Immunotherapy is a key and effective method for the treatment of cancer patients, which improves the treatment mode of cancer. Immune checkpoint inhibitors (ICIs) can block cytotoxic T lymphocyte antigen-4 (CTLA-4) or programmed death-1 (PD-1) pathway and inhibit the release of negative regulatory factors of immune activation to enhance anti-tumor response (1). To date, a number of large-scale randomized controlled trials (RCTs) have demonstrated that ICIs represented by programmed death-1 (PD-1)/programmed death-ligand 1 (PD-L1) inhibitors, whether used as monotherapy or as combination therapy, provide long-term survival and lasting benefits for patients with non-small cell lung cancer (NSCLC) (2–6).

However, the survival benefits are observed in only a small number of patients (15%-25%), and the majority of patients have primary or acquired resistance to ICIs. Considering the high cost of immunotherapy and immune-related adverse reactions, it is necessary to explore appropriate biomarkers to find patients suitable for immunotherapy and to achieve accurate treatment of lung cancer (7). Our previous meta-analysis has demonstrated that PD-L1 expression detected by immunohistochemical is an effective biomarker for predicting the efficacy of checkpoint inhibitors in NSCLC. Patients with high levels of PD-L1 expression are more likely to benefit from anti-PD-1/PD-L1 therapy (8). However, the detection of PD-L1 expression depends on the patient’s tissue sample, which is difficult to obtain and the sample size is usually very small. Moreover, in practical application, there are various antibody clones and assays, which provide challenges for the detection of PD-L1 expression (9). Tumor mutation burden (TMB) is another predictive biomarker of widespread concern. Whether TMB can clearly predict the efficacy of immunotherapy remains controversial. The KEYNOTE-158 study prospectively explored the relationship between high tissue TMB and the efficacy of pembrolizumab (anti-PD-1 antibody), and found that patients with high TMB had better response rates (10). In the exploratory analyses of KEYNOTE-021 and KEYNOTE-189, there was no significant correlation between TMB and the efficacy of immunotherapy (11, 12). Therefore, it is of great value to explore other economic and practical factors to predict the efficacy of immunotherapy. In some prespecified subgroups of RCTs, the effects of immunotherapy varied among patients with different clinical and molecular characteristics such as age, sex, race, Eastern Cooperative Oncology Group (ECOG) performance status (PS) score, and so on. For example, in IMpower 130, there was no significant difference in overall survival (OS) between the atezolizumab plus chemotherapy group and the chemotherapy group among male, patients aged <65 years and ≥65 years, current or previous smokers, never smokers, or with liver metastases (2). In the CheckMate 017 trial, nivolumab significantly improved survival in male, patients aged <75 years, and in the region of US/Canada or Europe, but not in female, patients aged ≥75 years, and in the rest-of-world region (4). Checkmate 227 found that no survival improvement of ICIs was observed in female, patients aged 65-74 years, ≥75 years, ECOG PS 1, never smokers, non-squamous NSCLC, with liver metastases or brain metastases (5). Thus, a pooled analysis of relevant RCTs is needed to further investigate whether clinical or molecular factors can predict survival in NSCLC patients receiving immunotherapy.

In this meta-analysis, we conducted a systematic review to comprehensively compare the efficacy of anti-PD-1/PD-L1-based therapy and non-ICI therapy in patients with different >clinical and molecular characteristics, and to identify people who are more likely to benefit from immunotherapy. We present the following article in accordance with the PRISMA reporting checklist.

Although immunotherapy has made a significant breakthrough in NSCLC, only a small number of patients benefit from ICIs. Therefore, it is of great value to explore appropriate biomarkers to guide the selection of patients suitable for immunotherapy. PD-L1 expression and TMB are currently the most widely studied biomarkers, but the detection process is complex and expensive, which brings challenges to cancer treatment. Some retrospective studies have also been conducted to try to explore novel biomarkers. Prat et al. found that PD-1 gene expression and 12 signatures tracking the activation of CD8 and CD4 T-cells, natural killer cells, and IFN were significantly associated with PFS (33). In the POPLAR trial, atezolizumab benefited survival in tumors with high expression of T-effector and IFN-γ gene signatures (3). Patients with higher ratios of central memory T cells to effector T cells had longer PFS (34). Neutrophil to lymphocyte ratio (NLR), pretreatment lactate dehydrogenase (LDH), lung immune prognostic index (based on derived NLR and LDH levels), C-reactive protein (CRP), and gut microbiome may also be potential biomarkers (35–39). However, these biomarkers have only been identified in retrospective or exploratory analyses of small samples, and their predictive value of efficacy needs to be further confirmed in prospective trials. In addition, for some biomarkers, such as NLR, LDH and CRP, there is still no uniform standard to define the relevant threshold. With the accumulation of the latest clinical data, we attempted to explore whether there are more available and cost-effective clinical and molecular pathologic factors to predict the efficacy of immunotherapy.

Our meta-analysis included 19 RCTs. We compared the survival of patients with different clinical and molecular characteristics (age, sex, histological type, ECOG PS score, smoking status, driver mutations, brain metastases, liver metastases, region and number of prior systemic regimens) who received immunotherapy with those who received non-ICI treatment. Meanwhile, we conducted pre-defined subgroup analyses according to the line of therapy, treatment regimen and target of ICIs to explore the role of ICIs in these populations. Our study found that age, EGFR mutation status, smoking status and number of prior systemic regimens could effectively predict the efficacy of immunotherapy. To the best of our knowledge, our meta-analysis is the most comprehensive study with the largest number of RCTS included, providing guidance for better identification of which patients are most likely to benefit from anti-PD-1/PD-L1 treatment.

In previous studies which investigated the relationship between age and the efficacy of immunotherapy, the cut-off age was mostly 65 years old. They found no statistical difference between the ICI group and non-ICI group in patients <65 years old and ≥65 years old (40, 41). However, it remains unclear whether elderly NSCLC patients aged ≥75 years will also benefit from immunotherapy. A multicenter retrospective study of patients aged ≥75 years found that the efficacy of ICIs in elderly patients was similar to that in young patients (42). Another study in Italy found that patients aged ≥75 years had lower median OS than patients aged <65 years or aged 65-74 years (43). Zheng et al. found that there was no significant difference in survival between the immunotherapy group and the chemotherapy group in patients older than 75 years (44). In our meta-analysis, we included 15 studies on the relationship between age and immunotherapy. Our study was the first to perform a more detailed division of age, and explore the efficacy of immunotherapy in patients aged <65 years, 65-74 years, and ≥75 years. We found that ICIs significantly improved OS compared with non-ICI treatment for patients aged <65 years and aged 65-74 years. However, in elderly patients ≥75 years old, immunotherapy did not significantly prolong the survival.

Due to the poor prognosis of NSCLC patients with distant metastasis (such as brain or liver metastases), the effect of immunotherapy on patients with different metastatic sites has been a research hotspot in recent years. Our combined analysis of six studies involving patients with asymptomatic brain metastases suggested that these patients obtained longer OS after receiving immunotherapy than non-ICI treatment. However, further subgroup analysis suggested that the survival benefit of ICIs was only observed in first-line combination therapy, but not in second or more line monotherapy. Therefore, early ICI-based combination therapy is recommended for patients with asymptomatic brain metastasis. In addition, patients with liver metastases also benefited from immunotherapy. Subgroup analysis showed that both first-line ICI-based combination therapy and second or more line anti-PD-1/PD-L1 monotherapy were associated with improved OS in patients with liver metastases. It is worth noting that PD-1 inhibitors significantly prolonged survival in patients with liver metastases compared with non-ICI therapy, while the survival benefit was not observed in patients receiving PD-L1 inhibitors. Similarly, a recent study also found that PD-1 inhibitors showed superior survival compared to PD-L1 inhibitors in cancer treatment (45). This may be because although both PD-1 inhibitors and PD-L1 inhibitors can block the binding of PD-1 to PD-L1, PD-1 inhibitors can also block the binding of PD-1 to PD-L2 (46). Previous studies suggested that PD-L2 expression was a predictor of the efficacy of ICIs independent of PD-L1 expression. Therefore, the clinical effect of immunotherapy may also be related to the blockage of the PD-1/PD-L2 pathway (46, 47).

In our meta-analysis, smoking status predicted the effect of immunotherapy. We found that survival benefits of immunotherapy were observed only in current or previous smokers, but not in never smokers. This may be because smoking is associated with high TMB, which makes it easier to benefit from ICIs (48). The number of prior systemic regimens also predicted the clinical outcome of immunotherapy. The survival benefit of ICIs was observed when the number of prior systemic regimens was 0 and 1, but it was not observed when the number of prior systemic regimens was 2. Furthermore, our study demonstrated that patients benefited from anti-PD-1/PD-L1 immunotherapy, regardless of sex (male or female), histological type (squamous or non-squamous NSCLC), ECOG PS (PS 0 or 1), and region (East Asia or America/Europe). Since most RCTs excluded patients with poor performance (PS ≥2), we did not investigate the role of ICIs in the population with PS ≥2. A recent meta-analysis, which included 19 clinical studies in real-world, found that PS ≥2 predicted worse survival in patients receiving immunotherapy (49). In the future, whether PS ≥2 is a predictor of poor immunotherapy efficacy remains to be further confirmed in RCTs. For women who received PD-1/PD-L1 inhibitors, there was a trend of PFS improvement compared with non-ICI therapy, but the difference between the two groups was not statistically significant. This may be because female have stronger immune escape mechanisms than male cancer patients, and thus are more likely to develop resistance to immunotherapy (50–52). In addition, our study indicated that women eventually achieved OS improvement, which further suggested the importance of subsequent treatment.

The relationship between driver mutation and anti-PD-1/PD-L1 therapy has been a hot topic. Our study found that EGFR mutation status was associated with the efficacy of ICIs. EGFR wild-type patients benefited from ICIs, while EGFR mutation-positive patients did not. This may be explained by the following reasons. Firstly, different from patients with wild-type EGFR, EGFR mutations influenced the anti-tumor immune response by regulating possible factors related to tumor microenvironment status (such as regulatory T cells, tumor-infiltrating lymphocytes, exosomes, etc.). Secondly, patients with EGFR sensitive mutations were more common among never-smokers, and they had significantly lower TMB than those with wild-type EGFR. Thirdly, previous studies showed that PD-L1 expression in EGFR mutant tumors was significantly lower than that in EGFR wild-type tumors, which led to poor response to anti-PD-1/PD-L1 therapy in EGFR mutant patients (53–56). Although patients with EGFR mutations generally respond poorly to ICIs, some patients may still benefit from immunotherapy. In IMpower150, EGFR mutation-sensitive patients (L858R and 19DEL) treated with atezolizumab plus bevacizumab and chemotherapy achieved an improvement in OS (18). In contrast, other retrospective studies suggested that uncommon EGFR mutations (G719X and exon 20 insertions) were positively associated with the survival benefits of immunotherapy. After disease progression during EGFR tyrosine kinase inhibitors (TKIs) treatment, patients without T790M mutations were more likely to benefit from subsequent immunotherapy (57, 58). In addition, ICI-based combination therapy (such as ICI in combination with chemotherapy or anti-angiogenic drugs) may be more effective than ICI alone in pre-treated EGFR mutant NSCLC patients (59). It has also been suggested that shorter duration of EGFR-TKI remission (<6 months) is associated with longer PFS in subsequent immunotherapy (58, 60). Smoking status may be a clinical predictor of the response to ICIs in EGFR-mutated NSCLC. An Italian study found that among patients with EGFR mutations, the median OS of current or previous smokers was higher than that of non-smokers (14.1 months vs. 5.6 months), although the difference was not statistically significant (P=0.12) (61). Yoshida et al. suggested that a higher Brinkman Index (≥600, defined as the number of cigarettes smoked per day multiplied by the smoking years) might be a favorable predictor for the efficacy of ICIs (58). In a word, it is currently difficult to use a single biomarker to screen potential populations of EGFR-mutated NSCLC who might benefit from immunotherapy. It is important to integrate multiple predictors to assess the outcome of immunotherapy in this population. In terms of KRAS mutation status, we did not have enough evidence to demonstrate its predictive value for the efficacy of immunotherapy. Although PFS improvement was observed in KRAS mutant patients, no statistical improvement in OS was observed in these patients. In addition, there was no statistical difference in survival between the two groups for KRAS wild-type patients. Notably, there were few studies on KRAS mutation status: 2 studies reported OS data and another 2 studies reported PFS data. In the future, it is necessary to conduct more studies to explore the relationship between KRAS mutation status and immunotherapy.

Our meta-analysis also has some limitations. First, some studies included patients with PD-L1 positive expression, which may overestimate the treatment effect of ICIs. Second, there were some differences among the included studies, such as line of therapy, treatment regimen, and target of ICIs, which may lead to heterogeneity. We used the random effect model to solve this problem and conducted subgroup analyses to explore the source of heterogeneity. At the same time, we also carried out sensitivity analyses, which confirmed the reliability of our conclusion. Third, our meta-analysis was based on the results of prespecified subgroup analyses of published RCTs, rather than studies that specifically analyzed the impact of a single clinicopathological factor on immunotherapy. There may be correlations between these clinicopathological factors, such as EGFR mutation status and its association in non-smokers. When we focus on a single feature, other confounders may influence survival outcomes.

In conclusion, age, smoking status, EGFR mutation status, and number of prior systemic regimens predicted the efficacy of immunotherapy. Patients aged <65 years or 65-74 years, receiving first-line or second-line treatment, current or previous smokers, and EGFR wild-type patients may benefit from immunotherapy. However, there was insufficient evidence to demonstrate the predictive value of KRAS mutation status for the efficacy of ICIs. PD-1/PD-L1 inhibitors improved the OS of NSCLC patients regardless of sex (male or female), histological type (squamous or non-squamous NSCLC), ECOG PS (PS 0 or 1), and region (East Asia or America/Europe). Patients with liver metastases also benefited from anti-PD-1-based therapy. In addition, first-line ICI-based combination therapy was recommended for patients with asymptomatic brain metastases. In the practical application of ICIs, the comprehensive consideration of these clinical and molecular biomarkers is helpful to better guide the treatment of NSCLC patients.

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

(I) Conception and design: PZ, YS. (II) Administrative support: TL, PZ. (III) Provision of study materials or patients: JX, MC. (IV) Collection and assembly of data: HL, QW, YX. (V) Data analysis and interpretation: YX, QW. All authors contributed to the article and approved the submitted version.

This work was supported by grants from the National Natural Science Foundation of China (grant number 81401903, 81572937 and 81572273); the 16th batch “Summit of the Six Top Talents” Program of Jiangsu Province (grant number WSN-154); China Postdoctoral Science Foundation 12th batch Special fund (postdoctoral number 45786); China Postdoctoral Science Foundation 64th batch (postdoctoral number 45786); Jiangsu Provincial Postdoctoral Science Foundation (grant number 2018K049A); the Natural Science Foundation of Jiangsu province (grant number BK20180139); Jiangsu Provincial Medical Youth Talent (grant number QNRC2016125); the Nanjing Medical Science and Technology Development Project (No. ZKX17044); and the Jiangsu Provincial Key Research and Development Program (No. BE2016721).

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.