Prospective comparison of two methods for assessing sarcopenia and interobserver agreement on retrospective CT images.

PURPOSE OF THE STUDY: To compare the relationships between two CT derived sarcopenia assessment methods, and compare their relationship with inter-rater and intrarater validations and colorectal surgical outcomes. STUDY DESIGN: 157 CT scans were identified across Leeds Teaching Hospitals National Health Service Trust for patients undergoing colorectal cancer surgery. 107 had body mass index data available, required to determine sarcopenia status. This work explores the relationship between sarcopenia, as measured by both total cross sectional-area (TCSA) and psoas-area (PA) and surgical outcomes. All images were assessed for inter-rater and intrarater variability for both TCSA and PA methods of sarcopenia identification. The raters included a radiologist, an anatomist and two medical students. RESULTS: Prevalence of sarcopenia was different when measured by PA (12.2%-22.4%) in comparison to TCSA (60.8%-70.1%). Strong correlation exists between muscle areas in both TCSA and PA measures, however, there were significant differences between methods after the application of method-specific cut-offs. There was substantial agreement for both intrarater and inter-rater comparisons for both TCSA and PA sarcopenia measures. Outcome data were available for 99/107 patients. Both TCSA and PA have poor association with adverse outcomes following colorectal surgery. CONCLUSIONS: CT-determined sarcopenia can be identified by junior clinicians, those with anatomical understanding and radiologists. Our study identified sarcopenia to have a poor association with adverse surgical outcomes in a colorectal population. Published methods of identifying sarcopenia are not translatable to all clinical populations. Currently available cut-offs require refinement for potential confounding factors, to provide more valuable clinical information.