Introduction
Combined hepatocellular cholangiocarcinoma (cHCC) is a rare subtype, accounting for only 0.4–14.2% of primary liver carcinoma (PLC) [
1‐
5]. Due to its rarity and a wide variety of pathological types, the research in cHCC has been tough and the clinicopathological characteristics and prognosis of cHCC still remain poorly understood.
It is well-acknowledged that the prognosis of cHCC is dismal, due to its high recurrence rate after hepatic resection [
2]. On the basis of previous studies [
1‐
5], most recurrences occur early after the hepatic resection for cHCC. Patients with cHCC suffered early recurrence (ER) rate of 57–75% and 6 months recurrence rate of about 40% or so [
4‐
9], higher than patients with HCC, similar to patients with ICC [
2,
10].
Nevertheless, most studies only concentrated on ER for patients with cHCC using a cut-off of 2 years, which consists with the cut-off for ER of hepatocellular carcinoma (HCC) and intrahepatic cholangiocarcinoma (ICC) [
4‐
8]. The ER rate and the relationship between ER and clinicopathological characteristics of cHCC have been researched in these studies [
4‐
8]. Given that patients with cHCC suffered earlier recurrence and a higher rate of recurrence than patients with HCC [
2,
3], it may not be appropriate to use a cut-off of 2 years to differentiate early and late recurrence for patients with cHCC or ICC [
11], aligning with patients with HCC.
In the field of other malignant tumors [
11‐
13], including HCC and ICC [
11,
13], researchers have noticed that patients who had VER(i.e., recurrence within 6 months after surgery) suffered a significant lower OS than those who did not have VER, and therefore brought forward the definition of VER and investigated the risk factors of VER. The research on VER would be of significance for patients with cHCC. As such, the tool we developed to predict VER in patients with cHCC after hepatic resection would help identify the patients at high risk for VER, therefore making a contribution to constructing individualized surveillance strategies following hepatic resection or recommending treatment postoperatively.
To the best of our knowledge, no study exists has investigated the prediction of VER of cHCC after surgical resection. Therefore, our study aims to illustrate the relationship between the clinicopathological characteristics and VER, develop, and validate a model predicting VER after hepatic resection for cHCC based on multi-institutional data sets. In addition, we have constructed an online calculator to promote the use of our model in clinical work (
https://chcrecurrence.shinyapps.io/myCHCVER2).
Discussion
cHCC is a subtype of primary liver carcinoma composed of HCC and ICC, with a high incidence of recurrence [
1‐
5]. Some researches display that patients with cHCC tend to have recurrence earlier than those with HCC, and similar to those with ICC [
2,
3,
9,
10,
24]. Several previous studies reported that more than half of patients with cHCC suffered a dismal prognosis of 2-year early recurrence [
4‐
8]. Similarly, 136 out of 221 patients (61.5%) experienced early recurrence in our study. Therefore, concentrating on ER with a cut-off of 2 years may not be appropriate to the recognition of the recurrence high-risk patients.
In the field of other malignant tumors, previous studies have brought forward the definition of VER and researched risk factors of VER [
11‐
13,
25]. Up to now, no one has proposed the definition of VER in cHCC. On the basis of our multiple center data, 43.4% of patients (96 out of 221) had recurrence within 6 months after the hepatic resection and 67.1% of those suffered the recurrence (96 out of 143) had recurrence within 6 months after the hepatic resection. Meanwhile, several previous studies proposed the definition of VER using a cut-off of 6 months in the research of HCC and ICC [
12,
13]. Given that cHCC consists of HCC and ICC, we deem it appropriate to determine the optimal cut-off of VER in cHCC as 6 months like most of the researches of HCC and ICC [
11‐
13]. In addition, several researches demonstrated that patients with PLC who had VER suffered a significant lower OS than those who did not have VER [
11,
13].
As such, this study concentrated on the recurrence happened within 6 months after the hepatic resection for patients with cHCC. We identified the independent risk factors of VER and constructed a prediction model integrating MiVI, MaVI, and CA19-9 > = 25 U/mL. Using the VER model, we stratified all patients into two groups with significantly discrete risk of VER (
P < 0.001). The high risk group consisting of 48.0% of all patients accounts for 71.9% of VER (Fig.
6A). The recognition of recurrence high-risk patients is of vital importance, because the clinician could suggest more vigorous surveillance strategy or appropriate anti-tumor strategies to them. To the best of our knowledge, our study is the first to define and predict VER in cHCC after hepatic resection. An easy-to-use website calculator and a nomogram were provided to facilitate the clinical use.
The clinicopathological features of cHCC resemble those of HCC and ICC [
3‐
5]. There is no existing predictive staging system that is commonly used for cHCC and applying the staging systems for HCC or ICC to cHCC may be problematic. From this perspective, we need a model to predict the recurrence or even VER for patients with cHCC. AJCC staging system is one of the traditional systems applied to the prediction of recurrence in primary liver carcinoma [
20], many previous studies have demonstrated the predictive value of AJCC staging [
26,
27]. Nevertheless, ROC curves displayed the superior accuracy of our VER model than AJCC 8th staging according to higher C-indexes (0.77 vs 0.64 in development cohort (
P = 0.012), 0.76 vs 0.64 in validation cohort (
P = 0.030)) (Fig.
3A, B). Coincidentally, DCA analyze showed that our VER model provided obvious superior net benefit than AJCC 8th staging (Fig.
4A, B). Therefore, our VER model is more practical and more powerful than AJCC 8th staging in clinical use. In addition, the calibration analyze of our VER model also displayed favorable outcome.
Due to the rarity of cHCC and the variety of definitions and pathological types of cHCC, the research on cHCC and its prognosis has been tough and scant. Previous researches demonstrated that the independent risk factors for the recurrence of cHCC include tumor recurrence, tumor size ,metastases, age, MiVI, MaVI, SN, regional organ invasion , elevated CA19–9, ALP, and CEA as well as GGT [
1‐
5]. A limited number of researches identified independent risk factors for the early recurrence of cHCC as follows: tumor size, tumor number, MiVI, MaVI, SN, lymph node metastasis, Midkine, DCP, CA19–9, and poor differentiation [
4‐
6,
28]. Additionally, some research suggest that because cHCC consists of mixed elements of both HCC and ICC, the risk factors for both HCC and ICC would be those for cHCC, which signifies the predictive value of the portion of HCC and ICC [
9,
29,
30]. Multi-variable logistic regression analysis identified MiVI, MaVI, and CA199 as independent factors for VER in our study, which is generally consistent with the previous reports.
According to the studies in HCC, MiVI, MaVI, and SN, the HCC related features, are related to the invasion behavior of HCC [
17‐
19,
31‐
33]. It is widely acknowledged that MiVI, MaVI, and SN are independent risk factors for intrahepatic metastasis, recurrence and survival in HCC [
17‐
19,
31‐
33]. In the studies in cHCC [
4,
28,
34,
35], some previous researches also demonstrated the relationship between MiVI, MaVI, SN, and RFS, which is consistent with the studies in HCC. However, this view still remains controversial because some researches did not identified MiVI, MaVI, and SN as independent risk factors in their analysis [
5,
6,
29,
36]. The cause of this controversy can be listed as follows: firstly MiVI, MaVI, and SN were proposed and deeply studied in HCC, actually the pathological features of MiVI, MaVI, and SN in cHCC have not yet been completely investigated and the definitions have not been formulated neither. Unlike tumor size, tumor number, and serum tumor markers (AFP, DCP, and CA19-9), which are readily available and quantifiable, the definitions of MiVI, MaVI, and SN are various or unmentioned in the researches of cHCC, and most researchers tend to define MiVI, MaVI, and SN in cHCC following the definitions in HCC. Secondly, the number of studies in the recurrence of cHCC is limited due to its own rarity and wide variety of pathological subtypes. Our research emphasized the predictive value of MiVI, MaVI for VER in cHCC, thus promoting further researches to determine the pathological features and definitions of MiVI and MaVI in cHCC.
Previous studies elucidated that the elevation of CA19-9 reflects extensive tumor burden, which signifies biologically aggressive behavior, poor differentiation and greater tumor volume of ICC, and correlates with bleak prognosis in ICC [
29,
37,
38]. Consistent with this conclusion, some researchers declared the elevation of CA19-9 to be an independent risk factor for RFS and OS in cHCC [
5,
29]. It is well known that CA19-9 is a biomarker for ICC components and the prognosis of ICC is worse than that of HCC [
2,
37,
38]. Ideally, more portion of ICC lead to worse prognosis for cHCC patients, i.e., ICC dominance (defined as ICC components more than 50% or 80% pathologically according to different studies) is an independent risk factor for recurrence and survival in cHCC. However, some researchers identified that no relationship was found between ICC dominance and the poor prognosis in cHCC [
5,
39]. Unfortunately, due to the lack of clinical data, we did not include this variable. More researches are needed to elucidate the relationship between ICC dominance and prognosis in cHCC.
Additionally, people developed VER tend to undergo major resection or lymph node dissection (Table
2), while logistics analysis showed no correlation between the performance of major resection or lymph node dissection and VER (Table
3). Consistent with our research in cHCC, some studies show that the performance of minor resection seems to improve the prognosis in patients with HCC, compared to that of major resection [
15,
40], because major resection may influence the postoperative liver function. But no statistically significant difference was found in both DFS and OS in patients underwent major or minor resection [
15,
40]. Meanwhile, some researchers recommend lymph node dissection for patients with cHCC, due to the high probability of lymph node metastasis. But no reliable research has been carried out to support the perspective [
29,
41]. Compared to HCC and ICC, management of cHCC is not yet standardized, and the choice of surgical procedure still remains controversial. More large-sample research were needed to standardize the treatment of cHCC.
There are three main limitations to our study. First, the clinical data were collected from two centers retrospectively and the sample size is limited, so the information bias, heterogeneity in patients’ characteristics, and different surgical levels between two centers should be taken into consideration. Previous studies identified some variables, AFP, DCP, ALP, tumor size, and tumor number, as independent risk factors for the recurrence of cHCC. Nevertheless, these variables of which the P value is small in univariable or multi-variable analysis were identified uncorrelated with VER of cHCC in other research. We need more multi-institutional, large sample-sized, and prospective studies to verify the conclusion. Second, we only enrolled patients from China. Nearly 85% of the enrolled patients had been infected by HBV while few of them had HCV infection or NAFLD. We also excluded those who did not received curative hepatic resection or prior anti-tumor treatments. So, the generalizability of the conclusion is limited. Third, due to the lack of clinical data, we did not include the survival data or the variables which were missing over 10%, such as HCC- or ICC-dominance.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.