Preoperative MRI to predict early recurrence after curative resection of single hepatocellular carcinoma

Abstract

Purpose: Hepatic resection is a primary treatment modality for hepatocellular carcinoma (HCC), but the high incidence of recurrence after hepatic resection remains its major drawback. Especially early recurrence within the first 2 years after resection of HCC is known to be associated with worse prognosis. The purpose of this study was to develop and validate a nomogram based on a multivariate model to preoperatively predict the risk of early recurrence after curative resection of single HCC, using various imaging findings of dynamic gadoxetic acid-enhanced and diffusion-weighted MRI.Materials and Methods: A total of 268 patients were included who underwent curative hepatic resection for single HCCs followed by gadoxetic acid-enhanced MRI from January 2008 to August 2011. We allocated 187 patients who underwent hepatic resection before September 2010 into the training cohort to derive a prediction model, and the remaining 81 patients were allocated into the validation cohort. All MR images of the training cohort were retrospectively reviewed by two radiologists by consensus, who determined whether the following eight MRI features were found in each HCC: ring-like arterial enhancement, peritumoral arterial enhancement, capsule, peritumoral hypointensity in the hepatobiliary phase (HBP), non-smooth tumor margin in the HBP, satellite nodules, intratumoral fat, and gross vascular invasion. Based on the multivariate logistic regression analysis, a nomogram was constructed. To validate a prediction model established from the results of the first review, the same radiologists analyzed the MR images of the validation cohort and assessed MRI features that had been independently associated with early recurrence in the training cohort.Results: Univariate and multivariate logistic regression analyses with 187 patients in the training set showed significant associations between early HCC recurrence and the following factors: ring-like enhancement (odds ratio [OR], 3.83; 95% confidence interval [CI], 1.39-10.52; P=0.01), peritumoral enhancement (OR, 2.64; 95% CI, 1.27-5.46; P<0.009), satellite nodule (OR, 4.07; 95% CI, 1.09-15.21; P=0.037), and tumor size (OR, 1.66; 95% CI, 1.31-2.09; P<0.001). The prediction model based on this result showed an area under the curve (AUC) of 0.791 in the training set. Internal validation with bootstrap sampling was performed to determine calibration accuracy, and the corrected AUC remained high (0.788) after bootstrapping. For the validation data set, discrimination was as good as for the training set, with an AUC of 0.783. The calibration was also good, with no significant difference (P=0.091) between the nomogram-predicted and the observed early recurrence rates.Conclusion: We developed and validated a nomogram that can be used to preoperatively estimate the risk of recurrence within 2 years after resection of single HCCs.