Immune checkpoint inhibitors (ICIs), including inhibitors of programmed cell death 1 (PD-1), programmed cell death ligand 1 (PD-L1), or cytotoxic T-lymphocyte protein 4 (CTLA-4), have dramatically changed the NSCLC treatment landscape over the past decade. Based on results from multiple global clinical trials, ICIs have been one of the standard first-line NSCLC treatments, as monotherapy or combined therapy (1–6). However, only a minority of NSCLC patients benefit from ICIs (7), and whether the magnitude of benefit is specific factor-dependent remains unclear. Given the adverse effects (8) and cost of ICIs, and the lack of access to the drug in low-to-middle income countries, it is crucial to identify subpopulations who can derive a larger relative benefit from immunotherapy.

Anti-PD-1/PD-L1 therapy works by blocking the PD-1/PD-L1 pathway, suggesting that high PD-L1 expression could be a reasonable biomarker predicting the efficacy. However, some PD-L1-negative patients also respond to anti-PD-1/PD-L1 therapy (9). Moreover, currently there is no unified standard for defining PD-L1 positivity. Whether there are differences in the relative benefit from ICIs over control therapies remains unclear when using different positive PD-L1 standards.

Sex, age and Eastern Cooperative Oncology Group (ECOG) performance status (PS) might correlate with immune response, thus affecting the efficacy of ICIs. We have explored the association of these three variables with the relative benefit of ICIs in solid tumors. Stratified analysis showed no difference in the survival advantage of immunotherapy among NSCLC patients grouped by sex, age, and ECOG PS (10). However, other clinicopathological and biomolecular characteristics were not analyzed, and several large clinical trials have published updated results in recent years, warranting an updated analysis.

NSCLC harboring epidermal growth factor receptor (EGFR) mutations or low tumor mutation burden (TMB) showed poor clinical outcomes with immunotherapy (11, 12). Histological type, smoking status and race are known to correlate with NSCLC EGFR mutation rate. NSCLC brain metastases showed an increased TMB and genomic instability in comparison with primary NSCLC (13). Therefore, these clinical features mentioned above may also be associated with therapeutic benefit of ICIs.

Given the lack of data on the above issues, we conducted a comprehensive meta-analysis to examine the potential association of the common clinicopathological and biomolecular features with relative advantages of immunotherapy in NSCLC.

The study was registered with PROSPERO (CRD42019123892), an international prospective register of systematic reviews, and performed following the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines (14). The need for institutional review board approval was waived by Drum Tower Hospital because this meta-analysis utilized publicly available data.

In this meta-analysis, we found that wild-type EGFR, high PD-L1 expression, and high bTMB were associated with a greater OS benefit from ICI therapy vs control therapies in patients with NSCLC. Clinicopathological features such as sex, age, ECOG PS, histological type, smoking history, or baseline brain metastasis status (previously treated) were not associated with OS benefit from ICI therapy. These results highlight the importance of EGFR, PD-L1 expression and bTMB as predictive markers for NSCLC immunotherapy. Moreover, our study suggests that the use of ICIs in NSCLC should not be restricted to certain factors such as sex, smoking history, race, etc.

A growing body of literature has illustrated sex­based differences in immune responses. Adult females mount stronger innate and adaptive immune responses than males (49), thus could potentially influence the efficacy of immunotherapy. In addition to sex, age and ECOG PS have also been reported to potentially influence the response to ICIs (50, 51). However, we did not find any statistically significant difference in OS benefit from ICIs in patients with different sex, age, or ECOG PS, both in our previous (10) and the current study. Many large-scale clinical trials of immunotherapy were carried out in different races and regions around the world, but there is no research on whether OS benefit from ICI vs control therapy differs by the patient’s race or region. We found that there was no difference in the degree of relative benefit from immunotherapy between races and regions. These findings suggest that ICI immunotherapy is beneficial in NSCLC patients regardless of sex, age, ECOG PS, race and region, and the use of ICIs should not be restricted to any subgroups based on these variables.

Brain metastases are associated with low quality of life and poor survival outcome. The blood-brain barrier extremely limits the therapeutic effectiveness of drugs. A phase 2 trial has demonstrated the positive role of pembrolizumab in treating NSCLC with brain metastases, as the first prospective study of immunotherapy focusing on brain metastasis. The brain metastasis response rate was 29.7% in the PD-L1 TPS≥1% population, with 4 patients achieved complete response and 7 achieved partial response, meeting the prespecified success criteria set for the trial (52). The subgroup analyses from some clinical trials did show that NSCLC patients with brain metastases could benefit from immunotherapies over control therapies (22, 34, 35, 40), whereas some other trials did not (6, 20, 33, 44). The majority of clinical trials with ICIs require that brain metastases be treated by surgical resection or radiation with a subsequent period of stability prior to enrollment. Pooled analyses have suggested that these patients with treated brain metastases still derive significant benefit with ICIs (53, 54). Here, we found NSCLC patients with baseline brain metastases could also benefit from the ICIs, and the magnitude of benefit was similar to that in patients without baseline brain metastases. The results suggest the possibility that ICIs may pass the (likely disrupted) blood-brain barrier and treat brain metastases effectively. However, the number, size, and location of brain metastases may influence the efficacy of ICIs, which needs to be analyzed in future studies.

We demonstrated ICIs resulted in an improved OS compared with controls among EGFR wild-type patients but not among EGFR-mutant patients, consistent with several previous publications (55, 56). Notably, a minority of patients with EGFR-mutant NSCLC could still benefit from ICIs. For example, NSCLC patients with L858 mutations had a similar outcome with ICI treatment compared with those with wild-type EGFR in a large retrospective study (11). Moreover, high PD-L1 expression in patients with EGFR-mutant NSCLC showed a trend toward better outcomes than those with low PD-L1 expression (57). Additionally, add immunotherapy to certain backbone therapies could significantly improve outcomes of NSCLC patients with TKI sensitive EGFR mutations (26). Therefore, NSCLC patients with sensitive EGFR mutations might still benefit from immunotherapy if there is high PD-L1 expression or when used in combination with chemotherapy (58–60). Mutation status in KRAS, TP53, MET, and ROS were associated with PD-L1 expression in patients with lung adenocarcinomas (61, 62). ALK rearrangement status may be associated with response to PD-1/PD-L1 pathway blockade (55). Given the small number of trials, we did not analyze the difference of survival benefit by other genetic mutations.

As the target of anti-PD-L1 antibodies, expression of PD-L1 has become the major initial predictor of benefit from immunotherapy. High tumor PD-L1 expression is correlated with an increased likelihood of response to anti-PD-1/PD-L1 antibodies. In this study, we grouped studies that used the same cutoffs of PD-L1 expression to analyzed the association with OS benefit. Pooled analysis shows that patients with TPS/TC<1% can also benefit from ICIs. The positive result may be partly attributed to some included trials reporting combined ICIs (CTLA4 antibody + PD-1/PD-L1 antibody). A greater OS benefit was observed in those with high PD-L1 levels when dividing subgroups by 50% tumoral PD-L1 expression, but no difference was found when using 1% as the cutoff. This suggested that NSCLC patients with a significantly higher PD-L1 expression may benefit more from ICIs than other groups, but that benefit isn’t restricted to patients with detectable PD-L1 expression. Quantifying PD-L1 expression on immune cells in addition to tumor cells seemed to strengthen the predictive value of PD-L1 for OS benefit, as immune cells expressing PD-L1 also play a key role in regulating antitumor immune response. There are some challenges in establishing PD-L1 expression as a reliable predictive biomarker. First, the attempts to take PD-L1 as a predictor of immunotherapy have yielded variable results using different cutoffs. Second, the antibodies and testing platforms varied, e.g., SP263 (durvamab) and SP142 (atezolizumab) on the Ventana platform, 22C3 (pembrolizumab) and 28-8 (nivolumab) on the DAKO platform, and 73-10 on the Abcam platform (avelumab). It is hard to reach a uniform standard for the detection of PD-L1 expression. The International Association for the Study of Lung Cancer Pathology Committee has made effort to harmonize and standardize testing for PD-L1 by immunohistochemistry (63). Third, PD-L1 expression may have a great temporal and spatial heterogeneity. In one study, the concordance rate for PD-L1 levels was only 67% between paired samples collected more than three months apart (64). In another study, assessment of different fields of view in the same patient sample showed discordant expression at a frequency of 25% (65). We have previously made significant effort to identified the heterogeneity of PDL1 between the tumor microenvironment of paired primary lung cancers and metastatic lesions (66–69). Finally, irresponsiveness and rapid disease progression can be observed in patients with high tumoral PD-L1 expression, while conversely responses can still occur in PD-L1 negative tumors (70). The CheckMate 227 trial showed a similar survival outcome between the high and low PD-L1 expression in NSCLC patient with a high TMB (71). The results indicated that combining biomarkers, such as PD-L1 and TMB, may increase the predictive efficiency.

As described previously, TMB could be a potential biomarker predicting better response to ICIs. Tumors with high TMB are more likely to generate neoantigens for immune recognition and tumor cell killing, thus resulting in stronger antitumor immune responses to PD-1/PD-L1 or CTLA-4 blockade. NCCN guidelines endorsed TMB as a predictor of immunotherapy based on the positive findings in advanced NSCLC from CheckMate 227 (4). CheckMate 227 also assessed the benefit of ICIs for patients with TMB^highPD-L1^low, and found that these patients could significantly benefit from ICIs (HR 0.51, 95%CI 0.30–0.87) (4). The findings suggested that combining TMB with PD-L1 can identify more patients who can potentially benefit from ICIs. This need to be verified in a larger sample size. The result in our study was consistent with previous publications, but our analysis is limited by the few trials reporting this analysis. A larger effort is needed before taking TMB as a convenient and affordable tool in clinical practice. The first challenge is the use of variable cutoffs for high TMB among different studies and between blood or tumor tissues (72–75).

Our study has several strengths. First, we performed an exceedingly comprehensive literature search and included most updated data. Second, we included multiple clinicopathological and biomolecular characteristics in the analysis, some of which such as race and region were analyzed for the first time. Third, we performed analyses with various cutoffs for PD-L1 and bTMB as well as different categorization of race and region. Additionally, we performed subgroups analyses to explore whether immunotherapy settings affected the association of ICI OS benefit with clinical variable such as sex, age and ECOG PS.

Our study also has limitations. First, we conducted the analysis based on published study-level data but not on individual patient-level data. Second, statistically significant heterogeneity was found among the studies for some subgroups. Third, the number of studies that were included in the subgroups analyses and in TMB analysis was small. Fourth, clinical trial enrollment criteria limit the range of many variables that are available for our analysis. For example, patients with an ECOG PS of 2 or lower typically are not included in trials, and patients with brain metastasis typically have undergone resection or received radiation. Finally, our results may not be generalizable to all patients and clinician experiences, because we did not include non-English studies and real-world studies.

This meta-analysis, which included an interaction test, demonstrated that the OS benefit from ICIs did not appear to differ on the basis of patients’ sex, age, ECOG PS, histology, smoking history, baseline brain metastasis, race, and region. These data suggest that the use of ICIs in NSCLC patients should not be restricted by those variables. However, EGFR status, PD-L1 expression, and TMB are important biomarkers that could potentially predict the efficacy of ICIs. The value of using multiple factors in predicting the efficacy of ICIs might be an important future research topic.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

Conception and design: FY, YW, and JM. Collection and assembly of data: FY, YW, and LT. Assessed the eligibilities of feasible studies: FY, YW, and LT. Statistical analysis: FY, YW, and LT. Wrote the first draft of the manuscript: FY and YW. Revised and edited the manuscript: all authors. Final approval of manuscript: all authors.

This research was funded by National Natural Science Foundation of China (No. 82002783), Nanjing Outstanding Youth Fund (No. JQX21001), The Double Innovation Talent Program of Jiangsu Province (2019).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.