Assessing Potential Factors Influencing the Efficacy of Immune Checkpoint Inhibitors with Radiation in Advanced Non-Small-Cell Lung Cancer Patients: A Systematic Review and Meta-Analysis

Objective Recent evidence suggests that combining radiotherapy (RT) with immune checkpoint inhibitors (ICIs) may result in better outcomes. In this study, we assessed the efficacy and safety of ICI plus radiation versus ICI alone and explored potential factors affecting its efficacy in advanced non-small-cell lung cancer (NSCLC) patients. Methods The databases including PubMed and Embase were searched to retrieve eligible studies comparing the efficacy and safety outcomes in advanced NSCLC patients after ICIs ± RT treatments. We performed subgroup analyses to identify potential prognostic factors from radiation details and study types. The odds ratio (OR) of objective response rate (ORR) and disease control rate (DCR), hazard ratio (HR) of progression-free survival (PFS) and overall survival (OS), and risk ratio (RR) of adverse events were used to represent the outcome effects. Results 26 eligible studies with 14192 cases were included. The results showed that the ORR (OR = 0.63, 95% CI: 0.42, 0.93; p = 0.02) and DCR (OR = 0.55, 95% CI: 0.36, 0.82; p < 0.01) of RT + ICIs groups were significantly higher than those of the ICIs alone group. The median PFS and OS for ICIs versus RT + ICIs were 2.2 versus 4.4 months and 9.0 versus 13.4 months, respectively. Patients in the ICIs plus RT group had a significantly better PFS (HR = 0.72, 95% CI: 0.64, 0.81; p < 0.01) and OS (HR = 0.74, 95% CI: 0.65, 0.83; p < 0.01) when compared to those in the ICIs group. In terms of adverse events, the risk of pneumonia was not significantly increased in patients treated with both ICIs and RT when compared to ICIs group alone (risk ratio = 0.89; 95% CI: 0.55, 1.44; p = 0.63). The correlation analysis found that PFS was significantly correlated with OS (p = 0.02). The subgroup analysis results showed that significant improvements in OS were observed in non-palliative RT group (HR = 0.29, 95% CI: 0.13, 0.65; p < 0.01) and extracranial RT group (HR = 0.70, 95% CI: 0.59, 0.83; p < 0.01). RT type could also be a prognostic factor associated with the OS (for conventional RT: HR = 0.68 and p = 0.22; for stereotactic body radiation therapy: HR = 0.77 and p < 0.01). However, concerning RT timing, the results showed a similar trend in reducing mortality risk (for previous RT: HR = 0.64 and p = 0.21; for concurrent RT: HR = 0.35 and p = 0.16). Conclusion RT plus ICIs is associated with improved survival for advanced NSCLC patients, especially for those with non-palliative RT. Further clinical trials are needed to validate its effect on survival outcomes.


Introduction
Lung cancer is one of the most common causes of cancer deaths worldwide [1]. Non-small-cell lung cancer (NSCLC) is the most common subtype of lung cancer, accounting for about 76% of all lung cancers [1]. In recent years, based on the understanding of tumor biology and the mechanism of occurrence and development, important progress has been achieved in the treatment of advanced NSCLC [2,3]. Among them, immune checkpoint inhibitors (ICIs) are considered one of the most promising agents [4,5]. Although the clinical administration of this therapy can bring signifcant efcacy and survival improvement for these patients, its efcacy is closely related to the expression of PD-L1 [3], tumor mutation burden [6], neutrophil/lymphocyte ratio [7], and body mass index [8]. Tere are still a considerable number of patients who fail to beneft from immunotherapy due to factors such as programmed death-ligand 1 (PD-L1) expression [3], epidermal growth factor receptor (EGFR) mutation status [9], low tumor mutation burden (TMB) [6], high neutrophil-lymphocyte ratio(NLR) [7], or low body mass index (BMI) [8]. Terefore, how to improve the efcacy of ICIs in advanced NSCLC is still under investigation.
Te key to improving the antitumor efcacy of immune checkpoint inhibitors lies in combination therapy [10][11][12][13]. Radiotherapy, chemotherapy, molecular targeted therapy, and other cancer treatment methods can have a synergistic efect to improve the efcacy of immunotherapy [10][11][12][13]. Studies have found that radiotherapy can enhance the efcacy of immunotherapy in locally advanced NSCLC [11,14]. Te PACIFIC study included 713 patients with locally advanced inoperable NSCLC [14]. Te median progression-free survival (PFS) of the group receiving durvalumab consolidation therapy after concurrent radiotherapy and chemotherapy was 17.20 months, while that of the control group was only 5.60 months (HR � 0.68 and p < 0.01); the median overall survival (OS) was 47.50 months, and that of the control group was 29.10 months (HR � 0.71 and p < 0.05) [14].
Due to the limitation of ICIs monotherapy, the application of radiotherapy combined with immunotherapy in advanced NSCLC is increasing in recent years, with palliative or non-palliative intents. However, the improvement in survival is not consistent. Tis raises several concerns about the factors afecting the efcacy of this combination strategy [15][16][17][18][19]. Besides, there are studies that fail to support that adding RT to ICIs can improve treatment outcomes when combining ICIs with RT [20][21][22][23]. A study screened 121 patients diagnosed with metastatic NSCLC and compared OS and PFS for patients with ICIs ± RT. Te results showed that no diference was found between patients receiving ICIs versus ICIs + RT in terms of median OS (16.7 months versus 16.2 months, p > 0.05) or PFS (9.3 months versus 10.7 months, p > 0.05). Tey concluded that the use of RT in addition to ICIs was not associated with improved OS or PFS in metastatic NSCLC patients [23]. Whether adding RT to ICIs could result in a signifcantly higher risk of adverse events is also not known.
Considering there are still inconsistent results about the efcacy and safety of combination therapy in advanced NSCLC and the concerns about factors afecting survival, we performed this systematic review and meta-analysis. By searching the relevant clinical studies of radiotherapy combined with immunotherapy for advanced NSCLC patients, we evaluated the efcacy and safety of RT + ICIs versus ICIs and explored the infuence of RT details, such as timing, sites, and types of radiotherapy, on the treatment outcomes in patients with advanced NSCLC by subgroup analyses.

Search Strategy.
Te PubMed and Embase databases were searched to identify eligible studies comparing the efcacy and safety outcomes in advanced NSCLC patients treated with ICIs versus RT + ICIs. Te preprint platforms, such as bioRxiv, were also searched to retrieve unpublished studies on the same topic mentioned above. References of important reviews were also identifed to further include eligible studies. Te search was conducted in the above databases until August 2022. No language limitation was applied during the search. Te search terms were as follows: "non-small cell lung cancer," "NSCLC," "radiotherapy," "stereotactic body radiation therapy (SBRT)," "SBRT," "stereotactic ablative radiotherapy," "immunotherapy," "immune checkpoint inhibitor," "nivolumab," "pembrolizumab," "anti-pd-1," "pd-1 inhibitor," "durvalumab," "atezolizumab," and "cytotoxic T lymphocyteassociatedantigen-4inhibitor." Te search terms were used in diferent combinations and were adjusted according to specifc databases. Te example of the search strategy is presented in Supplemental Table 2.

Inclusion and Exclusion
Criteria. Inclusion criteria were designed based on the previously registered protocol (#CRD42019120007) in "https://www.crd.york.ac.uk/ PROSPERO/." Te inclusion criteria were as follows: study types-retrospective or prospective studies; patients-subjects with a confrmed diagnosis of advanced/metastatic NSCLC by pathology or cytology; and interventions-ICIs were used with or without RT. RT was administrated according to the treatment goals described in the included studies. Comparisons of treatment efectiveness of ICIs versus RT + ICIs were reported in the eligible studies. Outcome indicators: these included ORR (overall response rate), DCR (disease control rate), PFS, OS, safety, and prognostic factors related to RT. Subgroup: subgroup analysis was introduced according to RT and study details. Its details are presented in the statistical methods section. Te defnitions of main outcome measures are detailed in a previously published study [4]. Animal experiments, duplicate publications, and literature with insufcient data were excluded. References such as comments and review articles were also excluded.

Literature Screening and Data Extraction.
Two reviewers (Dedong Cao and Dingjie Zhou) conducted the literature search, read the titles and abstracts of studies after the preliminary search, and then read the full text of eligible studies for further identifcation, independently. Te inconsistencies about the screening results were discussed with a third reviewer (Huilin Xu). Te data were extracted by three reviewers (Huilin Xu, Dedong Cao, and Dingjie Zhou), and the extracted information was as follows: (1) basic information of the studies, including the title, the frst author, region, the year of publication, age, and sex; (2) cancerrelated details, including the number of patients in each group and cancer stage; and (3) treatment details, including the timing of RT (before ICIs, concurrent with ICIs, and multiple time points), types of RT (conventional RT (CRT) and SBRT), and ICI details. Outcome data include the total number of patients, ORR, DCR, PFS, OS, and safety. Te reviewers (Dedong Cao and Huilin Xu) checked the extracted data with each other. Disagreements were resolved by re-examining the studies.

Quality Evaluation.
For RCT studies, the methods of Cochrane Handbook 5.1 were used to evaluate the quality of the included studies [25]. Five major aspects of bias, such as random, allocation, blinding, selective reporting, and other bias, were included in the assessment. For prospective and retrospective studies, literature quality evaluation was performed according to the Newcastle-Ottawa Scale (NOS) criteria [26], which included three major aspects: selection, comparability, and outcome. After the evaluation, a score ranging from 0 to 9 was calculated for each included study. For a study with six or higher stars, the quality was regarded as high.

Statistical
Methods. Meta-analysis was processed using the Comprehensive Meta-Analysis v3.0 software (Biostat Inc.) and RevMan 5.4. Te statistical methods of the metaanalysis were introduced as previously reported [27][28][29]. In brief, the odds ratio (OR) or risk ratio (RR) and its 95% confdence interval (CI) were used for the comparison of dichotomous data. Te median and its related 95% CI were used for continuous data. Te hazard ratio (HR) and its related 95% CI were used to present the survival benefts from treatments. For OR, RR, or HR, it was considered that the combination group had a better efect if the value was less than 1. Te chi-square test and I 2 were used for statistical heterogeneity analysis. According to the Cochrane Handbook, four levels of heterogeneity are classifed [25]. If I 2 is below 40%, it indicates that heterogeneity may not be important. I 2 between 30% and 60% indicates moderate heterogeneity, 50%-90% indicates signifcant heterogeneity, and 75%-100% indicates a greater signifcant heterogeneity. For analysis with low and moderate risk of heterogeneity, the fxed-efect model is used. If a signifcant heterogeneity is detected (I 2 > 50% and p < 0.1), the source of the heterogeneity will be analyzed and studies that are responsible for this diference will be excluded or subgroup analysis or sensitivity analysis will be performed. In addition, a randomefect model will be used for the meta-analysis. For insufcient data or signifcant heterogeneity, a descriptive analysis will be performed.
For the continuous data, the median/mean and its confdence interval were extracted as reported in the included studies. Te meta-analysis of median variables was performed using the methods as reported [27,29]. In addition, the median could be regarded as an estimate of the mean if the distribution of the data is symmetrical and therefore be used directly in the meta-analysis [25]. Also, the p value of the comparison between ICIs and RT + ICIs was either extracted as reported or calculated using the Review Manager 5.4.1 tool.
2.6. Subgroup Setting. Te primary endpoints of this metaanalysis were efcacy, survival, and safety of ICIs versus RT + ICIs in advanced NSCLC patients. Additional analyses were also conducted to evaluate the infuence of several factors related to RTon the outcomes of RT + ICIs versus ICIs. Subgroup analyses were used to assess the impact of study design (prospective and retrospective), disease condition (advanced and metastatic), and RT timing (prior, concurrent, and multiple) on the efcacy, survival, and safety outcomes. CRT referred to conventional radiotherapy, and SBRT was defned as delivering high-dose radiation (3 or higher Gy per fraction) to eliminate tumors in fewer treatment fractions than CRT. Te timing "prior" was defned as delivering RT before ICIs, "concurrent" was defned as RT during the treatment cycles of ICIs, and "multiple" was defned as RT used before, during, and/or after administration of ICIs. Te cumulative analysis was used to detect the dynamic trend of meta-analysis and evaluate the impact of a single study on the overall outcomes. Te sensitivity analysis was used to evaluate the reliability and stability of the pooled efect by removing one study each time. To detect publication bias, the funnel plot was drawn and Egger's test [25] was used if possible. Te cumulative meta-analysis is a sequence of meta-analyses, starting by analyzing a single study at the beginning and adding the rest of the included studies one by one to the analysis until all of them are included in the analysis [30]. It shows the dynamic trends of the overall estimate when adding every single study to the meta-analysis [30]. p < 0.05 was considered as there was statistical signifcance.

Basic Characteristics and Methodological Quality of the
Literature. Te basic characteristics of the included studies are shown in Table 1. Most of the included studies were retrospective studies, and only 4 prospective studies and 4 RCTs were included. Tese studies were mainly from America (n = 12), Australia (n = 3), Europe (n = 4), and Asia (n = 4). Te mean age of the involved patients varied across included studies. All studies reported the treatment strategy, and 8 out Journal of Oncology 3 of 26 described that SBRT was used to treat cancer. Tese studies described the treatment line, and 15 of them used the treatment in the ≥ 1st line setting and 11 of them in the ≥2nd line setting.. Studies were grouped into prior (n = 15), concurrent (n = 6), and multiple (n = 5) based on the timing of RT. All studies reported the diagnosis and stage of NSCLC. Te overall quality of the included studies was moderate. Only a few studies described methods of selecting and reporting patients and other sources of biases. In terms of NOS, eight studies were assigned with seven stars, ten with eight stars, and four with nine stars (Supplemental Table 3). Te main limitations were loss to follow-up rate and inadequate follow-up. According to the Cochrane Handbook methods, the most common bias was random bias and blinding bias (Supplemental Table 4). Subgroup analyses based on study type, disease stage, and RT timing were also performed (Supplemental Figure 1). For prospective and retrospective design, the OR of RT + ICIs versus ICIs was 0.82 (95% CI: 0.26, 2.62; p � 0.74) and 0.61 (95% CI: 0.40, 0.92; p � 0.02), respectively (Supplemental Figure 1A). With regard to disease types, the OR for comparing ORR of ICIs versus conventional RT (CRT) + ICIs was 0.58 (95% CI: 0.35, 0.96; p � 0.04) in the advanced disease group, and it was 0.64 (95% CI: 0.41, 0.99; p < 0.05) for the metastatic group (Supplemental Figure 1B). In terms of RT timing, the results showed that when adding radiotherapy concurrently with ICIs, the OR for ICIs versus RT + ICIs was 1.31 (95% CI: 0.19, 8.88; p � 0.28). When used radiotherapy before ICIs, the OR for ICIs versus RT + ICIs groups was 0.58 (95% CI: 0.38, 0.89; p � 0.01), and it was statistically signifcant (Supplemental Figure 1C). For patients treated with CRT, the OR for ICIs versus RT + ICIs was 0.71 (95% CI: 0.49, 1.04; p � 0.08), while it was 0.40 (95% CI: 0.16, 1.03; p � 0.06) for patients with SBRT (Supplemental Figure 2A). For patients that received ≥ frst-line treatment, the OR for ICIs versus RT + ICIs was 0.65 (95% CI: 0.47, 0.91; p � 0.01), while it was 0.73 (95% CI: 0.36, 1.50; p � 0.40) for patients that received ≥ second-line treatment (Supplemental Figure 2B).    Figure 3A). With regard to disease condition, adding RT to ICIs resulted in a signifcantly better DCR than ICIs alone (Supplemental Figure 3B) Figure 4B).

Correlation Analysis of PFS and OS.
To explore the association between PFS and OS, we performed a correlation analysis. A total of 9 pairs of PFS and OS from the RT + ICIs group were included. Te tabular result showed that the correlation Pearson r value was 0.77 (95% CI: 0.21, 0.95; p � 0.02), with a squared R of 0.59 (Figure 2(c)).

PFS.
Te PFS outcomes of diferent groups were reported in 9 studies (Figure 3). For patients receiving RT + ICIs, the risk of disease progression was signifcantly lower than that of the ICIs group (HR = 0.72; 95% CI: 0.64, 0.81; p < 0.01). After dividing studies into two groups by RT timing, we found that patients in the prior RT group had a signifcantly better PFS (Figure 3(a)), while it was not signifcant in the concurrent RT group (Figure 3(a)). Further analysis of concurrent group based on RT types still failed to fnd a signifcant diference in PFS (Supplemental Figure 5). After including studies using RT to treat extracranial sites, these patients could still beneft from RT in terms of PFS with a statistical signifcance (HR = 0.70; 95% CI: 0.59, 0.83; p < 0.01) (Figure 3(b)). We also assessed whether the types of RT could infuence the disease control benefts. As illustrated in Figure 3(c), although both non-palliative (HR = 0.66, p = 0.07) and palliative intent RT (HR = 0.77, p = 0.49) exhibited an obvious trend in reducing disease progression, it was not signifcant.

OS.
We extracted OS outcomes from 7 eligible studies (Figure 4(a)). Compared to ICIs alone group, the risk of death was signifcantly lower in the RT + ICIs group (HR = 0.74; 95% CI: 0.65, 0.83; p < 0.01). Te meta-analysis showed that both the prior RT group (Figure 4(a)) and the concurrent RT group (Figure 4(a)) were associated with improved OS. However, this survival beneft was not signifcant after dividing studies into two groups by RT timing (prior RT: HR = 0.64; concurrent RT: HR = 0.35; p > 0.05 for all). Te application to extracranial lesions resulted in a statistically signifcant improvement in OS compared to those in the ICIs alone group (HR = 0.65; 95% CI: 0.52, 0.80; p < 0.01) (Figure 4(b)). We performed another subgroup analysis to determine whether the types of RT could infuence the death risk benefts. As illustrated in Figure 4(c), patients in the non-palliative RT group had a signifcantly better OS (HR = 0.29; 95% CI: 0.13, 0.65; p = 0.002), but not those in the palliative intent RT group (HR = 0.78, p = 0.50). Next, the impact of RT types on OS was also evaluated (Figure 4(d)). By dividing the studies into CRT and SBRT groups, the meta-analysis suggested that patients receiving SBRT had a signifcantly longer OS than those in the ICIs alone group (HR = 0.77; 95% CI: 0.66, 0.90; p = 0.001). However, it was not signifcant in the CRT group (HR = 0.68; 95% CI: 0.37, 1.26; p = 0.22). For this diference, the RT biological efective dose (BED) may be the reason. Compared to low BED group, high BED was associated with a better OS in patients treated with RT + ICIs (Supplemental Figure 6). In addition, a correlation analysis between BED and OS of the RT + ICIs was performed (Supplemental Figure 7), and the result suggested that there was an obvious correlation between BED and OS (number of pairs = 3; Pearson r = 0.86; squared R = 0.75; p = 0.34).

Safety.
We summarized the safety data from included studies and performed meta-analyses. Te overall risks of any adverse events, grade 3 or higher adverse events, and pneumonitis were analyzed ( Figure 5). For ICIs versus RT + ICIs, the risk ratios of any adverse events ( Figure 5(a)), grade 3 or higher adverse events ( Figure 5(b)), and pneumonitis ( Figure 5(c)

Sensitivity, Cumulative, and Publication Bias
Assessments. As the analyses of ORR and DCR included most of the studies, sensitivity, cumulative, and publication bias assessments were performed using the data of ORR and DCR. Te sensitivity analyses of ORR were performed to examine whether the overall estimate could be signifcantly Journal of Oncology infuenced by the included studies (Supplemental Figure 8). Te results showed that the ORR was reliable when excluding each study at one time. Te ORR cumulative analysis results showed that the ORR and its associated 95% CI were stable and in favor of the RT + ICIs treatment (Supplemental Figure 9). To detect potential publication bias, the funnel plot was applied and Begg's test and Egger's test were conducted by using the data of ORR (Supplemental Figure 10). Te funnel plot of ORR showed that the risk of publication bias was low as the studies were located within the plot's range (Egger's test: p > 0.05; Begg's test: p > 0.05).
Te results of the sensitivity analysis (Supplemental Figure 11) were reliable after removing each study included in the DCR analysis, and the DCR estimates were within their fnal confdence interval. Te DCR cumulative analysis results (Supplemental Figure 12) showed that the DCR and its associated 95% CI were always in favor of RT + ICIs treatment. Te funnel plot of DCR showed that the risk of publication bias was moderate as the studies were located within the plot's range (Supplemental Figure 13).

Discussion
Te landscape of advanced NSCLC treatment has rapidly changed in recent years, especially after the emergence of ICIs. How to optimize the efcacy of this strategy is currently under investigation. In this study, the impact of RT on outcome and safety of advanced NSCLC patients treated with immunotherapy was assessed. Te results found that the efcacy and survival outcomes were improved when adding RT to ICIs, with acceptable safety. Subgroup analyses suggested that patients who received non-palliative RT or SBRT had signifcant improvements in OS. Of note, PFS may serve as an indicator of OS in patients treated with RT + ICIs. Tese were in accordance with previously reported studies that using RT with ICIs may be associated with improved PFS and OS in well-selected patients [16,31,50].
Compared to ICIs alone, whether the combination of RT and ICIs has a superior efect on the efcacy and survival of advanced NSCLC patients is still under debate. Although evidence from a large number of studies showed improved efcacy after combination therapy of RT and ICIs, other results were also reported. Several studies [20,21,51]     Journal of Oncology that RT failed to signifcantly improve the PFS in the setting of immunotherapy. Of note, the study of Cortellini et al. suggested that previous palliative RT was signifcantly associated with shortened PFS and OS (p < 0.05 for all) in metastatic NSCLC patients with PD-L1 expression ≥50% [20], indicating that the RT types and PD-L1 level may afect the survival outcomes. After combining all the eligible studies, the overall efect was in favor of the combination group, suggesting that RT could improve the ORR and DCR in these patients. Also, this was further validated by the cumulative analysis and sensitivity analysis. Some studies [20][21][22][23]38] did not support that the combination of RT and ICIs could have a better OS in treating patients with advanced NSCLC. In our analysis, the pooled estimate of OS was better in the RT + ICIs group in the setting of retrospective studies. Also, the OS of patients from the prospective studies was much longer for RT + ICIs than that of the ICIs group. Tese fndings suggest that the administration of RT can improve the efcacy and survival in advanced NSCLC patients with ICIs treatment.
Although the above fndings are promising, how to optimize the efcacy of RT + ICIs is unanswered. Based on the results of the PACIFIC study [14] and PEMBRO-RT [17], the RT timing and types may be associated with diferent treatment outcomes of immunotherapy. Terefore, several RCTs [52,53] on the topic of RT timing and types in advanced NSCLC are ongoing and without conclusions. Te defnition of RT timing varied between studies, leading to various cutof values of RT timing. Unlike other studies, the study of Kong et al. [22] was the main evidence supporting that prior RT was superior to concurrent RT in improving OS. Tey focused on the efect of thoracic radiotherapy on survival of stage IV NSCLC treated with and without immunotherapy. Patients who were treated with immunotherapy and thoracic radiotherapy concurrently had a worse OS (n = 177, median OS = 7.4 months, and p < 0.001), compared with those who had a thoracic radiotherapy history before immunotherapy ( n = 165, median OS = 12.2, months) [22]. All the above diferences indicate that more studies that directly compare the various RT timing on the efcacy of ICIs are needed.
In our analyses, we found that previous RT was associated with signifcant improvement in PFS. Te subgroup analysis of RT timing on OS suggested that neither prior RT nor concurrent RT was associated with signifcant improvements in OS, although there was a clear trend favoring RT + ICIs. Te limited number of included studies may be responsible for these results. Another concern about this combination strategy is the dose and fraction of radiotherapy. Indeed, the SBRT was associated a signifcantly longer OS but not CRT. Te BED may be responsible for this observation. Our correlation analysis suggested that BED was highly correlated with OS in patients treated with RT + ICIs. Moreover, this was in accordance with the fndings of Foster et al. [39]. Te impact of the types of RT on OS was also evaluated. Although patients may beneft from RT + ICIs treatment, patients who were treated with nonpalliative RT could have a signifcantly lower risk of death but not palliative RT. Patients who were suitable for nonpalliative RT may have a better disease burden or condition. Besides, they may receive higher dose of RT. Tese may result in the diference of OS.
Recently, a few meta-analyses [54][55][56] on the topic of ICI combination strategy have been published. Te study of Mo et al. [54] included nine RCTs and compared the PFS and OS benefts from the combination therapy, such as immunotherapy plus chemotherapy, double immunotherapy agents, or immunotherapy plus targeted chemotherapy. Te results of their study provided a clear suggestion that the risks of death and disease progression were signifcantly reduced when immunotherapy was combined with chemotherapy or    other treatment options in NSCLC patients [54]. For immunotherapy plus RT, the study of Kim et al. [55] included studies of NSCLC patients with brain metastases, and the results showed that the combination group had a better intracranial local efcacy than that of ICI monotherapy. Another study by Voronova et al. [56] evaluated the impact of the schedule of RT on efcacy outcomes of brain metastases when combined with ICIs. After including 40 studies with 4359 patients, they found that RT concurrent with ICIs was associated with a better survival rate than the sequential combination group [56]. Tough these fndings are in accordance with our study, there are several diferences. Te meta-analysis by Voronova et al. [56] focused on comparing RT versus combination therapy, and we focused on ICIs versus RT + ICIs. First, the subjects in our study were advanced NSCLC patients treated with either ICIs or RT + ICIs, and the primary endpoints were efcacy and survival. Second, the impact of RT timing and disease condition on ORR, DCR, PFS, and OS in advanced NSCLC was evaluated in the subgroup analyses. Tird, all relevant studies on the same topic were included, and subgroup analysis based on study design was performed, minimizing the risk of selection and inclusion bias. Nonetheless, this systematic review and meta-analysis evaluates the impact of RT sites, timing, and types on survival in advanced NSCLC patients treated with RT + ICIs versus ICIs alone. Te following limitations exist in the present metaanalysis. (1) Most of the included studies are retrospective analyses of ICIs versus RT + ICIs and may result in an increased risk of selection and reporting bias. Patients from most of the retrospective studies were likely to be diferent in disease conditions from those who did not receive radiation. Besides, the decisions of using radiotherapy as well as the timing and types of radiotherapy were associated with disease-related factors that would afect the outcome of these patients. Tese could introduce a high risk of selection bias.
(2) Tough the overall quality of the literature included in this meta-analysis is moderate, the risk of repeat reporting may exist among the included studies as few of the authors came from the same hospital, though they focused on different aspects of the disease and treatment. 11 of the studies were published on conferences and in abstract form, which may hamper the overall quality of our meta-analysis. (3) No language limitation is applied at the time of searching. However, the literature included in this study is only in English. It is not sure whether the results of the metaanalysis can be applied to all races/nations. (4) Te diferences between clinical characteristics, the defnition of RT timing, plan of RT, RT dose and fraction, reporting of survival, and the inconsistency in the disease stage and degree of some underlying conditions may increase clinical heterogeneity between studies. For example, the number of metastasis lesions, diferent metastatic organs, and disease burden can impact the treatment efectiveness of ICIs. Te varied doses and types of RT could result in a diferent tumor response. Terefore, a better analysis would be to obtain the source data from the previously published manuscripts to pool the data and analyze the data in a non-biased manner. (5) Not all the studies reported the primary endpoints, and thus a limited number of studies were included in the specifc analysis, such as meta-analyses of ORR and DCR. Tis may underestimate or overestimate the actual efectiveness of RT + ICIs. Nevertheless, this systematic review and meta-analysis may answer some concerns and provide evidence for clinical practice.
In conclusion, the present meta-analysis suggests a combination of RT and ICIs to serve as a promising treatment strategy for improving the treatment efcacy of advanced NSCLC patients. However, its impact on survival needs to be further determined.