Through a comprehensive search in PubMed, EMBASE, and Cochrane Library databases, a total of 694 citations were identified. Subsequently, after excluding 198 duplicate articles, our analysis removed 155 citations including case reports, reviews, conference papers, animal experiments, and research on children (Figure 1). Base on screening the title and abstract, an additional 322 citations were excluded, and finally 19 unique citations entered the full-text review. In order to retain the most recent and complete data, three studies based on the same population were eliminated after the further discussion by the two authors (Diyu Chen and Xiaode Feng). Eventually, 11 eligible cohort studies were included in this analysis.15,19,28-36
Characteristics of included studies and assessment of methodological quality
The 11 eligible retrospective-prospective cohort studies were carried between 2001-2020, including a total of 1031 patients. All studies were single-center studies, most of them (7/11) were performed in Asian populations (Japan, Korea, and India), and the rest (4/11) were based on Western populations (Germany, Italy, and USA). Seven studies were published before 2014, and four after 2014.Nine cohorts involved in our study represent the usage of PVE, seven of which conducted PVE before RH, while two of which conducted it not only before RH but also before LH. As for caudate lobectomy, all patients underwent caudate lobectomy in six of eleven papers, and the four studies only performed a caudate lobectomy in part of patients and the remaining one study describe unclearly about this. The specific characteristics and data of the included studies were shown in the Table 1.
After quality evaluation, the scores of our included studies ranging from 6 to 8, based on the Newcastle-Ottawa Scale. As shown in the e-Table 1, there were 3 Literatures with <7 points and 8 Literatures with ≥7 points.
Primary outcomes: overall survival
In order to evaluated the prognosis of patients in different surgery, we analyzed the overall survival data from ten cohorts (including 859 patients) in this study, and the data were visualized by forest plots in Figure 2. The pooled HR estimated based on the fixed-effect model and the random-effect model were both 1.27（95％CI，0.98-1.63; P =0.066）, indicating that the difference between LH and RH was not statistically significant. No significant difference was observed among studies in the estimates for OS (I2= 0%, P heterogeneity= 0.840). Subgroup analysis showed that the analysis results did not change due to the region, year of publication and the number of cases of left-side hepatectomy, and the differences between LH and RH remained not significant (Table 2).
1-, 3-, and 5-Year survival rates
1-year survival data included in five studies containing 576 patients, it was 79.0% (188 of 238) in the LH group and 78.7% (266 of 338) in the RH group；3-year survival data included in seven studies containing 662 patients, it was 46.2% (121 of 262) in the LH group and 49.0% (196 of 400) in the RH group; 5-year survival data included in eight studies containing 798 patients, it was 28.8% (95 of 330) in the LH group and 35.5% (166 of 468) in the RH group. The results of the pooled 1-year, 3-year, and 5-year survival rates for LH vs. RH are shown in Figure 3. Based on the random-effects model, the pooled RR for the 1-year survival rates was 1.01（95％CI，0.89-1.15;P = 0.835）; the pooled RR for the 3-, and 5-year survival rates calculated using the fixed-effects model were 0.94（95％CI，0.80-1.11;P = 0.49）, and 0.82（95％CI，0.67-1.01;P =0.067）, respectively.
The results indicated that there was no statistically significant difference in the 1-, 3-, and 5-year survival rates between LH and RH. All studies on 1-year survival had no obvious heterogeneity (I2= 48.3%, P heterogeneity= 0.102). No statistically significant heterogeneity was observed for all studies on 3-year and 5-year survival rates (I2= 0%, P heterogeneity= 0.519; I2= 0%, P heterogeneity= 0.643, respectively).
Subgroup analysis demonstrated that despite the different publication years and the number of cases of left-side hepatectomy, the results of 1-,3- and 5-year survival showed no obvious difference, which was the same as the results of 1- and 3-year survival under the subgroup of different regions. However, patients undergoing LH in western centers were associated with poor 5-year survival results (Table 2).
Overall postoperative morbidity and major postoperative morbidity
Five studies with 590 patients provide information on overall postoperative morbidity, with rates of 50.4% (132 of 262) in the LH group and 61.9% (203 of 328) in the RH group. As shown in Figure 4A, the pooled RR was 0.82(95％CI, 0.71-0.96; P = 0.014), and the overall morbidity of the LH group was significantly lower than that of the RH group. The heterogeneity between the studies was not obvious (I2= 13.9%, P heterogeneity= 0.323). Major postoperative morbidity was mentioned in five studies with 315 patients, major morbidity occurred in 34.5% (48 of 139) of patients in the LH group and 45.5% (80 of 176) in the RH group. The pooled RR was 0.73 (95％CI, 0.56-0.95; P = 0.020; Figure 4B). The results suggested that RH group had a higher risk of serious postoperative complications, and there was no heterogeneity among the studies (I2= 0%, P heterogeneity= 0.544).
Subgroup analysis indicated that LH was associated with reduced overall morbidity in post-2014, Western Central studies and less experienced centers (≤ 41cases). However, in Eastern Center and pre-2014 studies, there was no relationship between the two procedures and overall morbidity. All major morbidity data were collected from the studies published after 2014. The results of the Western Center studies and less experienced centers were consistent with the meta-analysis, but no significant differences were observed in the Eastern Center studies (Table2).
Post-hepatectomy liver failure and procedure-related mortality
Four studies reported the data about post-hepatectomy liver failure in 373 patients. In the LH and RH group, PHLF rate was 2.5% (4 of 161), and 12.7% (27 of 212) respectively. Figure 5A shows the pooled results of the fixed effects model, the pooled RR for PHLF was 0.22(95％CI, 0.09-0.56; P = 0.002). These results showed that performing LH could reduce the possibility of post-hepatectomy liver failure. Nine studies with 976 patients reported perioperative mortality. The mortality rates in the LH group and the other group were 3.9% (16 of 411) and 8.8% (47 of 535), respectively. As depicted in the forest plots, the pooled RR was 0.41(95％CI, 0.23-0.70; P = 0.001), LH significantly reduces perioperative mortality relative to RH (Figure 5B). For post-hepatectomy liver failure and postoperative mortality, no heterogeneity was observed between different studies (I2= 0%, P heterogeneity= 0.625; I2= 0%, P heterogeneity= 0.954, respectively).
In subgroup analysis, the results of the Eastern Center and less experienced centers showed that LH can reduce the incidence of PHLF. Regarding mortality, regardless of changes in region and publication year, LH was significantly associated with lower mortality. And in centers where LH was performed in more than 41 cases, the mortality rate was lower with LH (Table 2).
R0 resection rate
A total of 7 studies reported R0 resection rate of 885 HCCA patients. In the LH group, 70.8% (267 of 377) of patients achieved negative margin, while in the RH group, the data was 76.2% (387 of 508). The pooled analysis results showed that the RR of R0 resection rate was 0.95（95％CI，0.87-1.03; P = 0.179）without heterogeneity (I2= 0%, P heterogeneity= 0.607; Figure 6A). No statistical difference in R0 resection rate between LH and RH was identified. Subgroup analysis showed that the results of the Western Center were inconsistent with the meta-analysis, that is, a higher R0 resection rate could be obtained by RH (Table 2).
A total of 846 patients reported operating time in nine studies. Based on the fixed-effects model, there was a low level of heterogeneity between the studies (I2 = 45.1%, P heterogeneity =0.078). Considering I2 critical 50%, the random-effects model was used to pooled the studies in a more conservative way. As shown in Figure 6B, the pooled MD was 38.68(95％CI,7.41-69.95; P = 0.015), indicating that the operation time in the LH group was significantly longer than that in the RH group.
Figure 7 shows a funnel plot of OS. Neither the Begg's test nor the Egger's test found significant publication bias, that is, the P values for the outcome was greater than 0.05. Since the number of studies included in other endpoints in the meta-analysis was small, funnel plots, Begg's Test, and Egger's test were not performed to assess publication bias.