Study inclusion
Through systematic searching, a total of 1473 potentially relevant articles (including 867 duplicate studies) were identified. After removing the duplicates, we browsed the titles and abstracts of the remaining 606 articles and excluded 528 studies subsequently for these reasons: (1) cases, reports, reviews, comments, and editorials; or (2) irrelevant researches. Next, we conducted a full-text search of the remaining 78 literatures and read their contents carefully. After repeated confirmation by three researchers (SW. S, X. H and XB. F), most of them (n=63) were removed further due to these reasons: (1) had no controls; (2) contained subtentorial hemorrhage, inappropriate outcome indicators; (3) did not provide the full text. Finally, only 15 studies,15,18,24-36 including 2600 supratentorial ICH patients, met the inclusion criteria and were accepted in this meta-analysis. These studies included one RCT and fourteen OSs and the publication time ranged from 1998 to 2021. The characteristics of the included studies are listed in Table 1. The literature selection is presented in Figure 1, according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.37
In the study of Fu et al.,29 they divided the subjects (patients with thalamic hemorrhage breaking into the ventricle) into Anteromedial group and Posterolateral group based on the location of hemorrhage. Therefore, we designed them into group A (Anteromedial group) and group B (Posterolateral group) for comparison in the same way. Besides, the SA group was divided into two groups according to the volume of the hematoma (C1: 50-80 mL; C2: 20-49 mL) in the study of Chi et al.,25 so we designed it into two groups A and B in this study (Group A: NE vs SA with C1; Group B: NE vs SA with C2). In the study of Nishihara et al.,35 they mentioned some patients with cerebellar hemorrhage. Since these patients did not belong to the research object of this study, we removed the data of patients with cerebellar hemorrhage and re-extracted the effective data.
Table 1. Basic characteristics of included studies.
Reference
|
Design
|
Case, Gender(M/F)
|
Age(y)
|
HV(ml)
|
GCS score
|
Hematoma location
|
thrombolysis
|
Time to surgery(h), Follow-up (month)
|
Quality evaluation
|
NS
|
SA
|
NS
|
SA
|
Cai et al., 2017
|
OS
|
20, 11/9
|
22, 13/9
|
59.6±10.1
|
58.7±12.4
|
≥ 20
|
7.6/8.3
|
supratentorial
|
UK
|
NG, NG
|
★★★/★★/★★★
|
Chi et al., 2014
|
Group A
|
OS
|
144, 81/63
|
306, 227/79
|
62.8±9.2
|
57.3±10.1
|
≥ 20
|
≥ 5
|
supratentorial
|
NG
|
NG, 3-6
|
★★★/★★/★★
|
Group B
|
OS
|
144, 81/63
|
169, 101/68
|
62.8±9.2
|
63.3±8.63
|
≥ 20
|
≥ 5
|
supratentorial
|
NG
|
NG, 3-6
|
★★★/★★/★★
|
Cho et al., 2006
|
RCT
|
30, 19/11
|
30, 20/10
|
56.67±8.66
|
56.56±8.98
|
≥ 25
|
9.26/10.1
|
basal ganglion
|
UK
|
< 24, 6
|
5
|
Dong et al., 2019
|
OS
|
39, 17/12
|
42, 18/24
|
50.5
|
51.3
|
30-60
|
8/7
|
Basal ganglion, Lobar
|
UK
|
< 24, 3
|
★★★/★★/★★★
|
Du et al., 2021
|
OS
|
212, 135/77
|
343, 212/131
|
> 20
|
> 20
|
> 20
|
7.88
|
basal ganglion
|
UK
|
NG, 6
|
★★★/★★/★★★
|
Fu et al., 2018
|
OS
|
61, 30/31
|
56, 25/31
|
61.6±9.2
|
65.6±8.8
|
30-60
|
8.0/8.4
|
basal ganglion
|
UK
|
< 24, 6
|
★★★/★★/★★★
|
Fu et al., 2019
|
Group A
|
OS
|
33, 17/16
|
36, 18/18
|
57
|
62
|
10-30
|
7/7
|
thalamus + ventricle
|
UK
|
5-18, 6
|
★★★/★★/★★★
|
Group B
|
OS
|
35, 18/17
|
36, 17/19
|
60
|
63
|
10-30
|
8/9
|
thalamus + ventricle
|
UK
|
7-18, 6
|
★★★/★★/★★★
|
Guo et al., 2020
|
OS
|
105, 64/41
|
304, 184/120
|
< 60 (66.7%)
≥ 60 (33.3%)
|
< 60 (58.6%)
≥ 60 (41.4%)
|
> 20
|
NG
|
basal ganglion
|
UK
|
< 24, 6
|
★★★/★★/★★★
|
Kim et al., 1998
|
OS
|
8, 4/4
|
10, 7/3
|
53.5
|
53.2
|
20-105
|
NG
|
basal ganglion
|
UK
|
< 24, 6-10
|
★★/★★/★★★
|
Li YQ et al., 2017
|
OS
|
32, 14/18
|
36, 16/20
|
60.7±8.8
|
61.3±8.4
|
> 30
|
8.6/8.6
|
Supratentorial Lobar
|
UK
|
< 24, 6
|
★★★/★★/★★★
|
Li ZH et al., 2017
|
OS
|
58, 25/33
|
54, 25/29
|
61.8±9.9
|
59.7±7.5
|
> 30
|
8.6/8.4
|
basal ganglion
|
UK
|
< 24, 12
|
★★★/★★/★★★
|
Liu et al., 2020
|
OS
|
60, 39/21
|
99, 57/42
|
18-80
|
18-80
|
≥ 40
|
≤ 8
|
basal ganglion
|
UK
|
< 24, 6
|
★★★/★★/★★★
|
Mao et al., 2020
|
OS
|
63, 47/16
|
54, 35/19
|
56.48±9.70
|
60.19±11.92
|
> 30
|
9/10
|
basal ganglion
|
UK
|
> 6, 6
|
★★★/★★/★★★
|
Nishihara et al., 2007
|
OS
|
24, 17/7
|
17, 8/9
|
69.2
|
65.5
|
6.9-130
|
6-15
|
Putamen, thalamus, subcortical
|
UK
|
< 30 days, 6
|
★★★/★★/★★★
|
Zhang et al., 2019
|
OS
|
53, NG
|
45, NG
|
40-75
|
40-75
|
≥ 25
|
8.9/8.3
|
supratentorial
|
UK
|
≤ 72, 1
|
★★★/★★/★★★
|
NS: neuroendoscopic surgery; SA: stereotactic aspiration; RCT: Randomized Controlled Trial; OS: Observational Study; GCS: Glasgow Coma Scale; NG: not given; UK: urokinase; HV: Hematoma volume
Primary outcome
GFO
We could find that there were seven studies reporting the data about GFO, as shown in Figure 2. Significant heterogeneity was observed (I2 = 71%) and we used the randomized effect model to calculate the statistic result. The proportion of patients with GFO was 60.61% in the NS group, compared with 55.61% in the SA group. The pooled OR of GFO was 1.40 (95% CI: 0.88–2.24; P = 0.15), implying that there was no significant difference between NS and SA in improving the functional prognosis of patients with supratentorial ICH.
Hematoma evacuation rate
In Figure 3, seven studies contained the date about hematoma evacuation rate. Obvious heterogeneity was found between these studies (I2 = 98%), so the random-effects model was applied. The 95% CIs of these studies were on the right side of the baseline. The difference of the hematoma evacuation rate between the two groups was obvious (WMD: 34.51; 95% CI: 22.14–46.89; P < 0.00001), suggesting that NS was superior in clearing the hematoma.
Mortality
Seventeen studies mentioned the mortality, as shown in Figure 4. Due to a low heterogeneity in each study (I2 = 40%), we used the fixed-effects model to calculate. The mortality was 10.79% in the NS group, compared to 24.83% in the SA group. The pooled OR was 0.40 (95% CI: 0.32–0.51; P < 0.00001), implying that the difference of mortality between NS and SA was significant. NS could effectively reduce the postoperative mortality.
Secondary outcome
Operation time and blood loss
In terms of operation time, 8 studies were included to analyze (Figure 5A). Obvious heterogeneity was indicated from the result (I2 = 95%). Therefore, we used the random-effects model. The pooled results showed that the difference of operation time was significant (WMD: 39.76; 95% CI: 29.57–49.94; P < 0.00001), suggesting that SA could shorten the operation time obviously.
Four studies mentioned the blood loss volume, as shown in Figure 5B. The random-effects model was accepted because of the high heterogeneity in each study (I2 = 99%). The pooled WMD of blood loss volume was 68.37 (95% CI: 34.58–102.16; P < 0.0001), suggesting that SA could reduce blood loss better than NS during the operation.
Hospital stays and ICU stays
Four studies reported the hospital stays (n = 180 in the NS group and 219 in the SA group), as shown in Figure 6A. The test of heterogeneity showed little heterogeneity between studies (I2 = 0%). The fixed-effects model was applied. The pooled WMD was 1.29 (95% CI: 0.99–1.58; P < 0.00001). Meanwhile, we conducted the sensitivity analysis to evaluate the stability of this result. Finally, it was found that the P value would change to 0.11 (WMD: 0.77; 95% CI: -0.17 to 1.72; P = 0.11) if the study of Liu et al was removed. This result was different from the original one, which indicated that the pooled result was unstable. So, there was not enough evidence to support that SA could shorten the hospital stays better than NS.
In Figure 6B, only three studies mentioned the intensive care unit (ICU) stays (n = 86 in the NS group and 83 in the SA group). Obvious heterogeneity was observed between studies (I2 = 70%). The random-effects model was applied. The pooled WMD was –2.02 (95% CI: –5.29 to 1.25; P = 0.23), suggesting that there was no significant difference in ICU stays between NS and SA. Sensitivity analysis showed that this result was stable.
Complications
As shown in Figure 7, 11 studies mentioned rebleeding. Likewise, 4 studies mentioned digestive tract ulcer, 3 studies referred to epilepsy, 10 studies referred to intracranial infection, 7 studies referred to pneumonia. No significant heterogeneity (I2 < 50%) was apparent between these studies. The fixed-effects model was adopted for this analysis. We could find that there was no statistically significant difference between NS and SA in terms of postoperative rebleeding (OR: 0.57; 95% CI: 0.33–0.99; P = 0.05), digestive tract ulcer (OR: 1.37; 95% CI: 0.77–2.45; P = 0.29), epilepsy (OR: 1.17; 95% CI: 0.43–3.16; P = 0.76), pneumonia (OR: 0.93; 95% CI: 0.46–1.88; P = 0.85). However, obvious statistical differences could be discovered in terms of postoperative intracranial infection (OR: 0.42; 95% CI: 0.23–0.76; P = 0.004). It suggested that NS could have a positive effect on preventing intracranial infection.
Qualitative assessment and publication bias
The quality evaluation of the included studies appears in Table 1. one RCT conformed to five scores, based on the Cochrane Collaboration’s risk of bias tool.23 The funnel plots were slightly asymmetric for the primary outcomes by visual inspection. Therefore, we found that the publication bias was low regarding GFO (Figure 8), mortality (Figure 10), and moderate regarding hematoma evacuation rate (Figure 9).
Subgroup analysis
Considering several variables, including year of publication, Glasgow Coma Scale (GCS) score, age, hematoma volume, hematoma location, and follow-up, might affect the primary outcomes, therefore we performed the subgroup analysis. As shown in Table 2, our results showed that year of publication, age, hematoma volume, hematoma location, and follow-up did not affect the hematoma evacuation rate significantly. However, no studies mentioned the hematoma evacuation rate in the subgroup of Initial GCS score ≤ 8. As for the outcome indicator of GFO, the difference between NS and SA was statistically significant in the subgroup of Initial GCS ≤ 8 (OR: 1.52; 95% CI: 1.11–2.07; P = 0.008), suggesting that NS might be more advantageous in improving functional prognosis for patients with more severe condition at admission (GCS ≤ 8). Similarly, the significant difference in GFO was found in the subgroup of hematoma location (basal ganglion 100%) (OR: 1.43; 95% CI: 1.07–1.90; P = 0.01). It implied that NS was more effective in improving the prognosis for basal ganglia cerebral hemorrhage alone. However, as for hematoma elsewhere, the advantages between NS and SA needed to be carefully treated and further studied. In addition, there was no significant difference in the pooled data about GFO in the subgroups of publication time, mean age, hematoma volume and follow-up, suggesting that these factors had no significant effect on the postoperative prognosis. In the subgroups related to mortality, these factors including Year of publication, age, Hematoma volume and follow-up all had significant effects on mortality. In the subgroup of Year of publication before 2010, the combined data suggested that there was no significant difference in mortality between NS and SA (OR: 0.35; 95% CI: 0.07–1.63; P = 0.18). Similarly, no significant difference about mortality (OR: 0.70; 95% CI: 0.44–1.11; P = 0.13) was found in the subgroup of mean age > 60-year. Through further research, we were surprised to find that the mortality in the subgroup of mean age > 60-year was significantly lower than that in the subgroup of mean age ≤ 60-year (15.96% vs. 29.85%), however little change in NS group (11.60% vs. 10.74%). This outcome suggested that age might be an important factor influencing postoperative mortality in SA. Besides, the result showed no significant difference in mortality between NS and SA in the subgroup of Hematoma volume more than 50ml (OR: 0.68; 95% CI: 0.35–1.34; P = 0.27). However, both surgical groups had significantly higher mortality rates in the subgroup of Hematoma volume greater than 50ml than in the subgroup of Hematoma volume less than 50ml (NS: 15.47% vs. 2.69%; SA: 24.09% vs. 7.61%). It replied that Hematoma volume might be also an important factor affecting mortality. Finally, in the subgroup of Follow-up > 6-month, the data indicated that the variable of follow-up time also had an impact on mortality (OR: 1.04; 95%CI: 0.45–2.38; P = 0.93). However, due to the limited number of studies eligible for subgroup inclusion requirements (only two studies), the validity of the results suffered a bit of influence and further verification was needed.
Table 2. Subgroup analysis for primary outcomes
Subgroup
|
Good Functional Outcome
|
Hematoma Evacuation Rate
|
Mortality
|
N
|
OR (95%Cl)
|
P
|
N
|
WMD (95%Cl)
|
P
|
N
|
OR (95%Cl)
|
P
|
Year of publication
|
|
before 2010
|
2
|
1.12 (0.40-3.18)
|
0.83
|
2
|
21.97 (1.31-42.62)
|
0.04
|
3
|
0.35 (0.07-1.63)
|
0.18
|
after 2010
|
5
|
1.45 (0.85-2.49)
|
0.17
|
5
|
39.30 (26.71-51.88)
|
< 0.00001
|
14
|
0.44 (0.29-0.66)
|
< 0.0001
|
Initial GCS score
|
|
≤ 8
|
2
|
1.52 (1.11-2.07)
|
0.008
|
-
|
-
|
-
|
4
|
0.28 (0.20-0.40)
|
< 0.00001
|
> 8
|
3
|
1.30 (0.46-3.70)
|
0.62
|
6
|
34.74 (19.25-50.23)
|
< 0.0001
|
9
|
0.69 (0.48-0.99)
|
0.04
|
Age (y)
|
|
≤ 60
|
3
|
1.32 (0.97-1.81)
|
0.08
|
2
|
13.68 (9.56-17.80)
|
< 0.00001
|
8
|
0.33 (0.24-0.45)
|
< 0.00001
|
> 60
|
2
|
2.23 (0.85-5.83)
|
0.10
|
4
|
46.62 (39.34-53.91)
|
< 0.00001
|
5
|
0.70 (0.44-1.11)
|
0.13
|
Location
|
|
basal ganglion
|
4
|
1.43 (1.07-1.90)
|
0.01
|
4
|
31.28 (6.81-55.75)
|
0.01
|
8
|
0.35 (0.20-0.59)
|
< 0.0001
|
supratentorial
|
3
|
1.37 (0.44-4.30)
|
0.59
|
3
|
38.99 (25.81-52.17)
|
< 0.00001
|
9
|
0.64 (0.43-0.95)
|
0.03
|
Hematoma volume (ml)
|
|
≤ 50
|
3
|
1.03 (0.54-1.97)
|
0.93
|
4
|
34.55 (17.31-51.79)
|
< 0.0001
|
8
|
0.37 (0.16-0.88)
|
0.02
|
> 50
|
2
|
1.00 (0.50-2.02)
|
0.99
|
2
|
46.49 (38.52-54.45)
|
< 0.00001
|
4
|
0.68 (0.35-1.34)
|
0.27
|
Follow-up
|
|
≤ 6-month
|
6
|
1.42 (0.87-2.32)
|
0.16
|
6
|
33.37 (19.59-47.16)
|
< 0.00001
|
14
|
0.37 (0.29-0.48)
|
< 0.00001
|
> 6-month
|
1
|
1.11 (0.16-7.51)
|
0.91
|
1
|
41.50 (33.40-49.60)
|
< 0.00001
|
2
|
1.04 (0.45-2.38)
|
0.93
|
CI: confidence interval; OR: odd ratio; WMD: Weighted Mean Difference; GCS: Glasgow Coma Scale