Laparoscopic versus open major liver resection for hepatocellular carcinoma: systematic review and meta-analysis of comparative cohort studies

Hepatocellular carcinoma (HCC) represents the second deadliest malignancy around the world [1]. HCC is the unique type of malignancy for which the mortality rate continues to rise despite impressive advances in anticancer treatment [2]. Hepatic resection remains an essential treatment strategy for HCC. The initial application of laparoscopic liver resection (LLR) was in 1991 [3]. From then on, due to technological development and the improvement of laparoscopic liver surgery, the number of LLRs has increased dramatically in the past quarter-century [4]; indeed, LLR has been regarded as a great advance in modern liver surgery [5‐9]. LLR can be divided into two categories, i.e., (i) laparoscopic minor liver resection, which involves non-anatomic wedge resection, left lateral resections, and/or removal of anterior liver segments (4b, 5, 6), and (ii) laparoscopic major liver resection (LMLR), which includes removal of right and left hepatic hemispheres, trisectionectomy, and resection of posterior segments (1, 4a, 7, 8) [10]. Although surgical and oncological outcomes, including peri- and post-operative outcomes, overall survival (OS) and disease-free survival (DFS), are considered to be similar for LLR and open liver resection (OLR) in HCC, most studies only described laparoscopic minor liver resection [11‐15]. Ciria et al. [8] conducted one of the largest reviews of LLR compared with OLR, and suggested LLR might provide ameliorated short-term outcomes. However, the above study only focused on surgical outcomes and did not compare oncological outcomes. As technological support advances and experience in minimally invasive surgery grows, LLR has been developed from minor to major resection for the treatment of HCC [16, 17]. LMLR is increasingly practiced in high-volume and specialized centers. However, its application requires further evaluation. Therefore, an update on the worldwide situation is necessary to assess the current status of LMLR, especially focusing on its advantages and drawbacks comparatively to open major liver resection (OMLR).

Obviously, a prospective, randomized trial would be ideal for assessing the surgical and oncological outcomes of LMLR versus OMLR. Actually, according to the records of Clinicaltrial.​org, several prospective, randomized control trials on LLR versus OLR have been carried out, including NCT02014025, NCT01768741, NCT02526043, NCT00606385, NCT02014025, and NCT02131441. However, to our knowledge, no reports are currently available on these randomized controlled trials (RCTs).

Therefore, the current work primarily aimed to perform a systematic review of the global clinical evidence of LMLR versus OMLR for HCC by assessing reports published before March 2018. These reports were meta-analyzed to investigate perioperative and postoperative surgical outcomes as well as long-term oncological outcomes, comparing LMLR and OMLR.

The methods used in this study included a literature search. Eligibility criteria for studies, outcome measures, and statistical analyses followed the protocol recommended by Stroup et al., and Shamseer et al [18, 19]

The expanding range of LLR procedures, from non-anatomic wedge-, left lateral-, and anterior hepatic segment resections to sectionectomy, hemihepatectomy, trisectionectomy, and resection of difficult posterior segments, is regarded as mimicking OLR expansion [4, 10]. This expansion of LLR procedures is associated with both technological (instruments) and technical (skills) advances. In this period, two international consensus conferences have summarized the current status and future perspectives of LLR [9, 10]. Although multi-center and prospective, randomized studies would be ideal for assessing the effectiveness and safety of LMLR versus OMLR, an increasing amount of studies evaluating LMLR have been reported since 2009. However, minor resections constitute the vast majority of procedures in clinical practice. LMLR remains limited to very few centers and requires further evaluation and caution [4, 8, 15].

LLR has a particularly critical function in HCC treatment [32]. However, due to technical difficulties and the unique anatomical features of the liver, LLR remains somehow limited to a few high-volume and specialized centers [33, 34]. Berardi et al. reported a cohort study assessing perioperative and oncological outcomes from 4 European specialized centers [34]. The study revealed that the percentage of liver resections in which laparoscopy was applied yearly had increased from 5 to 43% during the past 15 years. They also found that perioperative and oncological outcomes have improved significantly with time and reached a stable level in the last few years [34]. However, HCC is commonly associated with chronic liver disease, cirrhosis, and/or impaired liver function, which might increase the risk of severe morbidity postoperatively and decrease the long-term survival rate [35]. Over the past quarter-century, there have been a number of studies evaluating perioperative and/or oncological outcomes of the LLR treatment in HCC patients. The laparoscopic technique and surgical care have been improved to establish LLR standardization [4, 6, 34]. These advances have remarkably increased the application of laparoscopic major liver resection (LMLR) in the last 10 years [6, 15, 31].

Studies or meta-analyses comparing laparoscopic methods to OLR all reported decreased blood loss, lower transfusion rate, reduced post-surgical morbidity, and decreased hospital stay, with comparable oncological outcomes [15, 35, 36]. However, the majority of trials focused on minor liver resection or failed to clearly differentiate between the outcomes of minor and major resections. Although the latest meta-analysis assessed short- and long-term outcomes between LMLR and OMLR, all retrospective trials comparatively evaluating LMLR and OMLR were included [37]. The combined results may be biased, so we conducted the current meta-analytical study of pooled perioperative and long-term LMLR and OMLR outcomes. Since no results of existing prospective randomized trials are currently available for analysis, totally 1173 patients from 9 retrospective trials were meta-analyzed. Our results demonstrated the technical feasibility and safety of LMLR in HCC patients. Our study included 9 published studies from major databases, comparing the short-term surgical and long-term oncological outcomes of LMLR and OMLR in the treatment of HCC. Only 9 studies were included in this meta-analysis because of the following plausible reasons. (1) We selected RCTs comparing LMLR with OMLR for HCC, and excluded studies reporting minor liver resection or with unavailable outcomes of major liver resection. (2) We included studies that analyzed the outcomes of LMLR in HCC, and excluded those in which LMLR was applied for recurrent HCC, hepatic metastatic cancer, simultaneous resection of the liver and other organs, or simultaneous resection of malignant and benign liver tumors. (3) We excluded studies that included less than 20 patients, considering the notion that studies reporting LMLR data in small samples might have limited reliability [15].

Application of LLR was rather delayed by technical challenges in keeping homeostasis at the transection plane and managing intraoperative bleeding from intrahepatic vessels [7, 13]. Intraoperative bleeding remains one of the most challenging issues in LLR, especially when major liver resection is performed in HCC complicated with chronic liver diseases or cirrhosis. Therefore, bleeding during LLR remains one of the most common reasons for selecting OLR. As shown previously, blood loss and perioperative blood transfusion negatively affect short-term surgical and long-term oncological outcomes [26, 27, 34, 38, 39]. In order to decrease bleeding and perioperative blood transfusion, some surgical techniques, such as the Glissonian approach, anatomic liver resection and selective clamping, have been proposed, which might exert reduced deleterious effects on postoperative liver function and yield more positive outcomes [40‐43]. Moreover, innovative methods, e.g. intraoperative ultrasonography, microwave-based coagulation, ultrasonic dissection, and argon beam coagulation, and the use of laparoscopic coagulation shears and endoscopic linear staplers, significantly help achieve appropriate homeostasis in LLR [17, 44]. In this study, although the number of patients that required blood transfusion was not significantly lower in the LMLR group, the volume of blood loss was markedly reduced, suggesting bleeding control could be well conducted in LMLR. Considering other intraoperative outcome measurements, the operative time was markedly prolonged after LMLR. These results were consistent with those reported by Laurent et al [45] The longer operation time may mainly be attributed to the “learning curve” effect, complexity and wide resection plane in LMLR [23, 36]. Despite longer operation duration and the use of special laparoscopic equipment in the LMLR group, the patients had markedly reduced blood loss and hospitalization duration. In this study we did not investigate whether the benefits were cost-effective. A retrospective analysis showed that laparoscopic major liver resection exhibits a high potential clinical outcome effect compared with open major liver resection with cost-effectiveness [46]. However, we expect a future randomized trial to assess the benefits and costs of both surgical methods.

In this study, the pooled data showed that postoperative morbidity rates were markedly reduced after LMLR compared with OMLR. Although Nomi et al. reported a total of 183 cases that underwent LMLR and confirmed that postoperative morbidity was comparable in both LMLR and OMLR groups, only 28 cases of OMLR were included in the study [47]. Takahara et al. published the data of a national clinical database in Japan, with postoperative morbidity comparable to that described in this meta-analysis [7]. Complication severity was assessed using the modified Clavien classification in most of the included studies (Table 3). Although the severity of complications following LMLR showed an increasing trend compared with the OMLR group, other studies showed that severity was similar in both groups [45, 48]. Given the retrospective nature of the included trials, it was difficult to review more detailed data of complications to obtain more meaningful results.

The risk of inadequate resection margin, potential risk of tumor seeding, et al., were the main concerns regarding LLR use for HCC treatment [15]. However, the application of anatomic resection and ultrasound scanning during laparoscopic liver resection could help delineate the cancerous lesions, achieving the intended margin. At the same time, the improvement of laparoscopic technology and the available equipment for reducing potential tumor seeding such as plastic bags for specimen removal, may help overcome all these limitations [49]. Although a meta-analysis conducted by Lin et al. [15], confirmed no differences in oncological outcomes associated with laparoscopic and open minor liver resections for liver cancer, we still expect future trials to explore any differences in long-term oncological outcomes between LMLR and OMLR for HCC. As shown above, although 1-year DFS was elevated after treatment by LMLR compared with OMLR, oncological outcomes were comparable in both groups. Guro et al. reported markedly higher recurrence rate within 1 year of OMLR compared with LMLR (40.0% versus 25.8%, P = 0.029) [29], but the other included trials found no marked differences in 1-year DFS between these two groups. Although long-term oncological outcomes are not better with the laparoscopic method, some studies showed that unexpected diagnosis of early HCC could only be achieved by laparoscopy [50]. In addition, trials assessing HCC only in patients with chronic liver diseases also demonstrated equivalent OS and the DFS after LMLR and OMLR, which suggests that the tumor recurrence rate for the liver parenchyma (or other tissues) is not elevated after LMLR. This is consistent with the results of the current meta-analysis.

The major shortcoming of the current report was that only retrospective non-randomized controlled trials were included for review and meta-analysis. Therefore, it was difficult to review enough data and information to obtain meaningful results. In most of the included studies, cases were assigned to either the LMLR or OMLR group according to their preoperative clinical data and tumor characteristics, so selection bias was inevitable. However, three included studies used the propensity score matching method to minimize bias [28, 30, 31]. The propensity score matching method is considered one of the most optimal tools for reducing selection bias in non-randomized studies [51, 52]. In addition, by focusing only on major liver resection, we might have missed a broader group of studies in which patients undergoing major liver resection represented only a subset of the entire population. However, the available data regarding major liver resection in these studies were very difficult to assess. In order to obtain additional relevant studies, we extended our search to “related articles,” and manually searched interesting references listed in the retrieved articles. Last but not the least, the small sample sizes of multiple trials also reduced data reliability. Although the methods recommended by Hozo and colleagues are mostly acceptable, they constituted a limitation in the present meta-analysis, mainly because the most important aspects of the analysis involved continuous variables and WMDs [18].

However, the data available from the included studies were published by high-volume and specialized centers that could perform LMLR as well as OMLR. At the same time, strict eligibility criteria were used to ensure the quality of included studies upon extensive search of the available literature. The Meta-analysis of observational studies in epidemiology (MOOSE) guidelines recommended by Stroup et al. [18] and the NOS were used for assessing study quality and risk of bias, and publication bias was minimal. Furthermore, the timing of this meta-analysis was inadequate since the global use of LMLR has increased dramatically in the past 10 years, as well as the amount of available data on LMLR and OMLR in HCC.

Following a meta-analysis of comparative cohort trials, our results revealed that in comparison with OMLR, LMLR may cause reduced bleeding, decreased postoperative morbidity, and shorter hospitalization in HCC; however, LMLR had prolonged operative time. The long-term oncological outcomes assessed were comparable in both groups. Retrospective studies and the small sample sizes of several included studies may decrease the reliability of these results. Therefore, large multi-center, prospective randomized trials are required to further assess the surgical and oncological outcomes of LMLR versus OMLR.