Hepatocellular carcinoma is the third most common causes of cancer-related death in the world [1]. Liver resection is considered the most widely used treatment for patients with hepatocellular carcinoma (HCC) [2]. Laparoscopic liver resection (LLR) has increasingly been adopted since it was first introduced in 1991 [3]. There has been a growing body of evidence that LLR is associated with lower mortality and morbidity rates in comparison with open liver resection (OLR) [4]. However, LLR is still a challenging approach for patients as well as surgeons since most HCC are developed on top of cirrhotic liver or chronic hepatitis [5,6,7]. Furthermore, LLR is associated with technical difficulties, relatively longer operative time and bleeding risk in parenchymal resection [8, 9]. Although many studies have been conducted comparing LLR versus OLR for HCC, data are still relatively controversial, given the recent advancements in laparoscopic devices and techniques over the last few years and the increase in surgeons’ experience [10,11,12,13,14,15,16]. The recently published meta-analysis comparing between LLR and OLR did not differentiate between propensity score-matched studies and unmatched studies, therefore, its results might have an inherent risk of confounding bias limiting the causal relationship between both interventions and the observed clinical outcomes [17, 18]. Moreover, several studies have reported favorable clinical outcomes associated with LLR for right HCC, however, the efficacy of LLR in right hepatectomy has not been well established [19, 20]. Therefore, we sought to conduct a systematic review and meta-analysis of studies comparing LLR versus OLR in patients with right and left HCC and reported clinical outcomes in propensity-score matched and unmatched cohorts, to mitigate the potential risk of confounding bias, to identify the safety and efficacy of LLR in terms of surgical and oncological outcomes.

Methods

This systematic review and meta-analysis is reported according to the Preferred Reporting Items for Systematic review and Meta-Analysis (PRISMA) [21]. MEDLINE and Cochrane Central Register of Controlled Trials database were systematically searched through July 2018 using the following search terms: (1) hepatocellular carcinoma, liver neoplasms, liver cirrhosis/complications, malignant liver tumor; (2) hepatectomy, open liver resection, open hepatectomy (3) laparoscopy, laparoscopic hepatectomy. We had no restrictions on studies’ design. We included prospective and retrospective observational studies and randomized controlled trials (RCTs). Studies had to fulfill the following prespecified criteria to be considered qualified for inclusion in our meta-analysis; (1) comparing in a head to head fashion between LLR and OLR in patients with right or left HCC; (2) reported oncological and surgical outcomes, as mentioned below, in both treatment groups; (3) included patients with primary HCC. In case of duplicate publications reflecting the same population, we included the report with the longer follow-up duration. Our outcomes of interest were operative time in minutes (min), amount of blood loss in milliliter (ml), rate of blood transfusion, length of hospital stay in days, 30-days mortality, morbidity, recurrence rate, resection margin in centimeter (cm), R0 resection, 1-year over-all survival (1-OS), 3-years overall survival (3-OS), 5-years overall survival (5-OS), 1-year disease-free survival (1-DFS), 3-years disease-free survival (3-DFS), and 5-years disease-free survival (5-DFS) rates.

Two reviewers (Meng and Xu) independently screened databases for relevant studies based on the abovementioned criteria. After title and abstract screening, the full text of the selected articles was evaluated for eligibility. The two reviewers independently extracted the relevant data in a standardized extraction form, and third reviewer’s opinion was sought in case of disagreements (Duan). We performed a subgroup analysis restricted to studies reported clinical outcomes in propensity-score matched cohorts to mitigate the potential risk of confounding bias in data gleaned from observational studies. We conducted a subgroup analysis focused on studies exclusively recruited patients with right HCC to investigate the role of LLR in comparison with OLR in right HCC.

This meta-analysis is exempt from the need for IRB approval.

Statistical analysis

For categorical variables, summary estimates were expressed as odds ratio (OR) with corresponding 95% confidence interval (CI), and for continuous variables, summary estimates were expressed as mean difference (MD) with corresponding CI. The OR of our outcomes of interest was calculated according to the DerSimonian-Laird random effects model [22]. Heterogeneity across studies was assessed using Q-statistic and I2 –statistic [23]. The I2 statistic describes the percentage of total variation across studies that is due to heterogeneity. Sensitivity analyses were performed using the one-study-removed method to show how the summary estimate changes if the study that has the largest effect size is removed. Egger’s regression test and visual inspection of funnel plots were used to assess for potential publication bias since studies with statistically significant results are more likely to be published than studies with non-significant findings [24]. The statistical level of significance was 2-tailed P < 0.05. All Analyses were performed using random-effects model. All analyses were performed using Comprehensive Meta-analysis version 3.0 software (Biostat, Inc., New Jersey, USA).

Results

All studies meta-analysis

The process of studies selection is displayed in Fig. 1. Out of 1004 studies screened, the full text of 72 articles was reviewed. Fifty-one studies were identified including a total of 6812 patients, 2786 patients underwent LLR and 4026 patients were subjected to OLR. Studies’ characteristics are shown in Table 1. Patients’ characteristics are shown in Table 2. The recruitment periods of the included studies was between 1990 and 2017. The follow-up duration of the included studies ranged from 3 to 78.5 months.

Fig. 1
figure 1

Flow chart showing the process of studies screening and selection

Full size image
Table 1 Characteristics of the included studies
Full size table
Table 2 Patients’ baseline characteristics of the included studies
Full size table

The mean of operative time in minutes was reported in 26 studies including 3664 patients. There was no significant difference between LLR and OLR regarding the operative time in min (MD 18.29; 95% CI − 1.58 to 38.15; P = 0.07; I2 = 91.73%), Fig. 2. The operative blood loss in ml was reported in 19 studies including 2112 patients. Operative blood loss was significantly lower in LLR in comparison with OLR (MD − 124.09; 95% CI − 188.21 to − 59.97; P = 0.001; I2 = 94.09%). The incidence of blood transfusion was significantly lower in LLR comparing with OLR (OR 0.45; 95% CI 0.30–0.69; P = 0.001; I2 = 55.83%).

Fig. 2
figure 2

Forest plot showing: A the mean difference of operative time in minutes in laparoscopic liver resection (LLR) in comparison with open liver resection (OLR), B The mean difference of blood loss in ml in LLR in comparison with OLR, and C The rate of blood transfusion in LLR) in comparison with OLR

Full size image

The mean of hospital stay in days, 30-days mortality rate and morbidity were significantly lower in LLR comparing with OLR (MD − 3.87; 95% CI − 4.86 to − 2.89; P = 0.001; I2 = 87.35%), (OR 0.32; 95% CI 0.16–0.66; P = 0.001; I2 = 0%), and (OR 0.42; 95% CI 0.34–0.52; P = 0.001; I2 = 39.64), respectively, Fig. 3.

Fig. 3
figure 3

Forest plot showing: A the mean difference of hospital stay in days in in laparoscopic liver resection (LLR) in comparison with open liver resection (OLR), B the 30-days mortality rate in LLR group comparing with OLR group, and C The morbidity rate in LLR in comparison with OLR

Full size image

Resection margin in centimeter (cm) and R0 resection were not significantly different between LLR and OLR groups (MD 0.04; 95% CI − 0.06 to 0.14; P = 0.41; I2 = 48.03%) and (OR 1.31; 95% CI 0.98 —1.76; P = 0.07; I2 = 0%), respectively. The recurrence rate was significantly higher in the OLR group than LLR group (OR 0.83; 95% CI 0.71–0.98; P = 0.03; I2 = 0%), Fig. 4.

Fig. 4
figure 4

Forest plot showing: A the mean difference of resection margin in CM in laparoscopic liver resection (LLR) in comparison with open liver resection (OLR), B the R0 resection rate in LLR in comparison with OLR, and C The recurrence rate in LLR in comparison with OLR

Full size image

The 1-OS and 5-OS were significantly higher in LLR comparing with OLR group (OR 1.45; 95% CI 1.06–1.99; P = 0.02; I2 = 25.59%) and (OR 1.36; 95% CI 1.07–1.72; P = 0.01; I2 = 14.88%), respectively, Fig. 5. 3-OS was not significantly different between both groups (OR 1.07; 95% CI 0.70–1.63; P = 0.77; I2 = 56.6). 1-DFS was significantly higher in LLR group comparing with OLR group (OR 1.42; 95% CI 1.032–1.972; P = 0.032; I2 = 55.57%), however, 3-DFS and 5-DFS did not differ significantly between both approaches (OR 1.349; 95% CI 1.349; 0.939–1.938; P = 0.105; I2 = 78.66%) and (OR 0.79; 95% CI 0.48–1.31; P = 0.36; I2 = 81.49%), respectively, Fig. 6.

Fig. 5
figure 5

Forest plot showing: A 1-year overall survival rate, B 3-years overall survival rate, and C 5-years overall survival rate

Full size image
Fig. 6
figure 6

Forest plot showing: A 1-year disease-free survival rate, B 3-years disease-free survival rate, and C 5-years disease-free survival rate

Full size image

Propensity score-matched studies subgroup analysis

Our subgroup analysis restricted to propensity score-matched studies did not show significant difference between both methods in terms of operative time in min (MD 11.64; 95% CI − 20.02 to 43.33; P = 0.47; I2 = 87.53%), blood loos in ml (MD − 95.62; 95% CI − 206.17 to 14.93; P = 0.09; I2 = 82.32), but significantly reduced blood transfusion rate with LLR than with OLR (OR 0.54; 95% CI 0.38–0.78; P = 0.001; I2 = 0%), Fig. 7.

Fig. 7
figure 7

Forest plot showing: A the mean operative time of laparoscopic liver resection (LLR) in comparison with open liver resection (OLR) in propensity score-matched studies, B The mean blood loss in ml in LLR in comparison with OLR in propensity score matched studies, and C The rate of blood transfusion in LLR in comparison with OLR in propensity score-matched studies

Full size image

The mean of hospital stay in days, 30-days mortality, and morbidity were significantly lower in LLR in comparison with OLR (MD − 4.306; 95% CI − 5.79 to − 2.81; P = 0.001; I2 = 62.68%), (OR 0.31; 95% CI 0.11–0.84; P = 0.02; I2 = 0%), and (OR 0.51; 95% CI 0.39–0.67; P = 0.001; I2 = 33.98%), respectively, Fig. 8. Resection margin in cm and R0 resection did not differ from all-studies analysis, Fig. 9. However, recurrence rate showed non-significant difference between both approaches when the analysis was limited to propensity-score matched studies (OR 0.93; 95% CI 0.74–1.16; P = 0.50; I2 = 13.86%).

Fig. 8
figure 8

Forest plot showing: A the mean difference of hospital stay in days in laparoscopic liver resection (LLR) in comparison with open liver resection (OLR) in propensity score matched studies, B The 30-days mortality rate of LLR in comparison with OLR in propensity score-matched studies, and C The rate of morbidity in LLR in comparison with OLR in propensity score-matched studies

Full size image
Fig. 9
figure 9

Forest plot showing: A the mean difference of resection margin in CM of laparoscopic liver resection (LLR) in comparison with open liver resection (OLR) in propensity score-matched studies, B The R0 resection rate of LLR in comparison with OLR in propensity score-matched studies, and C The recurrence in LLR in comparison with OLR in propensity score-matched studies

Full size image

1-OS showed a trend favoring LLR over OLR (OR 1.53; 95% CI 0.94–2.47; P = 0.09; I2 = 39.4%). 3-OS and 5-OS did not differ from all-studies analysis, Fig. 10. 1-DFS and 3-DFS showed a trend favoring LLR over OLR (OR 1.527; 95% CI 0.99–2.34; P = 0.05; I2 = 68.13%) and (OR 1.24; 95% CI 1.01–1.53; P = 0.04; I2 = 18.18%), Fig. 11. 5-DFS did not differ from all-studies analysis.

Fig. 10
figure 10

Forest plot showing: A 1-year overall survival rate in propensity score matched studies, B 3-years overall survival rate in propensity score-matched studies, and C 5-years overall survival rate in propensity score matched studies

Full size image
Fig. 11
figure 11

Forest plot showing: A 1-year disease-free survival rate in propensity score matched studies, B 3-years disease-free survival rate in propensity score-matched studies, and C 5-years disease-free survival rate in propensity score matched studies

Full size image

Right hepatectomy subgroup analysis

When the analysis was restricted to studies of right hepatectomy, operative time was significantly lower in the OLR group than LLR group (MD 135.05; 95% CI 47.83–222.27; P = 0.001; I2 = 70.57%). Blood loss showed a trend of reduced blood loss amount with LLR than with OLR group (MD 43.88; 95% CI − 162.54 to 9.48; P = 0.08; I2 = 45.05%). Hospital stay in days and morbidity were significantly lower in LLR comparing with OLR (MD -3.96; 95% CI − 6.19 to − 1.743; P = 0.001; II2 = 83.04%) and (OR 0.16; 95% CI 0.06–0.43; P = 0.001; II2 = 0%), respectively, Fig. 12. Resection margin did not significantly differ between both approaches (MD 0.03; 95% CI − 0.20 to 0.25; P = 0.82; I2 = 60.00%).

Fig. 12
figure 12

Forest plot showing: A The mean difference of operative time in minutes in laparoscopic liver resection (LLR) in comparison with open liver resection (OLR) in right hepatectomy, B The mean difference of blood loss in ML in LLR in comparison with OLR in right hepatectomy, C The mean difference of hospital stay in days in LLR in comparison with OLR in right hepatectomy, D The morbidity rate in LLR comparing with OLR in right hepatectomy, and E the mean difference of resection margin in CM in LLR in comparison with OLR in right hepatectomy

Full size image

Our sensitivity analysis using one-study-removal approach did not show any change in any outcomes of our interest, Supplemental Figs. 1, 2, and 3. Funnel plots of publication bias didn’t show any risk of publication bias with our outcomes of interest, Supplemental Fig. 4. Egger’s test showed a significant risk of publication bias with morbidity (Egger’s regression intercept − 1.37; 95% CI − 2.06 to − 0.68; P = 0.001), R0 resection (Egger’s regression intercept 0.88; 95% CI 0.18 –.57; P = 0.015) and 5-years DFS (Egger’s regression intercept 5.56; 95% CI 3.38–7.74; P = 0.001), Table 3. Egger’s test did not show any significant risk of publication bias with any other outcome.

Table 3 Egger’s test to assess for potential publication bias
Full size table

Discussion

Our meta-analysis showed that blood loss, blood transfusion rate, 30-days mortality, hospital stay in days and recurrence rate were significantly lower in the LLR group in comparison with OLR group. 1-OS, 5-OS and 1-DFS rates showed a significantly favorable outcome associated with LLR in comparison with OLR. There was no significant difference between both groups regarding operative time in min, 3-OS, 3-DFS and 5-DFS. In a subgroup analysis restricted to right hepatectomy, our meta-analysis did not find any significant difference between both groups in terms of operative time in min and resection margin in cm. There was a strong trend, albeit non-significant, favoring LLR over OLR regarding blood loss. Hospital stay in days and morbidity were significantly lower in LLR comparing with OLR. Our subgroup analysis restricted to studies reported clinical outcomes in propensity-score matched populations showed a non-significant difference regarding blood loss and recurrence rate between both approaches, however LLR was associated with a significantly reduced blood transfusion rate.

In consistent with previous studies, we found favorable outcomes associated with LLR regarding operative blood loss and blood transfusion [18, 25]. The reduction in bleeding with LLR may be explained by the less risk of hepatic vein or vena cava injury because of the meticulous parenchymal dissection provided by the laparoscopy modality and the hemostatic effect of pneumoperitoneum which might have controlled bleeding from hepatic vein branches [26]. Furthermore, the magnification provided by laparoscopy allows better identification of small blood vessels which might have reduced the risk of blood loss and subsequently the blood transfusion rate.

In line with previous meta-analyses, LLR was significantly associated with lower postoperative 30-days mortality and shorter hospital stay in days comparing with OLR group [18, 25]. This might be explained by the less manipulation of abdominal organs, smaller incision, decreased rate of complications, less severe pain, lower need for narcotic pain medications, and early ambulation in the LLR group in comparison with OLR group.

In contradiction with previous studies, our all-studies meta-analysis showed that LLR was significantly associated with decreased recurrence rate comparing with OLR [25], which came as no surprise because of the decreased operative blood loss in the LLR group. Katz et al. concluded that intraoperative blood loss is an independent predictable factor to the tumor recurrence and survival rates [27]. Furthermore, although resection margin status was not significantly different between both groups in our analysis, there were other different preoperative and postoperative factors, such as Child–Pugh classification, amount of blood loss, and resected liver volume between LLR and OLR groups which might have influenced the recurrence rate [14, 28,29,30,31]. Therefore, the recurrence rate between both groups was not significantly different after restricting the analysis to propensity score-matched studies.

The second international consensus conference rated the quality of evidence supporting the superiority of LLR as low since most of the evidence derived from observational studies comparing LLR versus OLR [8, 32, 33]. The jury of the conference strongly recommended launching studies comparing between LLR and OLR in a head-to-head randomized fashion. Currently, there are a few ongoing randomized controlled trials randomizing patients with HCC to LLR versus OLR (NCT01768741), (NCT00606385), and (NCT02526043). Hopefully, these studies can provide a valid, non-biased evidence regarding the superiority of either strategy.

Our meta-analysis demonstrated a favorable survival rate at 5 years in LLR in all-studies analysis, and this survival benefit persisted even after restricting the analysis to propensity score-matched studies. Previous studies that investigated the correlation between resection margin and the OS and DFS concluded that resection margin can significantly predict the prognosis of HCC [34, 35]. Our meta-analysis showed a non-significant difference between both groups in terms of resection margin, therefore, the difference in survival benefits was most probably driven by other postoperative clinical outcomes, such as blood loss, postoperative complications, morbidity and 30-days mortality. Furthermore, the more manipulation and compression of the tumor during OLR might have resulted in more dissemination of tumor cells through portal vein to systematic circulation and this could have impacted the survival rate [36]. Our results are inconsistent with Poon et al’s findings that concluded that the resection margin was not associated with postoperative recurrence pattern and subsequently survival rate [37].

Our study did not find a statistically significant difference between LLR and OLR regarding 3-OS which might be explained by the limited number of studies reported 3-OS in comparison to studies reported 1-OS. Although there was a limited number of studies reported 5-OS, this might be compensated by the relatively large number of events over at 5 years follow-up.

Our meta-analysis has several limitations; (1) none of the included studies were a randomized controlled trial which makes our study has a risk of confounding and selection bias, however we run a subgroup analysis focused on propensity score-matched studies to mitigate the inherent risk of confounding bias associated with observational studies. RCTs comparing between both resection approaches are needed to accurately identify the superiority of either strategy. (2) there was an inherent heterogeneity regarding the definition of clinical outcomes across the included studies. (3) Despite our extensive literature search, we found a limited number of studies recruited patients with right hepatectomy, therefore, the results of our study should be interpreted carefully regarding the safety and efficacy of LLR in comparison with OLR in right hepatectomy. Further studies investigating the clinical outcomes of both approaches in right hepatectomy are still needed. (4) We could not run a subgroup analysis based on tumor classification since most of the studies did not stratify the HCC into stages or based on Endmondson classification and report clinical outcomes accordingly. (5) Our results regarding morbidity, R0 resection, and 5-DFS should be interpreted with caution since our analysis found a potential publication bias with these outcomes. The potential publication bias might have arisen from that tertiary care centers, that are well equipped, are more likely to publish data about LLR than community hospitals.

In conclusion, LLR was significantly associated with shorter operative time, less blood loss, less blood transfusion rate, shorter hospital stay in days, lower 30-days mortality rate, and lower morbidity. There was no significant difference between LLR and OLR regarding resection margin. There was no significant difference between both groups in terms of 3-DFS, 5-DFS, 3-OS, nevertheless LLR was associated with a significantly higher 1-OS, 5-OS and 1-DFS rates. RCTs are needed to identify the efficacy and safety of LLR in comparison with OLR in patients with HCC. Further studies investigating LLR in right hepatectomy are needed.