Study selection and risk of bias assessment
A total of 397 studies were identified from PubMed and 380 studies were searched from Web of Science online database by the previously mentioned electronic search strategy up to December 31, 2019. Upon a manual search and inspection of the reference lists of other systematic reviews and meta-analyses identified 56 additional relevant studies. After exclusion of duplicates, irrelevant studies or unoriginal studies, there were 127 studies remained and assessed for eligibility by screening the full text. Finally, 20 full-text studies reporting comparative clinical outcomes of U-VATS versus M-VATS met the inclusion criteria and were suitable for meta-analysis. The PRISMA ﬂow chart describing the process of study selection is shown in Figure 1.
The studies selected for this meta-analysis were conducted in 6 countries which were published before 31st December, 2019. Among these 20 studies, four of them were prospective studies; one of them was RCT; and seven of them were PSM studies. This analysis included 4142 patients, of which 1869 patients underwent U-VATS and 2173 patients underwent M-VATS. The quality of the included studies was assessed by the NOS scale and scores ranged from 6 to 8. Table 1 summarized basic characteristics and demographics of the included studies.
In this meta-analysis, the comparison of perioperative outcomes between U-VATS and M-VATS was estimated by intraoperative outcomes (operation time, blood loss, number of resected lymph nodes, and conversion rate) and postoperative outcomes (drainage duration, length of postoperative stay, and pain in POD1). Table 2 summarized the overall outcomes of uniportal and multiportal group.
A total of 18 studies including 3699 patients provided comparative data on operative duration. The overall operation time was 146.48±55.07min and 171.70±79.40min in U-VATS and M-VATS group, respectively. The present meta-analysis revealed that there was no significant difference between U-VATS group and M-VATS group (SMD=-0.04, 95%CI = (-0.33, 0.26), P=0.81, Figure 2a). Random-effect model was used due to the high heterogeneity (I2= 94%, P<0.01). Subgroup analysis of PSM&RCT studies further confirmed the comparable operation time between two approaches (SMD=0, 95%CI = (-0.21, 0.22)).
Blood Loss was reported in 14 studies with a combination of 2964 patients. The overall blood loss was 74.49±109.03mL and 95.48±133.67mL in U-VATS and M-VATS group, respectively. The present meta-analysis indicated that there was no significant difference in blood loss between U-VATS group and M-VATS group (SMD=-0.14, 95%CI = (-0.35, 0.06), P=0.18, Figure 2b). Random-effect model was used due to the high heterogeneity (I2=86%, P<0.01). Subgroup analysis showed similar results in the PSM&RCT group (SMD=-0.22, 95%CI = (-0.54, 0.10)).
Number of Resected Lymph Nodes
Totally, 15 studies including 3009 patients reported the comparative outcomes of number of resected lymph nodes, which were 17.28±9.46 and 18.31±10.17 in U-VATS and M-VATS groups, respectively. There was no significant difference in number of resected lymph nodes between U-VATS and M-VATS group (SMD=0.03, 95%CI= (-0.08, 0.13), P=0.62, Figure 2c). These results were further confirmed in the PSM&RCT subgroup (SMD=0.03, 95%CI = (-0.08, 0.14)). Random-effect model was used due to the moderate heterogeneity (I2=25%，P=0.20).
In all, there were 13 studies including 2733 patients reporting conversion rate, which was defined as the rate of conversion to thoracotomy or need extra incisions. In U-VATS group, the total conversion rate was 6.18%, while the total value was 4.34% in M-VATS group. The meta-analysis result of conversion rate showed that there was no significant difference between U-VATS and M-VATS group (OR=1.27, 95%CI= (0.83, 1.94), Figure 2d). Fixed-effect model was used due to the low heterogeneity (I2=13%, P=0.32).
Drainage duration was defined as the period of time from the operation date to the extubation date. A total of 18 studies with a combination of 2743 patients provided comparative data on the length of drainage. The overall duration of drainage was 3.90±2.94 days and 4.44±3.12 days in U-VATS and M-VATS group, respectively. There was no significant difference between these two groups in drainage duration (SMD=-0.13, 95%CI = (-0.27, 0.02), P=0.09, Figure 3a). These results were further confirmed in the PSM&RCT subgroup (SMD=-0.12, 95%CI = (-0.30, 0.07)). Random-effect model was used due to the high heterogeneity (I2= 68%, P<0.01).
Length of postoperative stay
There were totally 12 studies including 2490 patients reporting length of postoperative stay. The overall postoperative hospital stay was 5.67±4.00 days in U-VATS group and 7.25±5.10 days in M-VATS group. The present meta-analysis indicated that the length of postoperative stay has no significant difference between U-VATS and M-VATS group (SMD=-0.37, 95%CI= (-0.81, 0.08), P=0.22, Figure 3b). According to subgroup analysis, length of postoperative stay in PSM&RCT group was (SMD=-0.01, 95%CI = (-0.18, 0.16)). By the virtue of moderate heterogeneity (I2=64%, P<0.01), random-effect model was applied to this analysis.
Pain in POD1
A total of 5 studies including 547 patients provided comparative outcomes on pain scoring in postoperative day 1 (POD1). All these five included studies utilized the visual analogue scale (VAS) to evaluate pain in POD1, and the overall value of VAS was 3.94±1.68 and 3.59±2.76 in U-VATS and M-VATS group, respectively. Since all these studies utilized the same method to assess pain in POD1, the effective valuables of pain in POD 1 were estimated by WMD. The present meta-analysis indicated that the value of VAS has no significant difference between U-VATS and M-VATS group (WMD=-0.78, 95%CI= (-1.61, 0.05), P=0.07, Figure 3c). Random-effect model was applied due to the high heterogeneity (I2=97%, p<0.01).
Funnel plots were utilized to graphically describe the publication bias of included studies in operation time, blood loss, number of resected lymph nodes, conversion rate, drainage duration, and length of postoperative stay. All funnel plots (See in Supplementary materials) showed a good symmetric distribution. Then Egger’s test and Begg’s test were used to quantize the publication bias, which demonstrated that there was no significant bias in each outcome.