CT colonography for assessing visceral obesity: clinical utility beyond colorectal screening
摘要
Visceral obesity is a significant health risk, and the visceral fat area (VFA) measured on CT is a reliable indicator. The current study investigated whether the CT colonography (CTC)-derived VFA can identify visceral obesity, accounting for imaging variations due to colonic insufflation.
MethodsWe retrospectively included patients who underwent both abdominal CT and CTC within 30 days. The waist circumference, VFA, and subcutaneous fat area (SFA) were measured at the umbilical and L1–L5 vertebral levels. Visceral obesity was defined as a VFA ≥ 100 cm² at the umbilical level on CT. The diagnostic performances of the CTC-derived VFA, BMI, and waist circumference were compared. In a secondary analysis, regression models were constructed to estimate the CT-derived VFA from the CTC measurements. Model performance was evaluated via 5-fold cross-validation and quantified with the mean absolute error (MAE).
ResultsFifty-eight patients were included (median age 51.0 years, IQR 43.0–58.8; 42 males, 16 females). Compared with CT, CTC revealed a significantly greater waist circumference (89.9 vs. 87.5 cm, P < 0.001) and a lower VFA (96.0 vs. 110.3 cm², P = 0.005), whereas there was no significant difference in the SFA. The umbilical slice position remained unchanged in 49 patients (84.5%), and no patient’s position shifted by two or more levels. The CTC-derived VFA demonstrated the highest diagnostic performance for visceral obesity, with a sensitivity and specificity of 0.90 and 0.93. BMI yielded 0.73 and 0.75, and waist circumference 0.90 and 0.79, respectively. Incorporating additional slices (e.g., L2) further reduced the MAE from 19.76 to 16.54 cm² (P = 0.013).
ConclusionThe CTC-derived VFA was shown to be a viable marker for identifying visceral obesity, thus potentially offering a secondary diagnostic resource without additional radiation exposure or costs. While single-slice models showed reasonable performance, multi-slice models demonstrated greater predictive accuracy.
Graphical abstract