Identifying critically ill patients with low muscle mass: Agreement between bioelectrical impedance analysis and computed tomography

Willem G.P.M. Looijaard, Sandra N. Stapel, Ingeborg M. Dekker, Hanna Rusticus, Sharon Remmelzwaal, Armand R.J. Girbes, Peter J.M. Weijs, Heleen M. Oudemans-van Straaten

    Research output: Contribution to journalArticleAcademicpeer-review

    13 Citations (Scopus)
    31 Downloads (Pure)

    Abstract

    Background & aims: 

    Low muscle mass and -quality on ICU admission, as assessed by muscle area and -density on CT-scanning at lumbar level 3 (L3), are associated with increased mortality. However, CT-scan analysis is not feasible for standard care. Bioelectrical impedance analysis (BIA) assesses body composition by incorporating the raw measurements resistance, reactance, and phase angle in equations. Our purpose was to compare BIA- and CT-derived muscle mass, to determine whether BIA identified the patients with low skeletal muscle area on CT-scan, and to determine the relation between raw BIA and raw CT measurements. 

    Methods: 

    This prospective observational study included adult intensive care patients with an abdominal CT-scan. CT-scans were analysed at L3 level for skeletal muscle area (cm2) and skeletal muscle density (Hounsfield Units). Muscle area was converted to muscle mass (kg) using the Shen equation (MMCT). BIA was performed within 72 h of the CT-scan. BIA-derived muscle mass was calculated by three equations: Talluri (MMTalluri), Janssen (MMJanssen), and Kyle (MMKyle). To compare BIA- and CT-derived muscle mass correlations, bias, and limits of agreement were calculated. To test whether BIA identifies low skeletal muscle area on CT-scan, ROC-curves were constructed. Furthermore, raw BIA and CT measurements, were correlated and raw CT-measurements were compared between groups with normal and low phase angle. 

    Results: 

    110 patients were included. Mean age 59 ± 17 years, mean APACHE II score 17 (11–25); 68% male. MMTalluri and MMJanssen were significantly higher (36.0 ± 9.9 kg and 31.5 ± 7.8 kg, respectively) and MMKyle significantly lower (25.2 ± 5.6 kg) than MMCT (29.2 ± 6.7 kg). For all BIA-derived muscle mass equations, a proportional bias was apparent with increasing disagreement at higher muscle mass. MMTalluri correlated strongest with CT-derived muscle mass (r = 0.834, p < 0.001) and had good discriminative capacity to identify patients with low skeletal muscle area on CT-scan (AUC: 0.919 for males; 0.912 for females). Of the raw measurements, phase angle and skeletal muscle density correlated best (r = 0.701, p < 0.001). CT-derived skeletal muscle area and -density were significantly lower in patients with low compared to normal phase angle. 

    Conclusions: 

    Although correlated, absolute values of BIA- and CT-derived muscle mass disagree, especially in the high muscle mass range. However, BIA and CT identified the same critically ill population with low skeletal muscle area on CT-scan. Furthermore, low phase angle corresponded to low skeletal muscle area and -density. 

    Trial registration: 

    ClinicalTrials.gov (NCT02555670).

    Original languageEnglish
    Pages (from-to)1809-1817
    JournalClinical Nutrition
    Volume39
    Issue number6
    DOIs
    Publication statusPublished - Jun 2020

    Fingerprint

    Dive into the research topics of 'Identifying critically ill patients with low muscle mass: Agreement between bioelectrical impedance analysis and computed tomography'. Together they form a unique fingerprint.

    Cite this