Recurrence of HCC after LT remains a major obstacle and associated with an unfavorable prognosis [20]. Several independent risk factors for post-LT HCC recurrence have been identified including number and size of tumors on preoperative imaging studies [11, 12, 21] and pre-LT serum levels of AFP at different cut-off points: 10,200,1000 ng/mL [22,23,24,25]. However, lack of agreement about accurate, reliable and robust validated tool especially for prediction of early recurrence of HCC in HBV predominant population, make appropriate risk stratification and doctor-patient communication challenging.
In the present study, 786 patients with HCC diagnosed by imaging who underwent deceased donor LT from centers distributed throughout the whole China were involved and the 5–8 score was developed and validated. Our predictive model is a simple and reliable tool that showed excellent stratification of HCC patients into three risk subgroups; low-, medium- and high-risk with predicted 2-year recurrence ranging from 5% in low-risk to 20.0% in the medium risk and 51% in the high-risk category. For the 253 patients (52.1%) in the low-risk group, the 2- year OS was 93%, significantly superior to the 83 and 57% of the medium- and high-risk groups.
Similar to our study, previous risk models [16, 26,27,28] have attempted to employ pre-LT AFP, largest tumor diameter and the number of nodules to predict the post-LT HCC recurrence, but the characteristics of their study population (western participants with predominantly HCV) was different. Our model is mainly based on the data of patients with HCC occurring in patients with HBV infection (96%) which is the most common cause of HCC in China. The high predictive accuracy of our preoperative model [Harrell’s C was 79%(95% CI: 0.73–0.86) and 74%(95% CI:0.66–0.82), respectively, in the training and validation cohorts] stems from the utilization of three preoperative factors with more accurate cut-off points that were strongly associated with HCC recurrence: pre-LT AFP(10–200, 201–1000, and > 1000 ng/mL), largest tumor diameter(4–6,6.1–8,> 8 cm), and the number of nodules (single vs multiple). Unlike previous studies, our model end-point was recurrence at 2-year which is the discriminative cut-off value for early and late recurrence of HCC, however, the precise cut-off may require genetic/genomic analyses recurrence. Early recurrence results from preexisting tumor cells while late recurrence or de novo tumor mainly arising due to new malignant clones [29,30,31]. Using the cut-off values in our study was a clinical decision based on the expert’s assessment and they achieved the lowest AIC values so better fit the model.
Furthermore, one of the main advantage of our score over previous mentioned models is its ability to discriminate three subgroups of HCC recurrence(i.e low,medium,high), while other models only classify HCC patients into two risk subgroups of HCC recurrence, high and low risk. The categorization defining only two subgroups at risk is not valuable, or at least less practical. For example, tumour recurrence risk divided into low (< 8%) and high (> 50%) will produces a big ‘grey zone’ through 8–50% of medium-risk individuals [32]. Our score solved this drawback by stratify the individuals at risk into three subgroups of risk of recurrence, low,medium and high. This can be translated clinically into an excellence of individuals selections for LT, a more reasonable organs allocation taking into account the donor offer, besides, and the opportunity to stratify risk for the development of upcoming adjuvant treatments [33].
One of the most interesting results comes when we compared our model with the conventional selection criteria by which our model can stratify patients within and outside the Milan and Hangzhou criteria. For instance, the subgroup of patients (25%) who were recognized as high risk of recurrence by Milan (out Milan), they carry a low risk according to our score with 2-year recurrence probability of 12%. While the subgroup of patients(4%) who were recognized as low risk of recurrence by Milan (in Milan), they carry a high risk according to our score with 2-year recurrence probability of 24%. Similarly, the subgroup of patients (12%) who were recognized as low risk of recurrence by Hangzhou (in Hangzhou), they carry a high risk according to our score with 2-year recurrence probability of 25%. The addition of our model to the conventional selection criteria may, therefore, allow us to capture accurately the patients with a higher risk of recurrence who were traditionally considered as the lowest risk group. External validation of this preoperative adjuvant model is mandatory to help in more accurate selection of HCC patients for LT and to avoid the high post-LT recurrence probabilities.
Moreover, one of the advantages of our model is when it was compared to the AFP model, about (11 and 32%) of patients recognized as high risk of recurrence by AFP model, but they had a low and medium risk according to our score with 2-year recurrence probability of 10, 21% respectively. Also, about 23% of the patients were defined as a low risk of recurrence by the AFP model, but a medium risk of recurrence was revealed by our score with 2-year recurrence probability of 20%. Another advantage of our model is the outcome of the competing risk analysis, the rates of HCC-unrelated death were similar (p = 0.120) and the rates of HCC-related death which mainly due to HCC recurrence were significantly different (p < 0.001), indicating that the 5–8 model is a powerful tool to pick up the HCC recurrence but not other causes of deaths and this explains the differences in survival rates according to the 5–8 model.
At this point, we need a further explanation for the clinical application of our model. The patients presented preoperatively with a single nodule, diameter of ≤4 cm and serum AFP of ≤10 ng/mL will fall in the low-risk group by getting the 5–8 score of 0 and this will effectively exclude the probability of post-LT HCC recurrence, so these patients can go directly for LT. However,The patients with largest nodule diameter of > 8 cm and serum AFP of > 1000 ng/mL will be categorized in the high-risk category (Table 3), so LT should be excluded or neoadjuvant therapy and close surveillance until the tumor biology and morphology could be brought down to a safer level. However, patients who belong to the medium HCC recurrence risk category (e.g. single nodule, diameter of 8 cm and serum AFP of 700 ng/mL), a careful selection for LT based on a personalized assessment and pre-LT downstaging therapy should be considered then early administration of mTOR inhibitor postoperatively are strongly recommended [34].
Our study has some limitations. First, its retrospective nature so we designed a prospective study to confirm the clinical utility of our model. Second, the characteristics of our study population (predominantly HBV infected male Chinese patients) so external validation in HCV predominant, non-Chinese populations is required to confirm the reliability of the 5–8 model. Third, the patients with vascular invasion were not excluded from our study. Fourth,some parameters, such as microvascular invasion and data regarding response to pre-LT neoadjuvant therapies are incomplete in the CLTR database and this precludes us from comparing our model with other prominent risk scores. Lastly, although the follow-up time was enough to pick up the events, it is relatively short in comparison with other studies.