If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The appropriate method of hepatectomy for hepatocellular carcinoma within University of California San Francisco (UCSF) criteria through neural network analysis
∗ Zheng J and Wang Ning contributed equally to this study and are co-first authors.
Jinli Zheng
Footnotes
∗ Zheng J and Wang Ning contributed equally to this study and are co-first authors.
Affiliations
Department of Liver Surgery, West China Hospital of Sichuan University, Chengdu, Sichuan Province, ChinaDepartment of Liver Transplantation Center, West China Hospital of Sichuan University, Chengdu, Sichuan Province, China
∗ Zheng J and Wang Ning contributed equally to this study and are co-first authors.
Ning Wang
Footnotes
∗ Zheng J and Wang Ning contributed equally to this study and are co-first authors.
Affiliations
Department of Liver Surgery, West China Hospital of Sichuan University, Chengdu, Sichuan Province, ChinaDepartment of Hepatobiliary Surgery, West China JinTang Hospital, China
Department of Liver Surgery, West China Hospital of Sichuan University, Chengdu, Sichuan Province, ChinaDepartment of Liver Transplantation Center, West China Hospital of Sichuan University, Chengdu, Sichuan Province, China
Correspondence Li Jiang, Department of Liver Surgery, Liver Transplantation Center, West China Hospital of Sichuan University, Chengdu, Sichuan Province, China.
Department of Liver Surgery, West China Hospital of Sichuan University, Chengdu, Sichuan Province, ChinaDepartment of Liver Transplantation Center, West China Hospital of Sichuan University, Chengdu, Sichuan Province, China
This study aimed to find effective treatments for the patient within UCSF criteria.
Methods
This study enrolled 1006 patients meeting UCSF criteria, undergoing hepatic resection (HR), divided into two groups: single tumor group and multiple tumors group. We compared and analyzed the risk factors between these two groups’ long-term outcomes, through log-rank test, cox proportional hazards model and using neural network analysis to identify the independent risk factors.
Results
The 1-, 3-, and 5-year OS rates in single tumor were significantly higher than multiple tumors (95.0%, 73.2% and 52.3% versus 93.9%, 69.7% and 38.0%, respectively, p < 0.001). The 1-, 3- and 5-year RFS rates were 90.3%, 60.7%, and 40.1% in single tumor and 83.4%, 50.7% and 23.8% in multiple tumors, respectively (p < 0.001). And tumor type, anatomic resection and MVI were the independent risk factors for the patient within UCSF criteria. MVI was the most important risk factor affecting OS and RFS rates in neural network analysis. The method of hepatic resection and the number of tumors were also affected OS and RFS rates.
Conclusion
The anatomic resection should be applied to the patient within UCSF criteria, especially for the patient was in single tumor with MVI-negative.
Introduction
Hepatocellular carcinoma (HCC) is the sixth most malignant tumor in the world and causes about 600,000 patients died every year,
ranking the third place of cancer-related death. The treatments of HCC are mainly including: liver transplantation (LT), hepatic resection (HR), radiofrequency ablation (RFA), transarterial chemoembolization (TACE) and so on. Although the LT is the optimal treatment for the patients with HCC in early stage,
especially for the patients within Milan criteria or University of California San Francisco (UCSF) criteria, the shortage of organs limits its feasibility.
UCSF criteria is one of the criteria for LT and the patients who met UCSF criteria could perform surgical resection. However, the study of HR for the patients within UCSF is limited. Terence C et al.
reported the patients meeting UCSF criteria undergoing HR could achieve a comparable long-term outcome comparing with Milan criteria. But they didn't analyze the different type of the tumor (single and multiple) and the sample was small. As the previous studies mentioned that the multinodular HCC was an important factor to predict the patients with microvascular invasion (MVI) resulting a poor prognosis.
Postoperative adjuvant transcatheter arterial chemoembolization for resectable multiple hepatocellular carcinoma beyond the Milan criteria: a retrospective analysis.
Validated nomogram for the prediction of disease-free survival after hepatectomy for hepatocellular carcinoma within the Milan criteria: individualizing a surveillance strategy.
but these studies didn't compare the long-term outcomes to single tumor. To the best of our knowledge, the surgical long-term outcomes of single and multiple tumors is unclear, especially for the tumors within UCSF criteria, which needs a further investigation. Furthermore, single tumor and multiple tumors also have different MVI status, MVI-positive and MVI-negative, which was a challenge for surgeons to make the decision.
UCSF criteria contained Milan criteria, meaning more patients would get benefits if we could figure out the independent risk factors for HCC within UCSF criteria, and recommend the optimal method to treatment with tumor within UCSF criteria. Thus, the study aims to compare the long-term outcomes of different tumor type within UCSF criteria and to find an effective treatment for the patients within UCSF criteria.
Patients and methods
This study was approved by the West China Hospital Ethics Committee and performed in accordance with the ethical guidelines of the Declaration of Helsinki.
Patients
The study was retrospect the patients with HCC in our center, department of liver surgery and liver transplantation center, West China Hospital of Sichuan University, from January 2005 to January 2020. The patients were enrolled as followings: 1) Age >18 years old, including male and female; 2) Meeting UCSF criteria; 3) Without other tumors therapy history, especially the recurrence of HCC; 4) Without other fatal disease, such as heart disease and respiratory insufficiency; 5) didn't receive the treatment of LT; 6) the Child-Pugh class within A. If the patient could not meet one of the followings, he would be excluded.
Methods
The patients were divided into two groups (single tumor group and multiple tumors group). The univariate analysis and cox proportional hazards model were carried out to identify the independent risk factors affecting the survival of HCC within UCSF. Finally, we performed the neural network analysis for independent risk factors to analyze the different situation for HCC within UCSF criteria.
Definitions and diagnostic criteria
The UCSF criteria was defined as a single tumor with the diameter ≤6.5 cm or the maximum diameter of 2–3 tumors ≤4.5 cm and the total tumor diameter ≤8.0 cm in patients without extrahepatic manifestations and vascular invasion.
The patients with HCC and MVI were diagnosed by a histopathological examination after hepatic resection.
Anatomical resection (AR) was followed the Glisson's Pedicles to resection the liver lesion. And local resection was also called non-anatomical resection (NAR), defined as the hepatic resection was focused on resection the tumor completely and left the volume of liver enough possibly to avoid the post-operative liver failure, without concerning about the Glisson's Pedicles.
All patients were informed about the treatments including LT, HR, RFA and TACE. The liver function, blood test, coagulation function and the imaging examination were under reviewed by the surgeons with more than 5-year experience in hepatectomy.
The procedures in HR
Surgical procedures in HR were similar to the previous studies.
Radiofrequency ablation combined with transarterial chemoembolization versus hepatectomy for patients with hepatocellular carcinoma within Milan criteria: a retrospective case–control study.
Briefly, surgeons opened the abdomen standing on the right side of the patient with a right subcostal incision, and extending the incision along the mid-line of the sternum. Every patient undergoing HR would have an Intraoperative ultrasonography to confirm the tumors boundaries and identify the relationship of the portal vein, hepatic artery and hepatic vein to give a guide line to hepatectomy, especially for AR, and to find whether there existed another lesion that we could not judge from imaging by naked eye. The AR was performed following to the Glisson's Pedicles and based on the patients' liver reserve function. When the surgeons were performing the AR, they would try to clap the artery or vein of liver to identify the liver segment. Taking the right lobe liver resection as an example, surgeons would clap the right branch of portal vein and hepatic artery to find an ischemic line to assist them in performing hepatectomy. For patient with multiple tumors, if the tumor located near lobe, the surgeon would perform AR. For example, the tumor located at VI segment and VII segment, respectively, the surgeon would perform AR.
Follow-up and treatment of recurrence
All patients consulted the doctor in the outpatient clinic after liver resection, routinely. A-fetoprotein (AFP) and hepatitis B virus deoxyribonucleic acid (HBV DNA) measures and abdominal ultrasonography were performed every 3 months. Contrast enhanced CT scans were performed every 6 months. When the recurrence was difficult to diagnose, contrast enhanced MRI or ultrasonography was performed. The patients with recurrence were treated as following: re-resection, RFA, LT, TACE and chemotherapy.
Statistical analysis
SPSS 22.0 statistical software (SPSS Inc, Chicago, IL, USA) was used to analyze the relevant data. The categorical data were presented as the number (percent) and compared using Pearson chi-square or Fisher's exact test. The continuous variables were expressed as the mean value ± SD and were analyzed by the T test. The overall survival (OS) rates and recurrence-free survival (RFS) rates were estimated by the Kaplan–Meier method, and the differences between groups were determined by the log-rank test. A Cox proportional hazards model was used to test potential predictors of survival. A 2-tailed P < 0.05 was considered statistically significant. The Neural network analysis was used to clarify the affection of the independent risk factors on OS and RFS rates. The neural network analysis exited randomness which couldn't be avoided, because the initial parameters were random at the beginning every time, meaning we could get thousands of predicting models. Certainly, we have performed the neural network analysis more than 100 times and picked out one of results in the maximal of AUC.
Results
The baseline of the patients
The study enrolled 1006 patients with HCC within UCSF criteria who had undergone HR, including 744 patients with single tumor and 262 patients with multiple tumors. Among the patients about 859 were male and 147 were female. Table 1 showed the characteristic of patients in two types of the tumor. The significant difference between the groups was just blood platelet (PLT) count (111.39 ± 53.14 versus 120.41 ± 56.59, p = 0.024). But the levels of PLT between the groups was in the normal range. And there were no deaths within 30 days after operation.
The median survival time of single and multiple tumors were 62.1 (range, 4.9–88.1) months and 48.0 (range, 4.0–83.0) months, respectively. During the follow-up period, 340 (45.70%) patients have passed away in single tumor group when 175 (66.79%) patients have died in multiple tumors group. The 1-, 3-, and 5-year OS rates in single tumor group were significantly higher than multiple tumors group (95.0%, 73.2% and 52.3% versus 93.9%, 69.7% and 38.0%, p < 0.001, Fig. 1a). There were 416 (55.91%) patients with tumor recurrence in single tumor group and 186 (70.99%) patients in multiple tumors group. The 1-, 3- and 5-year RFS rates were 90.3%, 60.7%, and 40.1% in single group versus 83.4%, 50.7% and 23.8% in multiple tumors group, respectively (p < 0.001, Fig. 1b).
Figure 1The OS and RFS for patients who were enrolled in this study. (a) The 1-, 3- and 5-year OS for patients with single and multiple tumor (95.0%, 73.2% and 52.3% versus 93.9%, 69.7% and 38.0%). (b) The 1-, 3- and 5-year RFS for patients with single and multiple tumor (90.3%, 60.7%, and 40.1% versus 83.4%, 50.7% and 23.8%). (c) The 1-, 3- and 5-year OS for patients in the anatomic resection and local resection (94.8%, 77.7% and 56.3% versus 94.7%, 66.3% and 38.4%). (d) The 1-, 3- and 5-year RFS for patients in the anatomic resection and local resection (91.0%, 66.9% and 41.8% versus 85.3%, 47.7% and 28.5%). (e) The 1-, 3- and 5-year OS for patients with MVI-positive and MVI-negative (92.1%, 62.5% and 34.8% versus 95.1%, 75.1% and 52.2%). (f) The 1-, 3- and 5-year RFS for patients with MVI-positive and MVI-negative (84.7%, 44.6% and 25.6% versus 89.5%, 61.8% and 39.1%).
According to univariate and multivariate analysis (Table 2), we found that the independent prognostic factors were not only the number of tumors, but also including anatomic resection and MVI (Fig. 1c–f).
Table 2Univariate and multivariate analysis of the prognostic factors for survival and recurrence
The neural network analysis for independent risk factors
The neural network analysis was conducted by independent risk factors (tumor type, anatomic resection and MVI) and the results showed that MVI was the most important risk factor affecting the patients' OS rates and RFS rates (Figs. 2d and 3d ). And the AUC was 0.672 and 0.635 of these predicted models, respectively. From Fig. 2a, we could discover that the independent risk factors have the different effect in different styles of UCSF criteria. As the showing of Fig. 2b, we found that anatomical resection on the patient with MVI-negative and single tumor had a better long-term outcome than non-anatomical resection, compared H (1) with H (3) and H (4/6) (1.243–0.757 > 0 > 0.995–1.005), and the results have shown in Table 3. Absolutely, the tumor located in different segments was an obstruct to perform anatomical resection, thus, the proportion of anatomical resection and non-anatomical resection was difference in H (2), H (5) and H (8). Following H (2) and H (5), it conveyed the information that when the patient was in multiple tumors with MVI-negative status, anatomical resection was better than non-anatomical resection (0.488–0.512 > 0.235–0.765, Table 3). On the other hand, few patients were in multiple tumors with MVI-positive in this study, it was difficult to analyze the effect of anatomical resection alone (H (7), H (8) and H (9)) in neural network analysis. However, from H (7) and H (9), we proposed the hypothesis that the number of tumors was the significantly risk factor affecting the OS rates in MVI-positive, and the resection methods also affected the OS rates in H (8), though the difference wasn't significant (Table 3). Similarly, following Fig. 3, we found that the patient with MVI-negative and single tumor could get more benefit from anatomical resection than non-anatomical resection (compared H (1) with H (2), H (4) and H (5), Table 3). Surprisingly, the patient with MVI-negative and multiple tumors couldn't prevent tumor recurrence form purely anatomical resection or non-anatomical (H (6) and H (7)), however, anatomical resection was stronger to restrain tumor recurrence (Table 3). As to the multiple tumors with MVI-positive, the patient undergone anatomical resection could have a lower risk of tumor recurrence than non-anatomical resection (H (8) and H (11)), however, the difference wasn't significant (Table 3). From H (3) and H (10), we could discover that the MVI-positive was the most important risk factor of tumor recurrence in multiple tumors.
Figure 2The Neural network analysis for survival within UCSF.
In the current study, we compared the surgical outcomes meeting the UCSF criteria by retrospecting a large cohort of HCC patients. We found that the patients with single tumor had better OS rates and RFS rates after HR than the patients with multiple tumors (Fig. 1a and b). According to Kaplan–Meier analysis and multivariate cox modeling analysis, the number of tumors (single tumor, Fig. 1a and b), the methods of HR (anatomic resection, Fig. 1c and d) and MVI (negative, Fig. 1e and f) were independent better prognosis factors for the patients (Table 2). Previous studies
Validated nomogram for the prediction of disease-free survival after hepatectomy for hepatocellular carcinoma within the Milan criteria: individualizing a surveillance strategy.
had reported that the number of nodules was an independent risk factor of recurrence, which was mainly caused by metastatic recurrence from the main tumor via the portal system.
Multiple tumors mean that the tumor would have invaded different sites of portal vein system, especially for the tumors located in different segments. Tumors located in different hepatic areas, meaning tumors were more possible intrahepatic spread for the tumor in different branch of portal vein resulting the recurrence higher and poor outcomes. And the anatomical resection couldn't resect the potential invaded segments completely so that MVI might lose its function in predicting for these patients with multiple tumors. However, from neural network analysis, we found that multiple tumors with MVI-negative could receive a better outcome, and the outcomes have shown in Table 3. Generally, the patients with multiple tumors may be metastatic by the portal system, though the surgical margin was informed the MVI-negative from histopathological examination. Therefore, the risk of recurrence in the patients with multiple tumors was higher, which was the major risk factor affecting the survival.
Comparison of the outcomes between an anatomical subsegmentectomy and a non-anatomical minor hepatectomy for single hepatocellular carcinomas based on a Japanese nationwide survey.
had shown that anatomical resection had better surgical outcomes than the non-anatomical resection. Therefore, the patient with single tumor has a better outcome in anatomical resection than local resection (Table 3). However, the anatomical resection for multiple tumors with MVI-positive was no significant difference compared to non-anatomical resection (Table 3). This was mainly correlated with the undetected invaded tumor, which has been mention above. Meanwhile, Matteo Donadon et al. had reported when the patients with multi-nodular hepatocellular carcinoma and the maximum tumor diameter less than 6 cm underwent hepatectomy, they could be better in survival.
As for the resection of multiple tumors, the previous studies had pointed out that resection could provide an acceptable long-term outcome for the patients with multi-nodular HCC.
A snapshot of the effective indications and results of surgery for hepatocellular carcinoma in tertiary referral centers: is it adherent to the EASL/AASLD recommendations?: an observational study of the HCC East-West study group.
Thus, for the patients with multi-nodular HCC, hepatectomy could be still as the first-line treatment. And following neural network analysis, the anatomical resection should be as the first methods of hepatectomy.
Absolutely, MVI status was of great important for the patients within UCSF criteria. MVI was an important risk factor of recurrence.
Postoperative adjuvant transcatheter arterial chemoembolization for resectable multiple hepatocellular carcinoma beyond the Milan criteria: a retrospective analysis.
Validated nomogram for the prediction of disease-free survival after hepatectomy for hepatocellular carcinoma within the Milan criteria: individualizing a surveillance strategy.
In our study, the patients with MVI-negative had a better outcome than MVI-positive. MVI was correlated with the tumors' diameter, and the previous findings figured out the tumors' diameter more than 4 cm was likely with MVI,
but the tumors' diameter more than 4 cm had no significant difference in survival and recurrence in this study. Furthermore, a survey was carried by Yan–Yan Wang et al. shown that the tumors' diameter more than 7 cm could be more accurate to predict the prognosis of HCC patients.
However, the relationship of MVI and tumors' diameter weren't analyzed in the current study, which needed a further search in future.
There were several factors affecting the long-term outcomes of HCC patients. The high level of AFP has been proved to have more aggressive behaviors in previous findings,
Alpha-fetoprotein and tumour size are associated with microvascular invasion in explanted livers of patients undergoing transplantation with hepatocellular carcinoma.
In the current study, the level of AFP (≥400 ng/mL) has no significant difference. The main reason was that we didn't analyze the AFP in different levels, respectively. The hepatic B virus (HBV) infection was the mainly reason for patients with HCC in China, and patients with HBV were more likely to develop into cirrhosis, which was also a risk factor in recurrence and bad prognosis.
The Cancer of the Liver Italian Program (CLIP) investigators A new prognostic system for hepatocellular carcinoma: a retrospective study of 435 patients.
The patients with HBV were no significant influence in both groups. This could explain why we should take antiviral therapy to control the viral replication and to slow down the progress of cirrhosis. And the other risk factor was gene,
which has been reported that p53-and AFP-positive were predictors for poor prognosis of HCC after hepatic resection. It is a shortcoming that we didn't do genetic testing between these patients.
The predicted models of neural network analysis for HCC within UCSF was credible with the AUC was stably in neural network analysis. The patient with HCC within UCSF might have many different clinical characters. Such as, the patient would display in multiple tumors with MVI-positive, multiple tumors with MVI-negative, single tumor with MVI-positive and single tumor with MVI-negative. Facing the different clinic features, the patient would obtain different benefits from different hepatic resection, anatomical resection or non-anatomical resection. Taking the patient with single tumor could get more benefit from anatomical resection, which was mentioned above, as an example. However, whether all patients with single tumor should have a treatment of anatomical resection, it was unclear, though it was significant in Table 3. Meanwhile, when we combined the results of neural network analysis in H (3), H (4) and H (6), we could predict that anatomical resection was better than non-anatomical resection in single tumor with MVI-negative. From neural network analysis, we could discover that the MVI was the most important risk factor affecting the OS rates and RFS rates, especially for multiple tumors. Generally, it's difficult to identify the situation of MVI before hepatectomy, however, with the techniques and radiology skills development, the radiologist could predict the MVI based on radiomics.
Therefore, combining with radiomics to estimate the MVI status and to make a decision on hepatectomy for multiple tumors within UCSF could reduce the posthepatectomy liver failure (PHLF) rates, which was because the patient with multiple tumors with MVI-negative could get a better outcome in anatomical (Table 3), and its incidence was about 7%.
Although we could measure the liver reserve function directly by ICG test, and reconstruct the liver 3D model by the technique of 3D reconstruction, which could provide a directly view of tumor volume and non-tumor liver volume, providing an evidence to surgeon before hepatectomy,
The technique of 3D reconstruction combining with biochemistry to build an equivalent formula of indocyanine green (ICG) clearance test to assess the liver reserve function.
we still should take the methods of hepatectomy into consideration. Because the anatomical resection would spend more time to complete the tumor resection and add the bleeding risk, which would add the risk to intraoperative hypothermia, leading to infection of incision and delaying the postoperative rehabilitation,
especially for the multiple tumors with MVI-negative. Thus, the neural network analysis predicted models could provide a reference for the surgeon to make a decision on hepatectomy.
The study also has its other limitations. Firstly, it was a retrospective study only a single center's experience with small sample included and the predicted models needed further research to prove the application. Secondly, we didn't compare the outcomes with liver transplantation, Milan criteria and radiofrequency ablation for the patients. Finally, the retrospective study has the biases, especially for the long-term outcomes follow-up. However, to the best of our knowledge, this study was the first research focusing on the outcomes of HR in the patients within UCSF criteria, analyzing the single and multiple tumors' outcomes, and by building predicted models to analyze the suitable treatment for these patients. Based on our study findings, the OS rates and RFS rates of HR were higher in the single tumor group than in the multiple tumors group.
Conclusion
The surgical resection obtained a better long-term outcome for single tumor within UCSF criteria. The anatomic resection should be applied to the patient within UCSF criteria, especially for the patient was in single tumor with MVI-negative.
Author's contribution
Author Contributions: Study conception and design: LJ and JYY; Acquisition of data: JLZ, WX and YH; Collected image picture and data: JSY and JLZ; Analysis and interpretation of data: LJ and JLZ; Drafting of manuscript: JLZ and NW; Critical revision: LJ; XYP and JLZ contributed in statistical analysis. All authors have read and approved the manuscript.
Funding
This study was supported by grants from the National Sciences (2012ZX10002-016) and Technology Major Project of China(2012ZX10002-017), provided by JY Yang, providing to collect the data. And the National Natural Science Foundation of China (81400636), Sichuan Province Key Research and Development Project (2019YFS0203), and the Key Project of Clinical Research Incubation in West China Hospital of Sichuan University (2020HXFH028), provided by L Jiang, to supported the study designing, data analysis and interpretation, writing the manuscript and scientific language editing, respectively.
Consent for publication
Not applicable.
Availability of data and materials
The data sets used during the current study are available from the corresponding author on reasonable request.
Competing interests
The authors declare that they have no competing interests.
Acknowledgments
Thanks for the funding supported by JY Yang and LJ. Author's information.
References
Zhao C.
Nguyen M.H.
Hepatocellular carcinoma screening and surveillance:practice guidelines and real-life practice.
Postoperative adjuvant transcatheter arterial chemoembolization for resectable multiple hepatocellular carcinoma beyond the Milan criteria: a retrospective analysis.
Validated nomogram for the prediction of disease-free survival after hepatectomy for hepatocellular carcinoma within the Milan criteria: individualizing a surveillance strategy.
Radiofrequency ablation combined with transarterial chemoembolization versus hepatectomy for patients with hepatocellular carcinoma within Milan criteria: a retrospective case–control study.
Validated nomogram for the prediction of disease-free survival after hepatectomy for hepatocellular carcinoma within the Milan criteria: individualizing a surveillance strategy.
Comparison of the outcomes between an anatomical subsegmentectomy and a non-anatomical minor hepatectomy for single hepatocellular carcinomas based on a Japanese nationwide survey.
A snapshot of the effective indications and results of surgery for hepatocellular carcinoma in tertiary referral centers: is it adherent to the EASL/AASLD recommendations?: an observational study of the HCC East-West study group.
Alpha-fetoprotein and tumour size are associated with microvascular invasion in explanted livers of patients undergoing transplantation with hepatocellular carcinoma.
The technique of 3D reconstruction combining with biochemistry to build an equivalent formula of indocyanine green (ICG) clearance test to assess the liver reserve function.
00040-0&id=gr1.jpg){kind=link}
00040-0&id=gr2.jpg){kind=link}
00040-0&id=gr3.jpg){kind=link}