Redefining Treatment Paradigms in First-line Advanced Non–Small-Cell Lung Cancer

Treatment of metastatic non–small-cell lung cancer (mNSCLC) has diversified remarkably in recent years. From a “one-size-fits-all” approach in terms of platinum doublet chemotherapy regardless of histologic subtype or molecular profile, with poor median overall survival (OS), the emergence of targeted therapy with tyrosine kinase inhibitors (TKI) has improved outcomes in molecularly defined subsets of non–small-cell lung cancer (NSCLC). The introduction of checkpoint inhibitors, enhancing the immune response to cancer, showed long-term responses and even the remote possibility of cure for another subset of patients. Their broad application, however, especially in combination with chemotherapy, poses the risk of a high financial burden to health care systems.

In this review, we present the current landscape of predictive biological factors defining three different scenarios within this disease population. First, established selection markers like histology and suitability for antiangiogenic agents. Second, druggable driver alterations like EGFR-mutations, anaplastic lymphoma-kinase (ALK) rearrangements, predominantly occurring with echinoderm microtubule–associated ligand 4 (EML4), and rearrangements involving ROS proto-oncogene (ROS1). Third, the impact of programmed-death-receptor-ligand1 (PD-L1) IHC staining in tumor- and surrounding immune cells and the impact of the number of somatic mutations within the tumor genome, so-called tumor mutational burden (TMB), on treatment decisions and opportunities.

Consequently, different treatment paradigms result: patients with treatable driver mutations, patients with a promising “immune signature” (either PD-L1 high, TMB high, or both) and those, who have overlapping features or lack them all. The current and future treatment options will be described, as well as areas requiring research to avoid a return to the “one-size-fits-all” approach in the treatment of mNSCLC.

The pathologic diagnosis of NSCLC has been transformed over the past decade: virtually every newly approved drug for NSCLC requires biomarker-driven patient selection.

The identification and subtyping of NSCLC on usually limited amounts of tumor in biopsy/cytology samples (1), makes tissue acquisition and handling and avoidance of waste important (2). Accurate subtyping of NSCLC into, effectively, squamous cell or adenocarcinoma, is required for several reasons, relating to drug efficacy/regulatory approval (e.g., pemetrexed, nintedanib) and possible toxicity (e.g., bevacizumab in squamous cell carcinoma). IHC is a key factor in determining the NSCLC histo-subtype in about a third of cases when morphology alone is not enough (3). The predominance of targetable addictive oncogenes in adenocarcinoma drives guidelines recommending the triage of cases for molecular testing (4).

A wide range of potential oncogenic drivers and drug targets may be found in adenocarcinoma (5). Relevant drugs vary in approval status and availability, and this, together with the possibility of usage in trials often determines testing strategy. In many health systems, there are only three or four such molecular targets with approved agents available and therefore considered mandatory for testing (4); activating and sensitizing mutations in exons 18–21 of EGFR, at the V600 codon in BRAF or when there is an ALK or ROS gene rearrangement. In a Western population these alterations would account for around 18% of adenocarcinoma patients; in East Asia around 60%. Testing strategies are highly variable and clinical outcomes are, to a large extent, agnostic of methodology. How a mutation is identified is less important, but it is critical that methodology can detect the required range with sensitivity high enough to allow for low mutant allele prevalence in clinical samples. For ALK testing, although demonstration of the rearrangement might seem the intuitive biomarker test of choice (in situ hybridization, multiplex PCR, or RNA sequencing), the presence of elevated levels of ALK protein in tumor cells bearing the rearrangement is crucial for drug efficacy (6). This may be a principle that follows for other rearrangements or gene copy number alterations.

As more targets/biomarkers become clinically important, so will the pressure to implement multiplex testing. Next-generation sequencing strategies offer such an approach and are rapidly being adopted into clinical practice in many countries. Multiplex testing potentially provides a huge amount of data on concurrent molecular alterations, but we have yet to understand the full significance of this for treatment outcomes when targeting, for example, an EGFR mutation or ALK gene arrangement.

Cancer immunotherapy with checkpoint inhibitors (CIT) has revolutionized treatment for advanced NSCLC where PD1 axis inhibitors are predominant. Here, it is logical that PD-L1 expression, as measured by IHC, is now established as the preeminent approved biomarker in many indications in NSCLC. In trials, PD-L1 IHC expression levels have almost never failed to positively correlate with improving outcomes. The plethora of PD-L1 assays, each allied in clinical trials to different drugs, has been a challenging issue and this is well discussed elsewhere (7–9). Other biomarkers are now emerging in this space; TMB as a surrogate for possible neoantigenicity is also associated with response to CIT, as are measures of tumor inflammation, as a surrogate for the presence of a tumor-directed immune response that may be reactivated by CIT (10). These biomarkers are being actively investigated in trials, more often as additional rather than alternative factors. Future biomarker testing for CIT will undoubtedly be complex.

Bevacizumab added to chemotherapy results in a modest gain in progression-free survival (PFS) and OS, especially in adenocarcinoma (11). However, molecular predictors for the efficacy of antiangiogenic agents are lacking: the most promising marker for the efficacy of bevacizumab, VEGF-A concentration in serum, proved to be prognostic only (12).

Every case of NSCLC tested will not provide a target for therapy according to current mandatory practice (EGFR, ALK, ROS1, BRAF, PD-L1). More (rare) targets are in development but a proportion of cases are likely to remain “biomarker negative,” pending new discoveries.

Activating EGFR mutation was the first driver mutation to introduce precision medicine for management of mNSCLC (13, 14). The advantage of EGFR TKIs over chemotherapy was established in the landmark IRESSA Pan-Asian Study (IPASS) that confirmed the improvement in response rate and PFS in the EGFR mutation–positive subgroup (15). To date there are three generations of EGFR TKIs including erlotinib and gefitinib as first-generation, afatinib and dacomitinib as second-generation, and osimertinib as third-generation drugs. Multiple randomized studies have demonstrated an improvement in efficacy compared with chemotherapy, with EGFR TKI–treated groups reaching a median PFS ranging from 9 to 14.7 months (15–22). Second-generation EGFR TKIs show irreversible receptor binding and also inhibit other ErbB family members (EGFR/HER1, HER2, and HER4; refs. 23–25); thus, these drugs would be expected to be more potent. Park and colleagues reported a slight improvement in PFS (median 11.0 vs. 10.9 months) when comparing afatinib with gefitinib in a randomized phase IIb study; however, a significant OS benefit was lacking (26). In contrast, the ARCHER 1050 study demonstrated improvement in both PFS (median 14.7 vs. 9.2 months) and OS (median 34.1 vs. 26.8 months) when dacomitinib was compared with gefitinib in patients without central nervous system (CNS) metastasis (27). A direct comparison between dacomitinib and afatinib has not, so far, been reported. Afatinib is the only TKI to be also trialed in rarer EGFR-mutations (28). Osimertinib has the advantage of inhibiting the T790M resistance mutation in addition to activating mutations. The first-line FLAURA study compared osimertinib with first-generation EGFR TKI and demonstrated superiority in PFS (median 18.9 vs. 10.2 months) as well as significant improvement in control of CNS metastasis (29).

Addition of chemotherapy or an antiangiogenic drug to an EGFR TKI may potentially improve treatment efficacy. NEJ026 is the first randomized phase III study that compared erlotinib with the combination of erlotinib and bevacizumab. The study reported median PFS of 13.3 and 16.9 months, respectively, but no significant difference in OS (47.4 vs. 47 months; ref. 30). A combination of EGFR TKI with chemotherapy was investigated in another Japanese randomized phase III study (NEJ009). The combination of pemetrexed, carboplatin, and gefitinib followed by pemetrexed and gefitinib maintenance showed significant improvement in both PFS (median 20.9 vs. 11.2 months) and OS (52.2 vs. 38.8 months; ref. 31) when compared with a sequential approach with gefitinib followed by chemotherapy upon progression.

ALK gene rearrangement can result in an addictively oncogenic driver fusion protein (32). Similar to EGFR mutation, ALK rearrangement is more common in younger, never-smoker patients with pulmonary adenocarcinoma. Solomon and colleagues reported the first randomized study comparing crizotinib with standard chemotherapy (33), which established a standard-of-care, mandating all patients with adenocarcinoma be tested for ALK rearrangement. Medium OS reaches beyond 48 months with crizotinib in these patients (34). Second generation, including ceritinib, alectinib, and third generation ALK TKIs (lorlatinib, brigatinib, and ensartinib), are specifically designed to inhibit the ALK tyrosine kinase and are associated with better CNS penetration (35–37). Ceritinib is superior to chemotherapy in patients with ALK-rearranged tumors, but this drug has not been compared directly with crizotinib (38). The ALEX study is the first randomized phase III study that compared alectinib directly with crizotinib, and the reported median PFS was 34.8 and 10.9 months, respectively (36). A more striking result from this trial was the significant difference in cumulative incidence of CNS progression at 1 year (alectinib 9.4%; crizotinib 41.4%). Similar efficacy is observed with brigatinib in the ALTA-1L study. While median PFS was not reached after 11 months of observation for brigatinib, median PFS with crizotinib was 9.8 months. Moreover 12-month PFS was 67% with alectinib and 43% with crizotinib (39). Intracranial response rate was 78% in patients with measurable CNS metastasis while cumulative incidence data for CNS progression are not available. Two other phase III studies on ensartinib (NCT02767804) and lorlatinib (NCT03052608), each in comparison with crizotinib, are ongoing.

There are at least 6 other potentially targetable driver alterations, which include ROS-1, RET, and NTRK rearrangements, BRAF V600E point mutation, MET Exon 14 skipping mutation, and HER-2 mutation.

To date, the FDA and European Medicines Agency (EMA) have approved drugs targeting ROS-1 and BRAF V600E. A total of three single-arm studies have confirmed the high tumor response rate (71.2%–80.0%) and prolonged PFS (median 9.1–19.2 months) for crizotinib in patients with ROS-1–positive lung cancer (40–42). However, the majority of patients enrolled in these trials had prior chemotherapy. With only 31 patients who were treatment-naive in these three studies, it remains debatable whether crizotinib should be offered as first- or second-line therapy. The combination of dabrafenib and trametinib is the only approved therapy for BRAF V600E–positive lung cancer. Results of the single-arm first-line study and the second/third-line study are similar. Response rate was 64.0% and 63.2%, and median PFS was 10.9 and 9.7 months, respectively (43, 44).

A number of promising drugs are being investigated in trials or off-label usage for other uncommon driver mutations, and these include LOXO-292 (NCT03157128) and BLU-667 (NCT03037385) for RET rearrangement; tepotinib (NCT02864992) for MET Exon 14 skipping mutation, and entrectinib (NCT03066661) and larotrectinib (NCT03213704) for NTRK rearrangement.

Use of the PD-1 checkpoint inhibitor pembrolizumab in patients with high PD-L1 expression in tumor cells (TPS ≥50% of cells), is now a standard-of-care for first-line treatment of mNSCLC, excluding patients with EGFR- or ALK-driver alterations. A median OS of 30 months confers a significant benefit in comparison with platinum-based chemotherapy, achieving 14.2 months of median OS (HR 0.63; P = 0.002) with less severe side effects under CIT (45, 46). The PD-L1 “high-expressors” consist of a “typical” lung cancer population: predominantly males with a smoking history in approx. 80% of cases (Fig. 1). In another trial pembrolizumab was equally effective as platinum chemotherapy in patients with TPS of 1%–49% (47). It is hypothesized that nonimmunogenic “cold” tumors, may be made immunogenic by using combinations of immunotherapy with cytotoxic chemotherapy that increase expression of neoantigens (48), Antiangiogenic therapy may reverse the immunosuppressive effect of the VEGF-mediated pathway (49), and several trials have been designed using such combinations. In PD-L1 unselected populations meanwhile six pivotal trials have shown significant improvement of PFS and, in most, OS in favor of a combination of platinum-based chemotherapy with an anti–PD 1/PD-L1 antibody. This benefit, although variable, has been observed independent of histology and across all levels of PD-L1 expression including PD-L1–negative tumors (Table 1; refs. 50–55).

Another approach combined the PD-1 checkpoint-inhibitor nivolumab with the cytotoxic T-lymphocyte–associated Protein 4 (CTLA-4) checkpoint inhibitor ipilimumab, in the phase III Checkmate 227 trial. In this six-arm trial, patients were stratified into two cohorts (PD-L1<1% and PD-L1 ≥1%, respectively). In the PD-L1–positive cohort, patients were randomized into three arms: platinum doublet chemotherapy, nivolumab + ipilimumab, or nivolumab alone. The three arms in the PD-L1–negative cohort were nearly similar; however, instead of nivolumab alone, platinum doublet chemotherapy was added. Initial analysis revealed significant improvement in PFS favoring the combination of nivolumab + ipilimumab compared with chemotherapy in the group of patients selected by a TMB ≥ 10 per megabase (called “TMB high”): There is a clear PFS benefit in patients with TMB high, regardless of PD-L1 expression (Table 2; ref. 56).

Although new therapeutic options are evolving, key questions remain, such as how to manage patients with overlapping biological factors (e.g., PD-L1 high/TMB low or vice versa) or patients lacking any predictive biomarker (Fig. 2).

Some data come from the Checkmate 227 trial. In those patients with a PD-L1 expression < 1%, an improvement of response and PFS was observed favoring the combination of chemotherapy and nivolumab compared with chemotherapy. However, an additional exploratory analysis revealed that this benefit was restricted to those patients who also had TMB high (Table 3; ref. 57). This study cohort was small and prospective validation of TMB is required, especially because TMB may have a strong prognostic effect independent of any predictive activity (new.bms.com).

The question whether patients with PD-L1 ≥ 50% need combination therapy with chemo and CIT or are sufficiently treated with pembrolizumab alone, is unanswered. The authors are unaware of any clinical data guiding this decision, that is, whether TMB could serve as a biomarker in this population.

The presence of an addictive driver alteration like EGFR or ALK, treatment should be with a target-specific TKI, regardless of PD-L1 expression or TMB. Studies have shown that monotherapy with checkpoint inhibitors do not work in these patients, either in pretreated or untreated populations (58–60). Furthermore, patients with EGFR mutation and high TMB have poorer outcomes compared with those with low TMB (61). The average TMB in patients with EGFR- or ALK-alterations cancers is below the median for all mNSCLC (62). However, combination therapy might work differently. While checkpoint inhibitors plus TKI combinations have not so far shown substantial improvement in efficacy compared with TKI alone, and were frequently associated with higher toxicity, patients with pretreated EGFR mutations or ALK translocations, who received the combination of atezolizumab, bevacizumab, and chemotherapy within the IMpower 150 study (52) appeared to derive some benefit. This area clearly requires further investigation.

Options for treating lung cancer, especially mNSCLC, have grown dramatically in recent years. Complete pathologic and biologic typing of the tumor is mandatory to guide first-line treatment. In the future, after proper histologic or cytologic diagnosis of the tumor type, Multiplex testing using next-generation sequencing is rapidly becoming the mainstay of molecular diagnostics, as it can provide data on molecular targetable alterations (e.g., ALK, EGFR, BRAF, ROS1, cMET, NTRK), putative prognostic factors (i.e., KRAS, p53), and TMB. PD-L1 expression by IHC should remain.

EGFR- and ALK- alterations should be treated with a corresponding TKI in first line. Patients without driver alterations and PD-L1 expression ≥50% should receive pembrolizumab. It is unclear whether the addition of platinum chemotherapy really adds benefit to either of these patient groups. In patients with oncogene-negative tumors with PD-L1 expression <50%, combination of chemotherapy and checkpoint inhibition is a treatment option; however, the role of TMB to guide treatment remains uncertain (Fig. 2). We clearly need more clinical data to decide whom to treat with ipilimumab and nivolumab and whom with chemotherapy plus a checkpoint inhibitor. Of course, there remains a subpopulation (most likely those with negative PD-L1 and low TMB) who will derive no benefit from CIT.

D.F. Heigener reports receiving speakers bureau honoraria from Roche, Boehringer Ingelheim, Lilly, Pfizer, MSD, Bristol-Myers Squibb, AstraZeneca, Abbvie, and Takeda, and is a consultant/advisory board member for Roche, Boehringer Ingelheim, Bristol-Myers Squibb, and AstraZeneca. K. M. Kerr is a consultant/advisory board member for AstraZeneca, Roche, MSD, Bristol-Myers Squibb, Boehringer Ingelheim, Pfizer, Novartis, and EMD Merck Serono. T.S.K. Mok reports receiving other commercial research support from AstraZeneca, Boehringer Ingelheim, Clovis Oncology, MSD, Novartis, Pfizer, Roche, SFJ Pharmaceuticals, and XCovery, speakers bureau honoraria from AstraZeneca, Roche/Genentech, Pfizer, Eli Lilly, MSD, Novartis, Bristol-Myers Squibb, Taiho, Takeda Oncology, PrIME Oncology, Amoy Diagnostics, and ACEA Pharma, holds ownership interest (including patents) in Sanomics Ltd. and Hutchison Chi-Med, and is a consultant/advisory board member for AstraZeneca, Roche/Genentech, Pfizer, Eli Lilly, Boehringer Ingelheim, Merck Serono, MSD, Novartis, SFJ Pharmaceuticals, ACEA Pharma, Vertex, Bristol-Myers Squibb, geneDecode Co. Ltd., OncoGenex, Celgene, Ignyta Inc., Cirina, Fishawack Facilitiate, Janssen, Takeda Oncology, Hutchison Chi-Med, OrigiMed, Hengrui Therapeutics, Sanofi-Aventis R&D, Yuhan Corporation, Loxo-Oncology, and Virtus Medical Group. F. V. Moiseyenko reports receiving speakers bureau honoraria from AstraZeneca, Takeda, Roche, Bristol-Myers Squibb, Boehringer Ingelheim, and Lilly, is a consultant/advisory board member for AstraZeneca and Takeda, and reports sponsorship of research held at St. Petersburg City Cancer Center by Biocad and AstraZeneca, as well as participation in educational activities held at St. Petersburg City Cancer Center and international educational activities sponsored by Sanofi, Pfizer, Novartis, MSD, Merck, and AstraZeneca. M. Reck reports receiving speakers bureau honoraria from Abbvie, Amgen, AstraZeneca, Bristol-Myers Squibb, Boehringer-Ingelheim, Celgene, Lilly, Merck, MSD, Novartis, Pfizer, and Roche, and is a consultant/advisory board member for Abbvie, Amgen, AstraZeneca, Bristol-Myers Squibb, Boehringer-Ingelheim, Celgene, Lilly, Merck, MSD, Novartis, Pfizer, and Roche. No potential conflicts of interest were disclosed by the other authors.