Intra- and interobserver variability in skeletal muscle measurements using computed tomography images

Abstract

PurposeThe progressive loss of skeletal muscle and function (known as sarcopenia) has been shown to be associated with various adverse outcome measures. Sophisticated measurements of body composition are increasingly being incorporated into research studies to stratify patients into those with or without sarcopenia, monitor treatment effects, and predict complications. A typical approach is to select axial image(s) at the mid-lumbar level and use semi-automated software to identify and quantify the skeletal muscle area. This area is then used to estimate whole-body parameters. This approach is somewhat subjective, and in this study we investigate its reproducibility, both within and between observers. Materials and methodsRepeated muscle measurements were performed on a cohort of 29 patients by 3 radiologists, to examine their intra- and interobserver reproducibility. Results and discussionMean muscle area for the cohort was 156 cm2, with a wide range (98 – 261 cm2). There was good intraobserver agreement between measurements, with a mean absolute difference between repeated measurements on the same patient of 0.98 cm2, and a measurement variability of 2.92 cm2. Much of the variability was shown to be due to the choice of a different slice when performing the repeated measurement. Averaging two slices provided a small but non-significant improvement in comparison to the single slice approach. Interobserver results showed good agreement, though there was a small bias for one observer, who measured slightly larger volumes compared to the other two. We conclude that the approach described provides reproducible skeletal muscle area measurements, and offer three specific recommendations to minimise variability.