*3.4. Subgroup Analysis*

Subgroup analyses according to sex (women vs. men), age (<65 years vs. ≥65 years), Eastern Cooperative Oncology Group performance status (ECOG-PS = 0 vs. ECOG-PS = 1), smoking status (never-smoker vs. current/former smoker), liver metastasis (yes vs. no), and PD-L1 expression (high vs. low vs. negative) were carried out. As shown in Figure 3, overall, the addition of PD-(L)1 blockade to chemotherapy significantly improved PFS in all the subgroups. Specifically, stratification according

to PD-L1 expression revealed a benefit across all PD-L1 strata with a strong reduction in the risk of disease progression in those patients showing high expression levels (HRpooled = 0.412, 95% CI: 0.34–0.5, *p* < 0.001). In terms of OS (Figure 3), although almost all subgroups benefited from the use of the PD-(L)1 inhibitor–chemotherapy combination, in certain cases, such as in never-smokers and PD-L1-low patients, results did not achieve statistical significance (HRpooled = 0.589, 95% CI: 0.335–1.069, *p* = 0.082; HRpooled = 0.819, 95% CI: 0.648–1.035, *p* = 0.093, respectively).

**Figure 2.** Forest plot of pooled hazard ratios for (**A**) progression-free survival (PFS) and (**B**) overall survival (OS) in patients who received programmed cell death-1 (PD-1)/programmed death ligand-1 (PD-L1) inhibitors plus chemotherapy vs. chemotherapy alone. HR, hazard ratio; CI, confidence interval.

**Figure 3.** Forest plot of hazard ratios for progression-free survival (PFS) and overall survival (OS) in the subgroup analysis. PFS, progression-free survival; OS, overall survival. HR, hazard ratio; CI, confidence interval; Curr., current.; <sup>a</sup> *p* < 0.001; <sup>b</sup> *p* = 0.006; <sup>c</sup> *p* = 0.003; <sup>d</sup> *p* = 0.082; <sup>e</sup> *p* = *0.007*; <sup>f</sup> *p* = 0.093; <sup>g</sup> *p* = 0.055.

Regarding patients with liver metastasis, a specific benefit with atezolizumab plus bevacizumab was observed both in terms of OS and PFS. Further details on the OS and PFS subgroup analyses are shown in Figures 4 and 5, respectively. Additional subgroup analyses based on the histology are available only for PFS and OS in Supplementary Figure S2.

**Figure 4.** Forest plot of hazard ratios for progression-free survival (PFS) in the different patient subgroups. CI, confidence interval; ECOG-PS, Eastern Cooperative Oncology Group performance status; HR, hazard ratio; IM., IMpower; KN, KEYNOTE.

**Figure 5.** Forest plot of hazard ratios for overall survival (OS) in the different patient subgroups. CI, confidence interval; HR, hazard ratio; ECOG-PS, Eastern Cooperative Oncology Group performance status; IM., IMpower; KN, KEYNOTE; CM., CheckMate.

#### **4. Discussion**

The optimal treatment strategy for advanced NSCLC has been the focus of several randomized clinical trials. Promising immunotherapy results in the second or later lines of therapy resulted in the approval of atezolizumab, pembrolizumab, and nivolumab [31–35]. Several clinical trials subsequently evaluated PD-(L)1 inhibitor–chemotherapy strategies in front-line treatment, some of which are included in this MA.

Our results demonstrate an overall benefit—both in terms of PFS and OS—of the addition of PD-(L)1 blockade. Although statistical significance was reached for the pooled HR for OS, substantial heterogeneity (I<sup>2</sup> of 52.07%) across the seven trials was also identified. Furthermore, it is worth mentioning that the most recent data were considered for this MA in the vast majority of cases and that, to date, this is the first analysis to include results from CheckMate-227 part 2 [21]. Positive efficacy results have also been reported by Tun et al. [36], who included almost the same trials as those analyzed in this study (CheckMate-227 data were collected from part 1 [37]). Other meta-analyses have also reported improvements in the efficacy of the combined strategy. Differences in these may be explained by the trials included therein, such as the study by Chen et al., in which comparisons of immune checkpoint inhibitors against chemotherapy were also considered [38]; the study by Shen et al. [39] with broader inclusion criteria (e.g., studies that directly or indirectly investigated the ORR, the disease control response (DCR), or some safety endpoints); or the meta-analysis by Addeo et al. [40], in which studies using avelumab and durvalumab were also considered. Thus, our results support the evidence that a combination strategy of PD-(L)1 inhibitor plus chemotherapy may be beneficial compared to chemotherapy alone. Indeed, to date, the combination of pembrolizumab or atezolizumab with platinum-based chemotherapy, with or without bevacizumab, are EMA-approved options available for first-line treatment of advanced/metastatic NSCLC wildtype tumors.

With respect to subgroup analyses, overall benefits were reported across the different categories. Specifically, analysis in terms of PD-L1 marker yielded a statistically significant improvement in PFS regardless of the level of PD-L1 expression. In the case of OS, improvements were observed in patients with high PD-L1 expression and patients negative for this biomarker, but not in those with low levels, probably because of the moderate–high heterogeneity recorded in the pooled analysis (I2 = 67.01%; *p* = 0.016). It is also important to note that the studies included utilized different PD-L1 assay methods, possibly representing an additional confounding factor to be considered. Other subgroup analyses also resulted in important outcomes. Thus, this meta-analysis demonstrated that patients benefited from additional immunotherapy regardless of their age. It should be noted that the impact of advanced age on the effectiveness of immune checkpoint inhibitors has not been strongly established so far, highlighting the importance of these findings. Interestingly, combinations with pembrolizumab yielded the lowest HR values in terms of both PFS and OS in several subgroups, including women, patients <65 years, and patients with ECOG-PS = 0, pointing to a potential benefit in these individuals. With respect to liver metastasis, in the IMpower150 trial [23,24], improvements were reported both in terms of PFS and OS, suggesting a specific benefit with the atezolizumab and bevacizumab combination. Indeed, although other atezolizumab trials previously reported outcomes in patients with liver metastases, data from IMpower130 [41] and IMpower132 [42] showed no survival benefit with atezolizumab plus chemotherapy, supporting the benefits of adding the antiangiogenic agent in the combination [23]. Despite the fact that the updated KEYNOTE-189 analysis showed a clinical benefit of pembrolizumab-containing regimens over chemotherapy alone in patients with liver metastases (median OS 12.6 vs. 6.6, OS HR 0.62, 12-month OS rate 51% vs. 3%) [43], this baseline characteristic, in contrast with the IMpower trials, was not a stratification factor in the study.

Most clinical trials do not include advanced NSCLC patients with driver mutations. IMpower150 was the only study to include this type of patient, showing a positive trend in OS probably due to the addition of bevacizumab to the combination strategy, as previously discussed [43]. However, this therapeutic strategy for patients with EGFR/ALK mutations should be further confirmed in prospective, randomized studies.

This meta-analysis also has some limitations. First, as mentioned, the PD-L1 assay methods were not consistent across different studies. Thus, while PD-L1 immunohistochemistry was read on both tumor cells and tumor-infiltrating cells in the atezolizumab studies (IMpower) [20,24,25,29,43], PD-L1 expression was only measured on tumor cells in the trials assessing pembrolizumab (KEYNOTE) and nivolumab (CheckMate-227), [17,21,22,28]. Second, six of the included trials only provided interim analysis of the OS [17,20,22,24,25,28,29,43], which may misrepresent overall efficacy. Finally, the subgroup analysis was limited by the available information (PFS subgroup analyses were not assessed in CheckMate-227), and consequently caution must be exercised when interpreting the results. In this regard, certain limitations were also found with the available data of three of the studies, IMpower131, IMpower132, and CheckMate-227 part 2, whose results have only been published as congress abstracts and personal communications to date [20,21,29]. Despite these limitations, our results confirm those obtained in individual studies and are in line with the outcomes obtained in similar meta-analyses.

In conclusion, treatment with PD-(L)1 inhibitors resulted in significantly longer OS and PFS in stage IV NSCLC patients compared with chemotherapy alone. As a result, immunotherapy–chemotherapy combinations may be considered as a first-line strategy for these patients.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/7/2093/s1: Figure S1: Forest plot of pooled odds ratios for overall response rate (ORR) in patients who received PD-(L)1 inhibitors plus chemotherapy vs. chemotherapy alone; Figure S2: Forest plot of pooled hazard ratios for (**A**) progression-free survival (PFS) and (**B**) overall survival in patients with nonsquamous or squamous NSCLC who received PD-(L)1 inhibitors plus chemotherapy vs. chemotherapy alone; Table S1: Treatments previously administered for nonmetastatic disease (data not available for atezolizumab studies); Table S2: Mutation status of IMpower150 [25,26] study patients; and Table S3: Characteristics of the patient population of the studies included in the meta-analysis.

**Author Contributions:** J.G.-G., J.R.-B., L.L.-M. and D.P.P. were responsible for data analyses and manuscript preparation and revision. F.J.A.-A., M.A.-G., M.d.C.A.-M., B.C.-B., J.C.-R., N.F.-N., J.L.F.P., M.L.-Q., D.P.P., L.C., P.R.-G., L.S.-C. contributed to manuscript preparation and approved the final version. All authors have read and agree to the published version of the manuscript.

**Funding:** This study was sponsored by Roche, Spain. Qualified researchers may request access to individual patient level data through the clinical study data request platform (https://vivli.org/). Further details on Roche's criteria for eligible studies are available here (https://vivli.org/members/ourmembers/). For further details on Roche's Global Policy on the Sharing of Clinical Information and how to request access to related clinical study documents, see here (https://www.roche.com/research\_and\_development/who\_we\_are\_how\_we\_work/clinical\_ trials/our\_commitment\_to\_data\_sharing.htm).

**Acknowledgments:** The authors would like to thank Almudena Fuster-Matanzo from Medical Statistics Consulting S.L. (Valencia) for providing scientific support and medical writing services. J.R.-B. is supported by a Río Hortega fellowship from the Institute of Health Carlos III (CM19/00087).

**Conflicts of Interest:** J.G.-G. reports advisory and consultancy honoraria from Roche, Merck, Bristol-Myers Squibb, AstraZeneca, Lilly, Pfizer, and Boehringer; speaker honoraria from Roche; and travel/accommodation/expenses support from Roche, Merck, Bristol-Myers Squibb, and Boehringer. J.R.-B. reports advisory and consultancy honoraria from Boehringer; speaker honoraria from Roche; and travel/accommodation/expenses support from Bristol-Myers Squibb, Merck Sharp & Dohme, Ipsen, and PharmaMar. F.J.A.-A. reports advisory and consultancy honoraria from Roche, Merck, Bristol-Myers Squibb, AstraZeneca, Pfizer, Janssen, Ipsen, Boehringer, Takeda, and Sanofi; speaker honoraria from Lilly; and travel/accommodation/expenses support from Roche, Bristol-Myers Squibb, AstraZeneca, Pfizer, Ipsen, and Boehringer. M.A.-G. reports advisory honoraria from Clovis Oncology and Tesaro, as well as speaker honoraria from Astrazeneca, PharmaMar, and Roche. M.d.C.A.-M. reports advisory and consultancy honoraria from Roche, Merck, Bristol-Myers Squibb, and Boehringer. B.C.-B. reports advisory honoraria from Boehringer and Sanofi; speaker honoraria from Roche, Merck, Bristol-Myers Squibb, and AstraZeneca; and travel/accommodation/expenses support from Roche, Lilly, and Boehringer. N.F.-N. reports advisory and consultancy honoraria from Bristol-Myers Squibb, Bayer, and Boehringer; speaker honoraria from Roche, Bristol-Myers Squibb, AstraZeneca, Bayer, Janssen, Boehringer, and Sanofi; and travel/accommodation/expenses support from Roche, Bristol-Myers Squibb, Lilly, Pfizer, Bayer, and Sanofi. J.L.F.P. reports advisory, consultancy, and speaker honoraria from AstraZeneca, Bristol-Myers Squibb, Boehringer, Kyowa, Lilly, Merck, MSD, Novartis, Pfizer, Pierre Fabre, Takeda, and Roche. M.L.-Q. reports advisory and consultancy honoraria from Roche, Merck, Bristol-Myers Squibb, GSK, Ipsen, Boehringer, Takeda, Sanofi, and Tesaro; speaker honoraria from Roche, Merck, AstraZeneca, Lilly, Janssen, Ipsen, and Boehringer; and travel/accommodation/expenses support from Roche, Merck, Lilly, Pfizer, Ipsen, Boehringer, and Takeda. L.S.-C. reports advisory and consultancy honoraria from Bristol-Myers Squibb and Boehringer, speaker honoraria from Roche, and travel/accommodation/expenses support from Roche and Pfizer. L.L.M. reports advisory and consultancy honoraria from Roche/Genentech, AstraZeneca, Boehringer Ingelheim, Merck Sharp and Dohme, Takeda, Lilly, Bristol-Myers Squibb, and Ipsen. L.C., P.R.-G. and D.P.P. are full-time employees of Roche Farma S.A. at the time the study was conducted. The rest of the authors declare no conflicts of interest.
