Abstract
OBJECTIVE: Computed tomography (CT) measured muscle density is prognostic of health outcomes. However, the use of intravenous contrast obscures prognoses by artificially increasing CT muscle density. We previously established a correction to equalize contrast and noncontrast muscle density measurements. While this correction was validated internally, the objective of this study was to obtain external validation using different patient cohorts, muscle regions, and CT series. METHODS: CT images from 109 patients with kidney tumors who received abdominal CT scans with a multiphase intravenous contrast protocol were analyzed. Paraspinal muscle density measurements taken during noncontrast, venous phase, and delayed phase contrast scans were collected. An a priori correction of -7.5 Hounsfield units (HU) was applied to muscle measurements. Equivalence testing was utilized to determine statistical similarity. RESULTS: In the sample of 109 patients (mean age: 63 years [SD: 14.3]; 41.3% female), densities in smaller regions of interest within the paraspinal muscles and the entire paraspinal muscle density (PS) in venous and delayed phase contrast scans were higher than in noncontrast. Equivalence testing showed that average corrected contrast and noncontrast muscle densities were within 3 HU for both muscle measures for the total patient sample, and for a majority of male and female subsamples. The correction is suitable for regions of interests of venous contrast (90% CI: -1.90, -0.69 HU) and delayed contrast scans (90% CI: 0.075, 1.29 HU) and within the PS measures of venous contrast (90% CI: -2.04, -0.94 HU) and delayed contrast scans (90% CI: -0.11, 0.89 HU). CONCLUSIONS: The previously established correction for contrast of -7.5 HU was applied in a new patient population, axial muscle region, muscle measurement size, and expanded on previously studied contrast phases. The correction produced contrast-corrected muscle densities that were statistically equivalent to noncontrast muscle densities. The simplicity of the correction gives clinicians a tool that seamlessly integrates into practice or research to improve harmonization of data between contrast and noncontrast scans.