The Long-To-Short-Axis Ratio and Multifocality are Associated With TP53 Mutation Status in Surgically Resected...

In hepatocellular carcinoma (HCC), the tumor protein 53 (TP53) gene is frequently mutated and the mutations have been associated with poor prognosis. We aim to retrospectively identify the relationship between TP53 mutation status, tumor size (long-axis diameter, short-axis diameter, and long-to-short-axis ratio [L/S ratio]), margin and multifocality in surgically resected HCC.

The image features and TP53 mutation data from 78 patients generated with National Cancer Institute's multi-institutional The Cancer Genome Atlas (TCGA)/The Cancer Imaging Archive databases were assessed. Binary logistic regression analyses were performed to identify independent factors of harboring TP53 mutation status. The final model was selected by using the backward elimination method.

TP53 mutations were found in 19 (31.5%) of 78 patients. TP53 mutation rates were significantly higher (a) in L/S ratio ≤ 1.2 14 of 41 [34.1%]) lesions than in L/S ratio >1.2 lesions (five of 37 [13.5%]) (p = 0.034) and (b) in nonmultifocality (17 of 54[31.5%]) than in multifocality lesions (two of 24 [8.3%]) (p = 0.028). On univariate logistic regression analysis, L/S ratio (≤1.20 vs >1.20. odds ratio [OR]: 3.319; p = 0.040; 95% confidence interval [CI]: 1.059-10.401 Area Under Curve (AUC) = 0.634) and multifocality (no vs yes OR: 5.054; p = 0.041; 95% CI: 1.065-23.986 AUC = 0.640) were associated with TP53 mutations. On multivariate logistic regression analysis, L/S ratio (≤1.20 vs >1.20 OR: 3.430; p = 0.040; 95% CI: 1.058-11.118) and multifocality (no vs yes OR: 5.232; p = 0.041; 95% CI: 1.072-25.526) were associated with TP53 mutations. The area under the receiver operating characteristic curve for predicting TP53 mutation status was 0.714 (95% CI: 0.590-0.837).

Our study focusing on identifying imaging aspects related to TP53 positive HCC. L/S ratio of HCC in combination with multifocality might be used to prognosticate TP53 mutation status. And the discriminatory power for this prediction model was good.