functional status was largely preserved. In summary, a multiparametric qMRI protocol shows promise for characterizing skeletal muscle pathology in patients with CKD prior to significant functional decline. Early identification of compromised skeletal muscle health could facilitate targeted exercise or pharmacologic interventions to mitigate muscle pathology prior to loss of muscle strength or performance. This approach may also yield new insights into changes in muscle function over time and following medical transitions, such as the initiation of dialysis.

Prior studies using MRI and ³¹P MRS found reduced muscle cross-sectional area and impaired bioenergetics in patients with CKD and ESKD (42-45). Our quantitative, multiparametric approach builds on this work. We found differences in a number of clinically relevant metrics that correlate with histological findings and physical function outcomes in other muscle disorders. For example, compared with control subjects, T1ρ relaxation times were prolonged in CKD patients. T1ρ employs an endogenous contrast method that is sensitive to interactions between extracellular matrix macromolecules, such as collagen and glycosaminoglycans, and protons of mobile water molecules (46, 47). Expansion of the extracellular matrix during fibrosis causes T1ρ times to lengthen: investigations of fibrosis in the heart, lung, and kidney reported elevated T1ρ times in each of these organs (48-56). Furthermore, in a swine model of myocardial infarction, assessment of the severity of fibrosis by T1ρ demonstrated excellent correlation with the gold standard late gadolinium enhancement MRI (52). In skeletal muscle, elevated T1ρ relaxation times correlate with expansion of the extracellular matrix and increased collagen content (9, 10, 57, 58); thus, T1ρ has been proposed as a non-invasive marker of fibrosis in skeletal muscle as well. Our prior work documented skeletal muscle fibrosis characterized by fibrillar collagen accumulation in patients with CKD, the severity of which correlated with eGFR reduction (8, 21). We also noted that fibrosis was dialysis-responsive, in that collagen content and other parameters improved after dialysis initiation (21). The T1ρ data presented here corroborate those findings and further support the distinction between muscle fibrosis and muscle atrophy suggested by our biopsy study: that in CKD, fibrosis occurs independently of atrophy, and that regulation of these processes is uncoupled following dialysis initiation, with regression of fibrosis despite worsening atrophy. Our results yield a further important insight: this fibrotic transformation appears to involve appendicular skeletal muscle diffusely. Thus, in addition to the benefit of being non-invasive, another advantage of qMRI over biopsy is the ability to assay a large volume of muscle and minimize sampling bias.

Another important finding was the elevation of fat fraction in CKD: IMCL in particular was higher based on ¹H-MRS data. This aligns with the myosteatosis that has been observed in patients with CKD stage 3 through 5 using ultrasound (59, 60). In that cohort, as in studies of muscular dystrophy and human aging, higher intramuscular fat content was associated with poorer physical performance (3, 60, 61). In the setting of CKD, intramuscular fat infiltration has multiple potential etiologies: fatty replacement accompanying fibrosis, possibly mediated by an expanded pool of fibro-adipogenic progenitor cells (8, 62); insulin resistance due to uremia, obesity, and type 2 diabetes (63); and downstream effects of neuropathy and denervation (64). Fat fraction correlated with both T1ρ and T2 among participants with CKD, consistent with the hypothesis that intramuscular fat deposition in CKD is linked to extracellular matrix expansion and fibrosis. Similarly, prior studies in genetic and acquired muscle disorders have reported correlations of fat fraction with T2 values (15, 16, 65), with the strength of the correlation possibly being greatest among patients with the most severe disease burden (15). This intramuscular fat deposition may be a component of the sarcopenic obese phenotype that becomes increasingly prevalent with progressive CKD and is associated with an increased risk of mortality (66-68).

¹H-MRS demonstrated lower concentrations of TMA and creatine in CKD muscle. Prior work suggests these findings could be clinically relevant. In patients with Duchenne muscular dystrophy, levels of these metabolites were lower in skeletal muscle compared with unaffected controls and correlated inversely with muscle function (19, 20). Molecules measured by the TMA and creatine peaks play key roles in muscle metabolism: the TMA peak captures carnitine and acetylcarnitine, which contain TMA moieties and can be distinguished at 7T but not 3T (69); the creatine peak captures both creatine and phosphocreatine (19). Thus, lower levels of TMA and creatine could indicate reduced muscle quality or impaired bioenergetics, which have been well documented in this patient population (44, 45). Future work is needed to investigate this hypothesis definitively; however, these findings suggest ¹H-MRS holds promise as a non-invasive tool for assessing muscle health and function in patients with CKD.

While strong evidence suggests microstructural alterations in CKD muscle (5), we did not observe significant differences in DTI markers between CKD patients and controls. This unexpected finding may be explained by the numerous factors that can alter water diffusion in skeletal muscle. Beyond changes in the macromolecular structure of the extracellular matrix, restriction of water diffusion can be impacted by muscle fiber diameter and density and by sarcolemma permeability (70). In addition, muscle fibers are composed of a mesh-like myofibrillar network, and sarcomere branching rates vary by fiber type (71). This suggests that water diffusion might be less restricted along the longitudinal axis of the fiber than previously thought, with the impact on fractional anisotropy dependent on the specific fiber type distribution within a given measurement area (voxel).

In this study, we applied unilateral multiparametric qMRI to the thigh and calf muscles on a clinical 3T MRI scanner. We selected a 16-channel transmit-receive (T/R) knee coil instead of the commonly used multi-channel body coil. The knee coil offers not only superior SNR but also lower heat deposition and improved B₁ uniformity with local transmission using circular polarization excitation. These advantages are particularly important for quantitative T1ρ and T2 mapping. Although a multi-channel body coil could have enabled more efficient bilateral muscle acquisition, the reduced B₀ and B₁ homogeneities would likely have compromised quantitative accuracy. To minimize the influence of muscle fat differences between study groups, active spectral selective fat suppression was applied during T2 mapping, T1ρ mapping, and diffusion-tensor imaging (ADC and FA). Fat fraction and T2* values were derived from a six-echo 3D gradient echo Dixon sequence (mDixon Quant) using iterative fitting of a multi-peak fat model to separate water and fat signals. This approach minimizes the impact of fat on the T2, T1ρ, ADC, FA, and T2* measurements. High-order B₀ shimming was performed for all scans to minimize B₀ inhomogeneity on quantitative measurements. Furthermore, a self-compensated T1ρ preparation module was implemented to reduce the influence of B₀ and B₁ inhomogeneity on T1ρ measurements (72, 73). Future work will focus on correcting the residual B₀ and B₁ inhomogeneities in T1ρ and T2 measurements to further improve quantification accuracy. Additionally, the vendor MRS sequence was used for metabolite quantification, but corrections for T1 and T2 relaxation times of metabolites and water were not performed, as these were not integrated into the sequence. Such corrections typically require separate scans to measure T1 and T2 values (74). While previous studies have applied T1 and T2 corrections using literature values (75, 76), reference values for metabolites in the thigh and calf muscles of CKD patients remain unavailable. Incorporating T1 and T2 corrections for each metabolite, as well as water T2, in the ¹H MRS measurements in future studies will further enhance their accuracy and address this important limitation (77).

Our multiparametric qMRI approach yielded a comprehensive assessment of muscle structure and metabolic health within a clinically acceptable timeframe. However, several limitations warrant consideration. First, our T1ρ and T2 mapping sequence was limited by the finite number of TSL and TE samples (four for each). Additionally, a mono-exponential decay model was employed in the fitting algorithm, which may not be ideal for the multi-component nature of muscle tissue. Furthermore, the observed muscle T2 values appear elevated. More advanced imaging techniques or post-processing methods, such as applying an extended phase graph fitting approach, could potentially enhance quantitative accuracy by mitigating the effects of B0 and B1 field inhomogeneities (78). Second, the protocol included only static qMRI measurements, precluding assessment of exercise-induced changes in blood flow and muscle bioenergetics. Such dynamic data could potentially enhance the identification of CKD patients at risk for functional decline. We chose not to incorporate these measures in this pilot study due to time constraints, the need for specific operator expertise, and limitations in hardware availability. Third, assessment of muscle CSA was limited by lack of anatomic matching of the slice selected for analysis and recognition that the maximal CSA for a particular muscle group was not necessarily captured. Fourth, the small sample size limited our ability to investigate the independent contributions of comorbidities. While this limits inferences regarding CKD-specific effects, it does not diminish the potential utility of qMRI metrics as biomarkers for physical function impairment. Future studies should incorporate additional metrics, including exercise response measurements, to develop a multiparametric protocol that maximizes prognostic ability while maintaining feasibility for clinical and research settings. Furthermore, larger studies are needed to delineate changes in qMRI metrics that are specific to CKD versus the contributions of related comorbidities such as diabetes, and to explore the influence of sex on the association of CKD with qMRI parameters. Fifth, the fat suppression technique employed did not suppress the olefinic lipid resonance, although this would have only a very minor effect on the results.

In conclusion, multiparametric qMRI imaging of skeletal muscle identified differences among CKD patients in metrics associated with fibrosis, fat infiltration, and metabolic dysregulation. Future work is needed to determine the extent to which these measures predict impairment of physical performance in this patient population.

Supplementary Material

Supplemental File. doi: 10.6084/m9.figshare.28856993

Supplemental Figure S2. doi: 10.6084/m9.figshare.28856993

Preserving Physical Function in CKD (PPF-CKD) Investigators

Albert Einstein College of Medicine/Montefiore Medical Center

Matthew K. Abramowitz, MD, MS

Maureen E. Brogan, MD

Mary J. Dominguez, MD

Sally Duran, BA

Molly C. Fisher, DO, MS

Ladan Golestaneh, MD, MS

Rimon Golovey, MD

Sonali Gupta, MBBS

Tanya S. Johns, MD, MHS

Michele H. Mokrzycki, MD, MS

Tina M. Neade, MD

Joel J. Neugarten, MD

William Paredes, BS

Qi Peng, PhD

Amanda C. Raff, MD

Stuart O. Rosenberg, MD

Gracy Sebastian, NP

Deep Sharma, MBBS

Milagros D. Yunes, MD

University of Kentucky

Christopher S. Fry, PhD

Alexander R. Keeble

Allison M. Owen, PhD

Lehman College, City University of New York

Brad J. Schoenfeld, PhD

University of Maryland School of Medicine

Christopher W. Ward, PhD

University of California, Berkeley

Moriel Vandsburger, PhD

DATA AVAILABILITY

Source data for this study are not publicly available due to privacy restrictions related to concerns over maintenance of participant confidentiality given the limited sample size. The source data are available to verified researchers upon request by contacting the corresponding author.