Improving Prediction of Early Recurrence in Perihilar Cholangiocarcinoma

Researchers developed and validated a noninvasive model that has the potential to preoperatively predict early recurrence for perihilar cholangiocarcinoma (pCCA), a type of hepatobiliary tumor. This model uses a magnetic resonance imaging (MRI)-based radiomics signature combined with clinical variables to form the prediction.

Zhao et al developed these models during a retrospective study on 184 patients (115 men and 69 women) with pCCA. They developed 3 models from a training set of 128 patients and validated these models in a separate testing set of 56 patients. These models included Model^radiomic, Model^clinic, and Model^combine.

Development of Prediction Models for pCCA

The investigators created the Model^radiomic using contrast-enhanced arterial and portal vein phase MRIs of the hepatobiliary system. To extract and model radiomics features from these MRIs, they applied correlation analysis, least absolute shrinkage and selection operator (LASSO) logistic regression (LR), and backward stepwise LR. Model^clinic consisted of preoperative clinical predictors selection and modeling and used univariate and multivariate backward stepwise LR. Lastly, Model^combine combined the Model^radiomics and Model^clinic predictors using a multivariate LR method.

Validation of Prediction Models for pCCA

Next, the researchers validated these constructed models using the Delong test to compare the area under the curves in the training set. Both Model^clinic and Model^combine demonstrated superior statistically significant performance when compared with the Model^radiomic and the Tumor-Node-Metastasis (TNM) system. The results from the training set were then compared with the validation testing. Both Model^clinic and Model^combine demonstrated statistically significant performance compared with the TNM system; however, only the Model^combine proved superior to Model^radiomic. Currently, the TNM system is the most widely used grading system to predict cancer recurrence and spread.

The authors suggest “preoperative accurate prediction of postoperative early recurrence can avoid unnecessary operation(s).” The driving purpose of developing this method of predicting early recurrence of pCCA is to provide advanced insights so that surgeons can make informed decisions based upon the poor prognosis that typically accompanies early recurrence following surgeries in these patients.

Reference

Zhao J, Zhang W, Zhu Y-Y, et al. Development and validation of noninvasive MRI-based signature for preoperative prediction of early recurrence in perihilar cholangiocarcinoma. J Magn Reson Imaging. Published online July 23, 2021. doi:10.1002/jmri.27846