We retrospectively evaluated 382 eCCA patients who underwent surgical resection at the National Cancer Center Hospital between January 2002 and March 2022. All the patients included in this study underwent macroscopic curative resection of conventional eCCAs or IPNBs that developed during EHBD. We excluded patients who had received any therapy before surgery and those with inadequate IES data. In addition, cases of hospital death after surgical resection, cases with unknown causes of death, or early death not due to eCCAs within 12 months after surgical resection were excluded. Finally, 305 patients were included in this study. For survival analyses, cases of carcinoma in situ of conventional eCCA were excluded. Supplementary Figure 1 describes the details of patient selection.

Surgical procedures were performed based on the location of the primary tumor. Among 305 patients, 140 (45.9%) underwent hepatectomy with extrahepatic bile duct resection, 124 (40.7%) underwent pancreaticoduodenectomy, 17 (5.6%) underwent combined hepatectomy and pancreaticoduodenectomy, and 24 (7.9%) underwent extrahepatic bile duct resection. Para-aortic lymph node sampling was performed when lymph node metastasis was suspected. In our hospital, adjuvant therapy is not routinely performed after surgery, although adjuvant S-1 therapy has become a standard of care according to the results of the JCOG1202 study since October 2021 (21). Only four patients underwent adjuvant chemotherapy with S-1 during the study period. Clinical and radiological follow-ups were scheduled on a 3-month basis for a few years after resection. The median follow-up period was 41.0 months for all 305 patients. Recurrence was confirmed by radiological examination and elevation of tumor markers. The date of recurrence was defined as the date on which clinicians confirmed recurrence in medical records. The census date was December 31, 2022. According to previous studies, early recurrence is defined as any recurrence within 12 months after surgery (22, 23).

To assess the incidence and extent of IES in BTCs, 102 cases of AVC, 174 cases of GBC, and 20 cases of CyDC were included, all of which were surgically resected at the National Cancer Center Hospital between January 2002 and March 2022. Supplementary Table 1 shows the demographics of the patients with GBC, AVC, and CyDC. AVCs raised in the common, bile, or pancreatic ducts of the ampulla of Vater were selected for this assessment.

All of the BTCs were examined pathologically and classified according to the WHO classification (2, 24, 25) and the International Union against Cancer (UICC) TNM classification 8th edition (6). We had some modification about tumor location as mentioned later. Macroscopic types of eCCA and the following histopathological variables were evaluated following the Japanese Society of Biliary Surgery (JSBS) classification (26): lymphatic, venous, and perineural invasions that were classified into negative (–), slightly positive (1+), moderately positive (2+), and markedly positive (3+) based on their event frequencies. For the survival and correlation analyses, “high” and “low” grades were determined based on these values; high grade to be combined with 2+ and 3+ and low grade to be - and 1+. According to the JSBS classification (27), the right and left hepatic ducts and their confluence were defined as the (peri)hilar duct (Bph), common hepatic duct, and common bile duct, and were divided into three portions as follows: superior (Bs), middle (Bm), and inferior (Bi) portions of the EHBD. Bs and Bm were defined as the respective portions in the upper and lower halves of the bile duct length from the confluence of the right and left hepatic ducts to the upper margin of the pancreas, and Bi was defined as the portion from the upper margin of the pancreas to the ampulla of Vater. In some analyses, we combined Bph and Bs eCCAs as perihilar eCCAs, and Bm and Bi eCCAs as distal eCCAs. IHBD was defined as the hepatic side of the bile duct from the third branch (e.g., segmental ducts 5 and 8) of the hepatic duct in this study. All patients with stage IV disease were diagnosed on the basis of para-aortic lymph node involvement. Surgically resected specimens were fixed in 10% formalin and cut into serial slices 5 mm thick. All sections were stained with hematoxylin and eosin for pathological examination. IES was defined as the intraepithelial spread of cancer cells beyond the invasive area. IES contained lesions corresponded to biliary intraepithelial neoplasia, high grade (BilIN-3), intraductal papillary neoplasm of the bile duct (IPNB), intracholecystic papillary neoplasm (ICPN), intraampullary papillary tubular neoplasm (IAPN), and cancerous duct. IES was diagnosed only when the cancer cells extended along the biliary tract. IES was not applied when cancer cells had stromal invasion beyond the biliary tract structure without extension along the biliary tract and the cancer cells re-entered the biliary tract mucosal epithelial layer. We defined that “intraepithelial extension of cancer cells on duodenum” was present when intraepithelial extension of cancer cells on ampullary common duct continued to extend to the duodenum epithelial layer. The length of the IES was described using a 5 mm scale in general.

Statistical analyses were performed using JMP version 12.2 (SAS Institute, Cary, NC, USA) and the StatView-J software version 5.0 (Abacus Concepts, Berkeley, CA, USA). Continuous data were expressed as median (range) and compared using the Mann–Whitney U test. Categorical variables were compared between groups using Pearson’s chi-square test or Fisher’s exact test, as appropriate. Relapse-free survival (RFS) was defined as the interval between the date of surgery and time of recurrence. Overall survival (OS) was calculated based on the time from surgery to death from any cause or last follow-up. Survival data were estimated using the Kaplan–Meier method and examined using the log-rank test. Factors found to be significant in the univariate analysis were subjected to multivariate analysis using the Cox proportional hazards model. Differences at P < 0.05 were considered statistically significant.

Details of the surgical and clinicopathological features are presented in Table 1 . Among 284 patients with invasive eCCAs, those with Bph eCCAs were significantly younger than those with other conventional eCCAs. The female ratio in Bph eCCAs was significantly higher than that in Bs and Bi eCCAs, and similar tendencies were found in Bm eCCAs and invasive IPNBs. The total tumor sizes of Bph eCCAs were significantly smaller than those of Bs and Bm eCCAs, whereas the sizes of the tumor area with stromal invasion of cancer cells (invasive tumor sizes) of Bi eCCAs and invasive IPNBs were significantly smaller than those of the other conventional eCCAs.

Approximately 80% of conventional invasive eCCAs belong to the nodular-infiltrating macroscopic type. In contrast, more than 85% of invasive IPNB cases were papillary types. The incidence of poorly differentiated adenocarcinomas in Bph eCCAs was significantly lower than that in other conventional eCCAs. Invasive IPNBs predominantly include papillary adenocarcinomas. The invasion depth tended to be lower in invasive IPNBs than in conventional eCCAs. The frequencies of portal vein or artery invasion of cancer cells were much higher in Bph and Bs eCCAs, as the portal vein or artery exists nearer to the EHBD in these regions compared to other sites. The lymphatic, vascular, and perineural invasion of tumor cells tended to be lower in invasive IPNBs than in conventional eCCAs. Venous invasion was significantly higher in Bph eCCAs than in other conventional eCCAs and invasive IPNBs. Perineural invasion was significantly lower in Bi eCCAs than in Bs and Bm eCCAs. The rates of residual tumor-free tumors (R0) were significantly higher in Bi eCCAs than in other conventional eCCAs. The rates of positive bile duct margins with invasive cancers and positive residual tumor status with invasive cancers were significantly higher in Bph and Bs eCCAs than in Bi eCCA and invasive IPNBs.

Patients with noninvasive IPNBs had significantly better survival than those with invasive eCCAs, including invasive IPNBs ( Supplementary Figure 2 ). The survival outcomes of invasive IPNBs were significantly better than those of Bm or Bph eCCAs, but not significantly different from those of Bs or Bi eCCAs. The survival curve of Bi eCCAs was similar to that of invasive IPNBs for both RFS and OS, and the 5-year and 10-year survival rates were similar. The survival outcomes of patients with Bi eCCAs were significantly better than those of patients with Bph and Bm eCCAs in terms of both RFS and OS.

IES was observed in 60.2% of invasive eCCAs ( Figure 1 ). The incidence of IES in Bph or Bs eCCAs was significantly lower than in Bm or Bi eCCAs or invasive IPNBs ( Figure 1D ). Table 2A summarizes the extent and distribution of IES in invasive eCCAs. Invasive IPNBs had the highest incidence (P <0.001) and longer duration of IES than invasive conventional eCCAs. The incidence peak of IES was at lengths of ≥10 and <20 mm in conventional eCCAs and at lengths of ≥20 and <30 mm in invasive IPNBs. The IES extended further to the liver-side in invasive eCCAs.

Patients with invasive eCCAs and IES showed significantly longer survival times than those without IES (P = 0.039) ( Figure 2A ). Similar survival associations were found in patients with invasive conventional eCCAs, but without statistical significance ( Figure 2B ). IES was not associated with outcomes in perihilar eCCAs ( Figure 2C ), although patients with IES showed significantly longer OS in distal eCCAs (P = 0.030) ( Figure 2D ). When survival was evaluated at each location in invasive conventional eCCAs and invasive IPNBs, significant differences in RFS (P = 0.039) and OS (P = 0.025) were observed in Bm eCCAs ( Supplementary Figure 3 ). Multivariate analyses of patients with distal eCCAs ( Table 3 ) revealed that the IES was not a significant predictor of RFS or OS. In distal eCCAs, the presence of IES significantly correlated with a lower female-to-male ratio, larger total tumor size, higher positive bile duct margins, and a higher positive residual tumor status ( Supplementary Table 2 ).

The IES on the EHBD reached the junction of the CyD and often extended beyond the junction and continuously into both the CyD and the other side of the EHBD. When invasive eCCAs in which cancer cells reached the junction of the CyD were selected and assessed, the incidence of IES along with CyD (CyD-IES) was lower than that of IES on EHBD in all locations of the conventional and invasive eCCAs ( Figure 1D and Table 2B ). 60.0% of Bi eCCAs had CyD-IES and 30% of the other conventional eCCAs had CyD-IES, and their CyD-IES extended rarely to the gallbladder ( Table 2B ). In contrast, 75.0% of IPNBs had CyD-IES, and 57.1% had IES on the gallbladder when IPNBs reached the border between the CyD and the gallbladder.

In patients with invasive eCCAs, as well as those with invasive conventional eCCAs, patients with CyD-IES showed significantly longer survival times for both RFS (P = 0.037 and P = 0.016) and OS (P = 0.006 and P = 0.004, respectively) than patients without CyD-IES ( Figures 3A, B ). Perihilar eCCAs had a low incidence of CyD-IES, and there were no associations between CyD-IES and patient outcomes ( Figure 3C ). Patients with CyD-IES in distal eCCAs had significantly longer survival than those without CyD-IES for both RFS (P = 0.043) and OS (P = 0.018) ( Figure 3 ). Multivariate survival analysis of patients with invasive conventional eCCAs revealed that the CyD-IES was not a significant predictor of RFS or OS ( Table 3 ). Multivariate survival analysis of patients with distal eCCAs also revealed that the CyD-IES was not a significant predictor of RFS or OS ( Supplementary Table 3 ). CyD-IES closely correlated with tumor location, perineural invasion, and early recurrence of distal eCCAs ( Supplementary Table 4 ).

Distal eCCA cases reaching the junction of the CyD were divided into three groups according to IES patterns: eCCAs with CyD-IES as group A, eCCAs with only IES as group B, and eCCAs without any IES or CyD-IES as group C ( Supplementary Figure 4 ). Group A contained 5 eCCAs with only CyD-IES and 42 cases with both CyD-IES and IES. Survival analyses revealed that group A had longer survival and group C had shorter survival, with significant differences in both RFS (P = 0.002) and OS (P < 0.001) ( Supplementary Figure 5 ). Group B also showed longer survival compared with that of group C for both RFS (P = 0.024) and OS (P = 0.034) ( Supplementary Figure 5 ). All survival rates were higher in group A than in group B.

Some properties of the IES on the EHBD of the eCCAs are described above and depicted in Figure 4 . Liver-side IES was ended on the IHBD in 60.4%, on the bile ducts of the borderline area between the IHBD and EHBD (i.e., hepatic ducts and the second branches) in 37.5%, and on the confluence of the right and left hepatic ducts in 2.1% of conventional eCCA cases who underwent hepatectomy with liver-side IES ( Table 4A ). In perihilar eCCAs, IES on EHBD was mostly ended on the EHBD in front of the IHBD, and extension of IES into the IHBD from the EHBD was very rare, whereas 95.5% of cases with IES on IHBD had stromal invasion in the IHBD area. In distal eCCAs, cases of IES on IHBD were rare, with longer IES. Liver-side IES ended on the IHBD in 12.5% of patients, on the confluence of the right and left hepatic ducts in 25.0%, and on EHBD distal to the confluence of the hepatic ducts in 37.5% of patients with IPNB who underwent hepatectomy with liver-side IES. When cancer cells reached the EHBD in front of the IHBD, IES on the IHBD was found in 61.7% of patients. Thus, the liver-side IES from the EHBD often ended in front of the IHBD and extended to the IHBD in a limited conventional eCCAs with a long length of IES or in a part of perihilar eCCAs with stromal invasion of the IHBD.

Duodenal-side IES ended in the ampullary common duct in 20.3%, in the ampullary bile duct in 42.2%, and in 37.5% of patients with conventional eCCA who underwent pancreatoduodenectomy with duodenal-side IES ( Table 4B ). Duodenal-side IES ended in the ampullary common duct in 30.8%, in the ampullary bile duct in 46.2%, and in 23.1% of patients with IPNB who underwent pancreatoduodenectomy with duodenal-side IES. Thus, the duodenal-side IES on EHBD and the ampullary duct were similar to one tissue without barriers, although it did not reach the duodenum.

The incidence of IES in CyDCs was 80.0%; IES on CyD was found in 25.0% of cases; IES on the gallbladder in 65.0%; IES on EHBD in 40.0% (liver-side in 30.0% and duodenal-side in 15.0%); and no IES on the IHBD, ampullary common duct, or duodenum. The IES from CyDCs extended in both directions of the gallbladder and EHBD in 40.0% of the cases. The incidence of IES in conventional CyDCs and ICPNs is shown in Figure 4 and Supplementary Table 5 . CyDCs had a high frequency of IES on both the gallbladder and EHBD, especially in ICPN showing a very high frequency of IES; their incidences were 53% and 27% in conventional CyDCs and 100% and 80% in ICPNs, respectively.

IES on the gallbladder was often observed in GBCs, although IES extending into the CyD and EHBD was not observed, which was found in 8.1% and 1.2% of GBCs, respectively. This series contained 30.2% ICPN cases, and ICPNs showed a similar incidence of IES on CyD as conventional GBCs, as shown in Figure 4A and Supplementary Table 5 . However, in conventional GBCs and ICPNs that reached the borders between the CyD and the gallbladder, IES on the CyD was found in 62.5% and 100% of cases, respectively ( Figure 4B ). Thus, GBCs often have an IES, although the IES extension is usually limited to the gallbladder. The low incidence of IES on the CyD in both conventional GBCs and ICPNs was presumed to be because the tumor cells did not reach the border between the gallbladder and the CyD.

The incidence of IES was investigated in AVCs raised in the ampulla of Vater, except for the ampullary duodenum, which was 60.8%, 54.0%, and 76.9% in total AVCs, conventional AVCs, and IAPNs, respectively. Three directions of the IES beyond the ampulla of Vater were found: the EHBD, main pancreatic duct (MPD), and duodenum. The incidences of these IES are shown in Figure 4B and Supplementary Table 5 . These IESs beyond the ampulla of Vater were usually of short length ( Supplementary Table 6 ); IES on EHBD with ≤5 mm in length was in 54.5% of conventional AVCs and in 58.3% of IAPNs; IES on MPD with ≤5 mm in length was in 57.1% of conventional AVCs and in 70.0% of IAPNs. In 98.0% of conventional AVCs, the extension of the IES ended at the border between the ampullary common duct and the duodenum ( Figure 5 ), whereas 30.8% of IAPN cases had the IES extending beyond the border into the duodenum ( Figure 4B ). There was no IES reaching the IHBD, CyD, or gallbladder, except in one case of IAPN that spread to the CyD. Thus, IES extensions into the EHBD and MPD were found in approximately 20% and 15% of cases, respectively, and short length in both conventional AVCs and IAPNs, although intraepithelial spread on the duodenum was found in approximately 1/3 of IAPNs but not in conventional AVCs.