Wang X, Zhu MX, Wang JF, Liu P, Zhang LY, Zhou Y, Lin XX, Du YD, He KL. Multivariable prognostic models for post-hepatectomy liver failure: An updated systematic review. World J Hepatol 2025; 17(4): 103330 [DOI: 10.4254/wjh.v17.i4.103330]
Corresponding Author of This Article
Kun-Lun He, Medical Big Data Research Center, Chinese PLA General Hospital, No. 28 Fuxing Road, Haidian District, Beijing 100853, China. hekunlun301dr@163.com
Research Domain of This Article
Gastroenterology & Hepatology
Article-Type of This Article
Systematic Reviews
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Xiao Wang, Ying-Dong Du, Department of Hepatobiliary Surgery, Chinese PLA 970th Hospital, Yantai 264001, Shandong Province, China
Xiao Wang, Ming-Xiang Zhu, Pan Liu, You Zhou, Xi-Xiang Lin, Kun-Lun He, Medical Big Data Research Center, Chinese PLA General Hospital, Beijing 100853, China
Ming-Xiang Zhu, Medical School of Chinese PLA, Chinese PLA General Hospital, Beijing 100853, China
Jun-Feng Wang, Division of Pharmacoepidemiology and Clinical Pharmacology, Utrecht Institute for Pharmaceutical Sciences, Utrecht University, Utrecht 358 4CG, Netherlands
Li-Yuan Zhang, China National Clinical Research Center for Neurological Diseases, Beijing 100853, China
You Zhou, School of Medicine, Nankai University, Tianjin 300071, China
Co-corresponding authors: Ying-Dong Du and Kun-Lun He.
Author contributions: Wang X, Zhu MX, Wang JF, Du YD, and He KL conceptualized and designed the research; Wang X and Wang JF developed the literature search strategy; Wang X and Zhu MX completed the literature screening and data extraction; Wang X, Zhu MX, Liu P, Zhang LY, Zhou Y, and Lin XX organized and counted the extracted data; Wang X and Zhu MX completed the quality assessment and performed the statistical analysis; Wang X, Zhu MX, Du YD and He KL wrote the paper; All the authors have read and approved the final manuscript. Wang X proposed, designed and conducted literature screening, completed the quality assessment, performed data analysis, and prepared the first draft of the manuscript. Zhu MX was also mainly responsible for data extraction, quality appraisal, and data analysis. Both authors have made crucial and indispensable contributions towards the completion of the project and thus qualified as the co-first authors of the paper. Both Du YD and He KL have played essential and indispensable roles in the study design, data interpretation, and manuscript preparation as the co-corresponding authors. He KL applied for and obtained the funds for this research project. He KL conceptualized, designed, and supervised the whole process of the project. He revised and submitted the early version of the manuscript with a focus on the predictive performance, quality assessment, and clinical application of current PHLF models. Du YD was instrumental and responsible for data re-analysis and re-interpretation, figure plotting, preparation and submission of the current version of the manuscript with a new focus on the PHLF models based on AI technology and their further development trends. This collaboration between He KL and Du YD is crucial for the publication of this manuscript and other manuscripts still in preparation.
Supported by The Science and Technology Innovation 2030 - Major Project, No. 2021ZD0140406.
Conflict-of-interest statement: The authors declare that they have no conflict of interest.
PRISMA 2009 Checklist statement: The authors have read the PRISMA 2009 Checklist, and the manuscript was prepared and revised according to the PRISMA 2009 Checklist.
Open Access: This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: https://creativecommons.org/Licenses/by-nc/4.0/
Corresponding author: Kun-Lun He, Medical Big Data Research Center, Chinese PLA General Hospital, No. 28 Fuxing Road, Haidian District, Beijing 100853, China. hekunlun301dr@163.com
Received: November 18, 2024 Revised: February 28, 2025 Accepted: March 21, 2025 Published online: April 27, 2025 Processing time: 161 Days and 5.4 Hours
Abstract
BACKGROUND
Partial hepatectomy continues to be the primary treatment approach for liver tumors, and post-hepatectomy liver failure (PHLF) remains the most critical life-threatening complication following surgery.
AIM
To comprehensively review the PHLF prognostic models developed in recent years and objectively assess the risk of bias in these models.
METHODS
This review followed the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guideline. Three databases were searched from November 2019 to December 2022, and references as well as cited literature in all included studies were manually screened in March 2023. Based on the defined inclusion criteria, articles on PHLF prognostic models were selected, and data from all included articles were extracted by two independent reviewers. The PROBAST was used to evaluate the quality of each included article.
RESULTS
A total of thirty-four studies met the eligibility criteria and were included in the analysis. Nearly all of the models (32/34, 94.1%) were developed and validated exclusively using private data sources. Predictive variables were categorized into five distinct types, with the majority of studies (32/34, 94.1%) utilizing multiple types of data. The area under the curve for the training models included ranged from 0.697 to 0.956. Analytical issues resulted in a high risk of bias across all studies included.
CONCLUSION
The validation performance of the existing models was substantially lower compared to the development models. All included studies were evaluated as having a high risk of bias, primarily due to issues within the analytical domain. The progression of modeling technology, particularly in artificial intelligence modeling, necessitates the use of suitable quality assessment tools.
Core Tip: Currently, with the exploration of new meaningful predictive variables and modeling methods, post-hepatectomy liver failure prognostic models are developing rapidly. However, for existing models, the issues of the analysis domain are the main reason for being assessed as having a high risk of bias. Therefore, researchers should also concentrate on validating existing models to realize their clinical application potential. Additionally, the development of new tools for evaluating model quality may be necessary, and future efforts must follow strict guidelines to maintain the integrity of the models.
Citation: Wang X, Zhu MX, Wang JF, Liu P, Zhang LY, Zhou Y, Lin XX, Du YD, He KL. Multivariable prognostic models for post-hepatectomy liver failure: An updated systematic review. World J Hepatol 2025; 17(4): 103330
Partial hepatectomy remains the principal treatment approach for liver tumors, particularly hepatocellular carcinoma (HCC)[1,2], largely due to the limitations of other treatment options such as liver transplantation or ablation surgery[3,4]. With advancements in medical technology and improvements in perioperative management, the criteria for surgical intervention have broadened significantly[5]. Although studies have shown that aggressive surgical resection leads to favorable prognostic outcomes[6,7], post-hepatectomy liver failure (PHLF) continues to be the most severe life-threatening complication following surgery[8,9]. Hence, accurately assessing the risk of postoperative liver function decompensation is crucial in the management of liver tumors.
The reported incidence of PHLF varies significantly across studies[10,11]. Preoperative liver function assessments or predictive models are crucial for clinicians to make informed treatment decisions and to evaluate if surgical resection will deliver the anticipated results[12]. In addition to traditional scores like the Child-Pugh score and the Model for End-Stage Liver Disease (MELD) score[13,14], numerous enhanced predictive models have been developed, incorporating various predictive variables to more accurately assess the risk of PHLF. However, many of these models suffer from methodological weaknesses, lack detailed descriptions, or do not undergo rigorous external validation, which casts doubt on their clinical utility[15].
Researchers continue to explore potential predictive factors to more effectively assess the risks and benefits of surgery. With the ongoing incorporation of artificial intelligence (AI) into various aspects of the medical field, including medical imaging[16], numerous AI-based models have been developed to aid clinicians in their decision-making processes. The identification of new, meaningful predictive variables and the applied integration of AI techniques could theoretically enhance the efficiency of these models. However, there are limited prospective reports on the clinical application of PHLF prediction models, and both the development and performance of these models require further scrutiny.
Although a prior review has evaluated prediction models for PHLF[17], a substantial number of new models incorporating novel predictive factors have emerged and been documented since then. Our study has comprehensively collated these recent multivariable prediction models for PHLF and impartially appraised the risk of bias. Compared to the original review, our goal is to enhance the current understanding of modeling techniques, confirm advancements in model quality, keep summarizing practical challenges, and investigate future possibilities for PHLF prediction models. This approach aims to provide a more current and detailed overview, ensuring that the latest innovations and their impacts on clinical practice are well understood.
MATERIALS AND METHODS
We conducted a systematic review to identify multivariate risk prediction models for the occurrence of PHLF, focusing on both the development and validation of these models. Our review adhered to the guidelines of the Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies[18] and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses[19] which guided the reporting process of our review. Ethical approval for the study was obtained from the Ethics Committee of the First Medical Center of the General Hospital of the People's Liberation Army in compliance with the Declaration of Helsinki.
Search strategy
We conducted a thorough literature search in two stages to identify relevant published studies. Initially, we performed an electronic search across the EMBASE, Ovid MEDLINE, and Web of Science Core Collection databases, restricting our focus to studies published between November 2019 and December 2022. The keywords and search terms were derived from the Supplementary materials provided in the systematic review by Mir et al[17] (Supplementary Table 1). Subsequently, in March 2023, we manually screened references and cited literature from all studies included to ensure the comprehensiveness and currency of our review.
Inclusion and exclusion criteria
A multivariate prognostic model that used at least two predictors to develop and/or validate the risk of PHLF, regardless of targeted patients, data sources, or study design, was considered for inclusion in our study. We established exclusion criteria as follows: (1) Studies not published in English; (2) Studies solely focused on identifying predictors; (3) Studies whose outcomes did not encompass PHLF; (4) Studies that evaluated the impact of only one factor without employing a model; (5) Studies with unsuitable analytical objectives, such as those not aimed at prognostication or those primarily exploring novel statistical techniques; and (6) Studies from which data were unobtainable.
Study screening and selection
All studies retrieved through our search strategy were imported into EndNote to identify and eliminate duplicates. Initially, titles and abstracts were screened to pinpoint relevant studies. Subsequently, full-text screening determined the final inclusion of eligible studies. The entire process of article screening and selection was conducted independently by two reviewers (Wang X and Zhu MX). Any disagreements between the reviewers were discussed until a consensus was reached.
Data extraction
Data were extracted across several categories including basic article information, study characteristics, population demographics, dataset details, model development, performance, evaluation, and the clinical application of the model. An information extraction form was designed with specific domains and items, and data from all included articles were extracted twice for accuracy. In cases where multiple models were presented in a single article, only the model demonstrating the highest predictive performance was selected for inclusion. Two reviewers (Wang X and Zhu MX) independently performed the data extraction for each article. Disagreements were resolved through discussion, and when necessary, a third reviewer (Wang JF) made the final decision.
Quality appraisal
We utilized the Prediction Model Risk of Bias Assessment Tool (PROBAST)[20], a tool specifically designed for evaluating diagnostic and prognostic prediction model studies, to assess the quality of each article included in our review. Responses to each question within the tool were categorized as “yes/Y”, “probably yes/PY”, “probably no/PN”, “no/N”, “no information/NI”, or “no application/NA”. Our analysis thoroughly evaluated the risk of bias across all four domains and the applicability across the first three domains of each study, assigning a rating of high (-), low (+), or unclear (?). Two reviewers (Wang X and Zhu MX) independently assessed the risk of bias and applicability, ultimately achieving consensus on each evaluation.
Statistical analysis
We conducted a narrative synthesis to encapsulate the models using descriptive statistics and data visualization techniques. Due to the heterogeneity among the selected studies, categorical data were summarized using frequencies and percentages, while the distribution of continuous data was evaluated and presented using the median and interquartile range (IQR). Additionally, graphical charts and tables were employed to display the results. All statistical analyses were carried out using R software, version 4.2.1.
RESULTS
After screening and eligibility assessment, a total of thirty-four eligible studies were included in our systematic review and quality appraisal (Figure 1). Detailed information on the included studies and the data extracted from them can be found in Supplementary Table 2. Additionally, a basic statistical analysis of the characteristics of the included models is provided in Supplementary Table 3.
Figure 1 Flow chart of the included studies for analysis.
PHLF: Post-hepatectomy liver failure.
Characteristics of the included studies
The quantity of models included in our review is comparable to that of the previous review[17], reflecting the rapid advancement in research on PHLF prediction models in recent years. Of the studies included, thirty-two developed new multivariate risk prediction models for PHLF. Among these, 19 (59.4%) conducted internal validation, 11 (34.4%) carried out formal external validation, and only 6 (18.8%) completed both internal and formal external validations. The remaining two studies focused on validating and assessing the performance of existing PHLF models.
The study cohorts originated from six different countries, predominantly from China, which represented the majority (24 out of 34, or 70.5%). The impact factor (IF) of the articles included in our study varied widely, ranging from 1.187 to 30.083, with a median of 4.413 and an IQR from 3.282 to 5.738. Detailed specifics of each study are provided in Tables 1 and 2, which also supported our evaluation of the models[21-56].
Table 1 Characteristics of studies included for analysis, n (%).
Population characteristics, data sources, and outcomes
The majority of the study populations (30/34, 88.2%) consisted of HCC patients, with a small number involving perihilar cholangiocarcinoma (PHCC) patients (2/34, 5.8%) or having no specific tumor restrictions (2/34, 5.8%). Over two-thirds of the studies (23/34, 67.6%) did not impose any age restrictions on participants. Almost all of the models (32/34, 94.1%) utilized private data sources, sourced from either single or multiple centers. The sample sizes in these studies varied from 101 to 7376 (median 363).
The International Study Group of Liver Surgery (ISGLS) definition and grading system[57] (28/34, 82.4%) was the most commonly used reference standard for the development or validation of PHLF prediction models, followed by the "50-50" Criteria[58] (4/34, 11.8%). The number of PHLF events reported in each study varied from 15 to 622 (median 81). Additionally, the percentage of PHLF events relative to the total sample size in each study ranged from 3.1% to 57.6% (median of 15.1%). Details of these findings are available in Table 1 and Supplementary Table 3.
Predictive variables of the PHLF models
Predictive variables were categorized into five groups: Patient or tumor characteristics/comorbidities, surgical records, indocyanine green (ICG)-based measures, blood test-based measures, and imaging-based measures. Total bilirubin emerged as the most frequently utilized predictive variable, as depicted in Figure 2. We also analyzed and illustrated the frequency and co-use of these variable types in model development, presented in Figure 3. Compared to a previous review[17], our included studies revealed the emergence of several new significant predictive variables derived from various clinical data types, as outlined in Supplementary Table 4. There were ten different combinations of variable types identified, with the most prevalent combination being “characteristics of the patient or tumor/comorbidities”, “surgical record”, and “measures based on blood tests”, accounting for 26.5% (9/34). Furthermore, techniques have been continually refined based on the combinations of predictive variables, including the adoption of AI techniques[27,40].
Figure 3 Types and frequencies of prediction variable combinations used in establishing post-hepatectomy liver failure prediction models.
ICG: Indocyanine green.
Statistical considerations and model performance
Table 2 displays the statistical methods utilized for the development and evaluation of models. Among thirty-two development studies, seventeen studies (53.1%) had reported employing a specific strategy for selecting variables during the modeling process. Regarding handling missing values, which was reported in twenty-seven studies (79.4%), the majority (16/34, 47.1%) excluded patients with incomplete data outright. One study utilized single imputation[25], another employed multiple imputation[46], and two studies did not address missing values at all[38,41]. The events-to-predictor variable (EPV) ratio for thirty-one development models varied from 4.8 to 69.1, except for one article that did not specify the number of events in the training set[52]. Among these, the distribution of models with an EPV of ≤ 10, between 10 and 20, and ≥ 20, was ten (32.3%), fourteen (45.2%), and seven (22.6%), respectively (Table 1).
Twenty-five studies (73.5%) included reports on the model calibration, and all these studies additionally provided assessments of discrimination. A variety of methods were used to report the calibration performance of the models. The discrimination was predominantly evaluated by comparing the area under the receiver operating characteristic curve (AUC) values. Among the nineteen studies that performed internal validation, the predominant method involved splitting the patient sample. Formal external validation was carried out in distinct patient populations across ten studies and in patients from different time periods in one study, as detailed in Table 2.
The AUC values for the training models varied from 0.697 to 0.956, with a median value of 0.845. Among the nineteen models that underwent internal validation, twelve (63.2%) achieved AUC values ranging from 0.8 to 0.9. Furthermore, of the eleven models that underwent external validation, six (54.5%) reported AUC values within the same range of 0.8 to 0.9. However, in the two studies that externally validated existing models, all reported AUC values were below 0.70 (Table 2 and Supplementary Table 3).
Assessment of risk of bias and applicability
Except for one study[23] that was assessed as having a high risk of bias in the predictor domain, all other studies were considered to have a low risk of bias concerning participants and outcomes. Nevertheless, the overall evaluation of the included studies indicated a high risk of bias, primarily due to concerns in the analysis domain. Each study was found to have a low concern regarding the applicability across all domains. Moreover, there was no significant correlation found between the IF of the journals and the risk of bias assessment outcomes of the models. A detailed summary of the risk of bias and applicability for each study is presented in Table 3 and Supplementary Table 5.
Table 3 Quality assessment of included studies using the prediction model risk of bias assessment tool.
This systematic review identified thirty-four models for predicting PHLF risk in liver tumors. Combined with the review by Mir et al[17], there is a clear increase in the number of studies over the past eight years. Initially, most studies originated from Japan, but China has recently emerged as the leading nation in developing PHLF models, as depicted in Figure 4. This trend can be attributed to two main factors: Firstly, PHLF remains a significant clinical concern that affects postoperative recovery and is a long-standing issue for medical professionals; secondly, China has the largest population of liver cancer patients[59].
Figure 4
Number of publications by year and country (including development and validation models).
Although half of the studies included were multi-center, nearly all were retrospective, failing to mitigate the risk of selection bias. Common practice in our modeling research was to include only complete datasets or exclude incomplete ones, yet distinctions between missing data completely at random and missing data at random, which are typically less problematic, were seldom made in the studies[60]. Due to the heterogeneity and diversity among clinical cases, the complete cases analyzed often do not fully represent or capture the characteristics of all participants. Multiple imputation, which uses repeated simulations to handle missing values, is now regarded as an effective strategy to address statistical losses and prevent selection bias[61,62]. Rigorous inclusion criteria and proper management of missing values are essential to prevent selection bias and are fundamental in choosing model parameters and developing high-performance models. Therefore, despite the potential difficulties and challenges, multi-center prospective studies are still crucial for developing robust multivariate PHLF prediction models.
Ongoing research in the field of PHLF model development continues to identify and incorporate new predictive variables that enhance model performance, as detailed in Supplementary Table 4. Imaging-related predictive variables have been increasingly utilized in model development. However, the wide variation in imaging techniques across different hospitals and the personal preferences of individual doctors call for a more thorough assessment of the universal applicability of these models. Among all the models reviewed, AI techniques were employed in the development of PHLF prediction models in only two studies, specifically using machine learning (ML)[40] and artificial neural networks[27]. However, AI models based on imaging have been successfully developed for the diagnosis and prognostic prediction of various diseases, demonstrating strong performance. Thus, identifying effective PHLF predictive variables and developing an image-based multimodal AI model represents a promising research avenue to enhance the generalizability and predictive efficiency of these models.
Although most studies have adopted the ISGLS criteria, the definition of PHLF varies widely. This variation in definitions leads to differing rates of occurrence, complicating the external validation of models and rendering comparisons of model performance unreliable. In addition, since many PHLF models are derived from retrospective cohort studies, it is essential to agree on a standardized definition of PHLF that minimizes the incidence of missing data as much as possible.
The model internal and external validation have improved significantly compared with previous review studies, which may be due to the inclusion of multi-center studies with large sample sizes. These models showed promising predictive abilities on their respective training or validation sets. However, the validation performance of these models is generally moderate, as indicated by AUC values below 0.7. The primary aim of research on predictive models is to ensure they are clinically applicable, making the process of validating these models potentially more critical than their initial development.
During our risk bias evaluation of the included studies, several issues were noted. First, there has been a notable improvement in the proportion of studies with low EPV rates compared to earlier reviews. However, the selection process for candidate predictors in most studies remains flawed. Typically, predictors are initially selected based on their statistical significance in univariate analysis, followed by a multivariate analysis to finalize the predictors for the model. This process equates the statistical significance of a single covariate with its predictive power and excludes valid potential predictors[63]. As a result, the risk of model overfitting and the presentation of misleading performance continue to be key concerns in assessing model effectiveness and practicality.
Secondly, the majority of the studies reviewed employed MLR to predict the binary outcomes of PHLF using short-term postoperative test results or complications recorded on or after the fifth day post-surgery. These studies often did not require comprehensive follow-up of patients, thus negating the need to address the complexities in the data, such as censoring, competing risks, and how control participants were sampled. However, prognostic models that factor in time variables, like survival time, necessitate the use of Cox regression for statistical modeling. As AI models continue to evolve in the medical field, selecting the appropriate modeling technique for different prognostic outcomes, based on the data characteristics, is crucial for designing valid research and enhancing the accuracy of models.
Lastly, certain assessment criteria in PROBAST are not suitable for AI models, particularly regarding the evaluation of model overfitting and the weighting of predictors. For image-based AI models, most criteria in the analysis section of PROBAST are also inappropriate. Although the Checklist for AI in Medical Imaging[64] has been advocated as a framework for evaluating these models, there is a clear need for a quality assessment tool specifically designed to gauge the risk of bias in AI models. This is an area that requires further attention and development.
As advancements continue in modeling techniques, including deep learning and ML, the integration of medicine and engineering becomes crucial for the clinical verification and implementation of prognostic models. It is essential to focus on the management of missing data, minimization of selection bias, selection of suitable statistical methods, and the creation of quality assessment tools specifically designed for AI models. To ensure that PHLF prediction models are clinically useful, standardized development and thorough validation must be prioritized. This approach will not only enhance the reliability of the models but also their applicability in real-world clinical settings.
Limitations
As an update to the previous review, there are several limitations of our study. Firstly, although we adopted the literature search strategy from the prior review and conducted additional screening, there is still a risk of missing relevant literature from other databases. Secondly, by restricting our inclusion to English-language studies, we might have overlooked significant research published in other languages. Thirdly, since the bulk of our study cohorts are based in China, regional variations in population characteristics could affect the effectiveness of the predictive variables used. Therefore, caution should be warranted when generalizing our findings to populations from different regions. Lastly, our study has not performed a rigorous quality assessment of AI models. With the increase of AI models and the update of quality evaluation tools, further exploration and summary are needed in future reviews.
CONCLUSION
As researchers persist in identifying new and meaningful predictive variables, it is equally crucial to concentrate on validating existing models to realize their potential clinical applications. The main reason these models are often assessed as having a high risk of bias stems from problems within the analysis domain. Thus, by addressing and mitigating analysis domain issues, researchers can reduce biases that compromise model performance. As scientific methodologies advance, there is a growing need for the development of new tools to evaluate model quality. Future research must comply with strict guidelines to ensure the integrity and quality of predictive models. Overall, ensuring adherence to comprehensive guidelines can promote the development of robust and effective predictive models, thus fostering better patient outcomes in clinical practice.
Footnotes
Provenance and peer review: Unsolicited article; Externally peer reviewed.
Peer-review model: Single blind
Specialty type: Gastroenterology and hepatology
Country of origin: China
Peer-review report’s classification
Scientific Quality: Grade A, Grade C
Novelty: Grade B, Grade C
Creativity or Innovation: Grade B, Grade C
Scientific Significance: Grade A, Grade C
P-Reviewer: Kobayashi K; Li S S-Editor: Li L L-Editor: A P-Editor: Zhao YQ
Mehta N, Bhangui P, Yao FY, Mazzaferro V, Toso C, Akamatsu N, Durand F, Ijzermans J, Polak W, Zheng S, Roberts JP, Sapisochin G, Hibi T, Kwan NM, Ghobrial M, Soin A. Liver Transplantation for Hepatocellular Carcinoma. Working Group Report from the ILTS Transplant Oncology Consensus Conference.Transplantation. 2020;104:1136-1142.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 39][Cited by in RCA: 145][Article Influence: 29.0][Reference Citation Analysis (0)]
Ho MC, Hasegawa K, Chen XP, Nagano H, Lee YJ, Chau GY, Zhou J, Wang CC, Choi YR, Poon RT, Kokudo N. Surgery for Intermediate and Advanced Hepatocellular Carcinoma: A Consensus Report from the 5th Asia-Pacific Primary Liver Cancer Expert Meeting (APPLE 2014).Liver Cancer. 2016;5:245-256.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 42][Cited by in RCA: 62][Article Influence: 6.9][Reference Citation Analysis (0)]
Kokudo T, Hasegawa K, Matsuyama Y, Takayama T, Izumi N, Kadoya M, Kudo M, Ku Y, Sakamoto M, Nakashima O, Kaneko S, Kokudo N; Liver Cancer Study Group of Japan. Survival benefit of liver resection for hepatocellular carcinoma associated with portal vein invasion.J Hepatol. 2016;65:938-943.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 237][Cited by in RCA: 360][Article Influence: 40.0][Reference Citation Analysis (0)]
Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Shamseer L, Tetzlaff JM, Akl EA, Brennan SE, Chou R, Glanville J, Grimshaw JM, Hróbjartsson A, Lalu MM, Li T, Loder EW, Mayo-Wilson E, McDonald S, McGuinness LA, Stewart LA, Thomas J, Tricco AC, Welch VA, Whiting P, Moher D. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.Syst Rev. 2021;10:89.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 2603][Cited by in RCA: 3992][Article Influence: 998.0][Reference Citation Analysis (33)]
Yugawa K, Maeda T, Nagata S, Shiraishi J, Sakai A, Yamaguchi S, Konishi K, Hashimoto K. Impact of aspartate aminotransferase-to-platelet ratio index based score to assess posthepatectomy liver failure in patients with hepatocellular carcninoma.World J Surg Oncol. 2022;20:248.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Reference Citation Analysis (0)]
Zhang D, Pan Y, Yang Z, Zeng H, Wang X, Chen J, Wang J, Zhang Y, Zhou Z, Chen M, Hu D. A Nomogram Based on Preoperative Lab Tests, BMI, ICG-R15, and EHBF for the Prediction of Post-Hepatectomy Liver Failure in Patients with Hepatocellular Carcinoma.J Clin Med. 2022;12:324.
[RCA] [PubMed] [DOI] [Full Text][Reference Citation Analysis (0)]
Prodeau M, Drumez E, Duhamel A, Vibert E, Farges O, Lassailly G, Mabrut JY, Hardwigsen J, Régimbeau JM, Soubrane O, Adam R, Pruvot FR, Boleslawski E. An ordinal model to predict the risk of symptomatic liver failure in patients with cirrhosis undergoing hepatectomy.J Hepatol. 2019;71:920-929.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 39][Cited by in RCA: 64][Article Influence: 10.7][Reference Citation Analysis (0)]
Wang J, Zheng T, Liao Y, Geng S, Li J, Zhang Z, Shang D, Liu C, Yu P, Huang Y, Liu C, Liu Y, Liu S, Wang M, Liu D, Miao H, Li S, Zhang B, Huang A, Zhang Y, Qi X, Chen S. Machine learning prediction model for post- hepatectomy liver failure in hepatocellular carcinoma: A multicenter study.Front Oncol. 2022;12:986867.
[RCA] [PubMed] [DOI] [Full Text][Cited by in RCA: 9][Reference Citation Analysis (0)]
Ye JZ, Mai RY, Guo WX, Wang YY, Ma L, Xiang BD, Cheng SQ, Li LQ. Nomogram for prediction of the international study Group of Liver Surgery (ISGLS) grade B/C Posthepatectomy liver failure in HBV-related hepatocellular carcinoma patients: an external validation and prospective application study.BMC Cancer. 2020;20:1036.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 10][Cited by in RCA: 4][Article Influence: 0.8][Reference Citation Analysis (0)]
Hobeika C, Guyard C, Sartoris R, Maino C, Rautou PE, Dokmak S, Bouattour M, Durand F, Weiss E, Vilgrain V, Beaufrère A, Sepulveda A, Farges O, Paradis V, Luciani A, Lim C, Sommacale D, Scatton O, Laurent A, Nault JC, Soubrane O, Ronot M, Cauchy F. Performance of non-invasive biomarkers compared with invasive methods for risk prediction of posthepatectomy liver failure in hepatocellular carcinoma.Br J Surg. 2022;109:455-463.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 10][Cited by in RCA: 9][Article Influence: 3.0][Reference Citation Analysis (0)]
Ding C, Jia J, Han L, Zhou W, Liu Z, Bai G, Wang Q. Developing and validating a nomogram based on skeletal muscle index and clinical scoring system for prediction of liver failure after hepatectomy.Front Oncol. 2023;13:1036921.
[RCA] [PubMed] [DOI] [Full Text][Reference Citation Analysis (0)]
Wang YY, Xiang BD, Ma L, Zhong JH, Ye JZ, Wang K, Xing BC, Li LQ. Development and Validation of a Nomogram to Preoperatively Estimate Post-hepatectomy Liver Dysfunction Risk and Long-term Survival in Patients With Hepatocellular Carcinoma.Ann Surg. 2021;274:e1209-e1217.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 23][Cited by in RCA: 50][Article Influence: 10.0][Reference Citation Analysis (0)]
Rahbari NN, Garden OJ, Padbury R, Brooke-Smith M, Crawford M, Adam R, Koch M, Makuuchi M, Dematteo RP, Christophi C, Banting S, Usatoff V, Nagino M, Maddern G, Hugh TJ, Vauthey JN, Greig P, Rees M, Yokoyama Y, Fan ST, Nimura Y, Figueras J, Capussotti L, Büchler MW, Weitz J. Posthepatectomy liver failure: a definition and grading by the International Study Group of Liver Surgery (ISGLS).Surgery. 2011;149:713-724.
[RCA] [PubMed] [DOI] [Full Text][Cited by in Crossref: 1224][Cited by in RCA: 1686][Article Influence: 120.4][Reference Citation Analysis (0)]
Cai J, Chen L, Zhang Z, Zhang X, Lu X, Liu W, Shi G, Ge Y, Gao P, Yang Y, Ke A, Xiao L, Dong R, Zhu Y, Yang X, Wang J, Zhu T, Yang D, Huang X, Sui C, Qiu S, Shen F, Sun H, Zhou W, Zhou J, Nie J, Zeng C, Stroup EK, Zhang X, Chiu BC, Lau WY, He C, Wang H, Zhang W, Fan J. Genome-wide mapping of 5-hydroxymethylcytosines in circulating cell-free DNA as a non-invasive approach for early detection of hepatocellular carcinoma.Gut. 2019;68:2195-2205.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited by in Crossref: 206][Cited by in RCA: 188][Article Influence: 31.3][Reference Citation Analysis (0)]
