Tumor Size, Not Small Vessel Invasion, Predicts Survival in Patients With Hepatocellular Carcinoma

Abstract

Objectives

The 8th edition American Joint Committee on Cancer (AJCC) staging system for hepatocellular carcinoma (HCC) has been criticized for failing to stratify patients. We aimed to reassess and modify the tumor staging criteria for HCC.

Methods

Three independent study cohorts were collected and analyzed.

Results

The initial cohort consists of 103 patients with HCC. By Kaplan-Meier survival analysis, the 8th edition failed to distinguish between T1b and T2. Only tumor size and large vessel invasion, but not small vessel invasion or other histopathologic parameters, predicted HCC survival. We modified the T staging criteria by eliminating small vessel invasion while emphasizing tumor size in the middle categories (T2 and T3), which achieved more even distribution of cases and significantly improved risk stratifications (P < .001). This modification was then validated in a cohort of 250 consecutive patients from Mount Sinai Hospital and an online Surveillance, Epidemiology, and End Results data set comprising 9,685 patients, which showed similar results. Small vessel invasion was not an independent prognostic factor in either validation cohort.

Conclusions

Our study showed that tumor size, but not small vessel invasion, predicts survival in patients with HCC. We suggest incorporating our modified T staging criteria in future AJCC revisions.

Introduction

Hepatocellular carcinoma (HCC) is the most common primary malignancy in the liver, the fifth most common malignancy worldwide, and the third leading cause of cancer-related deaths globally.^1,2 HCC can be attributed to many different risk factors, such as viral hepatitis, metabolic disorders, or toxins.^3-5 Current treatment choices include tumor resection, liver transplantation, radiofrequency ablation, and transcatheter arterial chemoembolization.^6,7 Optimal management remains controversial owing to the heterogeneity of HCC and underlying liver diseases.^2,8 It is essential to provide accurate pathologic analysis and tumor staging for patients with HCC, which should correlate directly with risk stratification, prognosis prediction, proper clinical management, and follow-up. Several staging and scoring systems can predict the clinical outcomes of HCC,⁹ among which the American Joint Committee on Cancer (AJCC) TNM staging system is widely accepted and most frequently used.^10-12

The 8th edition of the AJCC Cancer Staging Manual is the current version and has been in use for approximately 3 years.¹³ Compared with the 7th edition, the 8th edition includes substantial changes for HCC regarding the T criteria, although the N criteria remain the same. Briefly, T1 is subdivided into T1a and T1b based on a tumor size cutoff of 2 cm, where T1a includes solitary tumors 2 cm in size or smaller regardless of vascular invasion, while both T1b and T2 include solitary tumors greater than 2 cm, but T2 is distinguished from T1b by the presence of vascular invasion. In the 7th edition, T3 was subdivided into T3a (multiple tumors, at least 1 greater than 5 cm in size) and T3b (tumors involving a major branch of the portal vein or hepatic vein). The 8th edition does not subdivide T3: the previous T3a is now T3, and the previous T3b is upstaged to T4. Those changes were incorporated into the 8th edition based on a few studies performed in Asia, which has the largest population of patients with HCC because of the high prevalence rate of hepatitis B virus infection.¹²

Since its publication, this newest AJCC staging manual was criticized by several studies in Asia and Europe as well as studies using the online Surveillance, Epidemiology, and End Results (SEER) database, which suggested that the 8th edition performed similarly to or not better than the 7th edition.^14,15 Besides anatomic criteria, many researchers explored the possibilities of other histologic or biological parameters for prognostic significance and staging potential. For example, some found that macrotrabecular HCC, glypican-3 expression, and certain HCC subtypes may be associated with better or worse disease outcomes.^16-18 A recent study identified differentially expressed stage-specific genes in HCC that may enhance our understanding of the molecular determinants of HCC progression and serve as biomarkers that potentially underpin diagnosis and therapeutic targets.¹⁹

The aims of this study were fourfold: first, to comprehensively characterize the clinicopathologic and immunohistochemical features of HCC in an initial study cohort at our institution (University of Rochester Medical Center [URMC]); second, to assess the 8th edition AJCC T staging criteria compared with the 7th edition and evaluate the prognostic value of other histopathologic parameters on HCC survival in the initial study cohort; third, based on the results of the first and second aims, to propose modifications of the T staging criteria with improved prognostic functions, offering suggestions for future revisions; and fourth, to validate the proposed modifications in an independent study cohort from Mount Sinai Hospital (MSH) in New York and a large SEER database queried between 2010 and 2015. To achieve the best result in the initial study cohort, we strictly selected cases by slide review, careful chart review, and matched tissue microarray analysis for each case.

MATERIALS AND METHODS

Initial Study Cohort at URMC

We conducted a retrospective review of the URMC pathology database for all liver resection and total hepatectomy specimens carrying a final diagnosis of “hepatocellular carcinoma” between 2009 and 2019. Specimens with a diagnosis of other malignancies besides HCC, including any focus of concurrent intrahepatic cholangiocarcinoma or combined HCC-intrahepatic cholangiocarcinoma, were excluded from the study. Cases with biopsy diagnosis of HCC but no viable tumor cells on resection specimens were excluded. Cases without slides for review or with insufficient clinical data in patient medical records were also excluded.

Histomorphologic Evaluation

Data on demographic and clinicopathologic features, including age, sex, pertinent medical history, presurgical ablation/chemoembolization, type of surgical procedure, and survival status, were obtained through electronic health records. Follow-up (measured in months) was defined as the time from initial diagnosis (mostly from biopsies) to death or last clinical examination. If no death occurred, only those patients with at least 2 months of follow-up data were included in this study. The pathologic stage was reassessed based on both the 7th and 8th editions of the AJCC Cancer Staging Manual.^13,20 Any disagreement with original pathology reports was resolved by slide review on a multiheaded microscope until consensus was achieved.

Two liver pathologists independently reviewed all H&E-stained sections for diagnosis confirmation. The following features were recorded based on recommendations from the most recent World Health Organization Classification of Tumours, Digestive System Tumours²¹: (1) histologic patterns, including trabecular, pseudoacinar, and solid; (2) tumor grade (ie, well, moderately, or poorly differentiated based on cytoplasm and nuclear features); (3) specific subtypes, including steatohepatitic, clear cell, chromophobe, fibrolamellar, lymphocyte-rich, and scirrhous—specifically, the steatohepatitic variant is defined as more than 50% tumor nodule showing at least 3 “steatohepatitis-like” features (ie, steatosis, hepatocyte ballooning, Mallory-Denk bodies, inflammation, or pericellular fibrosis)²²; and (4) the presence of vascular invasion. Small vessel invasion is defined as microscopic intratumoral or peritumoral vascular invasion, while large vessel invasion is defined as a tumor in large portal vein branches, which is often macroscopically identified as tumor thrombi. Because of the critical role of vascular invasion in tumor staging, any suspicion for vascular invasion, if not documented in the original pathology report, was carefully evaluated and resolved by double scope slide review until consensus was reached.

Histologic features of hepatocellular carcinoma (HCC). A, HCC with small vessel invasion (arrows; H&E, ×100). This case also had large portal vein branch invasion. B, C, Small vessel invasion in 2 HCC cases that were not documented in the original pathology reports (arrows; H&E, ×100). D, Macrotrabecular HCC (H&E, ×100). E, Steatohepatitic HCC showing steatosis, inflammation, and fibrosis (H&E, ×200). F, Steatohepatitic HCC showing abundant ballooning tumor cells with Mallory-Denk bodies (H&E, ×200).

Tissue Microarray Construction and Immunohistochemistry

RESULTS

Validation Study Cohorts

Continuous variables (age, tumor size) were compared using 1-way analysis of variance. Categorical variables, such as sex, type of procedure, treatment modalities, stage groups (pT), IHC results, and patient outcomes, were compared using χ ² tests. Kaplan-Meier survival analysis was used to test for differences in overall survival (OS) and disease-specific survival (DSS) during the follow-up period, and log-rank tests were employed for statistical significance. The statistical analysis was carried out using SPSS statistical software, version 27 (IBM); Prism software, version 8.3.0 (GraphPad Software); or R software, version 4.0.5 (R Foundation for Statistical Computing). P values less than .05 were considered statistically significant.

Specimens from 103 patients with a diagnosis of HCC who fulfilled the study criteria included 69 (67.0%) total hepatectomy (transplant) and 34 (33.0%) resections Table 1 (initial cohort). The patients, predominantly men (n = 83 [80.6%]), had a median (range) age of 61 (35-83) years. Presurgical tumor ablation or chemoembolization was performed in 61 (59.2%) patients, among whom 45 (43.7%) underwent transplant while 16 (15.5%) underwent resection. Twenty-four (23.3%) patients had transplant without presurgical ablation/chemoembolization. Notably, a total of 16 (15.5%) cases had a reduced number of tumor foci because of presurgical treatment, among whom 6 had solitary viable tumor left leading to lowered tumor stage in 3 cases. Twenty-nine (28.2%) cases had concurrent lymph node dissection for histologic examination, and no lymph node metastases were identified (pN0) in any of those cases. Two (1.9%) cases had distant metastasis at the time of hepatectomy (M1; 1 to adrenal gland, the other to the diaphragm). On follow-up, 31 (30.1%) patients died at a median (range) 56 (0-137) months. Among them, 23 (22.3%) died of disease, while 8 (7.8%) died of other causes (eg, cardiac arrest).

Tumor size and focality were measured and recorded based on both macroscopic and microscopic examinations. Of 103 cases, 56 (54.4%) were solitary tumors, with a size range of 0.6 to 21 cm, among which 16 (15.5%) were 2 cm or less in size, 24 (23.3%) were greater than 2 cm to 5 cm or less in size, and 16 (15.5%) were greater than 5 cm in size. Twenty-one (20.4%) patients had 2 tumor foci, with a size range of 0.3 to 15 cm, and 26 (25.2%) patients had 3 or more tumor foci ranging from 0.5 to 15 cm in size. Small vessel invasion was identified in 31 (30.1%) cases, including 8 cases not documented in the original pathology report Figures 1A-1C. Large portal vein branch invasions were present in 4 (3.9%) cases, all documented in the original pathology report and confirmed by slide review.

Histologic and IHC features are summarized in Table 2. The background liver disease was recorded, combing both clinical history and histologic evaluation of the liver parenchyma away from the tumor. The majority of background livers were cirrhotic (n = 74 [71.8%]). Viral hepatitis comprised more than half of the underlying liver disease (n = 55 [53.4%]), predominantly hepatitis C virus (HCV) (n = 0 [48.6%]), followed by steatohepatitis (alcoholic and nonalcoholic; n = 27 [26.2%]). Histologic evaluation of the HCC showed that more than half of the tumors (n = 61 [59.2%]) were moderately differentiated. A large proportion of the tumors demonstrated mixed (≥2) growth patterns (n = 62 [60.2%]). Macrotrabecular growth pattern as a major (>50%) or minor (5%-50%) component was seen in 9 (8.7%) cases Figure 1D. Morphologically, most cases were conventional HCC (n = 70 [68.0%]), while 21 (20.4%) cases were subtyped as a steatohepatitic variant based on proposed diagnostic criteria Figures 1E-1F. Not surprisingly, the steatohepatitic variant of HCC was more frequently seen in patients with background steatohepatitis (12/27 [44.4%]) than HCV (6/50 [12%]; P < .005) or other etiologies (3/26 [12%]; P < .005). Other variants, such as clear cell, chromophobe, scirrhous, and lymphocyte rich, were rarely (each <5%) identified.

Immunohistochemically, 72 (69.9%) cases showed positive glypican-3 positivity with the following staining intensity: 1+ (n = 18 [17.5%]); 2+ (n = 29 [28.2%]); 3+ (n = 25 [24.3%]) Figure 2. Arginase 1 and Hep Par-1 were equally sensitive, detected in 98 (95.1%) cases. All tumors were positive for at least 1 of the above 3 IHC markers. CK7 was positive in 21 (20.4%) cases, while CK19 was only focally positive in 1 (1.0%) case. Overexpression of p53 (n = 4 [3.9%]) was rare, as with nuclear β-catenin positivity (n = 3 [2.9%]). CDX2 and SATB2 were negative in all 103 tumors tested on TMA.

The AJCC staging system for HCC incorporated tumor size, focality, and presence of small and large vessel invasion as staging criteria Table 3. Based on the AJCC 7th edition, the 103 cases were classified into T1 (n = 34), T2 (n = 62), T3a (n = 3), T3b (n = 3), and T4 (n = 1). Based on the AJCC 8th edition, they were classified into T1a (n = 16), T1b (n = 23), T2 (n = 57), T3 (n = 3), and T4 (n = 4). To evaluate the prognostic significance of the 8th edition AJCC TNM staging and compare it with the 7th edition, we performed Kaplan-Meier survival analysis for both OS and DSS. As shown in Figure 3, the 7th edition T staging criteria failed to separate T1 from T2 on both OS and DSS; the 8th edition T staging criteria separated T1a from T2 but failed to distinguish T1b from T2. After combining T1b and T2 into 1 stage, the overall risk was well stratified (P < .05; Supplemental Figure 1 [all supplemental materials can be found at American Journal of Clinical Pathology online]), but this led to an overwhelming number of cases in the T2 category (n = 80 [77.7%]).

Univariate and multivariate analysis revealed that the largest tumor size (regardless of the number of tumor foci) was a significant risk factor for DSS, especially in cases of solitary tumors (Supplemental Table 1, Supplemental Figure 2). Only large vessel invasion involving the portal vein branch, not microscopic small vessel invasion, affected DSS. Interestingly, patients who underwent transplant had much better outcomes than patients who underwent resection (P < .05), largely because of the significantly smaller tumor size in those patients who had undergone transplant (mean [range] 2.5 [0.4-12] cm) than in those who had undergone resection (mean [range] 6.3 [1.2-21] cm; P < .001). By multivariate analysis, transplant was not an independent prognostic factor. Similarly, patient age (≤65 vs >65 years), sex (female vs male), presurgical tumor ablation/chemoembolization, background liver disease (cirrhotic vs noncirrhotic, viral hepatitis vs steatohepatitis), tumor differentiation, growth patterns, HCC subtypes (steatohepatitic vs nonsteatohepatitic), and glypican-3 expression had no prognostic significance for OS or DSS (P > .05; Supplemental Table 1).

Statistical Analysis

Given the prognostic significance of tumor size and large vessel invasion and the fact that patients with multiple tumor foci had a similar prognosis as those with solitary tumors between greater than 2 cm and 5 cm or less (Supplemental Figure 2), we propose a new T staging system that keeps the 8th edition stage T1a as T1 (solitary, ≤2 cm) and T4 as T4, respectively, but modifying T2 as “solitary tumor >2 cm and ≤5 cm, or multiple tumors, none >5 cm,” and T3 as “any tumor focality, at least one >5 cm.” Using these new staging criteria, the initial cohort had not only a significantly increased number of cases in the T3 category (17 vs 3) but also much better stratification for each category on both OS and DSS Table 3 and Figures 4A-4B (P < .001).

Clinicopathologic Analysis of the Initial Study Cohort

The first validation cohort came from MSH and consisted of 250 consecutive specimens (103 transplants and 147 resections) Table 1. Although age, sex, lymph node status, and distant metastasis were similar to the initial cohort, a much higher rate of small and large vessel invasions was reported in this cohort (67.2% vs 34%; P < .001). Further analysis showed that the detection of vascular invasion correlated significantly with tumor size so that cases with no vascular invasion were significantly smaller (mean [range], 2.6 0.3-18.0] cm) than those with small vessel invasion (mean [range], 3.9 [0.6-21.0] cm) or large vessel invasion (mean [range], 7.9 [2.2-16.5] cm; P < .001). Thirty patients died of disease at a median follow-up of 145 months. By multivariate analysis, only tumor size and large vessel invasion, not small vessel invasion, were independent prognostic factors (Supplemental Table 2). The distribution of cases based on the AJCC 8th edition stage was as follows: T1a = 56 (22.4%), T1b = 28 (11.2%), T2 = 132 (52.8%), T3 = 11 (4.4%), and T4 = 23 (9.2%). Based on our modified staging criteria, the distribution was T1 = 56 (22.4%), T2 = 129 (51.6%), T3 = 42 (16.8%), and T4 = 23 (9.2%). Similar to the initial study cohort, the 8th edition staging failed to separate T1b from T2 in this cohort Figure 4C; using the modified staging criteria, the stages were successfully stratified Figure 4D (P < .0001).

The second validation cohort was collected from the SEER database and consisted of 9,685 patients with pathology-proven HCC and adequate data for tumor staging Table 1. Although some specific data points were missing, given the AJCC 7th edition staging information and data on tumor size, focality, and vascular invasion, we were able to derive the AJCC 8th edition staging and modified staging information for all cases Table 3. Similar to the initial study cohort and the MSH cohort, the 8th edition staging failed to adequately separate T1b from T2 and T3 from T4 in this cohort Figure 4E, but using the modified staging criteria, all stages were successfully stratified Figure 4F(P < .0001). Interestingly, only tumor size and large vessel invasion, not small vessel invasion, were significant prognostic factors in this data set (Supplemental Figure 3).

The AJCC tumor staging manual on HCC has been criticized for its unnecessary complexity and failure to adequately stratify patients.^24,25 The 8th edition HCC T staging system has 2 major flaws: First, it underestimates solitary tumor with large tumor size so that a patient with a large solitary tumor (>5 cm) will never be classified with T3 disease. Second, it assumes that vascular invasion has prognostic power in tumors larger than 2 cm but not in tumors 2 cm or smaller in size, so that a solitary 2.0-cm tumor with small vessel invasion is T1a, but a solitary 2.1-cm tumor with small vessel invasion will bypass T1b to become T2. This skip is difficult to justify and in fact has led to reversed risk stratification in our cohorts and other published studies.²⁶ In fact, the prognostic significance of small vessel invasion in HCC has long been controversial because of its interobserver and intraobserver variability.^27,28 Although some authors insisted on including it for staging,^{14,24,26,29,30} their studies were frequently based on online SEER database information^14,26 or data collected from multicentric patients^14,30 and so lacked consensus or consistency not only for HCC diagnosis but also in documenting vascular invasion, potentially leading to inconsistent tumor staging.²⁸ In addition, many of those studies were performed in Asian countries,^29-31 where the underlying liver diseases, treatment modalities, and patient care could have confounded the clinical course and disease outcomes.

Discussion

Other Histomorphology and IHC Characteristics in the Initial Cohort

In our initial study, we found that the anatomic parameters, such as tumor size and focality, are straightforward and easy to document with accuracy, but the report of intratumoral small vessel invasion is subjective because of marked peritumoral fibrosis and tumor retraction artifact, which is sometimes difficult to ascertain. Indeed, we identified 8 cases in which vascular invasion was not originally documented in the pathology report. Nonetheless, either keeping the original reported number of cases with vascular invasion or adding the 8 newly identified cases with vascular invasion, this parameter was not an independent prognostic factor on patient survival, echoing the same findings from another study performed in the United States.³⁰ Interestingly, the validation cohort from MSH showed a much higher detection rate of small vessel invasion, which appeared to be a prognostic factor in univariate analysis but in fact was the result of its significant correlation with tumor size. Given the significant intra- and interinstitutional variability and its high correlation with tumor size, it is reasonable to remove small vessel invasion for staging purposes. Large portal vein branch invasion, however, is often easily identified macroscopically, recognized histologically, and associated with poor prognosis, as expected.

Patient Outcomes and Survival Analysis in the Initial Cohort

Through our strict study design and detailed analysis, we showed that tumor size is a powerful predictive factor for patient survival, especially in cases of solitary tumor without large vessel invasion. The importance of tumor size was reflected in the fact that in our initial cohort, patients who underwent transplant had significantly better outcomes, largely because of the smaller tumor size in those patients, and that in the MSH validation cohort, the rate of small vessel invasion correlated significantly with tumor size, suggesting that it is truly the single most important factor for prognosis. It is reasonable to emphasize tumor size for T staging, but we took into consideration tumor multifocality, including satellitosis, so that those cases will be at least T2 depending on the largest tumor size and presence of large vessel invasion. Not surprisingly, our modified system not only ensured more even distribution of case numbers in the middle categories (T2 and T3) for both initial and validation cohorts but also achieved significantly improved risk stratification between each T category. Compared with the AJCC 8th edition staging and other proposed modifications,²⁶ our modified staging system is much easier to apply in practice.

Besides anatomic parameters, we evaluated other clinical and histopathologic factors for prognostic significance in our initial study cohort because some previous studies have indicated that certain histologic features, including coexisting hepatitis status, hepatic functional reserve, subtypes of HCC, or glypican-3 expression, may be associated with worse prognosis. Through our analysis, we did not find any other factors that had a significant prognostic effect on HCC survival, signifying the importance of keeping these histopathologic parameters for diagnostic purposes but not for staging criteria. Our TMA data also confirmed that SATB2 and CDX2 were not expressed in HCC, and only 1 case demonstrated focal CK19 positivity, suggesting that CK19-positive HCC was extremely rare in our patient population. Likewise, p53 mutation and β-catenin activation were rare events in HCC, rendering those factors neither diagnostic nor prognostic.

For both the AJCC 7th and 8th editions, the staging groups equal the T category if no lymph node or distant metastasis is identified.^13,20 We did not specifically evaluate the staging groups in our study, but given the low rate of regional lymph node and distant metastasis at the time of surgery in the initial and validation cohorts, we expect that the prognostic significance of the correspondingly modified staging groups (stage 1 = T1N0M0, stage 2 = T2N0M0, stage 3A = T3N0M0, stage 3B = T4N0M0, stage 4 = any T N1 and/or M1) would be similar to the modified T criteria, which can be reliably incorporated into future AJCC revisions. The low regional lymph node metastasis rate for HCC at both institutions proves that HCC predominantly spreads through an intrahepatic blood vessel rather than a lymphatic space.

Development of New T Staging Criteria in the Initial Cohort and Validation Cohorts

In summary, we report that tumor size, not small vessel invasion, is a powerful prognostic factor for HCC. We modified the AJCC 8th edition T1, T2, and T3 staging criteria by eliminating small vessel invasion while keeping the large vessel invasion in the highest tumor stage (T4) category. This modification is easy to apply and has achieved superior results in both our initial and validation cohorts. We propose incorporating this modified system into future AJCC revisions. Our study also confirmed previous observations that many histomorphologic parameters are important for recognizing and diagnosing HCC but not for prognosis. Future studies may expand the case numbers to validate our modified T stage criteria and identify other biomarkers with potential diagnostic or prognostic associations.

The authors thank the histology laboratory at University of Rochester Medical Center Department of Pathology for technical support.

Acknowledgments

References