Defining when to offer operative treatment for intrahepatic cholangiocarcinoma: A regret-based decision curves analysis.

Abstract

Regret-based decision curve analysis (DCA) is a framework that assesses the medical decision process according to physician attitudes (expected regret) relative to disease-based factors. We sought to apply this methodology to decisions around the operative management of intrahepatic cholangiocarcinoma (ICC). Utilizing a multicentric database of 799 patients who underwent liver resection for ICC, we developed a prognostic nomogram. DCA tested 3 strategies: (1) perform an operation on all patients, (2) never perform an operation, and (3) use the nomogram to select patients for an operation. Four preoperative variables were included in the nomogram: major vascular invasion (HR = 1.36), tumor number (multifocal, HR = 1.18), tumor size (>5 cm, HR = 1.45), and suspicious lymph nodes on imaging (HR = 1.47; all P < .05). The regret-DCA was assessed using an online survey of 50 physicians, expert in the treatment of ICC. For a patient with a multifocal ICC, largest lesion measuring >5 cm, one suspicious malignant lymph node, and vascular invasion on imaging, the 1-year predicted survival was 52% according to the nomogram. Based on the therapeutic decision of the regret-DCA, 60% of physicians would advise against an operation for this scenario. Conversely, all physicians recommended an operation to a patient with an early ICC (single nodule measuring 3 cm, no suspicious lymph nodes, and no vascular invasion at imaging). By integrating a nomogram based on preoperative variables and a regret-based DCA, we were able to define the elements of how decisions rely on medical knowledge (postoperative survival predicted by a nomogram, severity disease assessment) and physician attitudes (regret of commission and omission).