Differentiation of dual- and non-dual-phenotype hepatocellular carcinoma based on contrast-enhanced computed tomography and patient characteristics.

Background: Dual-phenotype hepatocellular carcinoma (DPHCC) is associated with a higher risk of recurrence, but little is known about its clinicodemographic characteristics or imaging features. This study aimed to assess whether contrast-enhanced computed tomography (CECT) and patient characteristics can facilitate preoperative differentiation of DPHCC and non-DPHCC. Methods: Features on CECT images and clinicodemographic characteristics were retrospectively analyzed from a consecutive series of hepatocellular carcinoma (HCC) patients between January 2020 and December 2020. Disease was confirmed based on surgical pathology, and CECT was performed within four weeks before surgery. Univariate and multivariate logistic regression analyses were performed to identify independent risk factors for DPHCC, and an imaging model based on CECT features and a combined model based on clinicodemographic characteristics and CECT features were constructed, respectively. Delong's test was used for comparison of the area under the curve (AUC) values between the two models. Kaplan-Meier survival analysis was used to assess overall survival (OS) in DPHCC and non-DPHCC groups. Results: A total of 29 patients with DPHCC and 140 with non-DPHCC were included in this study. DPHCC was significantly more prevalent among female patients and less prevalent among those infected with hepatitis B virus (HBV). CECT associated DPHCC with rim arterial phase hyperenhancement (APHE) and peripheral washout, whereas it associated non-DPHCC with non-rim APHE and non-peripheral washout. Multivariate logistic regression identified one independent CECT feature of DPHCC: rim APHE [odds ratio (OR) 11.040, 95% confidence interval (CI): 1.98-63.532]. The imaging model was constructed based on rim APHE-predicted DPHCC with an AUC of 0.562 (95% CI: 0.439-0.684). Multivariate logistic regression identified three independent clinicodemographic characteristics and CECT features of DPHCC: female sex (OR 4.519, 95% CI: 1.529-13.357), infection with HBV (OR 0.234, 95% CI: 0.084-0.654) and rim APHE (OR 15.016, 95% CI: 2.335-96.585). The combined model was constructed based on three independent predictors of DPHCC with an AUC of 0.716 (95% CI: 0.603-0.829). Delong's test showed that the AUC of the combined model was higher than that of the imaging model, and the difference was statistically significant (Z=3.207, P<0.05). The OS rates of the patients in the DPHCC and non-DPHCC groups were 68.7% and 77.2%, respectively. Kaplan-Meier survival analysis showed no statistical difference in OS rates between groups (P=0.362). Conclusions: The combined model established based on female sex, infection with HBV, and rim APHE by CECT can facilitate preoperative differentiation of DPHCC and non-DPHCC. DPHCC is more likely to cause death than non-DPHCC, suggesting that active postoperative management of patients with DPHCC is required.