Search results and quality assessment
A total of 440 studies were identified for the first time from PubMed by the search strategy previously established and 269 studies were identified via other sources or review. Subsequently, 11 studies were deleted for duplication with the help of Mendeley (Elsevier Inc.; Atlanta, GA, USA). The titles and abstracts of 270 studies were then screened for inclusion. The full texts of 36 studies were read and, finally, we included 10 non-randomized controlled trial (Non-RCT) studies that met the present meta-analysis criteria(3,10,13–16,18,19,21,22). The details of PRISMA flow diagram of the literature for meta-analysis is shown in Supplementary Figure S1(23).
In present analysis, 3 studies were based on already existed database(13,15,19) and the remaining 7 studies were based on retrospective studies(3,10,16–18,21,22). 2 studies conducted SBRT or RFA for transplant intent. 5 studies were performed in USA(10,13,15,18,19), 2 in Japan(12,16), 1 in Canada(3), 1 in South Korea(17), and 1 in China(22). Out of 2732 patients from the 10 included studies 859 patients were classified into the SBRT group, and the rest 1873 patients were classified into the RFA group. The Newcastle Ottawa Scale (NOS)(24) were used to assess the quality of non-randomized studies. Among all studies included, over 50% of which were medium-score studies because of inconsistent comparability, although the sections of selection and outcomes were relatively appropriate qualities. Therefore, we believed that the present meta-analysis possesses a medium-class of quality (Supplementary Table S1).
Among included studies, 2 studies were in favor of SBRT on local control(10,21), 2 studies preferred to RFA on OS(15,19) and others reported comparable outcomes between groups(3,13,16–18,22). Notably, according to baseline characteristics, several studies enrolled patients in SBRT group were prone to suffer from larger tumor diameter(3,10,16,19) and higher proportion of Child-Pugh C patients(18) (Table 1). These groups were evaluated for therapeutic efficacy in treating HCC patients. The details of the studies included in the present meta-analysis are listed in Table 1.
Table 1 Characteristics of include studies
|
|
|
|
|
|
|
|
|
|
|
Author
|
Design
|
Group
|
Number of patients
|
Number of tumor (1/≥2)
|
Age (years)
|
Sex (M/F)
|
Tumor size (cm)
|
Median dose (Gy)
|
Child-Pugh Class (A/B/C)
|
Median follow up (months)
|
|
Mohamed et al., 2016 (American)
|
Non-RCT
|
SBRT
|
23
|
14/9
|
57.5 (44-70.2)
|
20/3
|
NR
|
50 (45-60)
|
17/0/5
|
41 (7.3-77.9)
|
|
|
|
RFA
|
9
|
8/1
|
57.5 (44-70.2)
|
8/1
|
NR
|
|
9/0/0
|
41 (7.3-77.9)
|
|
Wahl et al., 2016 (American)
|
Non-RCT
|
SBRT
|
63
|
49/14*
|
62 (35-85)
|
54/9*
|
2.2 (0-10)
|
50 (27-60)
|
43/18/2*
|
27 (0.5-86.5)*
|
|
|
|
RFA
|
161
|
109/52*
|
60 (31-81)
|
117/44*
|
1.8 (0.6-7.0)
|
|
80/68/13*
|
50.9 (3.5-112.8)*
|
|
Sapisochin et al., 2017 (Canada)
|
Non-RCT
|
SBRT
|
36
|
17/19*
|
60.4 (56.4-64.8)
|
30/5
|
4.5 (2.9-5.8)*
|
36 (30-40)
|
22/14/0*
|
28.1 (14.9-64.7)*
|
|
|
|
RFA
|
244
|
156/88*
|
57.8 (53.5-62)
|
208/36
|
2.5 (1.9-3)*
|
|
158/68/8*
|
52.2 (21-90.7)*
|
|
Hara et al., 2019 (Japan)
|
Non-RCT
|
SBRT
|
106
|
94/12
|
74 (48-93)
|
71/35
|
1.8 (1.0-3.0)
|
37.5 (35-40)
|
104/2/0
|
33.7 (0.5-75.0)
|
|
|
|
RFA
|
106
|
93/13
|
75 (47-88)
|
76/30
|
1.7 (0.7-2.8)
|
|
105/1/0
|
29.9 (6.0-72.8)
|
|
Berger et al., 2017 (American)
|
Non-RCT
|
SBRT
|
157
|
NR
|
68.61 (11.74)
|
113/44
|
4.8 (4.8)
|
NR
|
NR
|
NR
|
|
|
|
RFA
|
627
|
NR
|
68.21 (10.00)
|
454/173
|
4.2 (4.6)
|
|
NR
|
NR
|
|
Duan et al., 2016 (China)
|
Non-RCT
|
SBRT
|
37
|
NR
|
NR
|
NR
|
1-5
|
NR
|
NR
|
NR
|
|
|
|
RFA
|
40
|
NR
|
NR
|
NR
|
1-5
|
|
NR
|
NR
|
|
Kim et al., 2019 (Korea)
|
Non-RCT
|
SBRT
|
95
|
95/0
|
63.0 (35.0-85.0)
|
80/15
|
2.4 (0.7-5.5)
|
60 (52-60)
|
90/5/0
|
21.9 (11.8-31.2)
|
|
|
|
RFA
|
95
|
95/0
|
67.0 (40.0-86.0)
|
83/12
|
2.1 (0.8-4.6)
|
|
90/5/0
|
21.6 (11.1-37.3)
|
|
Rajyaguru et al., 2018 (American)
|
Non-RCT
|
SBRT
|
275
|
190/85
|
65 (55-75)
|
194/81
|
2.5 (2.5-3.5)
|
45 (45-55)
|
NR
|
25.3 (14.1-41.0)
|
|
|
|
RFA
|
521
|
349/172
|
65 (55-75)
|
381/140
|
2.5 (2.5-3.5)
|
|
NR
|
25.3 (14.1-41.0)
|
|
Shiozawa et al., 2015 (Japan)
|
Non-RCT
|
SBRT
|
35
|
35/0
|
75.1 (67-83)*
|
24/11
|
2.86 (1.2-5)*
|
50.6 (7.8)
|
28/7/0
|
12.6 (6.8-35.5)*
|
|
|
|
RFA
|
38
|
38/0
|
68.7 (42-86)*
|
27/11
|
1.75 (0.7-2.9)*
|
|
31/7/0
|
18.7 (7.4-40.8)*
|
|
Parikh et al., 2018 (American)
|
Non-RCT
|
SBRT
|
32
|
NR
|
77 (72-81.25)
|
20/12
|
NR
|
NR
|
NR
|
NR
|
|
|
|
RFA
|
32
|
NR
|
79 (76-82)
|
22/10
|
NR
|
|
NR
|
NR
|
|
Total
|
|
SBRT
|
859
|
|
|
|
|
|
|
|
|
|
|
RFA
|
1873
|
|
|
|
|
|
|
|
|
SBRT, stereotactic body radiotherapy; RFA, radiofrequency ablation; NR, not reported. *Statistically significant (P < 0.05).
|
Local progression rates
6 out of the 10 studies illustrated 1-, 2- and 3-year LP rates(10,16–18,21,22). Our pooled results showed that the SBRT group had significantly better 1- and 3-year local control rates than the RFA group (OR 0.42, 95% CI 0.24 to 0.74, P = 0.003; OR 0.54, 95% CI 0.37 to 0.80, P = 0.002, respectively). However, the 2-year LP rate showed marginal benefit of SBRT group than RFA group (OR 0.67, 95% CI 0.43 to 1.05, P = 0.080). No heterogeneity was shown among the studies of 1-, 2- and 3-year local progression rates (χ2 = 2.93, I2 = 0%; χ2 = 0.33, I2 = 0%; χ2 = 2.66, I2 = 0%, respectively), so the fixed effect model was used to pool the outcomes (Figure 1).
Overall Survival
9 studies with 2700 patients compared OS rates of SBRT group with RFA group(3,10,13,15–17,19,21,22), 1 study with 280 patients were excluded as both SBRT and RFA were applied as bridge therapies before transplantation and thus the actual OS rates of SBRT or RFA might be affected by subsequent transplantation(3). Heterogeneity was low for the studies that reported 1- and 2-year OS rates (χ2 = 10.47, I2 = 33%; χ2 = 7.93, I2 = 24%, respectively), so the fixed effect model was used to pool the outcomes. However, high heterogeneities were observed in the studies that reported 3- and 5-year OS rates (χ2 = 20.54, I2 = 76%; χ2 = 12.88, I2 = 77%, respectively), so the random effect model was used to pool the outcomes (Figure 2). The pool results showed that 1- and 2-year OS rates of RFA group were better than SBRT group (OR 1.52, 95% CI 1.21 to 1.90, P = 0.0003; OR 1.66, 95% CI 1.38 to 2.01, P<0.00001, respectively). Whereas, there were no differences for 3- and 5-year OS rates in both groups (OR 1.44, 95% CI 0.90 to 2.33, P = 0.130; OR 1.35, 95% CI 0.81 to 2.26, P = 0.250, respectively).
Additionally, a secondary analysis was performed to control the potential report bias, and we enrolled the studies that reported outcomes of both LP and OS. As result, the 1-, 2-, 3- and 5-year OS rates indicated no significant difference between both groups (OR 0.96, 95% CI 0.59 to 1.57, P = 0.870; OR 1.35, 95% CI 0.89 to 2.03, P = 0.160; OR 0.97, 95% CI 0.28 to 3.36, P = 0.960; OR 0.65, 95% CI 0.37 to 1.16, P = 0.150, respectively) (Supplementary Figure S2).
Prognosis for treatment allocation
3 and 5 studies evaluated the results of treatment allocation as a prognostic factor for LP(10,16,17) and OS(13,15,17,19,21), respectively. The treatment allocation was not a prognostic factor for LP [HR 0.80, 95% IC 0.58 to 1.09, P = 0.160]. However, RFA group was more favorable than SBRT group for OS benefits (HR 1.43, 95% IC 1.23 to 1.80, P<0.00001). Heterogeneity did not exist among the studies (χ2 = 5.39, I2 = 63%; χ2 = 3.64, I2 = 0%, respectively), so the fixed effect model was applied to pool the outcomes (Figure 3).
Transplant and post-transplant pathological necrosis rate
3 and 2 studies reported the transplant rate and post-transplant pathological necrosis rate, respectively(3,10,18). There were no significant differences in transplant rate and post-transplant pathological necrosis rate for both SBRT and RFA (OR 0.57, 95% CI 0.32 to 1.03, P = 0.060; OR 0.49, 95% CI 0.13 to 1.82, P = 0.290, respectively). No significant heterogeneity was observed when the transplant and post-transplant pathological necrosis rates were evaluated (χ2 = 4.55, I2 = 56%; χ2 = 1.03, I2 = 3%, respectively) and the fixed-effect model was used (Figure 4).
Publication bias
The Begg’s funnel plot was used to evaluate the reliability of publication bias in our meta-analysis(25). The shape of 11 funnel plots was basically inverted and bilateral symmetry. Therefore, these results indicated that there was little publication bias among all comparisons in this meta-analysis (Supplementary Figure S3).