Prognostic Evaluation and Survival Prediction for Combined Hepatocellular-Cholangiocarcinoma Following Hepatectomy
Article information
Abstract
Purpose
This study aimed to assess prognostic factors associated with combined hepatocellular-cholangiocarcinoma (cHCC-CCA) and to predict 5-year survival based on these factors.
Materials and Methods
Patients who underwent definitive hepatectomy from 2006 to 2022 at a single institution was retrospectively analyzed. Inclusion criteria involved a pathologically confirmed diagnosis of cHCC-CCA.
Results
A total of 80 patients with diagnosed cHCC-CCA were included in the analysis. The median progression-free survival was 15.6 months, while distant metastasis-free survival (DMFS), hepatic progression-free survival, and overall survival (OS) were 50.8, 21.5, and 85.1 months, respectively. In 52 cases of recurrence, intrahepatic recurrence was the most common initial recurrence (34/52), with distant metastasis in 17 cases. Factors associated with poor DMFS included tumor necrosis, lymphovascular invasion (LVI), perineural invasion, and histologic compact type. Postoperative carbohydrate antigen 19-9, tumor necrosis, LVI, and close/positive margin were associated with poor OS. LVI emerged as a key factor affecting both DMFS and OS, with a 5-year OS of 93.3% for patients without LVI compared to 35.8% with LVI. Based on these factors, a nomogram predicting 3-year and 5-year DMFS and OS was developed, demonstrating high concordance with actual survival in the cohort (Harrell C-index 0.809 for OS, 0.801 for DMFS, respectively).
Conclusion
The prognosis of cHCC-CCA is notably poor when combined with LVI. Given the significant impact of adverse features, accurate outcome prediction is crucial. Moreover, consideration of adjuvant therapy may be warranted for patients exhibiting poor survival and increased risk of local recurrence or distant metastasis.
Introduction
The 2019 World Health Organization has defined combined hepatocellular-cholangiocarcinoma (cHCC-CCA) as a unique subtype of primary hepatic malignancy characterized by the coexistence of hepatocytic and cholangiocytic differentiation [1,2]. While the cellular origin of cHCC-CCA remains incompletely understood, hepatic progenitor cells, capable of differentiation into both hepatocytes and cholangiocytes, are recognized as pivotal in the development of cHCC-CCA [1].
Constituting a rare variant among primary hepatic malignancies, cHCC-CCA accounts for approximately 0.4%-14.2% of cases [3]. Given its rarity, the prognosis and optimal treatment strategies for cHCC-CCA have not been firmly established. Clinicians often face the challenge of extrapolating treatment approaches from more prevalent hepatic malignancies, such as hepatocellular carcinoma (HCC) or cholangiocarcinoma (CCA), where therapeutic paradigms are more well-defined [2]. However, the distinctive nature of cHCC-CCA, encompassing biological characteristics and clinical behavior of both HCC and CCA, necessitates a focused investigation into its prognostic factors and survival predictors.
Currently, no recommended adjuvant therapies exist for HCC following hepatectomy [4]. This absence of standard recommendations may be attributed to the characteristic tendency of HCC to recur primarily within the liver, rather than manifesting widespread metastasis. Given the liver-centric recurrence pattern of HCC, multiple local treatment options are available for salvage, and therefore, post-hepatectomy management typically involves observation with imaging. In contrast, intrahepatic CCA often displays a proclivity for distant metastasis, leading to the frequent recommendation of adjuvant chemotherapy for this subtype [5]. Notably, cHCC-CCA exhibits features resembling both HCC and CCA, thereby complicating the selection of appropriate adjuvant treatment options. As patients with cHCC-CCA experiencing poor survival and a heightened likelihood of distant metastasis may stand to benefit from adjuvant therapies, an essential prerequisite is the accurate evaluation and prediction of survival outcomes specific to cHCC-CCA.
In light of these considerations, the primary objective of this study is to systematically assess prognostic factors associated with the treatment outcomes of cHCC-CCA. Through this investigation, we aim to develop predictive models for survival based on these identified factors.
Materials and Methods
1. Patients
We have gathered a cohort of patients who underwent definitive hepatectomy at a single institution from 2006 to 2022. In surgical pathology, all individuals were diagnosed with cHCC-CCA. Patients who underwent liver transplantation as the definitive treatment for cHCC-CCA were deliberately excluded from the study. Furthermore, exclusions were applied to cases involving solitary HCC or CCA, palliative hepatectomy, and instances lacking comprehensive pathologic reports.
2. Clinicopathologic factors
We systematically gathered clinical information encompassing key variables such as age, sex, the presence of liver cirrhosis, and baseline liver function. In addition, preoperative and postoperative serum markers, including α-fetoprotein (AFP), protein induced by vitamin K absence or antagonist-II (PIVKA-II), and carbohydrate antigen 19-9 (CA19-9), were meticulously documented. Concerning preoperative serum markers, values obtained within the 3 months preceding the surgical procedure were included in the dataset. In instances where multiple values were available, the values nearest to the surgery date were utilized for analysis. Similarly, postoperative values were acquired within 3 months after surgery with values with nearest surgery date for analysis.
The pathologic findings were detailed by our pathologists and encompassed critical parameters such as tumor size, multiplicity of tumors, presence of tumor necrosis, tumor composition, lymphovascular invasion, perineural invasion, and histologic type. In the categorization of the tumor composition, we specifically classified it as either CCA cell dominant or HCC dominant, with dominance defined as the presence of a specific cell type comprising more than 50% of the tumor. The histologic type was reported based on patterns observed, including trabecular, compact (or solid), and mixed patterns. Notably, we further categorized the histologic type into two groups: compact pattern and non-compact pattern. The latter represents cases where no compact pattern (e.g., trabecular type or mixed type such as trabecular and acinar) is observed within the pathology. The width of the resection margin and tumor was collected and classified into two categories: (1) positive margin or close margin (≤ 2 mm) and (2) negative margin (> 2 mm).
3. Statistical analysis
In the survival analysis, the baseline for follow-up was defined as the date of surgery. Survival duration was defined as the time interval from the date of surgery to specific events, including progression, distant metastasis, intrahepatic progression, and death. The site of first recurrence was documented, with mixed recurrence indicating simultaneous recurrence within a 1-month timeframe. Recurrence sites were categorized as liver, regional, and distant metastasis. Regional recurrence referred to the occurrence of recurrence in lymph nodes recognized as regional lymph nodes in intrahepatic CCA.
For survival comparisons in graphical representations, log-rank analysis was employed. With the aim of assessing distant metastasis and identifying potential cohorts for adjuvant treatment, our focus was directed towards evaluating distant metastasis-free survival (DMFS) and overall survival (OS) in the context of factors associated with cHCC-CCA. The Cox proportional hazard model was utilized for both univariate and multivariate analyses of factors associated with survival. In the multivariate analysis, factors exhibiting a p-value less than 0.15 in the univariate analysis were included. All statistical analyses were conducted using the R project ver. 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria). A significance threshold of p < 0.05 was adopted to determine statistical significance. The nomogram developed in this study aims to predict 3-year and 5-year OS and DMFS, which incorporates the four factors that demonstrated the most significant impact on survival outcomes after multivariate analysis. Then, calibration curves for 5-year OS and DMFS were plotted, using ‘rms’ package in R.
Results
A total of 80 patients were included in the analysis. The mean age of the entire cohort was 59.3 years, with a predominance of males (72.5%). The majority of patients had underlying liver cirrhosis, all were categorized as Child-Pugh A, and only 13.8% had a grade 2 albumin-bilirubin index (ALBI). Approximately half of the patients (23 out of 47) exhibited elevated preoperative CA19-9 levels (> 37), while 26.4% (14 out of 53) continued to have elevated postoperative CA19-9 levels. Similar trends were observed for both PIVKA and AFP, with a reduction in the proportion of patients with abnormal values after surgery. S1 Table summarizes the serum marker of the patients. In terms of pathologic findings, the median tumor size was 3.1 cm (range, 1.1 to 15.0 cm). Tumor necrosis was present in more than half of the surgical specimens. Lymphovascular invasion was noted in 63.7% of patients, while perineural invasion were observed in 12.5% patients. Table 1 provides a summary of the clinicopathological characteristics of patients with cHCC-CCA.
Clinicopathological characteristics of patients with combined hepatocellular-cholangiocarcinoma
Recurrence following hepatectomy was observed in 52 cases (65.0%). Approximately more than half (34/52) of the patients experienced intrahepatic recurrence as the initial site of recurrence. Significantly, distant metastasis emerged prominently as a recurrent site, with a total of 17 out of 80 cases (21.3%) experiencing it as the initial site of recurrence. Additionally, regional recurrences, defined as the recurrence of regional lymph nodes akin to intrahepatic CCA, were noted in a total of 16 cases (20.0%). Table 2 provides a summary of the first recurrence sites among patients with cHCC-CCA.
Summary of site of first recurrence in patients who received definitive surgery for combined hepatocellular-cholangiocarcinoma
In the overall patient population, the median follow-up was 30.7 months (range, 1.4 to 129.7 months). The median progression-free survival (PFS) was 15.6 months, with estimated 3-year and 5-year PFS rates of 38.8% and 28.8%, respectively. The median survival for DMFS was 50.8 months, and the estimated 3-year and 5-year DMFS rates were 54.2% and 48.4%, respectively. Hepatic progression-free survival (H-PFS) closely mirrored the PFS, with a median of 21.5 months and estimated 3-year and 5-year H-PFS rates of 46.2% and 32.3%, respectively. Finally, the median OS was 85.1 months, and the estimated 3-year and 5-year OS rates were 66.1% and 56.4%, respectively. Fig. 1 describes the PFS, DMFS, H-PFS, and OS for all cHCC-CCA patients.
Progression-free survival (A), distant metastasis-free survival (B), hepatic progression-free survival (C), and overall survival (D) in patients with combined hepatocellular-cholangiocarcinoma.
In the univariate analysis exploring factors influencing survival, tumor size, tumor necrosis, lymphovascular invasion, perineural invasion and histologic type emerged as predominant factors for DMFS. Tumor size, tumor necrosis, lymphovascular invasion, perineural invasion and histologic type were identified as factors affecting OS in univariate analysis. Subsequently, factors with a p-value less than 0.15 in the univariate analysis were included in the multivariate analysis model. Tumor necrosis, lymphovascular invasion, perineural invasion and histologic type retained statistical significance in the multivariate analysis model for DMFS. Key factors influencing OS in the multivariate model encompassed postoperative CA19-9 levels, tumor necrosis, lymphovascular invasion and resection margin status. Remarkably, lymphovascular invasion was notably observed as the most significant factor impacting both DMFS and OS. Table 3 provides a summary of the results from the univariate and multivariate analysis for both DMFS and OS. Fig. 2 depicts survival curves for DMFS and OS, comparing factors such as tumor necrosis, lymphovascular invasion, histologic type, and postoperative CA19-9 levels. Significantly, lymphovascular invasion had a pronounced impact on OS, with a 5-year OS of 93.3% in cases without lymhpovascular invasion compared to 35.8% in cases with lymphovascular invasion.
Multivariate Cox regression analysis for factors with p-value less than 0.1 in DMFS and OS
Distant metastasis-free survival of patients according to tumor necrosis (A), lymphovascular invasion (LVI) (B), histologic type (C), and postop carbohydrate antigen 19-9 (CA19-9) level (D). Overall survival of patients according to tumor necrosis (E), lymphovascular invasion (F), histologic type (G), and postop CA19-9 level (H). N/A, not available.
Utilizing the factors identified through multivariate analysis, we constructed a nomogram to predict 3-year and 5-year DMFS and OS. Fig. 3A and B illustrate the developed nomogram for prediction of 5-year DMFS and OS. Additionally, calibration curves for the nomogram were plotted to assess the accuracy of survival predictions in comparison to actual survival outcomes (Fig. 3C and D). As anticipated, there was a strong correlation between actual survival and predicted survival, as evidenced by well-aligned calibration curves and ideal lines. The Harrell’s C-index, calculated as 0.809 for OS and 0.801 for DMFS, further confirmed the high correlation between the nomogram predictions and actual survival outcomes.
Nomogram for prediction of distant metastasis-free survival (A) and overall survival (B) for combined hepatocellular-cholangiocarcinoma, calibration curve for nomogram of 5-year distant metastasis-free survival (C) and 5-year overall survival (D). CA19-9, carbohydrate antigen 19-9; LVI, lymphovascular invasion; N/A, not available; PI, perineural invasion.
Eight patients (10%) within this cohort underwent chemotherapy within 3 months after surgery. Among them, two patients experienced immediate recurrences post-surgery, subsequently receiving chemotherapy as palliative care. Of the seven patients who eventually developed distant metastasis, five succumbed to the disease. Remarkably, only one patient maintained a disease-free status since the hepatectomy. Chemotherapeutic agents varied among patients; two were administered immunotherapeutic agents as for HCC, while six patients received either gemcitabine or 5-fluorouracil, similar to the approach for intrahepatic CCA.
Among the 26 patients experiencing intrahepatic recurrences alone, the majority (24 patients) underwent salvage treatment with either transarterial chemoembolization (TACE) or radiofrequency ablation, while two patients received sorafenib. These interventions proved effective, as evidenced by a median OS of 27.2 months (range, 0.7 to 111.2 months) after initial recurrence, with 12 patients still alive at the last follow-up (range, 11.6 to 111.2 months). Conversely, among the 26 patients who experienced recurrence elsewhere besides the liver, prognosis was poor. Although 18 patients received salvage or palliative chemotherapy, the majority (21 patients) ultimately succumbed to the disease, with a median survival of 11.2 months after initial recurrence (range, 1.9 to 42.4 months).
Additionally, we have identified local failure, which was defined as recurrence within neighboring hepatic lobes from the initial tumor site. Among the 80 patients included in the study, 29 experienced local failure. The cumulative incidence of local recurrence was evaluated, with rates of 24.4% at 1 year and 30.0% at 2 years (S2A Fig.). For factors related with local failure, tumor size (≥ 5 cm) and lymphovascular invasion was found to be related with an increased risk of local failure while other factors such as postoperative CA19-9, tumor necrosis, lymphovascular invasion, perineural invasion, resection margin status and histologic type did not have impact on local failure (S3 Table). Positive or close resection margin status and multifocal tumors were found to be associated with increased local failure after multivariate analysis (S2B and S2C Fig.).
Discussion
In this study, we observed a notable incidence of regional and distant metastasis as the initial recurrence. While the OS for the entire cohort of cHCC-CCA remained relatively high, lymphovascular invasion emerged as a crucial factor influencing overall survival. Specifically, the 5-year OS was 93.3% in cases without lymhpovascular invasion, contrasting with 35.8% in cases with lymhpovascular invasion. Additional factors associated with DMFS and OS included postoperative CA19-9 levels, tumor size, tumor necrosis, perineural invasion, and histologic type. Leveraging these insights, we developed a comprehensive nomogram for predicting OS and DMFS, demonstrating a high Harrell’s C-index.
cHCC-CCA is characterized by features of both HCC and CCA. Traditionally, the prognosis of cHCC-CCA is considered to lie between that of HCC and CCA [6-9]. However, our study revealed an extended survival in our cohort, with a median of 85.1 months and an expected 5-year OS of 56.4%, which can be compared to survival of HCC. This favorable outcome may be attributed to three main factors. Firstly, the optimal selection of surgical candidates resulted in patients with relatively good performance status and liver function, as evidenced by all patients having Child-Pugh Class A. Additionally, salvage treatments for liver recurrences, such as TACE, transarterial radioembolization, and radiotherapy, might contribute to the improved survival observed. While the median PFS and hepatic PFS were relatively modest at 15.6 and 21.5 months, respectively, patients with intrahepatic recurrences were effectively managed with multimodal therapeutic approaches. The majority of patients with intrahepatic-only recurrences were successfully treated with local therapies. This is evidenced by the median survival of 27.2 months observed in this group even after experiencing intrahepatic recurrence. Lastly, there was a substantial difference in survival between patients with and without lymphovascular invasion. The expected 5-year OS declined significantly, dropping from 93.3% to 35.8%. It is plausible that varying proportions of lymphovascular invasion in different studies could contribute to the prolonged survival observed in this particular study. This hypothesis is supported by the findings in our cohort, where solitary liver recurrence without regional or distant metastasis was not found to be associated with OS (data not submitted, p=0.93).
Lymphovascular invasion emerged as the most significant factor influencing both OS and DMFS in this study. According to the developed nomogram, lymphovascular invasion emerged as the most crucial factor influencing both OS and DMFS. Notably, vascular invasion is a well-established prognostic factor in both HCC and CCA [10,11]. Its importance is reflected in the American Joint Committee on Cancer (AJCC) staging for HCC and intrahepatic CCA, where the presence of vascular invasion leads to upstaging from T1 to T2 for solitary tumors [12]. For intrahepatic CCA, multiple studies indicate lymphovascular invasion as a poor prognosticator [13,14]. Given that cHCC-CCA embodies mixed features of both HCC and CCA, the prominence of lymphovascular invasion as a key factor in tumor outcomes is unsurprising. Previous studies have consistently identified vascular invasion as a one of determinant for prognosis in cHCC-CCA [6,15]. In our cohort, we observed that only three patients without lymphatic invasion exhibited positive vascular invasion, suggesting the importance of identifying any presence of lymphovascular invasion. Additional pathological factors, including tumor necrosis, perineural invasion, resection margin status and histologic type, were also found to significantly impact OS and DMFS in this study. Consequently, a comprehensive exploration of pathological findings is crucial for accurately predicting tumor outcomes. Future studies should incorporate these detailed pathologic findings to enhance understanding and prognostication of cHCC-CCA.
In the context of adjuvant radiotherapy, we focused on the evaluation of local recurrence. Local recurrence was defined as recurrence in neighboring lobes of the resection margin, which can potentially be covered by radiotherapy. Currently, there are no reported studies regarding adjuvant radiotherapy for cHCC-CCA. However, extrapolating from findings in CCA, adjuvant chemoradiotherapy could be considered for patients with microscopic margins (R1) or positive regional nodes [16-19]. Similar to CCA, our study found that margin status and tumor multiplicity were related to local recurrences of cHCC-CCA, suggesting the potential benefit of adjuvant radiotherapy in these cases. Therefore, given the results from studies of CCA and our current findings, patients with high-risk features such as close or positive resection margin status could be considered for adjuvant radiotherapy.
The potential role of adjuvant chemotherapy may be crucial for patients with high risk of distant metastasis and poor survival outcomes. However, survival analysis based on adjuvant treatment was not conducted in this study, primarily due to potential selection bias and the limited number of patients who received adjuvant treatment. While survival outcomes from adjuvant treatment remained notably poor, the key distinction between patients with and without adjuvant treatment was a selection bias toward adverse pathologic findings. Moreover, our study revealed that half of the patients experienced regional lymph nodes or distant metastasis as the initial site of recurrence after surgery. Given that hepatic recurrences can be managed with various locoregional treatments such as TACE, radiofrequency ablation, and radiotherapy, we emphasize the importance of addressing regional and distant metastasis concerns. Considering the absence of existing randomized trials regarding adjuvant treatment, we aimed to identify patient groups that could potentially benefit from adjuvant therapy. Patients with adverse findings, especially lymphovascular invasion, appear to be strong candidates for adjuvant treatment. Future studies should explore the role of adjuvant treatment in these specific groups. Furthermore, the establishment of well-defined chemotherapeutic agents for combined cHCC-CCA is lacking. Studies on unresectable cHCC-CCA have highlighted the potential role of gemcitabine-based chemotherapy, drawing parallels from established chemotherapy regimens for CCA [20-22].
This study is subject to several limitations. Primarily, being a single retrospective study, it is susceptible to potential selection bias. The scarcity of the disease entity contributes to a limited patient population, resulting in incomplete data, particularly regarding postoperative CA19-9 levels, introducing potential biases in survival analysis. Additionally, the study did not explore possible adjuvant treatment options for the poor prognosis group, as number of patients were limited to compare treatment outcome by with or without adjuvant treatment. Lastly, the nomogram developed has not undergone external validation, raising concerns about overfitting within our cohort. Despite these limitations, we believe that we have successfully identified a patient group with an extremely poor prognosis and developed a nomogram for predicting the group that may benefit from adjuvant treatment.
In conclusion, the prognosis of cHCC-CCA is notably poor when combined with lymphovascular invasion. Given the significant impact of adverse features on cHCC-CCA outcomes, accurate outcome prediction is crucial. Moreover, consideration of adjuvant therapy may be warranted for patients exhibiting poor survival and increased risk of distant metastasis.
Electronic Supplementary Material
Supplementary materials are available at Cancer Research and Treatment website (https://www.e-crt.org).
Notes
Ethical Statement
This study received approval from the Institutional Review Board of Seoul National University Hospital (No: H-2309-138-1470), and informed consent was waived, given the retrospective nature of the analysis.
Author Contributions
Conceived and designed the analysis: Chun SJ, Kim KS.
Collected the data: Chun SJ, Jung YJ, Lee KB.
Contributed data or analysis tools: Chun SJ, Jung YJ, Choi Y, Yi NJ, Lee KW, Lee KB, Kang HC, Chie EK, Kim KS.
Performed the analysis: Chun SJ, Jung YJ.
Wrote the paper: Chun SJ, Jung YJ, Choi Y, Yi NJ, Lee KW, Suh KS, Lee KB, Kang HC, Chie EK, Kim KS.
Conflict of Interest
Conflict of interest relevant to this article was not reported.
Acknowledgements
This work was supported by the National Research Foundation of Korea grant funded by the Korean Government (2022R1A2C-1092928).