Prediction of the lymph node status in patients with intrahepatic cholangiocarcinoma: analysis of 320 surgical cases

Interpretable machine learning-based clinical prediction model for predicting lymph node metastasis in patients with intrahepatic cholangiocarcinoma

OBJECTIVE

Prediction of lymph node metastasis (LNM) for intrahepatic cholangiocarcinoma (ICC) is critical for the treatment regimen and prognosis. We aim to develop and validate machine learning (ML)-based predictive models for LNM in patients with ICC.

METHODS

A total of 345 patients with clinicopathological characteristics confirmed ICC from Jan 2007 to Jan 2019 were enrolled. The predictors of LNM were identified by the least absolute shrinkage and selection operator (LASSO) and logistic analysis. The selected variables were used for developing prediction models for LNM by six ML algorithms, including Logistic regression (LR), Gradient boosting machine (GBM), Extreme gradient boosting (XGB), Random Forest (RF), Decision tree (DT), Multilayer perceptron (MLP). We applied 10-fold cross validation as internal validation and calculated the average of the areas under the receiver operating characteristic (ROC) curve to measure the performance of all models. A feature selection approach was applied to identify importance of predictors in each model. The heat map was used to investigate the correlation of features. Finally, we established a web calculator using the best-performing model.

RESULTS

In multivariate logistic regression analysis, factors including alcoholic liver disease (ALD), smoking, boundary, diameter, and white blood cell (WBC) were identified as independent predictors for LNM in patients with ICC. In internal validation, the average values of AUC of six models ranged from 0.820 to 0.908. The XGB model was identified as the best model, the average AUC was 0.908. Finally, we established a web calculator by XGB model, which was useful for clinicians to calculate the likelihood of LNM.

CONCLUSION

The proposed ML-based predicted models had a good performance to predict LNM of patients with ICC. XGB performed best. A web calculator based on the ML algorithm showed promise in assisting clinicians to predict LNM and developed individualized medical plans.

Serum Angiopoietin-Like Protein 4: A Potential Prognostic Biomarker for Prediction of Vascular Invasion and Lymph Node Metastasis in Cholangiocarcinoma Patients

Cholangiocarcinoma (CCA) is a tumor arising from cholangiocytes lining the bile ducts. Vascular invasion and lymph node metastasis are important prognostic factors for disease staging as well as clinical therapeutic decisions for CCA patients. In the present study, we applied CCA sera proteomic analysis to identify a potential biomarker for prognosis of CCA patients. Then, using bioinformatics tools, we identified angiopoietin-like protein 4 (ANGPTL4) which expressed highest signal intensity among candidate proteins in proteomic analysis of CCA sera. Expression of ANGPTL4 in CCA tissues was determined using immunohistochemistry. The results showed that ANGPTL4 was stained at higher level in CCA cells when compared with normal cholangiocytes. The high expression of ANGPTL4 was associated with lymph node metastasis and advanced tumor stage (p = 0.013 and p = 0.031, respectively). Furthermore, serum ANGPTL4 levels in CCA and healthy control (HC) were analyzed using a dot blot assay. And it was found that ANGPTL4 level was significantly higher in CCA than HC group (p < 0.0001). ROC curve analysis revealed that serum ANGPTL4 level was effectively distinguished CCA from healthy patients (cutoff = 0.2697 arbitrary unit (AU), 80.0% sensitivity, 72.7% specificity, AUC = 0.825, p < 0.0001). Serum ANGPTL4 level was associated with vascular invasion and lymph node metastasis (p = 0.0004 and p = 0.006), so that it differentiated CCA with vascular invasion from CCA without vascular invasion (cutoff = 0.5526 AU, 64.9% sensitivity, 92.9% specificity, AUC = 0.751, p = 0.006) and it corresponded to CCA with/without lymph node metastasis (cutoff = 0.5399 AU, 71.4% sensitivity, 70.8% specificity, AUC = 0.691, p = 0.01) by ROC analysis. Serum ANGPTL4 levels showed superior predictive efficiency compared with CA 19-9 and CEA for vascular invasion and lymph node metastasis. In addition, serum ANGPTL4 level was an independent predictive indicator by multivariate regression analysis. In conclusion, serum ANGPTL4 could be a novel prognostic biomarker for prediction of vascular invasion and lymph node metastasis of CCA patients.

The role of lymphadenectomy in the surgical treatment of intrahepatic cholangiocarcinoma: A review

Cholangiocarcinoma is the second most common primary tumor of the liver. The incidence and mortality of its intrahepatic form has been increasing over the past 2 decades. Currently, the only available curative treatment for intrahepatic cholangiocarcinoma is surgical resection. There is still no prospective evidence to support neoadjuvant systemic treatments in resectable disease, while adjuvant chemotherapy with Capecitabine is currently the only recommended systemic treatment after liver resection based on the results of randomised trial. Despite the implementation of perioperative treatments and improvements in resective surgery, intrahepatic cholangiocarcinoma remains a disease characterized by high incidence of recurrence and poor long-term survival. Lymph node metastases can be found in 45-65% of patients and are one of the most impacting prognostic factors after surgical resection. Preoperative imaging is not always sufficient in assessing lymph node status, thus hepatic pedicle lymphadenectomy can be important to ensure precise staging in surgical patients. An increasing trend in performing lymph node dissection during liver resection for intrahepatic cholangiocarcinoma has been observed in the last 20 years, although its actual efficacy compared to the potential complications remains debated. The current evidence on the prognostic role of the lymph node status, its preoperative predictability, the basis for a correct hepatic pedicle lymphadenectomy and its prognostic role in the surgical treatment of intrahepatic cholangiocarcinoma are presented.

Apparent diffusion coefficient value of mass-forming intrahepatic cholangiocarcinoma: a potential imaging biomarker for prediction of lymph node metastasis

PURPOSE

To compare the differences of MR features between mass-forming intrahepatic cholangiocarcinoma (IMCC) with and without lymph node metastasis (LNM) and to search for new imaging biomarkers for predicting LNM.

MATERIALS AND METHODS

The study included 91 patients with histopathologically confirmed single IMCC (20 patients with LNM and 71 patients without LNM). Findings of preoperative MR imaging including diffusion-weighted imaging (DWI) (b value 0, 500 mm²/s) were analyzed and apparent diffusion coefficient (ADC) values (b = 500 mm²/s) were calculated. Logistic regression analysis was performed to identify independent predictors of LNM. The diagnostic performance of independent predictors was assessed by receiver operating characteristic (ROC) and area under the curve (AUC) was compared.

RESULTS

Larger tumor size (p = 0.001), diameter of largest lymph node (LN) > 1 cm (p < 0.001), higher ADC value of primary IMCC lesion (ADC_IMCC value) (p = 0.001), and positive CA19-9 level (p = 0.018) were correlated with LNM. Multivariate logistic regression analysis demonstrated that ADC_IMCC value (odds ratio, 3.347; p = 0.001) and diameter of largest LN > 1 cm (odds ratio, 7.571; p = 0.004) were independent predictors of LNM. The AUCs for ADC_IMCC value, diameter of largest LN > 1 cm,and combined method (the combination of ADC_IMCC value and diameter of largest LN > 1 cm) were 0.782, 0.701,and 0.857, respectively. The AUC for combined method was significantly higher than that of diameter of largest LN > 1 cm (p = 0.033).

CONCLUSION

ADC_IMCC value can be a potential imaging biomarker for predicting LNM of IMCC, especially in combination with diameter of largest LN > 1 cm.

Radiofrequency Ablation vs. Surgical Resection for Small Early-Stage Primary Intrahepatic Cholangiocarcinoma

Aim: We aimed to compare the survival outcomes of radiofrequency ablation (RFA) and surgical resection (SR) for patients with small early-stage primary intrahepatic cholangiocarcinoma (ICC).

Methods: Patients with small (≤5 cm) and early-stage ICC were screened from the Surveillance, Epidemiology, and End Results (SEER) database. Overall survival (OS) and cancer-specific survival (CSS) rates between the SR and RFA patients were evaluated. The results were verified using an inverse probability-weighting model (IPTW).

Results: In total, 184 patients with small T1 stage ICC that received RFA or SR treatment were identified. The OS rates at 1, 3, and 5 years were 87.4, 73.3, and 61.5% for patients who underwent SR, respectively, and 89.9, 42.4, and 23.9%, respectively, for patients who received RFA. CSS rates at 1, 3, and 5 years were 91.5, 73.8, and 66.1%, respectively, for the SR group and 93.5, 53.4, and 30.0%, respectively, for the RFA group. The OS and CSS rates were evaluated to be significantly better in the SR group than in the RFA group after the multivariate Cox regression and IPTW analysis. Subsequently, the survival benefit of SR was also observed in the subgroup of patients with <4.5 or <4 cm early-stage ICC when compared with RFA.

Conclusion: Our results indicated that the SR provided a significantly better prognosis than RFA in patients with small and early-stage ICC. SR as the first-line treatment of primary early-stage ICC is still recommended. However, prospective randomized controlled trials with larger sample sizes are required to compare these modalities in the treatment of ICC.

Apparent diffusion coefficient value of mass-forming intrahepatic cholangiocarcinoma: a potential imaging biomarker for prediction of lymph node metastasis

PURPOSE

To compare the differences of MR features between mass-forming intrahepatic cholangiocarcinoma (IMCC) with and without lymph node metastasis (LNM) and to search for new imaging biomarkers for predicting LNM.

MATERIALS AND METHODS

The study included 91 patients with histopathologically confirmed single IMCC (20 patients with LNM and 71 patients without LNM). Findings of preoperative MR imaging including diffusion-weighted imaging (DWI) (b value 0, 500 mm²/s) were analyzed and apparent diffusion coefficient (ADC) values (b = 500 mm²/s) were calculated. Logistic regression analysis was performed to identify independent predictors of LNM. The diagnostic performance of independent predictors was assessed by receiver operating characteristic (ROC) and area under the curve (AUC) was compared.

RESULTS

Larger tumor size (p = 0.001), diameter of largest lymph node (LN) > 1 cm (p < 0.001), higher ADC value of primary IMCC lesion (ADC_IMCC value) (p = 0.001), and positive CA19-9 level (p = 0.018) were correlated with LNM. Multivariate logistic regression analysis demonstrated that ADC_IMCC value (odds ratio, 3.347; p = 0.001) and diameter of largest LN > 1 cm (odds ratio, 7.571; p = 0.004) were independent predictors of LNM. The AUCs for ADC_IMCC value, diameter of largest LN > 1 cm,and combined method (the combination of ADC_IMCC value and diameter of largest LN > 1 cm) were 0.782, 0.701,and 0.857, respectively. The AUC for combined method was significantly higher than that of diameter of largest LN > 1 cm (p = 0.033).

CONCLUSION

ADC_IMCC value can be a potential imaging biomarker for predicting LNM of IMCC, especially in combination with diameter of largest LN > 1 cm.

Failure patterns after curative resection for intrahepatic cholangiocarcinoma: possible implications for postoperative radiotherapy

Background

To explore the patterns of failures and areas at highest risk of recurrence for postoperative intrahepatic cholangiocarcinoma (IHCC), with the aim to guide IHCC adjuvant radiotherapy.

Methods

Patients with IHCC who had undergone radical surgery at our institution from July 2010 to August 2017 were retrospectively analyzed. The survival and prognostic factors were analyzed by univariate and multivariate analysis. All sites of recurrence were found out and classified as the surgical margin, regional lymph nodes, liver remnant and distant metastasis. According to the recurring area at highest risk, the target volume of adjuvant radiotherapy was proposed.

Results

The median follow-up time was 23.5 months (2–85 months). The median recurrence free survival (RFS) and overall survival (OS) were 12.1 months and 24.8 months, respectively. Seventy-three (73/127, 57.5%) IHCC patients developed tumor recurrence. Initial recurrences occurred in the potential postoperative radiotherapy (PORT) volume, remnant liver and distant sits were 46 (46/73, 63.0%), 36 (36/73, 49.3%) and 22 (22/73, 30.1%) cases, respectively. Of the 46 patients whose initial recurrence inside the potential PORT volume, 29 (29/73, 39.7%) developed recurrence only inside the potential PORT volume, including 13 tumor bed recurrences, 7 lymph node metastases, and 9 with both tumor bed recurrences and lymph node metastases. The most common lymph node metastases sites were nodes around the abdominal aorta, followed by lymph nodes along the celiac artery, the common hepatic artery, and in the hepatoduodenal ligament.

Conclusions

High proportion of the recurrences occurred only inside the potential PORT volume, implying adjuvant radiotherapy might improve the local-regional control. Surgical margins and lymph node stations No.16a2, 9, 8, 12, 13, and 14 are suggested to be included in the radiation volume.

A radiomics approach based on support vector machine using MR images for preoperative lymph node status evaluation in intrahepatic cholangiocarcinoma

Purpose: Accurate lymph node (LN) status evaluation for intrahepatic cholangiocarcinoma (ICC) patients is essential for surgical planning. This study aimed to develop and validate a prediction model for preoperative LN status evaluation in ICC patients.

Methods and Materials: A group of 106 ICC patients, who were diagnosed between April 2011 and February 2016, was used for prediction model training. Image features were extracted from T1-weighted contrast-enhanced MR images. A support vector machine (SVM) model was built by using the most LN status-related features, which were selected using the maximum relevance minimum redundancy (mRMR) algorithm. The mRMR method ranked each feature according to its relevance to the LN status and redundancy with other features. An SVM score was calculated for each patient to reflect the LN metastasis (LNM) probability from the SVM model. Finally, a combination nomogram was constructed by incorporating the SVM score and clinical features. An independent group of 42 patients who were diagnosed from March 2016 to November 2017 was used to validate the prediction models. The model performances were evaluated on discrimination, calibration, and clinical utility.

Results: The SVM model was constructed based on five selected image features. Significant differences were found between patients with LNM and non-LNM in SVM scores in both groups (the training group: 0.5466 (interquartile range (IQR), 0.4059-0.6985) vs. 0.3226 (IQR, 0.0527-0.4659), P<0.0001; the validation group: 0.5831 (IQR, 0.3641-0.8162) vs. 0.3101 (IQR, 0.1029-0.4661), P=0.0015). The combination nomogram based on the SVM score, the CA 19-9 level, and the MR-reported LNM factor showed better discrimination in separating patients with LNM and non-LNM, comparing to the SVM model alone (AUC: the training group: 0.842 vs. 0.788; the validation group: 0.870 vs. 0.787). Favorable clinical utility was observed using the decision curve analysis for the nomogram.

Conclusion: The nomogram, incorporating the SVM score, CA 19-9 level and the MR-reported LNM factor, provided an individualized LN status evaluation and helped clinicians guide the surgical decisions.

A serological scoring system to predict lymph node metastasis in patients with hepatocellular carcinoma

BACKGROUND

Lymph node metastasis (LNM)has widely been recognized as a poor prognostic indicator for hepatocellular carcinoma (HCC) patients. Preoperative prediction of LNM is important for clinicians to decide on treatment. This study was designed to develop a simple and convenient system to predict LNM.

METHODS

Consecutive HCC patients who were suspected to have LNM were divided into a training, an internal validation and an external validation cohort. The receiver operating characteristic (ROC) analysis was used to determine the threshold value of the preoperative serological variables. A nomogram visualization system model was then established.

RESULT

Of the 287 patients, there were 31 patients who had LNM (10.8%), and 21 of 203 patients (10.3%) were in the training cohort and 10 of 84 patients (11.9%) in the internal validation cohort. Sixteen of 176 patients (9.1%) in the external validation cohort had LNM. The serological indices including neutrophil/lymphocyte rate, age, platelet, prothrombin time, and total protein, were included in the nomogram. The areas of the ROC curve were 0.846, 0.679 and 0.738 in predicting LNM in the training cohort, the internal validation cohort and the external validation cohort, respectively.

CONCLUSION

The scoring system constructed using the preoperative serological variables predicted LNM in HCC patients.

Surgical resection of lymph node positive intrahepatic cholangiocarcinoma may not improve survival

BACKGROUND

Nodal positivity is a predictor of poor survival following resection for intrahepatic cholangiocarcinoma (ICC). The aim of this study was to evaluate the impact of surgical resection on survival in patients with lymph node (LN) positive ICC.

METHODS

An augmented version of the Surveillance, Epidemiology, and End Results program database was utilized to identify patients with LN-positive ICC without distant metastases from 2000 to 2014. Patients were stratified by treatment: chemotherapy alone or surgical resection with/without chemotherapy. Survival was evaluated using Kaplan-Meier and Cox proportional hazard models.

RESULTS

169 patients who underwent treatment for LN-positive ICC were identified. 88% underwent surgical resection and 12% underwent chemotherapy alone. The median survival for patients who underwent surgical resection was not different from patients treated with chemotherapy alone (19 months 95% Confidence Interval (CI) 17-33 versus 20 months CI 10-27, p = 0.323). A cox-proportional hazard ratio model demonstrated that black race was associated with worse survival (p < 0.05), while surgical resection was not independently associated with survival.

CONCLUSION

Surgical resection for patients with LN-positive ICC may not improve survival compared to chemotherapy alone. Pathologic LN evaluation should be performed prior to surgical resection, to improve patient selection and ensure receipt of optimal therapy.

Trends in use of lymphadenectomy in surgery with curative intent for intrahepatic cholangiocarcinoma

BACKGROUND

The role of routine lymph node dissection (LND) in the surgical treatment of intrahepatic cholangiocarcinoma (ICC) remains controversial. The objective of this study was to investigate the trends of LND use in the surgical treatment of ICC.

METHODS

Patients undergoing curative intent resection for ICC in 2000-2015 were identified from an international multi-institutional database. Use of lymphadenectomy was evaluated over time and by geographical region (West versus East); LND use and final nodal status were analysed relative to AJCC T categories.

RESULTS

Among the 1084 patients identified, half (535, 49·4 per cent) underwent concomitant hepatic resection and LND. Between 2000 and 2015, the proportion of patients undergoing LND for ICC nearly doubled: 44·4 per cent in 2000 versus 81·5 per cent in 2015 (P < 0·001). Use of LND increased over time among both Eastern and Western centres. The odds of LND was associated with the time period of surgery and the extent of the tumour/T status (referent T1a: OR 2·43 for T2, P = 0·001; OR 2·13 for T3, P = 0·016). Among the 535 patients who had LND, lymph node metastasis (LNM) was noted in 209 (39·1 per cent). Specifically, the incidence of LNM was 24 per cent in T1a disease, 22 per cent in T1b, 42·9 per cent in T2, 48 per cent in T3 and 66 per cent in T4 (P < 0·001). AJCC T3 and T4 categories, harvesting of six or more lymph nodes, and presence of satellite lesions were independently associated with LNM.

CONCLUSION

The rate of LNM was high across all T categories, with one in five patients with T1 disease having nodal metastasis. The trend in increased use of LND suggests a growing adoption of AJCC recommendations in the treatment of ICC.

Review to better understand the macroscopic subtypes and histogenesis of intrahepatic cholangiocarcinoma

Intrahepatic cholangiocarcinoma is macroscopically classified into three subtypes, mass-forming-type, periductal infiltrating-type, and intraductal growth-type. Each subtype should be preoperatively differentiated to perform the valid surgical resection. Recent researches have revealed the clinical, radiologic, pathobiological characteristics of each subtype. We reviewed recently published studies covering various aspects of intrahepatic cholangiocarcinoma (ICC), focusing especially on the macroscopic subtypes and stem cell features to better understand the pathophysiology of ICC and to establish the valid therapeutic strategy.