*3.2. Study Characteristics*

The specific characteristics of the studies included in the MA are summarized in Table 1. The control arm in all studies was platinum-based chemotherapy with pemetrexed (three studies [17,20,21,28]) or with nab-paclitaxel/paclitaxel (five studies [21–25,29]). In one three-arm trial (IMpower150 [24]), bevacizumab was added in the control and experimental arm. This study included two experimental arms: carboplatin, paclitaxel, and atezolizumab (arm A) and carboplatin, paclitaxel, bevacizumab, and atezolizumab (arm B) versus carboplatin, paclitaxel, and bevacizumab (arm C). No comparisons between arms A and C were performed because the HR may not reflect the actual effect of add-on immunotherapy (atezolizumab plus chemotherapy vs. bevacizumab plus chemotherapy).

**Figure 1.** Flow chart of study selection (up to 1 January 2020). NSCLC, non-small-cell lung cancer; RCTs, randomized controlled trials.



\* Patients with epidermal growth factor receptor (EGFR) or anaplastic lymphoma kinase (ALK) mutations excluded. PD-L1, programmed cell death-ligand 1; PFS, progression-free survival; OS, overall survival; ITT, intention-to-treat; WT, wildtype; +, plus (combination therapy).

Three studies tested anti-PD1 antibodies [17,21,22,28], and four studies tested anti-PD-L1 antibodies [20,23–25,29]. Patients with epidermal growth factor receptor (EGFR) or anaplastic lymphoma kinase (ALK) mutations were included in two clinical trials assessing atezolizumab [24,25]. Regarding the histology, four studies included patients with nonsquamous NSCLC [20,21,24,25], two included patients with squamous NSCLC [22,29], and one evaluated patients presenting both histological types [21] (note that the primary endpoint in CheckMate-227 part 2 was OS in nonsquamous mNSCLC patients only; however, both histological types were considered for this MA).

An all-comers design was used in all the studies, with NSCLC patients entering the trial regardless of their PD-L1 expression status. Stratification was performed based on this biomarker in all trials. Thus, subjects were classified as PD-L1-negative or PD-L1-positive, and within this group, investigators distinguished patients with high or low expression levels [20,25,29]. In the atezolizumab trials [20,23–25,29], levels were considered high (TC3 or IC3) when PD-L1 expression was recorded on at least 50% of tumor cells or at least 10% of tumor-infiltrating immune cells (TIICs) by immunohistochemistry; levels were considered low–intermediate, or TC1/2 or IC1/2, when expression was reported on at least 1% of tumor cells or TIICs and less than 50% of tumor cells or less than 10% of TIICs by immunohistochemistry; and PD-L1-negative status, or TC0 and IC0, was determined when expression was reported on less than 1% of tumor cells and TIICs. Similar criteria were followed in trials assessing pembrolizumab or nivolumab, but PD-L1 expression was only measured in tumor cells [17,21,22,28].

According to the eligibility criteria, none of the studies included patients who had received prior treatment for metastatic disease. However, in terms of therapy for nonmetastatic disease, although most of the studies included treatment-naïve patients, in those evaluating pembrolizumab (KEYNOTE-189 [17,28] and KEYNOTE-407 [22]) or atezolizumab [20,23–25,29], subjects had received previous therapies (Supplementary Table S1).

Coprimary endpoints for six clinical trials were PFS and OS [17,20,22–25,28,29]. The primary endpoint for CheckMate-227 part 2 was OS in nonsquamous NSCLC patients; PFS was assessed as a secondary endpoint [21]. Mature PFS data were reported in all the studies included in this MA [17,20–25,28,29], while final data for OS were available for only one of them [21]. Interim analyses were provided for the other six studies [17,20,22–25,28,29]. Both endpoints were evaluated in the intention-to-treat (ITT) population and specifically in the wildtype population (without EGFR or ALK mutations) in the IMpower130 [25] and IMpower150 studies [23,24] (see Supplementary Table S2 for the available information on patients with mutations in IMpower150). IMpower150 was the only study in which a subsequent subgroup analysis in patients with EGFR mutations or baseline liver metastasis was performed [23]. Additionally, one or both coprimary endpoints were analyzed according to different subgroups in all the studies included in the MA (PFS in six clinical trials [17,22–25,28,29] and OS in five studies [17,21,22,25,28,29]).

Patient population characteristics of all the studies included in the MA are shown in Supplementary Table S3.

#### *3.3. E*ffi*cacy Endpoints in the Overall Population*

Median PFS ranged from 4.8 to 6.8 months in the control arms and from 6.3 to 8.8 months in the treatment arms. Median OS ranged from 10.7 to 14.7 months in the control arms and from 14.2 to 22.0 months in the treatment arms. MA results demonstrated that the addition of a PD-(L)1 to chemotherapy was associated with improved PFS (PFS: HRpooled = 0.61, 95% CI: 0.57–0.65, *p* < 0.001, Figure 2A) and OS (OS: HRpooled = 0.76, 95% CI: 0.67–0.86; *p* < 0.001, Figure 2B) compared with chemotherapy alone. The objective response rate (ORR) was also significantly improved with the PD-(L)1 inhibitor–chemotherapy combination (odds ratio (ORpooled) = 2.12, 95% CI: 1.70–2.63, *p* < 0.001, Supplementary Figure S1). The best ORR values were obtained in the IMpower150 (ORR = 56.4%) trial for nonsquamous NSCLC and KEYNOTE-407 for squamous NSCLC (57.9%). Notably, in terms of both OS and ORR, there was significant heterogeneity across the six trials (I<sup>2</sup> = 52.07%, *p* = 0.03; I2 = 67.42%, *p* = 0.005).
