To establish a clinical prediction model for the recurrence after resection of hepatocellular carcinoma (HCC) and investigate its predictive value.

One hundred and ninety-five patients with HCC who underwent hepatectomy in the Third Affiliated Hospital of Sun Yat-sen University from January 2007 to December 2013 were enrolled in this prospective study. Among them, 172 were males and 23 females, aged 13-81 years old with a median age of 49 years old. The informed consents of all patients were obtained and the local ethical committee approval was received. The patients were followed up after surgery, and the tumor recurrence was recorded. Clinical parameters of tumor recurrence after resection of HCC were screened by Cox's proportional hazards regression model. Clinical prediction model for the recurrence after resection of HCC was established based upon the clinical parameters. The predictive value of this model was assessed by receiver operating characteristic (ROC).

The postoperative overall recurrence rate was 44.1%. The 1-, 2-, 3- year recurrence rate was respectively 30.7%, 38.9% and 44.1%. ALT>41.5 U/L, Pringle's maneuver, liver cirrhosis, tumor diameter≥4.35 cm, tumor number≥3, portal vein tumor thrombus and low differentiation were the independent risk factors of recurrence after resection of HCC (HR=1.687,-, 2.078, 1.759, -, 3.879, -; P<0.05). Clinical prediction model: personal prognosis index (PI)= 0.52×ALT-0.9×surgical instrument (1)-0.442×surgical instrument (2)+0.7× liver cirrhosis+0.56×tumor diameter -1.21×tumor number (1)-1.029×tumor number (2)+1.35×portal vein invasion-0.6×differentiation (1)-0.853×differentiation (2). The risk of recurrence was higher as the value of PI increased. The area under ROC curve was 0.74, the specificity for predicting recurrence was 85.3% and the sensitivity was 52.3%.

ALT>41.5 U/L, Pringle's maneuver, liver cirrhosis, tumor diameter≥4.35 cm, tumor number≥3, portal vein tumor thrombus and low differentiation are the independent risk factors of recurrence after resection of HCC. The recurrence prediction model established in this study yields good predictive effects, and has important significance for predicting the recurrence after resection of HCC.