To evaluate the association, agreement, and reproducibility of contrast-enhanced computed tomography (CT)-derived depth of invasion (rDOI) compared with pathological depth of invasion (pDOI) in lip squamous cell carcinoma.

This retrospective single-center study included 51 patients who underwent preoperative contrast-enhanced CT between January 2010 and April 2024. Two observers independently measured rDOI, tumor thickness, and maximum tumor diameter twice. Pathological measurements served as the reference standard. Correlation, Bland–Altman agreement, mean absolute error, classification performance at pDOI thresholds of >5 mm and >10 mm, and intraobserver and interobserver reproducibility were assessed.

Mean pDOI was 6.48 ± 4.67 mm; mean rDOI was 5.87 ± 4.11 mm for observer 1 and 6.77 ± 4.64 mm for observer 2. rDOI correlated significantly with pDOI (r = 0.685 and 0.781; both p < 0.001). Mean CT–pathology biases were −0.61 and 0.29 mm, but the 95% limits of agreement were wide. Mean absolute errors were 3.08 and 2.75 mm. Classification accuracy was 72.9% at the > 5-mm threshold and 83.3% at the > 10-mm threshold for both observers, although sensitivity varied at > 10 mm. Interobserver reproducibility was excellent for DOI, tumor thickness, and maximum tumor diameter.

CT-derived DOI was reproducible in measurable lip squamous cell carcinomas but was not interchangeable with pDOI. It should be used as an adjunctive preoperative estimate rather than as a stand-alone basis for DOI-based staging or surgical management.