Muscle Biomarkers in Colorectal Cancer Outpatients: Agreement Between Computed Tomography, Bioelectrical Impedance Analysis, and Nutritional Ultrasound

Abstract

Background: Muscle quality and mass in cancer patients have prognostic and diagnostic importance.

Objectives: The objectives are to analyze agreement between gold-standard and bedside techniques for morphofunctional assessment.

Methods: This cross-sectional study included 156 consecutive colorectal cancer outpatients that underwent computed tomography (CT) scanning at lumbar level 3 (L3), whole-body bioelectrical impedance analysis (BIA), point-of-care nutritional ultrasound^® (US), anthropometry, and handgrip strength in the same day. Measured muscle biomarkers were stratified by sex, age, BMI-defined obesity, and malnutrition using Global Leadership in Malnutrition (GLIM) criteria. Whole-body estimations for muscle mass (MM) and fat-free mass were calculated using two different equations in CT (i.e., Shen, and Mourtzakis) and four different equations for BIA (i.e., Janssen, Talluri, Kanellakis, and Kotler). Muscle cross-sectional area at L3 was estimated using the USVALID equation in US. Different cut-off points for muscle atrophy and myosteatosis were applied. Sarcopenia was defined as muscle atrophy plus dynapenia. Intra-technique and inter-technique agreement were analyzed with Pearson, Lin (ρ), and Cohen (k) coefficients, Bland-Altman analyses, and hypothesis tests for measures of central tendency.

Results: Intra-technique agreements on muscular atrophy (CT k = 0.134, BIA k = -0.037, US k = 0.127) and myosteatosis (CT k = 0.122) were low, but intra-technique agreement on sarcopenia in CT was fair (k = 0.394). Inter-technique agreement on muscular atrophy and sarcopenia were low. Neither CT and BIA (ρ = 0.468 to 0.772 depending on equation), nor CT and US (ρ = 0.642), were interchangeable. Amongst the BIA equations, MM by Janssen proved the best, with a 1.5 (3.6) kg bias, (-5.6, 8.6) kg LoA, and 9/156 (5.7%) measurements outside the LoA. Muscle biomarkers in all techniques were worse in aged, female, or malnourished participants. Obesity was associated with higher muscle mass or surface biomarkers in all techniques.

Conclusions: Bedside techniques adequately detected patterns in skeletal muscle biomarkers, but lacked agreement with a reference technique in the study sample using the current methodology.

Keywords: GLIM; bioelectrical impedance analysis; colorectal cancer; computed tomography; muscle mass; nutritional ultrasound®; obesity; sarcopenia; validation study.

Figure 1

Patient recruitment, exclusions, and final sample size. BIA: bioelectrical impedance analysis; CT: computed tomography; n = absolute frequency; US: nutritional ultrasound^®.

Figure 2

CT-based muscle biomarkers, by rows (from top to down): skeletal muscle area (A–D), skeletal muscle density (E–H), skeletal muscle index (I–L), and skeletal muscle gauge (D–P). Data are stratified in columns (from left to right) by sex (A,E,I), age (B,F,J), obesity (C,G,K), and malnutrition (D,H,L). Statistical significance for comparisons of central tendency measures is depicted as follows: p > 0.05 = ns; p ≤ 0.05 = *; p ≤ 0.01 = **; p ≤ 0.001 = ***; p ≤ 0.0001 = ****. AU: arbitrary uits; BMI: Body Mass Index; Cm: centimeters; GLIM: Global Leadership Initiative in Malnutrition; HU: Hounsfield units; L3-SMA: skeletal muscle area at L3; L3-SMD: skeletal muscle density at L3; m: meters; SMG: skeletal muscle gauge at L3; SMI: skeletal muscle index at L3M.

Figure 3

BIA-based muscle biomarkers, by rows (from top to down): Janssen et al. [34] MM (A–D), Talluri MM (E–H), and θ (I–L). Data are stratified in columns (from left to right) by sex (A,E,I), age (B,F,J), obesity (C,G,K), and malnutrition (D,H,L). °: degrees; θ: raw phase angle; MM: muscle mass. Statistical significance for comparisons of central tendency measures is depicted as follows: p > 0.05 = ns; p ≤ 0.01 = **; p ≤ 0.001 = ***; p ≤ 0.0001 = ****.

Figure 4

US-based muscle biomarkers, by rows (from top to down): rectus femoris muscle thickness (A–D), quadricipital muscle thickness (E–H), and quadricipital muscle area (I–L). Data are stratified in columns (from left to right) by sex (A,E,I), age (B,F,J), obesity (C,G,K), and malnutrition (D,H,L). Mm: millimeters. RF-MT: rectus femoris muscle thickness in mm; Quad-MT: quadricipital muscle thickness (rectus femoris plus vastus intermedius) in mm; RF-CSA: rectus femoris cross-sectional area in cm². Statistical significance for comparisons of central tendency measures is depicted as follows: p > 0.05 = ns; p ≤ 0.01 = **; p ≤ 0.001 = ***; p ≤ 0.0001 = ****.

Figure 5

Correlation matrices, based on Spearman’s correlation coefficients (r). Direct correlations are shown in shades of green, and inverse correlations are shown in shades of red. Correlation strength is proportional to color intensity. (A) BIA-measured (R_z, X_c, and θ) parameters, BIA-estimated FFM (Kanellakis, Kotler, FFM-Talluri) [36,37] and BIA-estimated MM (Janssen, MM-Talluri) [34], CT-measured parameters (SMA, SMD), CT-estimated whole-body MM (Shen) [2] and FFM (Mourtzakis) [3], anthropometry (Height, Weight), and handgrip strength (HGS). (B) US-measured (Quad-MT, RF-MT, RF-CSA), BIA-estimated FFM and FM, CT-measured parameters, anthropometry, and handgrip strength. Statistical significance for comparisons of central tendency measures is depicted as follows: p > 0.05 = ns; p ≤ 0.05 = *; p ≤ 0.01 = **; p ≤ 0.001 = ***.

Figure 6

Scatterplot for CT-based MM in kg using Shen [2] equation on the X-axis, and different BIA-based equations for muscle mass (MM) in kilograms (kg): Janssen [34] (A), and Talluri (D) on the Y-axis. Linear regression models are represented as a blue line, and a perfect regression is represented as a red line. Bland–Altman plots for CT-based MM in kg using Shen equation, in comparison with different BIA-based MM equations: Janssen (B), and Talluri (E). In all cases, averages of MM for both equations are presented on the X-axis; differences in MM for both equations are presented on the Y-axis. A linear regression model to detect dose-dependent bias is represented as a blue line. Please note how the Y-axis has the same scale and range for easier comparisons. Individual measurements are shown as black dots; biases are represented as horizontal dashed black lines; upper and lower limits of agreement are shown as horizontal solid red lines. The absence of differences (Y intercept = 0) between equations is represented as a horizontal solid black line. Boxplots with overlaid point geometry for CT-based MM in kg using Shen equation on the Y-axis, and different BIA-based equations for MM in kg: Janssen (C), and Talluri (F) on the X-axis. Please note how the Y-axis has the same scale and range for easier comparisons. Each participant is joint by dashed grey lines. Statistical significances for comparisons of central tendency measures is depicted as follows: p ≤ 0.0001 = ****.

Figure 7

Scatterplot for CT-based FFM in kg using Mourtzakis [3] equation on the X-axis, and different BIA-based equations for fat-free mass (FFM) in kilograms (kg): Kanellakis [36] (A), Kotler [37] (D), and Talluri (G) on the Y-axis. Bland–Altman plots for CT-based FFM in kg using Mourtzakis equation, in comparison with different BIA-based FFM equations: Kanellakis (B), Kotler (E), and Talluri (H). Boxplots with overlaid point geometry for CT-based FFM in kg using Mourtzakis equation on the Y axis, and different BIA-based equations for FFM in kg: Kanellakis (C), Kotler (F), and Talluri (I) on the X-axis. Statistical significances for comparisons of central tendency measures is depicted as follows: p ≤ 0.0001 = ****.

Figure 8

Scatterplot for CT-measured muscle CSA at L3 on the X-axis, and US-estimated muscle CSA at L3 on the Y-axis using the Fischer et al. equation [38] (A) and a similar regression equation with different ß coefficients (D). Bland–Altman plots for CT-measured muscle CSA at L3 on the X-axis, and US-estimated muscle CSA at L3 on the Y-axis using the Fischer et al. equation (B) and a similar regression equation with different ß coefficients (E). Boxplots with overlaid point geometry for CT-measured muscle CSA at L3 on the X-axis, and US-estimated muscle CSA at L3 on the Y-axis using the Fischer et al. equation (C) and a similar regression equation with different ß coefficients (F). Statistical significance for comparisons of central tendency measures is depicted as follows: p ≤ 0.0001 = ****.