Measurement of body composition: Agreement between methods of measurement by bioimpedance and computed tomography in patients with non-small cell lung cancer

BACKGROUND AND AIMS: Cancer cachexia is a muscle wasting syndrome which is promoted by the presence of inflammation and metabolic changes. Cancer cachexia has shown predictable for survival, progression of physical impairment as well as patient-reported outcomes. Being able to identify these patients with a clinically applicable method for measuring body composition could lead to early nutritional intervention, possibly preventing progression of cancer cachexia. This study aimed to investigate the agreement between body composition recorded with bioelectrical impedance analysis (BIA) and software analysis of CT scans of patients suffering from cancer with a special emphasis on muscle mass.

METHODS: This is a cross-sectional study of newly diagnosed patients with non-small cell lung cancer. All patients had muscle mass (MM) and fat mass (FM) measured by BIA with Tanita Segmental Body Composition Analyzer (BC-418) and by software analysis of CT scans at L3-level with VikingSlice. Secondary parameters (systemic inflammation, physical performance and electrolyte balance) were investigated for possible pitfalls of BIA. Patients who fulfilled best standard practice for BIA measurements, including 4 h of fasting, absence of oedema and no metallic implants, participated in a subgroup analysis. Agreement between method of measurements is investigated with the Bland and Altman method and pitfalls with paired t-test. A satisfactory agreement of BIA and CT measurements is established as ±10%.

RESULTS: A total of 60 patients participated in the study amongst which 37 participated in the subgroup analysis. The bias between methods of measurement for MM was 4.74 ± 11.55% (P = 0.002) with limits of agreement (LoA) of -17.91 to 27.38%. In the subgroup analysis, the bias of MM was 3.06 ± 10.71% (P = 0.091) and LoA were -17.94 to 24.06%. In the analysis of secondary parameters, significant differences between methods were found in patients with normal hand grip strength (1.83 kg, P = 0.023) and those performing under gender and age-specific cut-offs in the 30-s sit to stand test (5.15 kg, P = 0.009).

CONCLUSIONS: BIA and CT image analysis were not comparable for body composition measurements. BIA overestimated MM and underestimated FM with LoA outside that of the clinically assessed acceptable difference. Bias was reduced and LoA narrowed in the subgroup analysis however not to acceptable levels. This study finds it possible that patients with low physical performance are liable to larger overestimations of MM with BIA than those with normal physical performance.