Tumor grade and sex should influence the utilization of portal lymphadenectomy for early stage intrahepatic cholangiocarcinoma

Abstract

Background:

Portal lymphadenectomy for intrahepatic cholangiocarcinoma (ICC) is encouraged for staging purposes, though it is under-utilized for clinically early-stage tumors. We sought to determine if any factor knowable prior to resection influences rates of portal lymph node metastases.

Methods:

The Surveillance, Epidemiology, and End Results (SEER) Program (1973–2014) database was queried to identify patients with T1/T2 ICC undergoing resection. Patients were stratified by lymph node (LN) status. Patients deemed LN negative required examination of six or more LNs (AJCC guidelines).

Results:

One-hundred and fifty-two patients were included in the analysis (LN negative: 38, LN positive: 114). Patients with LN negative cancers experienced prolonged overall survival as compared to patients with positive LNs (median 77 months vs 19 months, respectively p < 0.001). Twelve patients had well-differentiated tumors (G1), 92 patients had moderately-differentiated tumors (G2) and 58 patients had poorly-differentiated tumors (G3). Tumor grade (OR 3.9, CI 1.1–13.7, p = 0.031) and male sex (OR 2.6, CI 1.1–6.1, p = 0.022) were associated with positive LNs on multivariable logistic regression analysis.

Conclusion:

Intermediate/High grade and male sex are associated with high rates of lymph node metastasis for patients with early-stage ICC, which portends abbreviated overall survival.

Introduction

Intrahepatic cholangiocarcinoma (ICC) represents a subtype of bile duct adenocarcinoma involving the small ducts within the liver, and accounts for approximately 10% of this cancer in the US.^1,2 Like many other gastrointestinal tract cancers, complete surgical resection offers the potential for cure, although most patients succumb to recurrence following tumor extirpation.^2–4 In order to increase the efficacy of surgery in the management of this disease, adjuvant chemotherapy has been utilized albeit with mixed results. Much of the adjuvant therapy used is based on combination therapy using gemcitabine with either capecitabine or cisplatin which was found to impart a survival advantage in advanced, unresectable cholangiocarcinoma.^5,6 Recently, capectibine after resection was demonstrated to impart a survival advantage of 17 months over placebo (BILCAP trial, 2018 ASCO Abstract #4006). Additionally, there is a paucity of data regarding how best to stratify the use of adjuvant therapy in order to minimize toxicity for those unlikely to benefit, while treating those at highest risk for recurrence.

Lymph node (LN) metastases, typically established on final pathology, carry a poor prognosis and likely influences the use adjuvant therapies.^7,8 Unlike other gastrointestinal solid organ tumors such as colon cancer, LNs are usually not removed en bloc with the specimen and typically require a separate dissection in addition to the hepatectomy. The most common sites for lymph node metastases with ICC include the pericholedochal, periportal and common hepatic lymph nodes, which are excised with a portal lymphadenectomy. The extent to which these sites are involved appears to correlate with the T-stage of the tumor. For example, Kitagawa et al. demonstrated a 33% LN positivity rate assoaciated with T2 cancers.⁸ Given the reported frequency of LN metatstases, debate surrounding a survival advantage imparted with LN dissection,^2,8,9 and concern regarding the risk of bile duct ischemia, bile duct leak or increased intra-operative blood loss,²,¹⁰ we hypothesized that an adequate LN dissection is underutilized for early-stage ICC.

With the aim of ensuring more accurate staging and prognostication, the most recent American Joint Committee on Cancer (AJCC) guidelines recommend that at least six lymph nodes be excised at the time of surgical resection.¹¹ In this study we aimed to identify the rate of lymph node positivity among early-stage ICC, as well as preoperative factors associated with lymph node metastases. With the identification of high-risk subgroups within the early-stage ICC patient population, we hoped to bolster the recommendation for a complete lymphadenectomy in those found to be at risk for metastatic disease.

Methods

Initial cohort selection

A retrospective cohort study was conducted using the Surveillance, Epidemiology, and End Results (SEER) Program database (1973–2014). Patients with ICC were identified using the International Classification of Disease for Oncology 3rd edition (ICD-O-3) and selecting the primary site code specific to intrahepatic bile duct (22.1) as well as the histology code for cholangiocarcinoma (8160). Klatskin tumor code (8162) was used to exclude those with more advanced disease so as to pare down our cohort to include only those patients with early stage resectable ICC. Only patients with pathologically confirmed primary ICC having undergone cancer-directed surgery (codes 20–90) were included. Nonoperative patients were also excluded from our analysis.

Demographic data were collected on all patients included in our analysis. Patients younger than 18 years of age or patients with no evidence of malignancy (T0) on final pathology were excluded from the study. To confirm tumor stage for all study patients, extent of disease (EoD) and collaborative staging (CS) codes provided by SEER were used to derive overall stage as well as individual T, N, and M classifications for each patient based on the AJCC Staging Manual, 8th edition. Only clinically early-stage and pathologically confirmed early-stage (T1 and T2) tumors were included in our analysis (Fig. 1).

Figure 1.

Schematic diagram of study design

Defining lymph node status

The SEER registry employs EoD (1988–2003) and CS (2004–2013) codes to document LN involvement based on clinical and pathology records. Patients were stratified by LN status after histopathologic examination. Based on AJCC guidelines, adequate lymphadenectomy is defined as excision of six or more lymph nodes. As such, patients were only included in our analysis if they had positive LNs detected or if they underwent examination of six or more LNs without evidence nodal involvement (Fig. 1).

Defining histologic grade

For cholangiocarcinoma, histologic grade (HG) is divided into four categories, namely, G1: well differentiated, G2: moderately differentiated, G3: poorly differentiated, and G4: undifferentiated. We stratified tumor grade into either “low-grade” including all G1 tumors or “intermediate/high-grade” including G2 and G3 tumors. G4 tumors were excluded given the low number of patients in this subgroup.

Survival analysis

Vital status and median overall survival were examined for all patients. Those without survival data and SEER cause-specific mortality classification were excluded. The primary endpoint of our study was median overall survival (OS). Overall and disease-specific survival time was calculated using the date of death, last date known to be alive, or the dataset follow-up cutoff date of November 2017.

Statistical analysis

Standard demographic and clinicopathologic data including age, sex, race, diagnosis year, tumor grade, tumor size, AJCC 8th-edition TNM classification, LN positivity status, and information on cancer-directed surgery were collected and analyzed. Descriptive statistics were calculated for all variables of interest. The association between LN positivity and a set of identified factors of interest (age, race, sex, T stage, and grade) was determined using Fisher’s exact test, Mehta’s modification to Fisher’s exact test,¹² or a Cochran—Armitage test for trend.¹³ Following the univariable analyses, those factors which were at least potentially associated with LN positivity (p < 0.10) were evaluated for their joint association and predictive ability using a multivariable logistic regression model analysis. The results of this analysis are presented as odds ratios and their associated confidence interval, along with a test for whether the odds ratio = 1.0.

Overall survival (OS) was calculated from the date of diagnosis to the date of last follow-up or death. Based on the SEER-provided survival times, Kaplan—Meier analyses and log-rank tests were used to determine the association between potential predictors and survival. The trend in the association for grade with survival was determined using a trend test.¹⁴ For the instances in which a parameter is originally divided into 3 or 4 groups and then the results of that analysis led to splitting the data into two groups to best identify the prognostic significance, the resulting p-value was multiplied by 2 or 3 to adjust for the number of implicit tests done to arrive at the two groups identified as having the greatest impact. Following these univariable analyses, a Cox proportional hazards model based on backward selection was constructed to determine the joint association between the factors of potential interest and survival. Except as stated above, all p-values are two-tailed and presented without correction.

Results

Patients

One hundred and fifty-two patients met inclusion criteria, of whom 82% were Causian (n = 125) and 59% were female (n = 89). When divided on the basis of histologic grade (HG), the proportion of patients with G1, G2, and G3 lesions was 8% (n = 12), 60% (n = 92), and 32% (n = 48), respectively. Lymph node metastases were identified in 114 (75%) patients, while 25% (n = 38) of patients had no nodal disease identified following adequate lymphadenectomy (Table 1).

Table 1.

Patient demographics and clinical characteristics

Patient Characteristics (n = 152)	n = (%)
Sex
Male	63 (41)
Female	89 (59)
Race
White	125 (82)
Black	9 (6)
Other	18 (12)
Age
<50	29 (19)
50–59	37 (24)
60–69	45 (30)
70–79	34 (22)
>80	7 (5)
T Stage
T0	3 (2)
T1	73 (48)
T2	76 (50)
Grade
G1	12 (8)
G2	92 (60)
G3	48 (32)

Survival

The median OS for the 152 patients analyzed was 24.0 months. The presence of lymph node metastases (LNM), high G3, male sex, and age greater 60 years were associated with significantly lower survival according to Kaplan—Meier analysis. Median OS was 19 months for patients with positive LNM versus 77 months for those without LNM (p < 0.001; Fig. 2). G1, G2, and G3 were associated with median OS of 73 months, 35 months, and 17 months, respectively (p = 0.001 by trend test). Female patients had median OS of 34 months compared to 18 months for male patients (p = 0.055).

Figure 2.

Overall Survival stratified by lymph node status. Patients without lymph node metastases (defined as no tumor on histological examination in at least six lymph nodes) experience significantly improved overall survival as compared to patients with lymph node metastases

A Cox proportional hazards regression, comprised of univariant variables with p < 0.1 (Table 2), demonstrated that all four factors identified on univariable analysis (LNM positivity, high HG, older age, and male sex) were jointly associated with decreased overall survival with Hazard Ratios (HR) of 2.62 (p = 0.002), 1.86 (p = 0.002), 1.75 (p = 0.010), and 1.71 (p = 0.015) respectively. Lymph node positivity was also found to be an independent predictor of OS by a likelihood ratio test (p = 0.001).

Table 2.

Multivariable cox survival analysis of univariable factors with p < 0.10

	Hazard Ratio (95% CI)	p value
Lymph Node Positivity	2.619 (1.446–4.743)	0.002
Sex	1.707 (1.111–2.622)	0.015
Age	1.753 (1.146–2.681)	0.01
Tumor Grade	1.861 (1.259–2.751)	0.002

Lymph node positivity

On univariable analysis, male sex 2.43 time more likely to have LNM compared to female patients (95%CI: 1.08–5.47). Intermediate/High tumor grade (G2-G3) was also associated with LNM, but this difference did not reach statistical significance on univariable analysis (p = 0.074) (Table 3).

Table 3.

Univariable and multivariable analysis of potentially knowable preoperative factors for lymph node positivity

	Univariable		Multivariable
	Odds Ratio	95% Confidence Interval	Odds Ratio	95% Confidence Interval
Non-White Race	1.33	0.53–3.35	–	–
Male Sex	2.43	1.08–5.47	2.64	1.15–6.09
High Grade	3.38	1.02–11.19	3.93	1.13–13.66
T Stage	1.48	0.75–2.92	–	–
Age	1.10	0.72–1.43	–	–

Given that sex and histologic grade demonstrated an association (p < 0.10) with LNM on univariable analysis, these factors were incorporated into a multiple logistic regression model (Table 3). Both male sex and intermediate/high tumor grade were significantly associated with LNM with odds ratios (OR) of 2.64 (p = 0.022) and 3.93 (p = 0.031), respectively. When the model was converted into a classification rule to determine whether a combination of sex and grade status could be used to predict LNM, this classification rule was able to correctly identify 79% (30/38) of patients without LNM, but only 43% (49/114) of patients with LNM.

Discussion

Intrahepatic cholangiocarcinoma (ICC) remains a devastating disease for which surgery offers the greatest chance for long-term survival. According to a meta-analysis by Mavros et al., OS for patients with ICC ranges from 18 to 33 months with a mere 21% of patients alive at 5 years.¹⁵ Prognosis is driven by several factors including patient age, tumor size, vascular invasion, tumor multiplicity, as well as lymph node metastases (LNM). For patients with confirmed LNM, OS ranged from 5 to 22 months with no survivors reported at 5 years.¹⁶ In our analysis, similar trends were observed. While median OS was 24 months for our cohort as a whole, the presence of LNM resulted in a median OS of 19 months compared to 77 months for those without LNM. According to the most recent AJCC guidelines, a lymph node dissection (LND) should include no fewer than 6 nodes when performed for cholangiocarcinoma. However, the use of LND varies widely with only a quarter of patients in the US undergoing the recommended six-node sampling,¹¹ perhaps secondary to the documentation of an association with increased post-operative complications.^15,17,18 None-the-less these disparate survival outcomes based upon LN status highlight the importance of performing a portal LN dissection for patients with early stage cholangiocarcinoma, which will be useful for both prognosis and stratification of patients for adjuvant therapy.

Currently, there is no consensus regarding the diagnostic evaluation of patients with suspected ICC. This results in a high degree of variability in terms of patient workup and assessment. While lesions with hallmark malignant features can be resected without further evaluation, smaller lesions are often biopsied for confirmation using tissue examination.¹⁹ In our analysis, we chose only to include patients with T1 and T2 tumors. These smaller, more confined lesions are of particular interest as they offer the greatest diagnostic ambiguity. Multiple approaches are available for the purposes of tissue diagnosis including percutaneous biopsy, endoscopic ultrasound, diagnostic laparoscopy, and endoscopic retrograde cholangiopancreatography. Though each of these presents advantages and challenges, all provide potentially valuable information that could guide treatment. For example, confirmation of cholangiocarcinoma should prompt operative intervention and, as our data supports, the identification of an intermediate or high tumor grade should prompt the surgeon to consider portal lymphadenectomy.

Examination of the cohort demographics revealed that a majority of patients meeting our inclusion criteria were white and female. Interestingly, female sex was associated with a lower rate of LNM and a median survival advantage of 16 months when compared to the male patients in our cohort. Previous studies have demonstrated that males are significantly more effected with ICC with an incidence of 9.8 per 100,000 compared to 6.9 per 100,000 in females.²⁰ While disparity exists in incidence, data supporting a more aggressive phenotype in males is controversial. In a study by Lang et al., male sex was significantly associated with worse overall survival, although only when uncontrolled for resection margin.²¹ Conversely, when male sex was reviewed in a large meta-analysis, no study was able to demonstrate a decreased overall survival.¹⁵ We were able to demonstrate that there is a moderate association of lymph node positivity with being male, which was in turn associated with decreased overall survival. Our strict definition of lymph node negatively sheds light on the importance of proper staging in male patients. Given the propensity for lymph node metastases in male patients with ICC, we recommend portal lymphadenectomy for adequate staging.

There are several limitations to the study that should be considered. In our analysis, only 8% of patients were found to have G1 tumors. While it has been demonstrated that the majority of ICC histologically is well to moderately differentiated,²² our sample size may be too small to accurately define this patient population. This analysis was done using the SEER database, which is retrospectively collected where errors in reporting and documentation are possible. Long term patient follow-up may have inaccuracies as patients may travel between SEER reporting facilities, potentially effecting survival analysis. Additionally, SEER offers limited information regarding adjuvant therapy, limiting the ability to control for different treatment modalities between each cohort. Next, SEER is unable to distinguish between those operations that were undertaken for curative intent versus those that were strictly palliative. This forces the assumption that all major resections were of curative intent in our analysis. Finally, SEER captures only 30% of cancer cases which can potentially skew the data, potentially over or underestimating the true incidence of LN positivity.

For patients with early-stage ICC, we found that intermediate and high-grade tumors and male sex were associated with a greater propensity for LNM, and those with high grade tumors had distinctly worse overall survival when compared to adequately staged patients with grade 1 and grade 2 tumors. Therefore LND should be considered as part of the resection for these patients. LND is underutilized for patients with intra-hepatic cholangiocarcinoma in general, and we suspect even more so for clinically early-stage tumors. Although there are limitations to SEER-based studies, we hope that our investigation will be useful to influence the adoption of guidelines recommending portal LND for patients with ICC. We believe the information will be valuable not only to the treating medical oncologist, but also for stratification of patients in future trials as new therapies come to trial. This study has no funding to disclose.