Trial selection
Overall, 2,315 citations were identified by the researchers and 66 potentially eligible articles were retrieved in full text. We excluded 55 reports, but included 4 additional studies from other sources, resulting in 11 publications describing 10 RCTs published and 6 historical-matched control trials (HMCTs) between 2004 and 2018 (Figure 1).
Main Characteristics of studies
1,288 participants who are conformed HGG by clinical, radiological or MRI evidence were including in this study. 67 participants were diagnosed AA/AO/Others (WHO grade Ⅲ). 1,221 participants were diagnosed GBM (WHO grade Ⅳ).
We cover AdvHSV-tk+ GCV/DC vaccine/TGF-β2 inhibitor Trabedersen/Cpg-ODN/ recombinant nonpathogenic polio–rhinovirus chimera (PVSRIPO) in this context, and focus on the need for intimately understanding the efficacy and safety of immunotherapy in HGG adults. We classified these study into viral therapy (AdvHSV-tk+ GCV/PVSRIPO), DC therapy, and immunopotentiators (Trabedersen/Cpg-ODN). Thereof 624 participants were provided with immunotherapy. Mean study sample size was 81, participants ranging from 13 to 250.
In viral therapy studies, five were without TMZ as a kind of SOC and six were with TMZ. Totally 931 participants were recruiting in viral therapy studies (393 patients in experimental arm and 538 patients in control arm). Three studies reported PFS and eleven studies reported OS . Details could be found in Table 1. As for DC therapy, there were three studies in DC therapy. And all studies used TMZ in SOC. 90 participants were recruiting in DC therapy (43 patients in experimental arm and 47 patients in control arm) (Table 2). Immunopotentiator(IP) was used in two studies with TMZ as one of SOC as experimental arm (88 patients in experimental arm and 179 patients in control arm (Table 1).
Primary endpoints
OS
We apply random-effect to assess efficacy of immunotherapy through HR of OS. A total of 1,288 participants in 16 studies were included in this meta-anlysis (524 patients treated with immunotherapy, 764 patients treated with standard of care). It shows that IMT could decrease the risk of death by 38% compared with SOC (HR = 0.62, 95% CI 0.48-0.81; p = 0.0003). Substantial heterogeneity was found (tau2=0.15, p < 0.01; I2= 64.7%) (Figure 1B). Moreover, three subgroups were divided according to IMT type (Figure 1B). Subgroup of DC therapy and immunopotentiator both have no heterogeneity (respectively, I2=0, p=0.58; I2=0, p=0.40). Yet viral therapy presents comparatively high heterogeneity (I2=69%, p < 0.01).
PFS
With respect to PFS, we used random-effect to assess efficacy of immunotherapy through HR of it. We pooled six trials, a total of 316 participants (113 patients treated with IMT, 203 patients treated with SOC). It shows that IMT could decrease the risk of recurrence by 41% compared with SOC (HR = 0.59, 95% CI 0.39-0.89; p = 0.012). Significant heterogeneity was found (tau2 = 0.166, p = 0.018; I2=62.9%) (Figure 1A). Besides, two subgroups were divided according to intervention type (IMT with SOC or IMT alone), which decrease heterogeneity significantly (Figure 1B). It demonstrates that IMT alone (HR = 0.30, 95% CI 0.17-0.52; I2=0) may be better to prolong PFS than IMT with SOC (HR = 0.79, 95% CI 0.57-1.10; I2=33.3)
ORR
We used fixed-effect to assess efficacy of IMT by ORR. Only one trial reported ORR. But it reported ORR according to different type of HGG. So we divided it into 3 studies. It shows that tumor lesion in IMT group was significant compared to SOC (RR=3.12; 95% CI 1.09-8.95; p = 0.034). Little heterogeneity was found (tau2 = 0.581, p = 0.232; I2=31.5%) (Figure 1C).
AE
As for safety of immunotherapy regimen, there are 6 studies reported some adverse events (AE) according to NCI-CTC including neutropenia (59.2%), liver function tests alterations (40.6%), and seizure (22.4%), lymphopenia(21.9%) have a relatively high incidence. In addition to them, other adverse events also existed. We used random-effect to assess safety of immunotherapy through RR of occurrence rate of AE. It shows significant difference between immunotherapy arm and SOC arm (RR = 1.67; 95% CI 1.15-2.44; p < 0.0001) and substantial heterogeneity (tau2=0.400, p=0.001; I2=52.5%) (Figure 1D). It reveals that immunotherapy existed a certain potential safety problem.
Secondary endpoints
Considering substantial heterogeneity within studies, we explore the source in several ways. We dedicated to decrease the heterogeneity and make our context homogeneity.
Galbraith radial plot
We drew Galbraith radial plot and found some studies including Ji(1)(2)/cho2012/Westphal2013/Rainov2000 fell outside the 95% confidence interval line. Thereof, Ji(1)(2) and cho2012 carried out in China, Westphal2013 and Rainov2000 was in a background that TMZ has not been widely applied. It hinted that recruiting area and TMZ applying may have some effect in heterogeneity between studies. (Figure 1E)
Main factors influencing efficacy
Besides exploring the source of heterogeneity, we also try to find out factors influencing efficacy of IMT. (Table 4)
As mentioned above, subgroups divided according to IMT type manifested a statistically difference among viral therapy, DC vaccine and immunopotentiator (p < 0.05).
In addition, we divided all studies reported OS in terms of recruiting area(Europe/China/USA) and deleting two multi-area studies[19, 20], which significantly decreased heterogeneity. What’s more, it also suggested that recruiting area may be a main factor for efficacy of IMT (p < 0.05; Figure 2A). Forest plot before deleting two studies can be found in Supplementary Fig. 4.
Besides, subgroup analysis in accordance with TMZ applying was done(TMZ/Not TMZ/Not TMZ partly) as galbraith radial plot indicating. After subgroups divided, heterogeneity was decreased respectively. And pooled HR effective value was statistically different among subgroups (p < 0.05).
Furthermore, IMT was defined as immunotherapy with or without SOC treatment. The former means IMT was used as an adjuvant, while the latter means IMT was used as a new therapy. We also found no matter as an adjuvant or as a new therapy, immnunotherapy has a certain effect to prolong the OS of HGGs. But when it was used alone, the confidence interval is larger.
At last but not least, whether studies are randomized-controlled or double-bind trials doesn’t have statistically significance to effective value (p > 0.05). Gliomas whether recurrent or not, and whether Grade III or IV are does not have statistically significance to effective value, too(p > 0.05) (Figure 2B).
Sensitivity analysis
Sensitivity analysis was done to assess which study contributes to influence our current results. We didn’t find any study which have a potential impact on results, and the results did not materially change by deleting any study at a time (Figure 4C).
Publication bias
Publication bias was explored with an inverted funnel plot, which showed a significant asymmetry around 95% CI and Begg’s test showed a significance of bias (p < 0.001). We used trim-and-fill method to identify the publication bias. However, the random effect value became larger (HR = 0.92; 95% CI 0.71-1.20).