Combined Immunotherapy with Chemotherapy versus Bevacizumab with Chemotherapy in First-Line Treatment of Driver-Gene-Negative Non-Squamous Non-Small Cell Lung Cancer: An Updated Systematic Review and Network Meta-Analysis

Background: A network meta-analysis was conducted to summarize randomized control trials and updated results to evaluate the efficacy and safety profiles of existing first-line therapies for advanced non-squamous non-small cell lung cancer (NSCLC) patients without known driver gene mutations. Patients and Methods: Eligible studies were identified following a systematic search of the Cochrane Library, PubMed, Embase, Web of Science, Wanfang Data, and the China Knowledge Resource Integrated Database from January 2000 to December 2021. Results: Nineteen trials involving 8176 patients with driver-gene-negative advanced non-squamous NSCLC were included. For patients with driver-gene-negative advanced NSCLC, immunotherapy + chemotherapy (IC) significantly prolonged overall survival (OS) (hazard ratio (HR), 0.80; 95% confidence intervals (CI): 0.67–0.95) and progression-free survival (PFS) (HR, 0.68; 95% CI: 0.53–0.86) compared with bevacizumab + chemotherapy (BC), with a similar objective response rate and incidence of ≥3 treatment-related adverse events (TRAEs) (risk ratios (RR), 0.98; 95% CI: 0.79–1.21/RR, 0.89; 95% CI: 0.61–1.28; respectively) compared with BC. IC yielded a superior PFS rate (HR, 1.59; 95% CI: 1.05–2.38) compared to BC in the subgroup of patients < 65 years old. Conclusions: Currently, IC is a more efficient first-line therapy for driver-gene-negative advanced non-squamous NSCLC patients, with prolonged PFS and OS, as well as a comparatively lower risk of ≥3 TRAEs compared to BC.


Introduction
Lung cancer is one of the dominant lethal malignancies in the world. Non-small cell lung cancer (NSCLC) accounts for nearly 85% of lung cancer [1]. Common driver genes of NSCLC include epidermal growth factor receptor (EGFR), anaplastic lymphoma kinase (ALK), the proto-oncogene 1 (ROS), receptor tyrosine kinase (ROS1), B-Raf proto-oncogene, serine/threonine kinase (BRAF), MET proto-oncogene, receptor tyrosine kinase (MET) (exon 14 skipping mutation), neurotrophic receptor tyrosine kinase 2 (NTRK), Ret protooncogene (RET), etc., which were defined in the National Comprehensive Cancer Network (NCCN) guidelines of NSCLC in 2021. The most prominent driver gene mutations were sensitive EGFR mutations, followed by ALK rearrangements. For decades, the standard first-line regimen for advanced NSCLC with no actionable driver gene mutations was platinum-based chemotherapy (CT). The survival outcome has been greatly improved with immunotherapy as a newly introduced treatment into the regimen for driver-gene-negative
The inclusion criteria of eligible studies were as follows: (1) phase 2/3 randomized controlled clinical trials; (2) previously untreated NSCLC patients; (3) stage III B/IV according to TNM stage (AJCC version 7.0); (4) patients without known driver gene mutations (EGFR mutations or ALK rearrangements); (5) RCTs comparing an anti-angiogenic combined therapy to other treatment or an immunotherapy combined therapy to other treatment. The exclusion criteria were: (1) non-RCT studies; (2) patients not treated with first-line treatment; (3) data on disclosures from the same study at different times.
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist with the extension for NMA [9].
The full protocol is published on the PROSPERO website in December 2021 with the PROSPERO registration number CRD42021290355.

Data Extraction and Quality Assessment
Data extraction and cross-checking were undertaken independently by two investigators. The following information was then recorded in an Excel spreadsheet: the research number, the first author, the year of publication, the type of study design, the inclusion and exclusion criteria, the sample size of patients in each group, the follow-up time, and the objective response rate (ORR), based on the Response Evaluation Criteria in Solid Tumors (RECIST; criteria versions 1.0 and 1.1 according to the different publication years), progression-free survival (PFS), OS, and treatment-related adverse events, (TRAEs) based on the National Cancer Institute Common Toxicity Criteria for Adverse Events (versions 2.0, 3.0, and 4.0 according to the different publication years).
The potential risks of bias in trials were assessed independently by two researchers based on RCT's Cochrane risk of bias assessment: (1) Method of generating random sequences; (2) allocation sequence concealment; (3) implementation of blinding; (4) the completion of results; (5) selective reporting assessment; (6) Other biases. These risks of bias were graded as three levels: low risk, high risk, and unclear risk. If there were disagreements, a third researcher was involved in the discussion to resolve them and explain the reasons (Supplemental Figure S1).

Statistical Analysis
OS and PFS outcomes were expressed as hazard ratios (HR) with their 95% confidence interval (95% CI). ORR and TRAEs were calculated using the risk ratios (RR) and their 95% CI as a measure of association. A 95% CI, excluding 1, was considered statistically significant. In terms of PFS and OS, outcomes with HR < 1 would suggest better survival outcomes. ORR outcomes with RR > 1 would suggest better efficacies, whereas toxicity outcomes with RR < 1 would suggest better toxicity profiles.
This NMA was carried out to identify indirect comparisons between experimental groups. Risk of bias assessment was carried out using Review Manager (version 5.3 for Windows; Cochrane Collaboration, Oxford, UK). The network relation graphs in this NMA were conducted using STATA (Version 26.0; StataCorp LP, College Station, TX, USA). Network meta-analyses were performed using the GeMTC R package (version 4.1.2; R Foundation for Statistical Computing, Vienna, Austria). Random-effects models were conducted due to slightly different treatment modalities in each eligible trial. The parameters of the R were set as follows: 4 chains, 50,000 iterations, a thinning interval of 1 for OS and PFS, and a thinning interval of 10 for ORR and ≥3 TRAEs to minimize autocorrelation. Ranking probabilities of treatments were assessed utilizing the surface under the cumulative ranking (SUCRA) scores to show the likelihood of therapies in best-to-worst order (score of 0-1 and 1 is the best).
The study identification and selection process is demonstrated in Figure 1. The baseline characteristics and outcome measurements of the eligible trials are listed in Table 1.
The study identification and selection process is demonstrated in Figure 1. The baseline characteristics and outcome measurements of the eligible trials are listed in Table 1.
A presentation of the network of OS, PFS, ORR, and ≥ three TRAEs is provided in Figure 2. Indirect comparisons among IC, BC, BIC, DI, DIC, and CT are connected.          Figure 3B). Forest plots of RRs for ORR and ≥3 TRAEs are presented in Supplemental Figures S4 and S5.   Figure 3B). Forest plots of RRs for ORR and ≥3 TRAEs are presented in Supplemental Figures S4 and S5.

Subgroups of Various Clinicopathological Characteristics
In the Eastern Cooperative Oncology Group (ECOG) score = 0 subgroup, the PFS of IC was longer than that of CT (HR, 0.56; 95% CI: 0.42-0.75). In the ECOG score = 1, IC and BC both had significantly longer PFS ( Figure 4).

Rank Probabilities
The Bayesian ranking probabilities and corresponding SUCRA of the various interventions in different populations are shown in Supplemental Table S2. Treatments

Rank Probabilities
The Bayesian ranking probabilities and corresponding SUCRA of the various interventions in different populations are shown in Supplemental Table S2. Treatments with the probability of being ranked first in OS are as follows: DIC (0.44), BIC (0.33), IC (0.14), DI (0.09), BC (0.00), CT (0.00). BIC was ranked the best therapy in terms of PFS and ORR with a probability of 0.63 and 0.91, respectively. BC was associated with the highest probability of ranking first for ≥3 TRAEs (0.73), followed by IC (0.27) and CT (0.00).
In the ECOG PS = 0/1, smokers/non-smokers, patients < 65 years old subgroups, treatments with the probability of being ranked first in OS are as follows: BIC, IC, BC, CT. In patients ≥ 65 years old, treatments with the probability of being ranked first in OS are as follows: BC, IC, CT. In the male subgroup, BIC ranked first in OS, followed by BC, IC, and CT. In the female subgroup, IC ranked first in OS, followed by BIC, BC, and CT. OS of BC ranked first, followed by IC and CT for patients with liver metastases. In terms of PFS, in the ECOG PS = 0/1, smokers/non-smokers, patients ≥ 65/< 65 years old subgroups, the male/female subgroup, patients with liver metastases, treatments with the probability of being ranked first are as follows: BC, IC, CT (Supplemental Table S3).

Discussion
As a result, this updated NMA enrolled 19 available studies including 8176 patients with driver-gene-negative advanced non-squamous NSCLC, which demonstrated the superior efficacy of IC as the first-line treatment for driver-gene-negative advanced nonsquamous NSCLC compared to BC in the aspect of OS and PFS. Moreover, IC showed similar efficacy and incidences of ≥3 TRAEs compared with BC. Consequently, the present study provided a theoretical basis for selecting a more effective modality for patients with driver-gene-negative advanced non-squamous NSCLC.
Vascular endothelial growth factor (VEGF) expressed by cancer cells could promote angiogenesis, chemotaxis, and vasodilation, and further sustain tumor growth. Antiangiogenetic drugs inhibit the combination of VEGF and vascular endothelial growth factor receptor (VEGFR), block the activation of downstream pathways, degrade the existing tumor vascular system and inhibit the formation of new blood vessels. Moreover, they also improve the efficacy of CT due to their anti-vascular permeability. The combination of BEV and CT in the treatment of NSCLC has received much attention in the era of CT. In phase 3 ECOG 4599 trial, 878 patients with treatment-naive advanced non-squamous NSCLC were treated with carboplatin and paclitaxel plus BEV or carboplatin and paclitaxel [10]. OS was significantly increased with BC compared to CT (median OS (mOS) 12.3 vs. 10.3 months, HR: 0.79, p = 0.003). PFS also increased (median PFS (mPFS) 6.2 vs. 4.5 months, HR: 0.66, p < 0.001) along with ORR of 35% and 15% (p < 0.001), respectively. Based on this research, the FDA approved BEV as the first-line treatment of advanced or metastatic non-squamous NSCLC in 2016 [6]. The results of the BEYOND study for Chinese patients with NSCLC were similar to those of the ECOG 4599 trial. The mOS of patients with NSCLC in the BEV combined with paclitaxel/carboplatin group was extended by 6.6 months and the mPFS was extended by 2.7 months, which indicated that BEV combined with paclitaxel/carboplatin in the treatment of Chinese patients with NSCLC was better than CT alone [11]. However, in the AVAiL, JO19907, PRONOUNCE, and ERACLE trials, BC was not observed to have significant OS benefit compared with CT alone [12][13][14][15][16]. Therefore, the position of BC as the first-line treatment in the advanced non-squamous NSCLC is now facing challenges.
The importance of immunotherapy in the treatment of NSCLC has put great significance on some previous studies. In addition, the discussion on how ICIs should be added to the regimen was discussed and a few trials are ongoing to evaluate the efficacy and safety profiles of different interventions. This study confirmed that the use of immunotherapy in driver-gene-negative advanced non-squamous NSCLC would significantly benefit the outcome in terms of survival, which was consistent with several previous studies [17,18]. Based on the Checkmate 227 part one B trial, KEYNOTE-021G trial, and KEYNOTE-189 trials, the combination of nivolumab + CT and pembrolizumab + CT were approved to significantly improve the PFS and OS in patients with driver-gene-negative advanced non-squamous NSCLC [19][20][21][22]. At the same time, three PD-1 inhibitors (carrelizumab, sintilizumab, and tislelizumab) produced in China combined with CT have achieved similar favorable results [23][24][25][26]. In addition to PD-1 inhibitors, PD-L1 inhibitors (atezolizumab and sugemalimab) combined with CT also showed amazing results in the treatment of patients with driver-gene-negative advanced non-squamous NSCLC [27][28][29]. The preliminary results in terms of DIC and DI, nivolumab combined with ipilimumab, with or without CT, subsequently have been considered the most effective novel dual immunological applicability to treat NSCLC [19,30,31]. Since the data of the non-squamous NSCLC subgroup were not provided in the MYSTIC, CHOICE-01, Empower-Lung 3, and POSEIDON studies, these four studies were not included in this NMA [32][33][34][35]. There was no significant difference in OS and PFS among the four ICI-based therapies (BIC, IC, DI, and DIC). However, the improvement in ORR was statistically significant in BIC vs. DIC and IC vs. DIC. One reason may be related to the relatively small number of research studies that met the inclusion criteria of this meta-analysis on DI, DIC, and BIC.
There was not enough evidence for a direct comparison between IC and BC. The only RCT we were able to find with sufficient data was IMpower150, a multicenter, randomized, phase three trial which evaluated the efficacy of first-line atezolizumab + BEV + CT (ABCP) in patients with metastatic non-squamous NSCLC in contrast to atezolizumab + CT (ACP) and BEV + CT (BCP) [36]. In the 4-year updated results of OS in the IMpower150 study, patients with metastatic non-squamous NSCLC in the ACP group versus the BCP group were numerical, but not statistically significant (HR, 0.84; 95% CI: 0.71-1.00). Unfortunately, this study did not report the complete results of PFS, ORR, or ≥3 TRAEs between ACP and BCP subgroups for comparison. In our study, IC appeared to have longer OS (HR, 0.80; 95% CI: 0.67-0.95) and PFS (HR, 0.68; 95% CI: 0.53-0.86) compared with BC, with a similar ORR and ≥3 TRAEs incidence (RR, 0.98; 95% CI: 0.79-1.21/RR, 0.89; 95% CI: 0.61-1.28; respectively) compared with BC, which suggested that IC was a better choice as first-line treatment in driver-gene-negative advanced non-squamous NSCLC (Figure 3). In the SURCA rank, treatments with the probability of being ranked first in OS are as follows: DIC (0.44), BIC (0.33), IC (0.14), DI (0.09), BC (0.00), and CT (0.00) (Supplemental Table S2). Subgroup analyses showed that IC yielded superior PFS (HR, 1.59; 95% CI: 1.05-2.38), but not OS than BC in the subgroup of patients < 65 years old. In ECOG, smoking, gender, age, patients ≥ 65 years old, or liver metastases subgroups, no statistical differences were observed between IC and BC ( Figure 4). In the SURCA rank of subgroups, an improved OS of patients who were treated with IC compared to patients who were treated with BC was observed in the ECOG PS = 0/1, smokers/non-smokers, and patients < 65 years old, and female subgroups. An improved OS of patients who were treated with BC compared to patients who were treated with IC was observed in patients ≥ 65 years old, male, and patients with liver metastases subgroups (Supplemental Table S3). Although when SUCRA prediction contradicts NMA results, the HR estimation of NMA should be given priority due to the non-absolute prediction of SUCRA for treatment strategy ranking, the SURCA ranking results could also serve as good references and help us to screen for patients with driver-gene-negative advanced non-squamous NSCLC who are most likely to benefit from IC or BC.
There were several limitations of this NMA. First, due to the time span of this NMA of more than 10 years, the expression of PD-L1 and tumor mutational burden were not reported in earlier studies (e.g., ECOG 4599, AVAiL, JO19907, and ERACLE trials), which might lead to a partially-biased conclusion. Second, of the 19 studies we included, none of the trials containing IC provided detailed data of OS of patients with a different stage (stage III B or stage IV), so we did not compare the survival outcomes of IC or BC in patients with different stages. Thirdly, different kinds of immunotherapeutic agents in combination with CT or BEV were included in this NMA, which might affect the outcome. Fourth, not all treatment modalities were compared in the subgroup analysis due to the limited data of subgroup analysis in some RCTs. Real-world prospective studies are warranted to validate the reliability of this conclusion.

Conclusions
In conclusion, this NMA suggested that IC is a better efficient first-line therapy for patients with driver-gene-negative non-squamous advanced NSCLC, with prolonged PFS and OS and comparatively lower risk of ≥3 TRAEs in comparison to BC.
Supplementary Materials: The following supporting information can be downloaded at: https://www. mdpi.com/article/10.3390/jcm11061655/s1, Figure S1: Quality assessment: risk of bias according to Cochrane Collaboration's tool. Figure S2: Forest plot of hazard ratios (HRs) for OS in NMA. Figure S3: Forest plot of hazard ratios (HRs) for PFS in NMA. Figure S4: Forest plot of risk ratios (RRs) for ORR in NMA. Figure S5: Forest plot of risk ratios (RRs) for ≥ 3 TRAEs in NMA. Table S1: Search Strategy. Table S2: Rank probabilities with SUCRA value for different outcomes in 6 kinds of first-line treatments for patients with non-squamous NSCLC. Table S3: Rank probabilities with SUCRA value for different outcomes in 6 kinds of first-line treatments for subgroup patients with non-squamous NSCLC.
Author Contributions: Data collection, Y.C. and X.W.; statistical analysis, Y.C.; writing-original draft preparation, Y.C. and X.W.; writing-review and editing, Y.C. and X.W.; supervision, J.D. and H.B. All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: The datasets developed and analyzed during this study are available from the corresponding author upon reasonable request.