2.1 Literature screening results and process
Based on the search strategy, 1,702 papers were initially obtained. After using EndnoteX9 to filter out irrelevant papers, 1,158 papers remained. Excluding 1,130 non-RCT documents such as reviews, the full text of the remaining 28 documents that met the criteria for analysis was thoroughly reviewed. Seven documents were excluded due to incomplete content or inaccessible outcome indicators, leaving 21 documents for the Meta-analysis. The literature screening process is illustrated in Fig. 2.
2.2 Basic characteristics of the included studies
Among the 21 papers finally included, there was 1 master's thesis and 20 journal papers; 4 were in English and 17 in Chinese; the publication years were from 2021 to 2023.A total of 2,487 patients with advanced NSCLC were included in the 21 papers, with 1,312 cases in the experimental group and 1,175 cases in the control group. The detailed baseline characteristics are shown in Table 1.
2.3 Evaluation of the quality of literature
All 21 papers were randomized controlled trials. Among them, 5 explicitly mentioned randomized control and allocation method concealment (Zhang et al.2022; Zhou et al, 2021; Lu et al,2023; He et al 2021; Lan et al.2022), such as the randomized envelope method and centralized randomized system method, which were defined as 'low risk'. 16 trials did not explicitly mention allocation method concealment, but only mentioned randomization, which was defined as 'unclear'. 3 trials were double-blind evaluations, defined as 'low risk' (Zhang et al.2022; Zhou et al, 2021; Lu et al,2023), while 18 trials were non-blinded evaluations, defined as 'high risk'. One trial had shedding cases but complete data, defined as 'low risk'. The remaining 20 trials with complete data and no selective reporting were also considered 'low risk' (Yang et al.2022). All studies with complete study objectives and no reporting bias were defined as 'low risk'. Other biases were not described in detail and were defined as 'unclear'. The quality of the literature is shown in Fig. 3a-3b.
2.4 Safety meta-analysis results
2.4.1 TRAEs at any level
Two studies were included in the analysis of TRAEs(Zhou et al, 2021; Lu et al,2023). The heterogeneity test revealed no statistically significant difference between the studies (P=0.94, I²=0%), thus the fixed-effects model was utilized for effect size combination. The pooled results indicated that there may be no significant variance in adverse reaction incidence between the combination group and the control group (RR=1.00, 95%CI: 0.98-1.02, P>0.05). The Meta-analysis results can be observed in Fig.4.
Out of the 21 included studies, 15 reported various categories of adverse events. This study will examine adverse reaction categories across all included literature. Arrhythmia and hypoproteinemia were excluded from the analysis due to being reported in only one study, preventing I2 calculation in Revman.
The study findings suggested no significant disparity between the intervention group and the control group in adverse event occurrence related to Platelet count decreased, Neutrophil count decreased, White blood cell count decreased, anemia, and hepatic impairment, with no statistically significant differences (P > 0.05). The intervention group exhibited slightly lower occurrence rates of gastrointestinal reactions like nausea, vomiting, and hair loss compared to the control group, with no statistically significant differences (P > 0.05). Conversely, the intervention group may have slightly higher incidence rates than the control group in pneumonia, albuminuria, skin rash, and infusion reaction adverse events, with no statistically significant differences (P > 0.05). Notably, the intervention group had a significantly higher probability of hypothyroidism compared to the control group (RR = 5.14, 95% CI: 2.36-11.21, P < 0.05). The Meta-analysis results are displayed in Fig.5-16.
2.4.2 ≥ 3 TRAEs
Four studies analyzed grade ≥3 treatment-related adverse events in advanced NSCLC patients treated with Sintilimab in combination with chemotherapy(Zhou et al, 2021; Lu et al,2023; Song 2021;Cao 2022). A statistically significant heterogeneity was observed among the studies (P=0.0002, I²=84%), and a random-effects model was used to combine the effect sizes. The results indicated that the intervention group may have a slightly lower incidence of grade ≥3 treatment-related adverse events compared to the control group, but the difference was not statistically significant (RR=0.81, 95% CI: 0.56-1.17, P>0.05). The meta-analysis results are presented in Fig.17. This study focused on analyzing literature concerning grade ≥3 adverse reactions. The findings revealed that, in comparison to the control group, the combination group had a slightly higher incidence of grade ≥3 adverse reactions related to Platelet count decreased, which was not statistically significant (P>0.05). There were no significant differences between the combination group and the control group in grade ≥3 adverse reactions related to Neutrophil count decreased and anemia (P>0.05). Additionally, the combination group had a lower incidence of grade ≥3 adverse reactions related to nausea, vomiting, other gastrointestinal(GI) reactions, and hepatic function impairment compared to the control group, with no statistically significant differences (P>0.05). The results of the meta-analysis can be seen in Fig.18-23.
2.5 Effectiveness Meta-analysis Results
2.5.1 Median overall survival
Three studies provided HR values and 95% CIs for mOS (Zhang et al.2022; Zhou et al, 2021; Lu et al,2023). After detecting possible heterogeneity among the studies (P=0.08, I2=60%), a random-effects model was used to combine the effect sizes. Subsequently, a random-effects model was selected for the meta-analysis of the three papers, as depicted in Fig.24a.
Further investigation into the sources of heterogeneity revealed no significant heterogeneity (P=0.62, I2=0%), leading to the utilization of a fixed-effects model for meta-analysis. This analysis indicated that the combination group exhibited a longer time to mOS and a 37% reduced risk of death in advanced NSCLC patients compared to the control group, with statistical significance (HR=0.63, 95% CI:0.50-0.79, P < 0.05), as illustrated in Fig.24b.
2.5.2 Median progression-free survival
HR values and 95% CIs for mPFS were extracted from two papers for analysis (Zhou et al, 2021; Lu et al,2023). A heterogeneity test indicated potential differences between the studies (P=0.11, I2=61%), prompting the use of a random-effects model for combining effect sizes. The findings revealed that the combination therapy group exhibited a longer time to mPFS and a 38% reduced risk of disease progression in NSCLC patients compared to the control group, with a statistically significant result (HR = 0.62, 95% CI:0.46-0.82, P < 0.05). These results are visually presented in Fig. 25.
2.5.3 Objective response rate
Eighteen papers reported ORR data(He et al.2021; Lan et al.2022; Zhang et al.2023; Zhang 2023; Ye 2022; Yang et al.2022; Jin et al.2022; Ma et al.2021; Wu 2022; Wang et al.2022; Song 2021; Shao and Li 2023; Zheng and Li 2022; Hu et al.2022; Hong 2023; Chen et al.2022; Cao 2022; Liang and Wei 2021). A heterogeneity test indicated no statistical heterogeneity among the studies (P=0.81, I2=0%), therefore, effect sizes were combined using a fixed-effects model. The results demonstrated a significant improvement in ORR for NSCLC patients in the combination group compared to the control group, with a statistically significant difference (RR=1.74, 95% CI: 1.55-1.94, P<0.05). The results of the meta-analysis can be seen in Fig.26.
2.5.4 Disease control rate
Nineteen papers reported data on DCR(Zhou et al.2021;He et al.2021; Lan et al.2022; Zhang et al.2023; Zhang 2023; Ye 2022; Yang et al.2022; Jin et al.2022; Ma et al.2021; Wu 2022; Wang et al.2022; Song 2021; Shao and Li 2023; Zheng and Li 2022; Hu et al.2022; Hong 2023; Chen et al.2022; Cao 2022; Liang and Wei 2021), and the heterogeneity test indicated no statistical heterogeneity among the studies (P=0.45, I2=1%). The effect sizes were combined using a fixed-effects model. The findings revealed that the combination group significantly improved the DCR of NSCLC patients compared to the control group, with a statistically significant difference (RR=1.26, 95% CI: 1.19-1.34, P<0.05). The results of the meta-analysis can be seen in Fig.27.
2.6 Publication bias
The number of meta-analysis documents for 2 indicators (ORR and DCR) exceeds 10, with statistically significant results. A funnel plot was utilized to assess distribution shape and detect publication bias. The findings revealed a convergence towards the tip and symmetrical distribution in both plots, falling within the 95% CI, suggesting minimal publication bias (Fig. 28-29)