How to Predict Recurrence After Resection of Hepatocellular Carcinoma

Discussion

The current study evaluated the performance of three commonly used HCC recurrence risk scores in the population of patients with resected HCC. Our results demonstrate that the well-known HCC risk recurrence scores AFP-model, RETREAT, and CRS perform well at differentiating patients with short versus long time intervals to tumor recurrence. In identifying patients who will experience tumor recurrence as opposed to those who will not, the accuracy of all scores remained very low. Instead, surrogates of loco-regional tumor invasion such as T stage, number of lesions, and small and large vessel infiltration were associated with tumor recurrence after HCC resection.

Since the Milan criteria were first introduced in 1996, they have remained the most widely used criteria for selecting HCC patients for LT, as they stratify patients according to their risk of tumor recurrence (8). However, more recently, the Milan criteria have been criticized for being too stringent and inappropriately limiting access to LT (11). Several alternative risk prediction models have been established to stratify the risk of recurrence in HCC patients potentially eligible for LT (6, 9, 10, 16). Multiple of these studies succeeded in demonstrating that OS and recurrence-free survival after LT were not impaired if eligibility criteria were extended for larger tumor sizes (6, 16).

Since LT is the most favorable curative-intent treatment for HCC, most recurrence scores focus on OS and recurrence-free survival after LT. Scores, however, are heavily dependent on the ethnicity of the underlying cohorts since aspects of HCC tumor biology vary largely between different regions of the world, most considerably between Asia and Western countries (5). More recent scores such as the AFP-score and the RETREAT-score take into account additional variables such as serum AFP levels and microvascular invasion or size of the largest nodule (8, 9, 17). Current studies demonstrate that novel scoring systems such as the AFP-score even outperform the MILAN criteria in predicting OS and recurrence-free survival after LT (12, 17).

Primary resection is considered as an alternative potentially curative option if technically feasible and the liver remnant function is considered sufficient. However, as in other cohorts, the rate of advanced liver disease was high in our cohort, with high MELD scores in a large proportion of patients, and background fibrosis or cirrhosis present in 48.2% of patients. Balancing liver function and oncological radicality remains a challenge in cirrhotic HCC patients. Global 5-year recurrence free and OS rates are 56.2% and 35.2% after HCC resection (5). Long-term outcomes, recurrence rates and potential recurrence risk factors are largely understudied in patients undergoing primary resection for HCC. Only few studies have evaluated risk predictors of recurrence in resected HCC patients to date (11, 18).

In the current study, we aimed to assess the accuracy of recurrence risk scores derived from Western populations in predicting time to recurrence and the event of recurrence in HCC patients after resection. The AFP-score and the RETREAT-score were developed to predict recurrence in LT patients, while the CRS is dedicated to predicting recurrence after resection of an HCC. All scores performed well in differentiating between short- and long-term recurrence-free survival. However, none of the scores was able to predict whether or not tumor recurrence would occur. Baseline and histopathological parameters that determine long-term outcomes after LT are supposedly different from those for primary resection. However, the CRS dedicated to patients who underwent resection of their HCC performed poorly. Only 5.1% of the patients in our cohort were classified as high-risk for recurrence based on the CRS; however, 51.9% of the patients recurred. Therefore, this score did not withstand external validation in our cohort and criteria seem too strict to identify a large proportion of HCC resection patients at risk for recurrence.

The three scores incorporated different variables and it is unclear whether all of these variables are of equal importance in HCC patients with curative-intent resection. In order to find predictors of actual tumor recurrence after HCC resection, we screened for potential preoperative and histopathological variables in our patient cohort. Formerly established risk factors, such as HCV, HBV or underlying cirrhosis (19, 20) failed to impact current recurrence rates most likely because antiviral therapy has become widely available in the United States. Interestingly, parameters associated with an increased risk of tumor recurrence were all surrogate markers of loco-regional tumor spread: Advanced T stage, multiple concurrent lesions, and small and large vessel invasion. The fact that T stage, but not tumor size, was a predictor of tumor recurrence suggests that factors, such as vessel infiltration and multiple disease foci, as incorporated in the 8th AJCC staging system T stage classification, are probably more important in triggering tumor recurrence than tumor size alone. HCCs with these features of loco-regional spread may be less amenable to curative local resection, and would benefit from LT, especially since patients recurred intrahepatically more often than extrahepatically. Furthermore, we found a higher risk of intrahepatic as compared to extrahepatic recurrence in patients with advanced T stage 3-4 and multiple lesions. These patients at high risk of intrahepatic recurrence, may fare better with LT than resection to avoid early intrahepatic recurrence after resection. Notably, the location of the tumor recurrence was rarely at the resection margin (3.2%), but rather in the liver remnant. This finding was confirmed in another recent study by Tsilimigras et al. evaluating 756 patients who underwent HCC resection. The study found that intrahepatic recurrence occurred in more than 70% of patients, compared to less than 30% for extrahepatic recurrences (21).

Few other studies aimed to identify predictors of HCC tumor recurrence, specifically after primary resection. Ivanics et al. developed a dynamic recurrence model for different stages of primary recurrence and sequential recurrences after resection for HCC in 486 patients (18). The most important determinants of recurrence included tumor size, satellite lesions, and small vessel invasion. Small and large vessel invasion was also identified as an important recurrence predictor in other HCC resection cohorts (20, 22). Xu et al. identified both small and large vessel invasion as risk factors of intrahepatic recurrence after HCC resection in 734 patients from China. Further predictors were cirrhosis, multiple tumors, satellite nodules, and tumor size greater than 5 cm (22). Another study, however, found that large but not small vessel invasion is a predictor of disease-specific survival in a cohort of 103 patients and two large validation cohorts (23). Importantly, small vessel invasion is prone to interrater variability and large registry cohorts in particular may not be consistent in the evaluations of this parameter. At the Massachusetts General Hospital, histopathological parameters were reviewed by the same specialized liver pathologists over the years relying on standardized criteria. Further studies from high volume centers with consistent pathologic review may demonstrate that both small vessel invasion and large vessel invasion are predictors of recurrence after HCC resection.

Our study identified multiple predictors of recurrence after curative-intent resection for HCC. A novel scoring system could unfortunately not be identified due to the limitation in patient numbers. However, the predictors of tumor recurrence identified in our study can help to guide HCC therapy decision making. Among the predictors in our study, T stage, large vessel infiltration, and multiple tumor lesions can be evaluated preoperatively by radiological imaging. Contrast-enhanced multidetector computed tomography (MDCT) or magnetic resonance imaging (MRI) can identify local invasion, and recent studies are developing artificial intelligence-based image processing and machine-learning feature extraction in order to identify surgically relevant HCC characteristics (24, 25). With this preoperative information, we may be able to better guide patients to resection rather than transplantation. Identifying predictors of recurrence preoperatively may aid in clinical decision making for HCC patients. If predictors of HCC recurrence after resection are absent, primary resection may be curative. In patients with one or more predictors of tumor recurrence, we suggest vigilant post-operative follow-up for detection of early recurrence. Additional recurrence detection strategies, such as regular measurements of AFP, des-γ-carboxy-prothrombin (DCP) or novel liquid biopsy approaches with circulating tumor DNA should also be considered. Overall recurrence rates of HCC remain high, and multiple attempts to prevent recurrence such as adjuvant therapy have yielded conflicting results. While HBV-targeted therapy in LT patients has been proven to reduce post-LT recurrence rates, its role in curative-intent resection remains unclear (26). Adjuvant therapy such as sorafenib have yielded discouraging results in prospective settings, and no superior alternative substance is currently available (27, 28). Identifying effective adjuvant therapies remains one of the challenging goals in curative-intent resection of HCC.

Limitations of this study include its retrospective nature and that it is a cohort from a single institution with institutional selection biases and regional transplantation constraints. Due to sample size limitations, subgroup analyses of score performance in different HCC stages were not feasible. As a limitation of this retrospective study, information on pre-resection TARE or drug therapy was not available. Furthermore, the median tumor distance to the resection margin was relatively small, at 0.88 cm. This corresponds with the rate of background fibrosis or cirrhosis of the liver, which was 48.2%. Balancing radical oncological resection and preservation of liver function can be particularly challenging in patients with liver fibrosis or cirrhosis. However, smaller resection margin distance failed to translate into high local recurrence rates (1.7%).