Introduction

Cholangiocarcinoma is an uncommon but fatal malignancy of the biliary tract. Over 90% of them were adenocarcinoma. Cholangiocarcinoma can be classified histologically into mass-forming, periductal infiltrative and intraductal growth type [1], and by tumour location, it can be classified into intrahepatic and extrahepatic cholangiocarcinoma. Regardless of the type and location, the prognosis of these tumours remains poor; the 5-year overall survivals are in the order of 30–40% in most series [2,3,4,5,6]. Resection is the only hope of cure; margin status and presence of metastatic lymph node are the two main determinants for recurrence and survival. Preoperative recurrence risk stratification is essential for individualized surgical planning and oncological treatments; however, method for such purpose is scarce. Carcinoembryonic antigen (CEA) [7] and carbohydrate antigen (CA) 19-9 [8,9,10] were found to be independent factors associated with survival after resection. However, these markers lack specificity as elevation of markers is commonly seen in benign conditions, while it remains normal even in patients with advanced cholangiocarcinoma [11, 12]. Recent research has been focusing on the role of PET/CT in the management of cholangiocarcinoma. It has been shown that PET/CT is superior to CT scan in detecting distant metastasis in cholangiocarcinoma [13]; however, the reported accuracy of PET/CT in metastatic lymph node detection was variable [14,15,16]. In addition, evidence about the prognostic value of PET/CT in patients with resectable cholangiocarcinoma is limited [17, 18]. This retrospective study served to further elucidate the accuracy of PET/CT in metastatic lymph node detection and the implication of maximum standardized uptake value in the prognosis of resectable cholangiocarcinoma.

Method

Consecutive patient with hepatectomy performed for cholangiocarcinoma with curative intent from January 2010 to March 2015 in Queen Mary Hospital was recruited. In our centre, working diagnosis and management plan of patients with resectable cholangiocarcinoma were made in a multidisciplinary meeting. In this study, all diagnoses were confirmed by experienced pathologist through identification cholangiocarcinoma with or without special immunohistochemical staining (i.e. cytokeratin-7, cytokeratin 20 and TTF-1) in the surgical specimen. Cholangiohepatocellular carcinoma, metastatic adenocarcinoma, carcinoma of the ampulla of Vater, carcinoma of distal common bile duct and pancreatic cancer were excluded from the study. Demographic, biochemical, radiological and operative data of patients were extracted from our prospectively maintained database. PET/CT report of each patient was reviewed; data such as the presence of hypermetabolic lymph node (Fig. 1), SUV-max (maximal standard uptake value) (Fig. 2) and metabolic size of the primary tumour were retrieved.

Fig. 1
figure 1

Regional and distant lymph node detected by PET scan

Full size image
Fig. 2
figure 2

Intrahepatic cholangiocarcinoma appears as hypermetabolic lesion on PET scan

Full size image

PET/CT predicted lymph node status was correlated with final pathological results; sensitivity, specificity, positive and negative predictive values of PET lymph node prediction were then calculated. Accuracy of the PET/CT in the detection of metastatic regional lymph node and the correlation between SUV-max and oncological outcomes were determined by the area under the receiver operator characteristic (ROC) curve. Cut-off value of SUV-max was determined by the point on ROC curve which corresponds to the highest value of sensitivity and specificity. Continuous variables including SUV-max, CEA, CA19-9, alanine aminotransferase (SGOT) and tumour size were analysed using Mann–Whitney U test or independent t test wherever appropriate, while categorical parameter including the presence of PET-positive node, pathological lymph node status and presence of lymphovascular permeation was analysed with Chi-square test/Fisher’s exact test whenever appropriate. Factors found to have a significant association (P value <0.05) in univariate analysis were put into multivariate analysis for the identification of independent factors. Kaplan–Meier analysis was used for survival calculation, and the difference of survival was compared using log-rank test. All data were presented as median with a given range unless otherwise specified. P value of <0.05 was considered statistically significant, and all P values were two tails in this article. Statistical Package for the Social Sciences (SPSS) version 20.0 was used for the analytical work.

The PET/CT imaging protocol

Patients in this study had undergone a PET/CT examination within 2 months of hepatectomy. The PET/CT protocol had been previously described [19]. In brief, patients would be kept fast for 6 h before examination. After normoglycemia checked, 18-F FDG with the dosage of 6.3 MBq/kg of body weight (330–520 MBq) was injected. Limited whole body PET/CT (from skull base to mid-thigh) was performed at 60 min after injection of 18-F FDG. Data acquisition with an integrated in-line PET/CT scanner (Biograph LSO or Biograph 16 LSO HI-REZ; Siemens) started with non-contrast CT scanning (130 kV, 110–115 mA, 2-mm pitch and 1-s tube rotation) followed by PET of a 2-min emission acquisition time and a 16.2-cm axial field of view per position. All PET/CT images were interpreted and reported by radiologists with extensive experience in the field of nuclear medicine. SUV-max was compared with the baseline liver parenchymal and lymph node metabolic activity. The metabolic size of lesion and whether the metabolic activities of lymph nodes were up to a suspicious level rested on the reporting radiologist’s discretion.

Technical aspect and follow-up of hepatectomy

The technical details of hepatectomy in our centre had been described in another report [20]. In short, all patients would have indocyanine green clearance (ICG) test and CT volumetric assessment before major liver resection. The maximum ICG-R15 (ICG retention in 15 min) for major liver resection was 22%, while the future liver remnant should be no less than 25% of the estimated standard liver volume by HKU formula in usual circumstances [21]. Major hepatectomy referred to resection of more than 3 Couinaud’s liver segments. Frozen section would be taken at resection margin of bile duct whenever possible, until clear margin status is achieved. Margin status is classified into three categories according to pathologist comments (R0 margin width over 1 mm or above, R1 margin width less than 1 mm, R2 margin involved by tumour). Patient could be discharged home around 1–2 weeks depending on individual progress of recovery. Outpatient follow-up would be scheduled at 2 weeks after discharge, then 1 month later and 3 monthly thereafter. Biochemical tests including liver function test, CEA and CA 19-9 would be monitored, and contrasted cross-sectional imaging was arranged 3 months and then every half yearly after operation. Recurrence is defined histologically by liver or lymph node biopsy or radiologically by the presence of intrahepatic or extrahepatic tumour.

Results

There were 66 patients (39 intrahepatic cholangiocarcinoma, 15 hilar cholangiocarcinoma, 9 gallbladder cancer and 3 carcinoma of common bile duct) eligible for the study. In this series, 28 patients were female and 38 were male; the median age was 66 year old. In total, 37 (56.1%) of them had one or more medical comorbidities before hepatectomy. Hepatitis B status was positive in thirteen patients, and one patient was a hepatitis C carrier. Majority of the patients were non-cirrhotic, and the median preoperative haemoglobin, bilirubin, albumin and CEA level was 12.4 g/dl, 10 μmol/l, 40 g/l and 3.2 ng/ml, respectively (Table 1).

Table 1 Baseline demographic and biochemical characteristics of the whole study population
Full size table

Operative findings and tumour characteristics

Majority of patients underwent right hepatectomy, followed by left hepatectomy (Table 2). The median operation time and blood loss were 469 min and 887 ml, respectively. None of the patient required intraoperative blood transfusion. The median hospital length of stay was 13 days. There were four hospital mortalities (6.1%). The median tumour size was 5 cm. R0 resection was achieved in 78.8%, while R2 and R1 resection was found in 18.2 and 3% of the patient, respectively. Lymphovascular permeation was found in more than half of our patients (43.9%). Out of the 54 patients with pathological lymph node examination, metastatic adenocarcinoma was detected in 21 patients (31.8%). Most of our patients had either TNM stage I or II disease (Table 3), and 29 patients were given adjuvant chemotherapy after the operation. In total, 38 patients developed recurrence upon the date of analysis. The 1- and 3-year disease-free survival was 47.2 and 27.1%, respectively, while the 1- and 3-year overall survival was 67.4 and 39.2%, respectively. The median follow-up time was 27 months.

Table 2 Operative findings and outcomes
Full size table
Table 3 Tumour characteristics after pathological examination
Full size table

Diagnostic and prognostic ability of PET/CT

The median metabolic size of tumour was 42 mm (2.2–130). In total, 22 patients were suggested to have metastatic lymph node on PET/CT, and the rest of them had either “reactive” lymph nodes or “no metastatic lymph node”. Pathological examination of lymph node was performed in 54 patients, and the pathological assessment of lymph node was correlated with preoperative PET findings. The sensitivity, specificity, positive predictive value and negative predictive value were 66.7, 78.8, 66.7 and 78.8%, respectively (P = 0.001). The area under receiver operating characteristic (ROC) curve for metastatic LN prediction was 0.727 (P = 0.005, 95% CI 0.583–0.871) (Fig. 3). When the analysis was repeated in intrahepatic cholangiocarcinoma subgroup, the accuracy of metastatic LN prediction increased to 81.3% (P = 0.011, 95% CI 0.635–0.990) (Fig. 4).

Fig. 3
figure 3

Receiver operating characteristic curve showing the predictive value of PET/CT for metastatic lymph node

Full size image
Fig. 4
figure 4

Receiver operating characteristic curve showing the predictive value of metastatic lymph node in intrahepatic cholangiocarcinoma

Full size image

Prognostic significance of primary tumour SUV-max in cholangiocarcinoma patient

The median SUV-max of the primary tumour was 7.8 (2.5–20.5). Univariate analysis showed that white cell count (P = 0.006), albumin (P = 0.02), carcinoembryonic antigen (P = 0.037), lymphovascular invasion (P = 0.042), pathological lymph node status, TNM staging (P = 0.003) and SUV-max (P = 0.016) were factors associated with disease-free survival. In multivariate analysis, TNM staging (P < 0.001 OR 1.881, 95% CI 1.318–2.685) and SUV-max of tumour (P = 0.007 OR 1.16, 95% CI 1.04–1.29) were the only independent factors of disease recurrence (Table 4). Concerning overall survival, SUV-max (P = 0.001), leucocyte count (P = 0.003), albumin (P = 0.001), aspartate transaminase (P = 0.025), CEA (P = 0.006), metabolic tumour size (P = 0.019), pathological tumour size (P = 0.017), presence of nodal metastasis (P = 0.035) and TNM staging (P = 0.006) showed association in univariate analysis; multivariate analysis revealed that presence of metastatic lymph node (P = 0.037 OR 3.211 95% CI 1.070–9.634) and SUV-max of tumour (P = 0.012 OR 1.145, 95% CI 1.030–1.273) were independent preoperative factors for overall survival (Table 5).

Table 4 Univariate and multivariate analysis for factors associated with disease-free survival
Full size table
Table 5 Univariate and multivariate analysis for factors associated with overall survival
Full size table

Prognostic cut-off value of SUV-max of primary tumour and subgroup analysis

Statistically significant association between disease-free, overall survival and SUV-max value was demonstrated in the whole population analysis (AUC 0.66.9, P = 0.022 95% CI 0.538–0.800, and AUC 0.698, P = 0.006 95% CI 0.571–0.826, respectively) (Fig. 5a, b). In subgroup analysis of patients with early disease, i.e. TNM stage I or II, SUV-max was shown to have high predictive value for disease-free and overall survival (AUC 0.747, P = 0.006 95% CI 0.600–0.894, and AUC 0.741, P = 0.012 95% CI 0.590–0.892, respectively) (Fig. 6a, b). SUV-max cut-off of 8 was determined with ROC curve, and we found that this SUV-max cut-off had a high predictive accuracy for disease-free survival (AUC 0.723, P = 0.013 95% CI 0.567–0.879) and overall survival (AUC 0.721, P = 0.018 95% CI 0.561–0.881) (Fig. 7a, b). Kaplan–Meier analysis showed that patients with primary tumour SUV-max value over 8 had significantly lower 3-year disease-free and overall survival (21.2 vs. 63.2%, P = 0.004, and 29 vs. 74% P = 0.048, respectively) (Fig. 8a, b).

Fig. 5
figure 5

a Receiver operating characteristic curve showing the predictive value of tumour SUV-max to disease-free survival. b Receiver operating characteristic curve showing the predictive value of tumour SUV-max to overall survival

Full size image
Fig. 6
figure 6

a Receiver operating characteristic curve showing the predictive value of tumour SUV-max to disease-free survival in patients with stage I or II disease. b Receiver operating characteristic curve showing the predictive value of tumour SUV-max to overall survival in patients with stage I or II disease

Full size image
Fig. 7
figure 7

a Receiver operating characteristic curve showing the predictive value of tumour SUV-max over 8 with respect to disease-free survival. b Receiver operating characteristic curve showing the predictive value of SUV-max over 8 with respect to overall survival

Full size image
Fig. 8
figure 8

a Kaplan–Meier curve showing the disease-free survival in patients with stage I or II cholangiocarcinoma. b Kaplan–Meier curve showing the overall survival of patients with stage I or II cholangiocarcinoma

Full size image

Resectable ICC was defined as R0 or R1 resection margin in the final pathology. In the 32 patients with resectable ICC, the median SUV-max of this group was 9, and the median disease-free survival and overall survival were 9.5 and 14 months, respectively. ROC curve showed that the value of SUV-max of primary tumour predicts survival outcomes (Fig. 9a, b). Patients with primary tumour SUV-max less than 8 were shown to have significantly better 3-year disease-free survival (55.6 vs. 13.4%) and a tendency of better 3-year overall survival (67.7 vs. 34.8%) (Fig. 10a, b).

Fig. 9
figure 9

a Receiver operating characteristic curve showing the predictive value of SUV-max to recurrence in patients with resectable ICC. b Receiver operating characteristic curve showing the predictive value of SUV-max to overall survival in patients with resectable ICC

Full size image
Fig. 10
figure 10

a Kaplan–Meier curve showing the difference in disease-free survival of ICC patients with SUV-max cut-off of 8. b Kaplan–Meier curve showing the difference in overall survival of ICC patients with SUV-max cut-off at 8

Full size image

Discussion

This study conveyed two important messages about the role of PET/CT in management of cholangiocarcinoma; firstly, it is an accurate imaging modality to predict the presence of metastatic lymph node in patients with cholangiocarcinoma, and the predictability is even higher in patients with intrahepatic cholangiocarcinoma. In a small series of intrahepatic cholangiocarcinoma, Park et al. found that the sensitivity and specificity of PET/CT in detecting metastatic lymphadenopathy were 80 and 92%, respectively [16]; in contrast, Kluge R et al. doubted this role as only 2 out of the 15 node-positive cholangiocarcinoma patients were detected by PET/CT in their series [15]. There seemed to be an intrinsic metabolic difference between the metastatic lymph node from intrahepatic and extrahepatic cholangiocarcinoma. Secondly, SUV-max value of the primary tumour is an independent factor for survival outcomes in all cholangiocarcinoma patients, and SUV-max cut-off of 8 is a prognostic indicator for patients with early disease (TNM stage I and II). These findings are useful in the perioperative management planning. Although conflicting is the role of neoadjuvant therapy in the management of cholangiocarcinoma in terms of local disease control and long-term survival benefits [22, 23], it should be considered in patients who have high primary tumour SUV-max (cut-off > 8) and presence of PET-positive lymph nodes. In case of intrahepatic cholangiocarcinoma where lymphadenectomy is not a routine procedure, the presence of PET-positive regional lymph node should indicate lymph node dissection so as to facilitate pathological staging and reduce the local recurrence rate in case of genuine nodal disease [24, 25]. For patients with low SUV-max and absence of suspicious hypermetabolic lymph node on PET/CT, aggressive surgery to obtain R0 resection and a wider negative margin (over 1 cm) could benefit survival particularly in patients with intrahepatic cholangiocarcinoma [26]. Furthermore, aggressive hepatic operation with radical lymphadenectomy is less justified in patients who have high primary tumour SUV-max, predicted distant lymph node spread and marginal physiological reserve. Despite the fact that the role of adjuvant treatment in resectable cholangiocarcinoma remains to be defined [27,28,29], SUV-max can serve as an extra point of consideration, in addition to the conventional tumour characteristics, before contemplating adjuvant treatment.

Apart from SUV-max, metabolic tumour volume (MTV) and total lesion glycolysis (TLG) are the two more sophisticated PET parameters that were shown to reflect viable tumour bulk [30]. In pancreatic cancer, it had been reported that MTV and TLG are superior prognostic parameters when compared to cancer antigen 19.9 (CA 19.9), tumour size and SUV-max [31]. However, these calculations are not routinely performed as MTV and TLG require manual tumour mapping and that could tedious and time-consuming. Use of automated volumetric study and the adaptive threshold for SUV-max calculation (tumour to background metabolic gradient) were suggested solutions for such drawbacks [32]. Nonetheless, the role of MTV and TLG in the management of cholangiocarcinoma remains to be elucidated as there is very limited study in this context [33].

There are a few weaknesses in the present study. Firstly, retrospective nature of the analysis inevitably confounded by selection bias and missing data. Consecutive patient recruitment and use of multivariate analysis in this study would have alleviated this inherent weakness of retrospective study; secondly, low incidence of cholangiocarcinoma limited the case volume of study, and this is the Achilles heel in the study of uncommon disease. Furthermore, inter-observer variability in the PET interpretation and SUV calculation could not be excluded. It has been reported that scanner calibration, synchronization between machine and injector, partial volume effect, imaging reformation protocol, patient body weight and serum glucose level, image acquisition time and definition of region of interest can all influence the precision of SUV-max [33,34,35]. Nonetheless, the findings of the current study provide important information to future multicenter study or meta-analysis on the area of PET and cholangiocarcinoma.

Conclusion

PET/CT is a reliable imaging modality for metastatic lymph node detection in cholangiocarcinoma. Tumour SUV-max is an independent factor for oncological outcomes in patients with resectable disease. For patients who have TNM stage I or II cholangiocarcinoma, tumour SUV-max over 8 is associated with significantly inferior disease-free and overall survival even after curative resection.