Laparoscopic versus Open Surgery for Hepatocellular Carcinoma: A Meta-Analysis of High-Quality Case-Matched Studies

Objective To present a meta-analysis of high-quality case-matched studies comparing laparoscopic (LH) and open hepatectomy (OH) for hepatocellular carcinoma (HCC). Methods Studies published up to September 2017 comparing LH and OH for HCC were identified. Selection of high-quality, nonrandomized comparative studies (NRCTs) with case-matched design was based on a validated tool (Methodological Index for Nonrandomized Studies) since no randomized controlled trials (RCTs) were published. Morbidity, mortality, operation time, blood loss, hospital stay, margin distance, recurrence, and survival outcomes were compared. Subgroup analyses were carried out according to the surgical extension (minor or major hepatectomy). Results Twenty studies with a total of 830 patients (388 in LH and 442 in OH) were identified. For short-term surgical outcomes, LH showed less morbidity (RR = 0.55; 95% CI, 0.47~0.65; P < 0.01), less mortality (RR = 0.43; 95% CI, 0.18~1.00; P = 0.05), less blood loss (WMD = −93.21 ml, 95% CI, −157.33~−29.09 ml; P < 0.01), shorter hospital stay (WMD = −2.86, 95% CI, −3.63~−2.08; P < 0.01), and comparable operation time (WMD = 9.15 min; 95% CI: −7.61~25.90, P = 0.28). As to oncological outcomes, 5-year overall survival rate was slightly better in LH than OH (HR = 0.66, 95% CI: 0.52~0.84, P < 0.01), whereas the 5-year disease-free survival rate was comparable between two groups (HR = 0.88, 95% CI: 0.74~1.06, P = 0.18). Conclusion This meta-analysis has highlighted that LH can be safely performed in selective patients and improves surgical outcomes as compared to OH. Given the limitations of study design, especially the limited cases of major hepatectomy, methodologically high-quality comparative studies are needed for further evaluation.


Introduction
Although the incidence of hepatocellular carcinoma (HCC) has decreased, HCC is still the fifth most common malignancy and the third leading cause of cancer-related death worldwide [1]. Since laparoscopic hepatectomy (LH) was first reported in 1996 [2,3], this treatment has been considered a landmark development in the progress of surgical treatment. However, the majority of HCC patients usually have cirrhosis and hypohepatia. Because of this, hepatectomy increases the risk of developing significant postoperative complications including ascites, hepatic failure, encephalopathy, and portal vein thrombosis [4]. There are some controversial aspects of LH for HCC including complications, postoperative recovery, and long-term survival outcomes.
During the last 6 years, a number of meta-analyses that compare LH with open hepatectomy (OH) for HCC have been published [5][6][7][8]. Although randomized controlled trials (RCTs) are the most ideal tools for meta-analysis, no RCTs on this topic have been yet conducted. These meta-analyses included the available nonrandomized comparative studies (NRCTs) to overcome the paucity of RCTs. Therefore unreliable results and little strong evidence had been presented. On the other hand, there was evidence that estimates derived from high-quality NRCTs may be similar to those derived from RCTs [9]. Also, when comparing surgical procedures, pooling of high-quality NRCTs could be as accurate as pooling of RCTs [10]. In addition, several comparative studies on this topic have been published in the last 3 years and none of the published meta-analyses included studies published 2.4. Subgroup Analysis. Because the different levels of hepatectomy can lead to different outcomes, and major hepatectomy is a technically dependent and time-consuming procedure, subgroup analyses were carried out according to surgical extensions. Included studies were assigned to 3 subgroups: minor hepatectomy, mixed hepatectomy, and major hepatectomy.

Statistical
Analysis. The risk ratio (RR) was utilized to analyze the dichotomous variables, and the weighted mean difference (WMD) was utilized to assess the continuous variables. If the study provided medians and ranges instead of means and standard deviations (SDs), we estimated the means and SDs as described by Hozo et al. [12]. Heterogeneity was evaluated by Cochran's statistic and Higgins 2 statistic [13]. If data was not significantly heterogeneous ( > 0.05 or 2 < 50%), the pooled effects were calculated using a fixed model. Otherwise, the pooled effects were calculated using a random model. The hazard ratios (HRs) of a 5-year overall survival rate (OS) and a 5-year disease-free survival rate (DFS) were used with a generic inverse variance metaanalysis. The log HR and its SE were estimated using the method introduced by Tierney et al. [14]. According to the overall morbidity, potential publication bias was determined by carrying out an informal visual inspection of funnel plots. A two-tailed value of < 0.05 was considered significant. All statistical tests were performed with Review Manager version 5.1 (The Cochrane Collaboration, Oxford, England).

Study Characteristics.
A total of 830 patients were included in the analysis with 388 undergoing LH (46.8%) and 442 undergoing OH (53.2%). The characteristics of these included studies are summarized in Table 1. Studies were well matched in terms of age, gender, ASA classification, body mass index (BMI), tumor size, and surgical extension. Eight studies reported only minor hepatectomy, and three studies focused on major hepatectomy, whereas the remaining nine studies included both minor and major hepatectomy. The majority of studies graded morbidity according to the Clavien-Dindo classification, with the study by Lee et al. being the only exception [24]. The assessments of the NRCTs are illustrated in Table 2. Each trial received more than 12 points (the maximum possible score is 16) and was considered to be of the highest quality (see Supplementary Materials). Articles retrieved for full-text evaluation (n = 63) Articles for quality assessment (n = 26)  Articles suitable for meta-analysis    [21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39][40]. Statistically significant between-study heterogeneity was identified in all subgroups ( < 0.01, 2 = 87.2%). There was no significant difference between the groups' operation times (Table 3). However, in the subgroup of major hepatectomy, the overall effect size of the mean operation time was significantly longer in LH than that in OH (WMD = 77.93 min, 95% CI: 40.45∼115.41, < 0.01).

Tumor Size.
Only one study did not report tumor size [30]. There was trifling heterogeneity between subgroups, mainly due to the major hepatectomy subgroup ( = 0.35, 2 = 4.2%) ( Table 3). Meta-analysis showed that the tumor size of OH was longer than that of LH with a marginal difference (WMD = −0.19 cm; 95% CI: −0.41∼−0.03, = 0.09), which was mainly due to smaller tumors in LH than those in OH in the major hepatectomy subgroup (Table 3).
Although statistical significant between-study heterogeneity was identified in each subgroup, there was no evidence of heterogeneity between subgroups ( = 0.63, 2 = 0%). On    pooling the results, the margin distance was longer in the LH group than that in the OH group (WMD = 2.61 cm; 95% CI: 1.06∼4.17, < 0.01) ( Table 3).

Overall Survival Rate and Disease-Free Survival Rate.
Summary of follow-up time, recurrence, and long-term survival rates is listed in Table 4. Nineteen studies reported the detailed long-term outcomes. Among them, the data for 5-year OS rates can be extracted from nine studies and the data for 5-year DFS rates can be extracted from ten studies. The follow-up periods in six studies were less than five years. The survival data of three studies cannot be extracted due to a technical problem with figures. Unfortunately, none of the three major hepatectomy studies can be included in our survival analysis [37,39,40]. In all, the pooled 5-year OS rate           was slightly better in LH than in OH (HR = 0.66, 95% CI: 0.52∼0.84, < 0.01) (Figure 4(a)). The 5-year DFS rate was comparable between groups (HR = 0.88, 95% CI: 0.74∼1.06, = 0.18) (Figure 4(b)).

Publication Bias.
The study by Lau et al. was outside the funnel [30], and the remaining representative plots were distributed symmetrically. We believed such publication bias was acceptable in the studies ( Figure 5).

Discussion
This meta-analysis selected and summarized high-quality literature that compared the short-and long-term outcomes of LH and OH for the treatment of HCC. All of the studies had case-matched design and were of high quality according to the modified MINORS scale. For short-term surgical outcomes, LH exhibited advantages in terms of blood loss, hospital stay, overall postoperative morbidity, and mortality, whereas no statistically significant differences were identified regarding operation time. As for oncological outcomes, R0 and survival rates of LH were also not inferior to OH. To date, there have been several meta-analyses comparing LH to OH for HCC [5][6][7][8]. The results have demonstrated that LH is comparable to OH regarding the operation time and postoperative mortality and is associated with less blood loss, as well as a shorter hospital stay (        meta-analyses included all available research [5,6,8] but had some limitations. Pooling of low-quality studies could undermine the strength of results, whereas selectively pooling high-quality NRCTs could strengthen the power of results [10]. Patients' characteristics and surgical extension have a major impact on the surgical outcomes of hepatectomy. Previous meta-analyses pooled studies, which did not balance the combined factors of tumor size, location, the severity of cirrhosis, and other underlying liver diseases between LH and OH. These factors would have influenced the decision of surgeons and patients and further influenced the major factors of both short-and long-term outcomes. In addition, previous meta-analyses studied LH confined to minor resection. With the accumulation of surgical techniques, major resection of LH has become more commonly performed, but various efficacy and safety concerns for the procedure are warranted. Furthermore, since the publication of previous meta-analyses, several notable clinical observational studies have become available and some of them are from China, where HCC has the highest prevalence in the world [8,19]. Therefore, our comprehensive meta-analysis will contribute to a more systematic and objective evaluation of the safety and HCC treatment of LH.
Several previous studies have demonstrated that LH can be feasible and beneficial for minor resections or nonanatomical resections of peripheral HCC. This is in accordance with our study that showed minor resection of LH with similar operation time and less blood loss than OH. However, minor hepatectomy is insufficient for large lesions or those located in posterosuperior liver segments to ensure an adequate resection margin and eliminate intrahepatic recurrence. Major hepatectomy is more frequently performed with a curative intent for multifocal or large size HCC or those with a high propensity to invade the portal vein branches [41][42][43]. Laparoscopic major hepatectomy is, because of the same steps and principles used in laparotomy, technically demanding. Mobilization of a heavy as well as fragile organ, excisions of bulky parenchyma, and major vascular dissection with its associated risk of major vessel injury are all considered risky under laparoscopy. As expected, the present study revealed longer operation times in laparoscopic major hepatectomy. Furthermore, unlike minor resections, the blood loss of laparoscopic major hepatectomy was not superior to its open counterpart.
Patients with HCC and concurrent cirrhosis tend to have higher incidences of postoperative complications and of greater severity. Therefore, the decreased complications in the LH group should be our most striking finding. In detail, postoperative ascites and liver failure tend to decrease in LH. Postoperative decompensation after hepatectomy occurs more frequently in patients with liver cirrhosis or portal hypertension, even for limited resections. The minimization of surgical incision and the subsequent preservation of abdominal wall circulation and lymphatic flow can explain fewer ascites and liver failure in LH. Moreover, a small incision limits the evacuation of ascites through the wall and decreases the risk of infection, thus facilitating wound healing. Laparoscopic surgery also decreases the manipulation of abdominal organs and exposure of bowels, which will also contribute to reduced ascites. Since refractory ascites and progressive liver insufficiency are major causes of severe postoperative morbidities, reduced severe postoperative morbidities and mortality could be expected. Major surgery was often thought to be unsuitable for those with severely impaired pulmonary function due to a higher risk of postoperative respiratory complications. Hepatectomy involving multiple systems, especially the water and electrolyte balance, is a major risk factor for medical complications. It was observed from the reviewed studies that respiratory complications were the most common medical complications, mainly pulmonary infection, followed by cardiovascular complications. Improved preservation of liver functions in LH maintains enough albumin synthesis and decreases the pleural effusion. The pain caused by large incisions, as well as the use of tension sutures and abdominal bandages after laparotomy, can make it difficult for patients to cough. Earlier postoperative ambulation in the laparoscopic group also helped to reduce respiratory complications and promote the postoperative recovery of gastrointestinal function. In accordance with other laparoscopic surgeries, LH achieved enhanced postoperative recovery. The postoperative hospitalization of LH decreases by more than two days. This can be explained by the milder surgical trauma of LH and subsequent faster bowel recovery. Less postoperative morbidities also contribute to shorter length of hospitalization.
The oncologic results of LH for HCC remain a matter of debate. Adequate surgical margins independently improve the long-term oncological outcomes. Our analysis showed that LH could achieve enough surgical margins (more than 2 cm) as OH. The 5-year OS and DFS also showed that LH was comparable to OH. However, the results warrant prudent interpretation because of the discrepancies among the pooled studies, such as tumor size, tumor number, and status of the vascular invasion. Other biases lie in other factors including preoperative TACE and postoperative adjuvant therapies. Unfortunately, none of the three major hepatectomy studies can be included in our survival analysis. Thus, welldesigned RCTs, that balance all potential factors, preferably containing major resection are needed to confirm our results.
In the process of our research and manuscript review, two similar articles by Sotiropoulos et al. were published [44,45], which also had limitations. Examples include pooling the low-quality studies together, failing to evaluate extension on surgical outcomes, and one paper only investigating studies conducted in Europe [45]. Besides, since these studies were published, several clinical observational studies have become available. Therefore, our comprehensive meta-analysis will contribute to a more systematic and objective evaluation of this subject.
The major limitation of this study was that all included studies are NRCTs and of retrospective design. NRCTs have potential biases that limit an unequivocal conclusion, even though we exclusively included the case-matched studies to minimize the selection biases. Another limitation is the lack of studies on laparoscopic major hepatectomy. The analysis was based on only three pooled studies. Little is known about how these results would hold for a larger sample size, which is particularly important as a fair number of patients with HCC are treated with open major hepatectomy. In addition, data from several studies are extracted using the methods reported by Hozo et al. and Tierney et al., which are not completely accurate and result in bias. Moreover, it is quite possible that surgical teams undertaking research and publishing their results are more experienced and more skillful than others. Publication bias was inevitable since one plot was outside the funnel. The bias would be overcome only with the collection of more reports.

Conclusions
This meta-analysis has highlighted that LH can be safely performed in select patients and improves surgical outcomes when compared to OH. The data indicate that laparoscopic minor hepatectomy is acceptable with less blood loss, less postoperative morbidity, shorter hospitalization, and comparable operation times and oncological outcomes. The role of laparoscopic major hepatectomy is promising in terms of decreasing postoperative morbidity and recovery, but the technique also has drawbacks in prolonged operation time. Given the heterogeneity of the patient groups, the limitations of study design, and the small sample size, it is likely that patients have potential to benefit from LH, but further well-designed studies are needed to accurately select them.