PDF(1923 KB)
基于因果森林模型的腹腔镜肝血管瘤术式个体化选择：单中心回顾性研究
吴浩俊, 徐洪卫, 彭榆富, 文宁远, 羊宇波, 梁彬, 吴年恒, 魏永刚
中国实用外科杂志 ›› 2026, Vol. 46 ›› Issue (3) : 361-367.
PDF(1923 KB)
PDF(1923 KB)
基于因果森林模型的腹腔镜肝血管瘤术式个体化选择：单中心回顾性研究
Individualized selection of laparoscopic procedures for hepatic hemangioma based on a causal forest model: a single-center retrospective study
目的 基于因果森林（CF）模型的个体化因果推断方法，比较腹腔镜肝血管瘤剥除术与肝实质切除术在术中出血控制中的差异，识别关键效应修饰因子并形成术式选择参考。 方法 回顾性分析2023年11月至2025年11月于四川大学华西医院肝脏外科接受腹腔镜手术治疗的196例肝血管瘤病人的临床资料。根据手术方式，将病人分为肝血管瘤剥除术组（111例）与切除术组（85例）。采用CF模型估计个体处理效应（ITE）与平均处理效应（ATE），并基于核心变量在CF模型中所有分裂点的中位数作为阈值进行亚组分析。随后构建多元线性回归模型进行解释性分析，以术中出血量为主要结局变量。 结果 剥除术111例、切除术85例；总体ATE为-2.4 mL，95%CI为-86.1~81.4，P>0.05，提示两术式总体出血差异无统计学意义；ITE范围为-53.3~52.6 mL，显示明显个体异质性。变量重要性分析显示门静脉期增强比率（PER）贡献最高；当PER≤32.58%时病人更倾向从剥除术获益，当PER>32.58%时切除术更优。线性模型R²=0.544，与CF预测结果相关系数为0.74。 结论 CF模型可有效揭示腹腔镜肝血管瘤术式选择中的个体化出血差异，PER可作为术前分层与术式匹配的重要量化指标，为个体化手术决策提供可解释的定量依据。
Objective To compare the differences in intraoperative bleeding control between laparoscopic hepatic hemangioma enucleation and hepatic resection, to identify key effect modifiers, and to provide a reference for surgical procedure selection, based on the individualized causal inference method of the causal forest (CF) model. Methods The clinical data of 196 patients with hepatic hemangioma who underwent laparoscopic surgery at the Department of Liver Surgery, West China Hospital of Sichuan University between November 2023 and November 2025 were retrospectively analyzed. According to the surgical procedures, the patients were divided into the hepatic hemangioma enucleation group (111 cases) and the hepatic resection group (85 cases). The CF model was adopted to estimate the individualized treatment effect (ITE) and average treatment effect (ATE), and subgroup analysis was performed using the median values of all split points for the key variables within the CF model as the thresholds. Subsequently, a multiple linear regression model was constructed for explanatory analysis, with intraoperative blood loss as the primary outcome variable. Results There were 111 cases of enucleation and 85 cases of resection. The overall ATE was -2.4 mL (95%CI:-86.1-81.4,P>0.05), indicating no statistically significant difference in overall bleeding between the two surgical procedures. The ITE ranged from -53.3 to 52.6 mL, demonstrating obvious individual heterogeneity. Variable importance analysis showed that the portal venous phase enhancement ratio (PER) contributed the most. When PER ≤ 32.58%, patients tended to benefit more from enucleation; when PER > 32.58%, resection was superior. The R² of the linear model was 0.544, and its correlation coefficient with the CF prediction results was 0.74. Conclusion The CF model can effectively reveal individualized bleeding differences in the selection of surgical procedures for laparoscopic hepatic hemangioma. PER can serve as an important quantitative indicator for preoperative stratification and procedure matching, providing an interpretable quantitative basis for individualized surgical decision-making.
肝血管瘤 / 因果森林 / 个体化 / 手术决策 / 门静脉期增强比率
hepatic hemangioma / causal forest / individualization / surgical decision-making / portal phase enhancement ratio
| [1] |
European Association for the Study of the Liver. EASL clinical practice guidelines on non-invasive tests for evaluation of liver disease severity and prognosis - 2021 update[J]. J Hepatol, 2021, 75(3):659-689. DOI: 10.1016/j.jhep.2021.05.025.
|
| [2] |
Background: Knowledge of the natural history and management of hepatic hemangiomas is lacking. The aim of this study was to investigate the natural history of hemangiomas and to elucidate the factors that determine tumor growth and optimal management. Methods: A total of 211 adult patients were enrolled, with follow-up for more than three years. Follow-up was performed with repeated ultrasonography (US) and laboratory tests for liver function and coagulation factors (platelets, prothrombin time (PT), fibrinogen, thrombin–antithrombin III complex (TAT), D-dimer, and fibrin and fibrinogen degradation products (FDP)). Results: Tumor size decreased in 38.9% of patients, showed no change in 31.3%, and increased in 29.8%. The incidence of a size increase was very high in patients under 40 years of age and decreased gradually with age, whereas the incidence of a size decrease increased with age and increased markedly over 60 years of age. The incidence of an increase in size decreased gradually with size enlargement, whereas the incidence of a decrease in size increased markedly with tumor size and further increased rapidly when hemangiomas became larger than 60 mm. Values of TAT, D-dimer, FDP, and Mac-2 binding protein glycosylation isomer (M2BPGi) were closely related to the change in size of hemangiomas. Conclusions: Hemangiomas in older patients (>60 years of age) and larger tumors (>60 mm in size) had a tendency to decrease in size, resulting from the reduction in coagulation disorders and the progression of liver fibrosis. Therefore, the majority of patients with hemangiomas can be safely managed by clinical observation.
|
| [3] |
李双喜, 胡宗凯, 赵静, 等. 肝血管瘤治疗指征及治疗策略[J]. 中华肝脏外科手术学电子杂志, 2023, 12(5): 504-510. DOI:10.3877/cma.j.issn.2095-3232.2023.05.007.
|
| [4] |
国际肝胆胰协会中国分会肝血管瘤专业委员会. 肝血管瘤诊断和治疗多学科专家共识(2019版)[J]. 中国实用外科杂志, 2019, 39(8): 761-765. DOI:10.19538/j.cjps.issn1005-2208.2019.08.01.
|
| [5] |
白雪莉, 陈伟, 梁廷波. 肝血管瘤规范化治疗及其值得注意的问题[J]. 中国实用外科杂志, 2013, 33(9): 755-758. DOI:CNKI:SUN:ZGWK.0.2013-09-013.
|
| [6] |
Surgical intervention is the most accurate method for the treatment of hepatic hemangioma. The advantages of laparoscopic surgery on quality of life should be clarified by prospective studies.The sample sizes of the laparoscopic and open surgery groups were calculated based on previous retrospective literature. Intraoperative and postoperative parameters were prospectively collected and analyzed. Quality of life in both groups was predicted by a mixed linear model.Sixty patients were enrolled in the laparoscopic surgery group and open surgery group. The laparoscopic group had a longer operation time (P = 0.040) and more hospitalization expenses (P = 0.001); however, the Clavien-Dindo classification and comprehensive complication index suggested a lower incidence of surgical complications in the laparoscopic group, with P values of 0.049 and 0.002, respectively. After mixed linear model prediction, between-group analysis indicated that the laparoscopic group had little impact on role-physical functioning and role-emotional functioning; in addition, within-group analysis showed a rapid recovery time on role-physical functioning and role-emotional functioning in the laparoscopic group. Quality of life in both groups recovered to the preoperative level within 1 year after the operation.The advantages of laparoscopic hepatectomy for hepatic hemangioma were fewer postoperative complications, lower impact on quality of life and faster recovery from affected quality of life.© 2022. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
|
| [7] |
|
| [8] |
Minimally invasive techniques have increasingly been adopted for liver resection. This study aimed to compare the perioperative outcomes of robot-assisted liver resection (RALR) with laparoscopic liver resection (LLR) for liver cavernous hemangioma and to evaluate the treatment feasibility and safety.
|
| [9] |
Liver resection or enucleation has been the basic treatment for liver hemangioma. However, there were few reports about laparoscopic surgery (LS) of hemangioma. The intention of this study is to explore the indication and efficacy of LS for laparoscopic hepatectomy (LH) and develop an opinion of these modern developments.Forty-four patients with LH underwent LS, with hemihepatic vascular occlusion (HVO group n = 24) or modified vascular occlusion (MVO group n = 20), and were retrospectively reviewed, including patients' demography, surgical technique, tumor size and location, blood loss, operation time, complications, modes of hepatic vascular occlusion and changes in postoperative liver function, and the difference in patients demography and operative outcome between HVO and MVO groups were compared as well.There were no deaths. The mean operating time was 162 minutes, intraoperative blood loss was 335 mL, blood transfusion rate was 9.1%, postoperative complication rate was 18.2%, and length of hospital stay was 7.3 days. Although the tumor size in the HVO group was significantly larger than that in the MVO group, there were no differences concerning operating outcomes, length of stay, and postoperative serum alanine transaminase (ALT), aspertate aminotransferase (AST) level between the HVO and MVO groups.LS was feasible for LH with hepatic vascular occlusion with zero mortality and low complication rate.
|
| [10] |
This study aimed to determine the feasibility of the extracapsular enucleation method for giant liver hemangiomas by infrahepatic inferior vena cava (IVC) clamping and the Pringle maneuver to control intraoperative bleeding under laparoscopic hepatectomy.From January 2012 to January 2016, 36 patients underwent laparoscopic extracapsular enucleation of giant liver hemangiomas. Patients were divided into two groups: infrahepatic IVC clamping + Pringle maneuvers group (IVCP group, n = 15) and the Pringle maneuvers group (Pringle group, n = 21). Operative parameters, postoperative laboratory tests, and morbidity and mortality were analyzed.The mean size of liver hemangiomas was 13.3 cm (range 10-25 cm). Infrahepatic IVC clamping + the Pringle maneuvers with laparoscopic extracapsular enucleation significantly reduced intraoperative blood loss (586.7 vs 315.3 mL, p < 0.001) and transfusion rates (23.8 vs 6.7%, p = 0.001), compared with the Pringle maneuver alone. The gallbladder was retained in both groups. The mean arterial pressure (MAP) in Pringle group remained virtually stable before and after clamping of hepatic portal, while it was significantly decreased after IVC clamping in IVCP group than that pre-clamping (p < 0.001). The heart rate of all patients was significantly increased after clamping when compared to pre-clamping heart rates (p < 0.001). Once vascular occlusion was released, MAP returned to normal levels within a few minutes. There were no significant differences in postoperative complications between two groups. The vascular occlusion techniques in both groups had no serious effect on postoperative of hepatic and renal function.Extracapsular enucleation with infrahepatic IVC clamping + the Pringle maneuver is a safe and effective surgical treatment to control bleeding for giant liver hemangiomas in laparoscopic hepatectomy.
|
| [11] |
Laparoscopic treatment has been increasingly adopted for giant hepatic hemangioma (HH), but the role of liver resection or enucleation remains uncertain. The aim of this study is to compare the laparoscopic resection (LR) with laparoscopic enucleation (LE) for HH, and to provide evidence on how to choose the most suitable approach for HH.A retrospective analysis of HH patients underwent laparoscopic treatment between March 2015 and August 2022 was performed. Perioperative outcomes were compared based on the surgical approaches, and risk factors for increased blood loss was calculated by logistic regression analysis.A total of 127 patients in LR group and 287 patients in LE group were enrolled in this study. The median blood loss (300 vs. 200 mL, P < 0.001) was higher in LE group than that in LR group. Independent risk factors for blood loss higher than 400 mL were tumor size ≥ 10 cm, tumor adjacent to major vessels, tumor occupying right liver or caudate lobe, and the portal phase enhancement ratio (PER) ≥ 38.9%, respectively. Subgroup analysis showed that LR was associated with less blood loss (155 vs. 400 mL, P < 0.001) than LE procedure in patients with high PER value. Both LR and LE approaches exhibited similar perioperative outcomes in patients with low PER value.Laparoscopic treatment for HH could be feasibly and safely performed by both LE and LR. For patients with PER higher than 38.9%, the LR approach is recommended.© 2024. The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.
|
| [12] |
Open enucleation (OE) is often performed for giant liver hemangioma (LH) because of its advantage in maximum preservation of functional liver parenchyma. Laparoscopic enucleation (LE) has been applied to LHs more frequently for its potential advantages in postoperative recovery and blood loss. However, to date, LE is still a difficult and complex surgical technique especially when the hemangioma is located in the right hemi liver. The aim of this study was to analyze whether LE is superior to OE for LH in the right hemi liver.Demographics and perioperative data of patients who underwent LE or OE for LH in the right hemi liver between May 2013 and July 2020 were collected. To decrease the selection bias, patients who underwent OE in first 2 years and those underwent LE in next 5 years by a same operation team were included. The data of sex, age, body mass index (BMI), American Society of Anesthesiologists (ASA) score, largest tumor size, and removed tumor number were enrolled in the propensity score matching (PSM) method to compensate for differences in the baseline characteristics between LE and OE groups. The perioperative outcomes were compared between 2 matched groups after PSM method.A total of 110 patients (36 LE 74 OE) were matched by age, sex, BMI, ASA grade score, largest tumor size, removed tumor number and tumor location. Finally, 34 patients in each group were retained after PSM. There were no significant differences in operative time, estimated blood loss, amount of autologous transfusion, morbidity grade and the levels of alanine aminotransferase (ALT) and aspartate aminotransferase (AST) on postoperative day 1 or 3 or 5. LE was associated with a significantly higher rate of use of the Pringle maneuver (P<0.001), shorter time to oral feeding (P<0.001) and shorter postoperative length of stay (P<0.001).For LHs in the right hemi liver, the perioperative safety of LE is not inferior to OE, and LE seems to achieves a faster recovery from surgery compared with OE.2022 Annals of Translational Medicine. All rights reserved.
|
| [13] |
The Medical Subject Headings (MeSH) thesaurus used by the National Library of Medicine defines logistic regression models as "statistical models which describe the relationship between a qualitative dependent variable (that is, one which can take only certain discrete values, such as the presence or absence of a disease) and an independent variable." Logistic regression models are used to study effects of predictor variables on categorical outcomes and normally the outcome is binary, such as presence or absence of disease (e.g., non-Hodgkin's lymphoma), in which case the model is called a binary logistic model. When there are multiple predictors (e.g., risk factors and treatments) the model is referred to as a multiple or multivariable logistic regression model and is one of the most frequently used statistical model in medical journals. In this chapter, we examine both simple and multiple binary logistic regression models and present related issues, including interaction, categorical predictor variables, continuous predictor variables, and goodness of fit.
|
| [14] |
|
| [15] |
Comparative effectiveness research (CER) using observational databases has been suggested to obtain personalized evidence of treatment effectiveness. Inferential difficulties remain using traditional CER approaches especially related to designating patients to reference classes a priori. A novel Instrumental Variable Causal Forest Algorithm (IV-CFA) has the potential to provide personalized evidence using observational data without designating reference classes a priori, but the consistency of the evidence when varying key algorithm parameters remains unclear. We investigated the consistency of IV-CFA estimates through application to a database of Medicare beneficiaries with proximal humerus fractures (PHFs) that previously revealed heterogeneity in the effects of early surgery using instrumental variable estimators.IV-CFA was used to estimate patient-specific early surgery effects on both beneficial and detrimental outcomes using different combinations of algorithm parameters and estimate variation was assessed for a population of 72,751 fee-for-service Medicare beneficiaries with PHFs in 2011. Classification and regression trees (CART) were applied to these estimates to create ex-post reference classes and the consistency of these classes were assessed. Two-stage least squares (2SLS) estimators were applied to representative ex-post reference classes to scrutinize the estimates relative to known 2SLS properties.IV-CFA uncovered substantial early surgery effect heterogeneity across PHF patients, but estimates for individual patients varied with algorithm parameters. CART applied to these estimates revealed ex-post reference classes consistent across algorithm parameters. 2SLS estimates showed that ex-post reference classes containing older, frailer patients with more comorbidities, and lower utilizers of healthcare were less likely to benefit and more likely to have detriments from higher rates of early surgery.IV-CFA provides an illuminating method to uncover ex-post reference classes of patients based on treatment effects using observational data with a strong instrumental variable. Interpretation of treatment effect estimates within each ex-post reference class using traditional CER methods remains conditional on the extent of measured information in the data.© 2022. The Author(s).
|
| [16] |
胡安宁, 吴晓刚, 陈云松. 处理效应异质性分析——机器学习方法带来的机遇与挑战[J]. 社会学研究, 2021, 36(1):91-114,228.
|
利益冲突 所有作者均声明不存在利益冲突
/
| 〈 |
|
〉 |
