Construction and Validation of a Nomogram for Predicting the Risk of Deep Vein Thrombosis in Hepatocellular Carcinoma Patients After Laparoscopic Hepatectomy: A Retrospective Study

Background: The incidence of deep vein thrombosis (DVT) in hepatocellular carcinoma (HCC) patients after laparoscopic hepatectomy (LH) is unclear, and there is no effective method for DVT risk assessment in these patients. Methods: The data from the total of 355 consecutive HCC patients who underwent LH were included. A DVT risk algorithm was developed using a training set (TS) of 243 patients, and its predictive performance was evaluated in both the TS and a validation set (VS) of 112 patients. The model was then used to develop a DVT risk nomogram (TRN). Results: The incidence of DVT in the present study was 18.6%. Age, sex, body mass index (BMI), comorbidities and operative position were independent risk factors for DVT in the TS. The model based on these factors had a good predictive ability. In the TS, it had an area under the receiver operating characteristic (AUC) curve of 0.861, Hosmer-Lemeshow (H-L) goodness of fit p value of 0.626, sensitivity of 44.4%, specificity of 96.5%, positive predictive value (PPV) of 74.1%, negative predictive value (NPV) of 88.4%, and accuracy of 86.8%. In the VS, it had an AUC of 0.818, H-L p value of 0.259, sensitivity of 38.1%, specificity of 98.9%, PPV of 88.9%, NPV of 87.4%, and accuracy of 87.5%. The TRN performed well in both the internal and the external validation, indicating a good clinical application value. The TRN had a better predictive value of DVT than the Caprini score (p < 0.001). Conclusion: The incidence of DVT after LH was high, and should not be neglected in HCC patients. The TRN provides an efficacious method for DVT risk evaluation and individua-lized pharmacological thromboprophylaxis.

Automated summary

This study developed a DVT risk prediction model (TRN) based on data from a large number of hepatocellular carcinoma patients who underwent laparoscopic hepatectomy, demonstrating good predictive value in both internal and external validation.