Minimally Invasive Anatomic Liver Resection for Hepatocellular Carcinoma Using the Extrahepatic Glissonian Approach: Surgical Techniques and Comparison of Outcomes with the Open Approach and between the Laparoscopic and Robotic Approaches

Simple Summary Surgical techniques and outcomes of minimally invasive anatomic liver resection (AR) for hepatocellular carcinoma (HCC) are undefined. In 327 HCC patients undergoing 185 open (OAR) and 142 minimally invasive (MIAR; 102 laparoscopic and 40 robotic) ARs, perioperative and long-term outcomes were compared, using propensity score matching. After matching (91:91), compared to OAR, MIAR was significantly associated with longer operative time; less blood loss; a lower transfusion rate; lower rates of 90-day major morbidity, bile leak or collection, and 90-day mortality; and shorter hospital stay. On the other hand, laparoscopic and robotic AR cohorts after matching (31:31) had comparable perioperative outcomes. Postoperative overall and recurrence-free survivals of newly developed HCC were comparable between OAR and MIAR or between laparoscopic and robotic cases. MIAR was technically standardized using the extrahepatic Glissonian approach. MIAR was safe, feasible, and oncologically acceptable and would be the first choice of AR in selected HCC patients. Abstract Surgical techniques and outcomes of minimally invasive anatomic liver resection (AR) using the extrahepatic Glissonian approach for hepatocellular carcinoma (HCC) are undefined. In 327 HCC cases undergoing 185 open (OAR) and 142 minimally invasive (MIAR; 102 laparoscopic and 40 robotic) ARs, perioperative and long-term outcomes were compared between the approaches, using propensity score matching. After matching (91:91), compared to OAR, MIAR was significantly associated with longer operative time (643 vs. 579 min, p = 0.028); less blood loss (274 vs. 955 g, p < 0.0001); a lower transfusion rate (17.6% vs. 47.3%, p < 0.0001); lower rates of major 90-day morbidity (4.4% vs. 20.9%, p = 0.0008), bile leak or collection (1.1% vs. 11.0%, p = 0.005), and 90-day mortality (0% vs. 4.4%, p = 0.043); and shorter hospital stay (15 vs. 29 days, p < 0.0001). On the other hand, laparoscopic and robotic AR cohorts after matching (31:31) had comparable perioperative outcomes. Overall and recurrence-free survivals after AR for newly developed HCC were comparable between OAR and MIAR, with potentially improved survivals in MIAR. The survivals were comparable between laparoscopic and robotic AR. MIAR was technically standardized using the extrahepatic Glissonian approach. MIAR was safe, feasible, and oncologically acceptable and would be the first choice of AR in selected HCC patients.


Introduction
Anatomic liver resection (AR) is a hepatectomy procedure, where an anatomically determined liver territory that is supplied by the arbitrary Glissonian or portal pedicles (GPs) is completely and optimally resected. Therefore, AR includes both major liver resection of three or more Couinaud's segments as well as parenchyma-preserving resection represented by mono-or bi-(sub)segmentectomy and monosectionectomy, and their combinatory resection. AR is expected to achieve both high curability and functional safety in liver resection for malignancy. In particular, parenchyma-sparing AR (PSAR), such as isolated segmentectomy and sectionectomy, is considered to confer benefits for resection of hepatocellular carcinoma (HCC), which is characterized by intra-portal vein tumor spread and accompanying impaired hepatic functional reserve [1][2][3].
The safety and feasibility of minimally invasive AR (MIAR) has been partly established in previous studies [4][5][6][7][8][9], though MIAR, particularly PSAR, is still not technically standardized and is regarded as a group of procedures suitable for expert hands or for high volume centers [5,6,8]. Furthermore, perioperative and long-term outcomes of MIAR have not been fully studied nor have they been compared to those of open AR (OAR) in detail. Surgical HCC patients have been shown to obtain oncological benefits from AR, if technically and functionally applicable, compared to non-anatomic liver resection (NAR) [2,3,9]. Therefore, technical standardization of MIAR and evaluation of its surgical results in HCC patients in comparison to those of OAR, can serve to reinforce or revise surgical strategies for HCC.
In this single-center study on 327 consecutive AR cases of HCC, including 185 OAR and 142 MIAR cases, we present our standardized surgical techniques for MIAR and compare perioperative outcomes between OAR and MIAR, as well as between laparoscopic and robotic AR, using propensity score matching (PSM) analyses. In addition, long-term outcomes after AR for newly developed HCC were compared between OAR and MIAR, as well as between laparoscopic and robotic AR. Based on the results, we discuss the technical, surgical, and oncologic aspects of MIAR in surgical management of HCC.

Terminology and Definition of AR
The terminology for liver anatomy and procedures of hepatectomy was primarily based on the Brisbane 2000 Terminology of Liver Anatomy and Resections [10] and Couinaud's classification [11]. Additionally, the anatomic subsegmentectomy was determined according to the PAM-meeting classification [12]. Thus, AR included trisectionectomy, bisectionectomy including right or left hemihepatectomy and central bisectionectomy, monosectionectomy, (sub)segmentectomy, and their continuous combination. Anatomic subsegmentectomy was defined as the resection of an isolated liver territory that is supplied by the third (or fourth) order division GPs or by its combination with the adjacent GPs smaller than sectional GPs. Isolated total caudate lobectomy was included in the segmentectomy and resection of the Spiegel lobe by dividing the left caudate GP at its root and was defined as subsegmentectomy [13]. Left lateral sectionectomy was included in sectionectomy because it was performed using the extrahepatic Glissonian approach, where GPs to Sg2 and Sg3 were isolated and divided extrahepatically before starting parenchymal dissection. In this study, we excluded AR cases with biliary or vascular reconstruction, those with concomitant extrahepatic procedures, and live donor hepatectomy.

Surgical Indications for MIAR for HCC
At our institution, AR was the first choice of hepatectomy procedure for HCC, if appropriate and applicable, irrespective of the surgical approaches; open, laparoscopic, or robotic. Selection of AR and the type of hepatectomy were based on the location, number, and size of tumors as well as patients' hepatic functional reserve, mostly according to the so-called Makuuchi criteria, and physical status [14].
Selection of OAR or MIAR was based on the operative difficulties depending on the tumor and patient characteristics (Ban's criteria) [15] and the surgeons' capability. It was also dependent on the chronological era, social background, and surgeons' preference. Basic indications for MIAR were in accordance with the following conditions: (1) tumors with a diameter ≤15 cm, without limitation of tumor location; (2) five or fewer excision sites; and (3) no need for biliary or vascular reconstruction. Several carefully selected cases with reconstructive procedures were performed on a clinical-study-based practice.
Selection of the surgical approaches was greatly affected by the medical paying system. Until 2015, before the start of national insurance coverage of MIAR in our country, the open approach was the first choice for AR in most patients, while from 2016, the laparoscopic approach had the priority, if indicated. Until March 2022, robotic liver resection was a practice at patients' own expense in Japan, which significantly affected selection of the laparoscopic or robotic approach. After starting reimbursement of robotic liver resection from April 2022, robotic AR was selected at the surgeons' or patients' preferences as well as depending on the machine availability in the hospital.

Baseline Data Collection
Patient, tumor, and surgical baseline data were retrospectively collected from the patients' medical charts. The patient baseline data included age, sex, body mass index (BMI), American Society of Anesthesiology-Performance Status (ASA) score, presence of diabetes, serum biomarkers, etiology of background liver disease (hepatitis B (HBV), hepatitis C (HCV) or non-B and non-C (NBNC)), indocyanine green (ICG) retention rate at 15 min (ICGR15), Child-Pugh grade, and histologically proven cirrhosis (postoperative evaluation). Tumor characteristics included location, number and size of the tumors, tumor stages according to the classification of the Liver Cancer Study Group of Japan [16], and serum tumor markers (alpha-fetoprotein (AFP); des-gamma-carboxy prothrombin (DCP)). Posterosuperior (PS) segments were defined as segments Sg1, Sg4a, Sg7, and Sg8 and were regarded as "locally difficult segments". The other segments were classified as anterolateral (AL) segments. Pathologic tumor stages and differentiation grades were determined postoperatively. The surgical baseline data included types of AR (left lateral sectionectomy, segmentectomy, sectionectomy, hemihepatectomy, or trisectionectomy) and repeat hepatectomy.

Surgical Techniques for AR
The surgical techniques of AR described below were consistently used either in OAR or MIAR and in any types of AR, irrespective of location of the target anatomic liver territory [5,6,[17][18][19]. Inflow control was based on the extrahepatic Glissonian approach, where GPs to the target territory to be resected were first isolated and clamped or divided extrahepatically before parenchymal dissection was started.
Then, the optimal amount of parenchyma was resected along the demarcation line or the intersegmental plane that was visualized by macroscopic discoloration or by the ICG negative staining method [5,6,17]. During parenchymal dissection, the landmark hepatic veins (HVs) were exposed from their root side in the cranial-to-caudal direction (HV root-at first one-way parenchymal dissection). The extrahepatic Glissonian approach and HV root-at first one-way parenchymal dissection were facilitated by the anatomical background of Laennec's capsule at the hilum and major hepatic veins, as well as the hilar 'Gate theory', as described previously [6,13,18,19].
Several examples of the extrahepatic Glissonian approach to isolate the target GPs for MIAR are described in Figure 1. In left hemihepatectomy ( Figure 1A), extrahepatic isolation of the left Glissonian pedicle above the hilar plate (G-UP) was facilitated by dissecting Gates I and III and passing forceps and a tape between the gates (Figure 1Aa). In isolated segmentectomy 8 ( Figure 1B), to isolate the target GP of Sg8 (G8) extrahepatically (Figure 1Ba), the right anterior section GP (G-ant) was first isolated after cystic plate cholecystectomy in which the cystic plate was resected along with the gall bladder. By exerting traction of G-ant downward, we isolated Sg5 GP (G5) extrahepatically, and finally, G8 was isolated extrahepatically, using the tape switching method (subtraction method) [6,13,19]. In right posterior sectionectomy ( Figure 1C), the GP of the posterior section (G-post) was isolated extrahepatically by passing a tape between Gates V and VI (Figure 1Cg). Through selective occlusion of these target pedicles, the isolated ischemic anatomic territories were identified macroscopically as discolored areas or by using an ICG negative staining technique ( Figure 1A(b),B(e),C(h)).
traction of G-ant downward, we isolated Sg5 GP (G5) extrahepatically, and finally, G8 was isolated extrahepatically, using the tape switching method (subtraction method) [6,13,19]. In right posterior sectionectomy ( Figure 1C), the GP of the posterior section (Gpost) was isolated extrahepatically by passing a tape between Gates V and VI ( Figure  1Cg). Through selective occlusion of these target pedicles, the isolated ischemic anatomic territories were identified macroscopically as discolored areas or by using an ICG negative staining technique (Figure 1Ab,Be,Ch).
Selective occlusion of inflow to the anatomic territory to be resected was followed by HV root-at first one-way parenchymal dissection, during which the already clamped target pedicles were exposed and divided on the liver dissecting plane. For parenchymal dissection, Cavitron Ultrasonic Surgical Aspirator (CUSA ® ) was used in open and laparoscopic cases, and a clamp-crushing method and ultrasonic coagulating shears were used in robotic cases. Pringle maneuver was not used routinely but was applied on demand. Parenchyma was dissected in the cranial-to-caudal direction, and AR was completed (Figure 1Ac,1Bf,Ci). Selective occlusion of inflow to the anatomic territory to be resected was followed by HV root-at first one-way parenchymal dissection, during which the already clamped target pedicles were exposed and divided on the liver dissecting plane. For parenchymal dissection, Cavitron Ultrasonic Surgical Aspirator (CUSA ® ) was used in open and laparoscopic cases, and a clamp-crushing method and ultrasonic coagulating shears were used in robotic cases. Pringle maneuver was not used routinely but was applied on demand. Parenchyma was dissected in the cranial-to-caudal direction, and AR was completed ( Figure 1A(c),B(f),C(i)).

Perioperative Data
Intraoperative outcomes were evaluated by operative time, blood loss, liver parenchymal dissection time, transfusion (of any blood elements), use of the Pringle maneuver, and open conversion (in MIAR). Postoperative outcomes were evaluated by the serum maximum levels of maximum total bilirubin (TB), aspartate aminotransferase (AST) and C-reactive protein (CRP), the serum minimum level of prothrombin time (PT), morbidity graded according to the Clavien-Dindo (C-D) classification [20], mortality, R0 resection, and the length of postoperative hospital stay (LOS). Overall and major complications were defined as those within 90 postoperative days of any C-D grades and those of ≥C-D grade IIIa, respectively.

Statistical Analysis
Continuous data were expressed as median with range (baseline data) or interquartile range (perioperative data) and compared using the Kruskal-Wallis test. Categorical data were compared using the Pearson's chi-square test. In some comparative studies, 1:1 PSM was conducted to reduce biases. In comparison of perioperative outcomes between the OAR and MIAR cohorts, the following 11 variables were matched for PSM: age, sex, ASA class (I or II/≥III), ICGR15 (<13.0%/≥13.0%), etiology of underlying liver disease (HBV/HCV/NBNC), tumor number (single/multiple), tumor size (<4.0/≥4.0, cm), tumor location (AL/PS), tumor stage (I or II/≥III), types of resection (left lateral sectionectomy/segmentectomy/sectionectomy/≥hemihepatectomy), and previous hepatectomy (yes/no). In comparison of perioperative outcomes between the laparoscopic and robotic cohorts, age, sex, presence of cirrhosis (yes/no), tumor number (single/multiple), tumor size (<3.0/≥3.0, cm), types of resection, and previous hepatectomy (yes/no) were matched. In analyses of long-term outcomes, we examined newly developed HCC cases only.
The PSM method was the nearest neighborhood method with a caliper width of 0.20. The standard mean deviation (SMD) was calculated for all studied variables, and an SMD <0.20 was confirmed for all matched variables, which indicated appropriate matching. The postoperative overall survival (OS) and recurrence-free survival (RFS) were analyzed only in patients with newly developed HCC, using the Kaplan-Meier method. p <0.050 was considered statistically significant. Statistical analyses were performed using JMP ® software ver. 14.0 (SAS Institute, Cary, NC, USA).

Results
Between 2010 and October 2022, we performed 667 liver resections for HCC, including 306 open, 279 laparoscopic, and 82 robotic resections, at our institution. Among these 667 cases, we retrospectively reviewed the consecutive 327 AR cases, consisting of 185 OAR and 142 MIAR (102 laparoscopic and 40 robotic) cases, where AR using the extrahepatic Glissonian approach was performed without biliary or vascular reconstruction or concomitant resection of extrahepatic organs.

Comparison between OAR and MIAR
Patient and Tumor Baseline Data Patient and tumor baseline data were compared between OAR and MIAR (Table 1). Before PSM (185 OAR and 142 MIAR cases), compared to OAR, MIAR was significantly associated with higher BMI, lower ASA class, lower ICGR15, lower Child-Pugh class, smaller tumor size, lower AFP and DCP levels, and more favorable tumor stage. Further, compared to OAR, MIAR was significantly associated with a lower rate of sectionectomy or more extensive procedures, a lower proportion of major resection, and a higher rate of the repeat hepatectomy setting. After 1:1 PSM (91:91), the OAR and MIAR groups were comparable in terms of all studied variables.

Comparison between Laparoscopic and Robotic AR for HCC
In the next set of analyses, we compared baseline data and perioperative outcomes between the two MIAR approaches: laparoscopic and robotic.

Patient and Tumor Baseline Data
Baseline data were compared between the laparoscopic and robotic AR groups (Table 3). Before PSM (102 laparoscopic vs. 40 robotic cases), compared to laparoscopic AR, robotic AR was significantly associated with a lower rate of cirrhosis, smaller tumor number and size, a lower AFP level, and a higher rate of repeat hepatectomy setting. After 1:1 PSM (31:31), all studied variables were comparable between the laparoscopic and robotic AR groups.

Perioperative Outcomes
Perioperative outcomes are shown in Table 4. Before PSM, laparoscopic and robotic AR groups had comparable outcomes, except for the significantly higher rate of Pringle maneuver application (35.0% vs. 17.7%, p = 0.026), the higher maximum AST level (767 vs. 546 IU/L, p = 0.026), and the lower minimum PT level (60% vs. 64%, p = 0.032) in the robotic AR group. After PSM, both groups had comparable perioperative outcomes.

Long-Term Outcomes after AR for Newly Developed HCC
In the next set of analyses, we studied long-term outcomes after AR in 276 patients with newly developed HCC, who underwent 163 OARs and 113 MIARs (89 laparoscopic and 24 robotic ARs), respectively, and compared OS and RFS between OAR and MIAR and between laparoscopic and robotic AR.

Comparison of Long-Term Outcomes between OAR and MIAR Patient and Tumor Baseline Data
Baseline data were compared between 163 patients undergoing OAR and 113 patients undergoing MIAR for newly developed HCC (Table 5). Before PSM (163:113), compared to OAR, MIAR was significantly associated with higher BMI, a lower ASA class, lower ICGR 15, smaller tumor number and size, more favorable tumor stages, and higher AFP and DCP levels. In addition, types of resection were significantly more extensive, with a higher proportion of major resection in OAR than in MIAR. After 1:1 PSM (76:76), all studied variables were comparable between OAR and MIAR.

Long-Term Outcomes
Comparing survival rates between the unmatched patients undergoing MIAR (n = 113) and OAR (n = 163), MIAR patients had statistically longer OS (Figure 2A

Comparison of Long-Term Outcomes between Laparoscopic and Robotic AR Patient and Tumor Baseline Data
Patient and tumor baseline data were compared between the unmatched 89 laparoscopic AR and 24 robotic AR cases (Table 6). Before PSM, compared to laparoscopic AR, robotic AR was significantly associated with a lower rate of histology-proven cirrhosis (12.5% vs. 42.7%, p = 0.006) and a lower rate of tumors ≥4.0 cm (25.0% vs. 48.3%, p = 0.041). One-to-one PSM identified 22 laparoscopic and 22 robotic matched cases. After PSM, all evaluated patient and tumor characteristics were comparable between the laparoscopic and robotic AR cohorts.
5-year rate: 84.4% vs. 50.0%) and longer RFS ( Figure 2B, p < 0.0001; 5-year rate: 47.1% v 25.1%) than OAR patients. For the matched cohorts after PSM, compared to OAR (n = 76 MIAR patients (n = 76) were associated with statistically longer OS ( Figure 2C, p = 0.00 5-year rate: 78.9% vs. 54.6%) and longer RFS ( Figure 2D, p = 0.010; 5-year rate: 41.9% v 31.1%).   The details of recurrence and its treatments are shown in Table 7. In recurrent cases in the entire cohort, patterns of recurrence (intrahepatic and/or extrahepatic), the rate of extrahepatic recurrence, and recurrent organs were comparable between OAR and MIAR. Compared to OAR, MIAR was associated with significantly longer RFS (median: 15.4 vs. 7.6 months, p = 0.007) and a significantly lower rate of the first recurrence within 1 postoperative year (37.3% vs. 62.2%, p = 0.003). Pathologic tumor stages were more advanced in recurrent cases in the unmatched OAR cohort (p = 0.008), while they were comparable between the matched OAR and MIAR cohorts. Pathologic tumor differentiations were comparable between the recurrent cases in the unmatched and matched cohorts. Perioperative morbidity, which has been suggested to affect tumor recurrence in previous studies [21,22], was comparable between OAR and MIAR, both in the unmatched and matched cohorts.

Comparison of Long-Term Outcomes between Laparoscopic and Robotic AR
Comparison between the unmatched cohorts of laparoscopic (n = 89) and robotic (n = 24) approaches showed comparable OS ( Figure 3A) and RFS ( Figure 3B). Similarly, OS ( Figure 3C) and RFS ( Figure 3D) were comparable between the matched laparoscopic AR (n = 22) and robotic AR (n = 22) cohorts.
The details of recurrence and its treatments are shown in Table 7. In recurrent cases in the entire cohort, patterns of recurrence (intrahepatic and/or extrahepatic), the rate of extrahepatic recurrence, and recurrent organs were comparable between OAR and MIAR. Compared to OAR, MIAR was associated with significantly longer RFS (median: 15.4 vs.  OAR: open anatomic liver resection; MIAR: minimally invasive anatomic liver resection; C-D: Clavien-Dindo classification; *MTA: molecular targeted agent; immunotherapy**: immunotherapy using immune checkpoint inhibitors; Bold: statistically significant. Regarding the treatments for cancer recurrence, in the unmatched cohorts, compared to OAR, MIAR tended to be associated with the higher rates of recurrent organ resection (45.1% vs. 31.5%, p = 0.094) and repeat hepatectomy (46.9% vs. 30.8%, p = 0.052). Systemic pharmacological treatment using molecular targeted agents (MTAs), such as sorafenib and lenvatinib, or immunotherapy using immune checkpoint inhibitors (ICIs), such as atezolizumab, were performed at a similar rate between the OAR and MIAR cohorts with recurrence (17.7% vs. 17.1%). On the other hand, in the matched cohorts, patterns of recurrence, RFS, timing of the first recurrence, and the rate of resection or systemic pharmacological treatment for recurrent tumors were comparable between OAR and MIAR ( Table 7).

Times of Surgery and Associated Factors
To further investigate factors that potentially affect the long-term outcomes after resection of newly developed HCC, we studied the times or era of surgery because significant changes occurred socially in the strategies of HCC treatment as well as in the control of HCV using direct acting antivirals (DAAs) during the study period. We first examined the trends of application of OAR or MIAR throughout the study period ( Figure 4). As shown, the ratio of OAR cases in all AR cases significantly decreased, and OAR was gradually replaced by MIAR over the years (p < 0.0001). Further, when the study period was divided into the early (2010-2015) and late (2016-2022) periods, MIAR was significantly more frequently performed than OAR in the late period in a comparison between the unmatched (n = 95 (84.1%) vs. n = 40 (24.5%), p < 0.0001) and matched (n = 61 (80.3%) vs. n = 14 (18.4%), p < 0.0001) cohorts.
cant changes occurred socially in the strategies of HCC treatment as well as in the c of HCV using direct acting antivirals (DAAs) during the study period. We first exam the trends of application of OAR or MIAR throughout the study period (Figure shown, the ratio of OAR cases in all AR cases significantly decreased, and OAR was ually replaced by MIAR over the years (p < 0.0001). Further, when the study perio divided into the early (2010-2015) and late (2016-2022) periods, MIAR was signifi more frequently performed than OAR in the late period in a comparison between th matched (n = 95 (84.1%) vs. n = 40 (24.5%), p < 0.0001) and matched (n = 61 (80.3%) v 14 (18.4%), p < 0.0001) cohorts. At our institution, eradication of HCV using DAAs became active from around which was coincidental with a remarkable change in the surgical approach from O MIAR. To exclude the impacts of HCV eradication on postoperative recurrence an patic functional status and to study the simple impact of different approaches on sur we examined the long-term outcomes in 184 AR cases without associated HCV infe i.e., 55 HBV and 129 NBNC cases. One-to-one PSM identified the 42 OAR and 42 M matched cases from the 107 OAR and 77 MIAR unmatched cases. The surviva showed that OS in MIAR, with a 5-year rate of 85.2% and not-reached median sur was significantly longer than OS in OAR, with a 5-year rate of 43.3% and the media vival time of 48.9 months (p = 0.001). Furthermore, RFS in MIAR, with a 5-year rate o and the median time of 40.4 months, was significantly longer than RFS in OAR, wit year rate of 21.8% and a median time of 7.6 months (p = 0.004). At our institution, eradication of HCV using DAAs became active from around 2016, which was coincidental with a remarkable change in the surgical approach from OAR to MIAR. To exclude the impacts of HCV eradication on postoperative recurrence and hepatic functional status and to study the simple impact of different approaches on survival, we examined the long-term outcomes in 184 AR cases without associated HCV infection, i.e., 55 HBV and 129 NBNC cases. One-to-one PSM identified the 42 OAR and 42 MIAR matched cases from the 107 OAR and 77 MIAR unmatched cases. The survival data showed that OS in MIAR, with a 5-year rate of 85.2% and not-reached median survival, was significantly longer than OS in OAR, with a 5-year rate of 43.3% and the median survival time of 48.9 months (p = 0.001). Furthermore, RFS in MIAR, with a 5-year rate of 47.4% and the median time of 40.4 months, was significantly longer than RFS in OAR, with a 5-year rate of 21.8% and a median time of 7.6 months (p = 0.004).

Discussion
In this study, we described our surgical techniques of MIAR, including laparoscopic and robotic AR, and compared perioperative and long-term outcomes in HCC cases between OAR and MIAR as well as between laparoscopic and robotic AR, using PSM-based analyses. Our surgical techniques for AR in all studied cases were consistently based on the extrahepatic Glissonian approach and HV root-at first one-way parenchymal dissection, irrespective of the open, laparoscopic, or robotic approach. Furthermore, AR was primarily performed and managed only by surgeons with appropriate skills at our institution. These study settings may have reduced biases, such as surgical techniques and surgeons' experience, which were latent in previous comparative studies.
In the current study, we first audited perioperative outcomes in 142 MIAR cases and compared them to 185 OAR cases. A PSM (91:91)-based analyses showed that compared to OAR, MIAR was significantly associated with less blood loss, a lower transfusion rate, lower postoperative TB and CRP levels, lower rates of 90-day overall and major morbidity and bile leak or intraabdominal collection, a lower rate of 90-day mortality, and shorter LOS. On the other hand, a significantly higher postoperative AST level was observed in the MIAR compared to the OAR group. Next, we compared perioperative outcomes between matched laparoscopic and robotic AR cohorts (46:46), which showed comparable results. Finally, we studied the long-term outcomes after AR for newly developed HCC. PSM-based analyses matching patient and tumor characteristics and types of resection showed comparable or potentially superior OS and RFS in the MIAR cohort compared to those in the OAR cohort, as well as comparable outcomes between the laparoscopic and robotic approaches.
As for technical aspects of MIAR, we believe, from the current and previous results, that we can standardize MIAR by using the extrahepatic Glissonian approach and HV root-at first one-way parenchymal dissection [4][5][6]13,18,19,23]. Furthermore, the consistent usage of these standardized techniques, irrespective of OAR and MIAR, may reduce technical biases, being an advantage of this study setting. These techniques for AR, which were originally developed in OAR at our institution, have been safely and effectively translated into MIAR with acceptable curability, as shown by the favorable perioperative and long-term outcomes. In addition, as the most recent surgical platform, robotics was technically applicable to AR with surgical outcomes comparable to those of the laparoscopic approach. The most important advantage of the extrahepatic Glissonian approach for AR is accurate determination of the anatomic liver area to be resected before starting parenchymal dissection, which leads to accurate, safe, and optimal AR. On the other hand, one of the disadvantages of this approach is that although it is applicable in any type of AR for any liver segments, there are cases where extrahepatic isolation of peripheral subsegmental GPs is unsafe or impossible, depending on the liver anatomy. In this study, PSAR accounted for over 80% of OAR or MIAR cases. Although minimally invasive accurate PSARs are more demanding than standard right or left hemihepatectomy [6,24,25], they were technically standardized and performed safely in this series. Therefore, MIAR including PSAR can provide HCC patients with benefits of not only minimally invasiveness but also safety and curability.
Previous studies comparing open and minimally invasive liver resections have shown less blood loss, decreased morbidity, and shorter LOS in the latter [23,[26][27][28][29]. Laparoscopic and robotic liver resections were shown to have comparable perioperative outcomes [23,30]. Results of the current study are in line with those of these studies, though the cohort setting was different; very few studies exclusively selected AR cases for the cohort. However, recent large studies have demonstrated advantages of the robotic approach on perioperative outcomes over the laparoscopic approach in selected types of AR [31][32][33].
Several studies have shown comparable long-term outcomes after resection of HCC between the open and minimally invasive approaches [26,28]. Only one study showed a survival advantage of the laparoscopic over the open approach, though significant biases were suggested in the study setting [29]. In our results, surprisingly, long-term outcomes after MIAR for newly developed HCC appeared to be more favorable than those after OAR ( Figure 2). Both OS and RFS were statistically longer in the MIAR than in the OAR cohort, not only before, but also after PSM. The longer OS and RFS in the unmatched MIAR cohort could be largely explained by the more favorable tumor characteristics in this cohort represented by lower tumor stages, lower serum tumor markers, the need for less extensive hepatectomy, and a lower rate of early recurrence (Tables 5 and 7). Furthermore, the better patients' physical conditions in this cohort represented by the lower ASA scores and ICGR 15 (Table 5) might have beneficial impacts on postoperative survivals. Additionally, a higher rate of resection of recurrent tumors in the MIAR cohort (Table 7) could have contributed to the longer OS.
On the other hand, it is worth noting that even the matched MIAR cohort after PSM had significantly longer OS and RFS than the OAR cohort. To seek the scenarios behind such potentially better survivals in MIAR, we addressed several relevant points. First, we examined the details of recurrence and found that the patterns of recurrence, the incidence of extrahepatic metastasis and pathologic tumor differentiation, and stages were comparable between the cohorts (Table 7). Second, we examined the times or era of application of OAR or MIAR because the drastic changes in the surgical approach from OAR to MIAR at our institution ( Figure 4) coincided with the changing trends in the pharmacological treatment for HCC recurrence toward MTAs or ICIs, along with significant success of HCV eradication by using DAAs. Our results showed, however, that these changes in non-surgical factors were comparable between the matched OAR and MIAR cohorts. Furthermore, OS and RFS in the HCC cohort excluding HCV-infected patients were still longer in MIAR than in OAR, suggesting that HCV eradication had minor impacts on the survival differences between MIAR and OAR in this study.
Other factors potentially contributing to the more favorable RFS in MIAR include theoretical immunological advantages conferred by its lower invasiveness. The advantages can stem from factors including less physical damage, less disturbance in the perioperative metabolic and nutritional status, and the resultant early recovery after surgery. In our results, such lower invasiveness in MIAR was evidenced by less blood loss, the lower levels of postoperative TB and CRP, and shorter LOS, compared to OAR (Table 2). Furthermore, lower rates of perioperative morbidity and blood transfusion in MIAR ( Table 2) may have had beneficial impacts on its long-term outcomes, in line with previous studies on resected cancers including HCC [21,22,34,35].
Another point to be taken into consideration is that in most previous studies comparing open and minimally invasive hepatectomy, both AR and NAR were included. Since NAR tended to be performed for HCC patients with impaired hepatic functional reserve, inclusion of both AR and NAR cases in the study cohorts and the case number ratio of each cohort may have affected the entire comparative survival data on which both oncologic and hepatic functional factors had significant impacts.
Collectively, our data were not enough to explain the differences in the long-term outcomes between the matched OAR and MIAR cohorts, and in the first place, the relatively small sample size in each cohort precludes definite conclusions. Nonetheless, our results suggest that the long-term outcomes after AR for newly developed HCC were at least comparable between OAR and MIAR. Larger studies incorporating more evaluation facets, such as post-recurrence treatment modalities, patient nutritional status and immunity characteristics and tumor genetic information, are necessary to investigate the impact of a minimally invasive approach to AR on long-term outcomes of resected HCC.
In view of our results showing mostly better perioperative outcomes in MIAR than in OAR, as well as comparable or potentially more favorable long-term outcomes in MIAR, it is reasonable to suggest that MIAR would be the first choice for a surgical approach to HCC, at least by an expert. On the other hand, despite the expected functional merits of robotics providing surgical dexterity, the robotic approach did not show advantages in surgical outcomes over the laparoscopic approach to AR in this study. However, the sample size was small, particularly in the robotic group, and techniques of robotic liver surgery are still developing. Nonetheless, the robotic platform may have potential advantages in perioperative outcomes in complex anatomic resection, such as less blood loss, decreased open conversion rate, and decreased morbidity, as suggested in other study settings [31][32][33]. Larger studies are warranted to investigate the potential differences in outcomes between laparoscopic and robotic AR.
There are several limitations to this study. First, this is a retrospective, observational, single-center study, though PSM was conducted to reduce biases. Second, the sample sizes in each PSM-matched cohort are relatively small. Third, long-term outcomes should be carefully interpreted because of the abovementioned small sample size, small number of matched variables in PSM-based analyses, and potential biases from 'difficult-to-match' factors, including selection of approach, learning curve of techniques, patient immunological status, era-dependent development of DAAs for HCV eradication, and advances in post-recurrence pharmacological therapy for HCC.

Conclusions
Although MIAR is technically demanding, particularly for HCC, because of potential underlying liver dysfunction, it can be technically standardized by the extrahepatic Glissonian approach and HV root-at first one-way parenchymal dissection. Furthermore, MIAR for HCC was safe, feasible, and oncologically acceptable, with perioperative outcomes mostly superior to those in OAR and with comparable or potentially more favorable long-term outcomes. A laparoscopic or robotic approach would be the first choice for AR in selected surgical HCC patients, at least by experts. Further larger studies are warranted to investigate potential advantages of MIAR for HCC in terms of long-term outcomes, as well as perioperative and long-term advantages of the robotic or laparoscopic approach over the counterpart.  Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Written informed consent has been obtained from the patient(s) to publish this paper.
Data Availability Statement: The datasets are not available for public access due to patient privacy concerns but are available from the corresponding author on reasonable request.