The approach to first-line treatment of advanced non-small cell lung cancer (NSCLC) has been evolving rapidly, particularly with respect to patients with lung adenocarcinoma.
H. Jack West, MD
The approach to first-line treatment of advanced non—small cell lung cancer (NSCLC) has been evolving rapidly, particularly with respect to patients with lung adenocarcinoma. In the past decade, one of the first results with major clinical implications was the recognition of superior overall survival (OS) among patients with advanced nonsquamous NSCLC who received cisplatin/ pemetrexed compared with cisplatin/gemcitabine. That was not the case in patients with advanced squamous NSCLC.1
We have also seen the addition of bevacizumab (Avastin) to chemotherapy based on a positive phase III ECOG 4599 trial of carboplatin/paclitaxel alone or with bevacizumab,2 although the modest magnitude of OS benefit along with absence of any OS benefit in another randomized phase III trial3 has led to variable adoption of bevacizumab. As a chemotherapy backbone in the United States, carboplatin is widely favored in light of its very good therapeutic index relative to chemotherapy, which has led to the carboplatin/pemetrexed regimen becoming widely adopted, with or without bevacizumab.
Although any of a wide range of platinum-based doublets can achieve relatively comparable efficacy in advanced lung adenocarcinoma, carboplatin/pemetrexed has become my favored regimen based on the convenient schedule, lack of hair loss, and generally favorable tolerability. I recently led a lung cancer expert summit with 12 national leaders in thoracic oncology; before the live meeting, I asked for specific practice patterns in a survey, which was notable for the complete consensus in favor of a platinum/pemetrexed combination (11 favoring carboplatin, 1 favoring cisplatin), despite the array of acceptable options.4 I have historically been inclined to add bevacizumab for patients with a good performance status and no contraindications, largely based on the impressive efficacy of this regimen demonstrated in the AVAPERL trial of carboplatin/pemetrexed with or without bevacizumab initially, followed by pemetrexed/bevacizumab maintenance,5 but I have been less inclined to favor including bevacizumab for senior patients, based on a subset analysis of the original ECOG 4599 trial that revealed no efficacy benefit and disproportionately greater adverse effects with bevacizumab among patients older than 70 years.6 The views of the aforementioned panel of experts also convey a level of ambivalence about including bevacizumab, as revealed in Figure 1. 7
However, these heavily trodden chemotherapy options and converging standards have undergone an upheaval, as immune checkpoint inhibitors have been studied and assumed a larger role in first-line treatment of advanced NSCLC. Initially studied and then approved as second-line therapies after demonstrating significantly greater efficacy compared with docetaxel in the second-line setting (or potentially later), anti—PD-1 and PD-L1 checkpoint inhibitors have generally been well tolerated and demonstrated remarkably prolonged responses in a subset of patients. These results led to the obvious question of whether immunotherapy agents would provide even greater benefit in the first-line setting, whether for a selected and enriched population or possibly a broader one, potentially replacing or augmenting the efficacy of conventional chemotherapy as initial therapy. Dozens of trials of immunotherapy agents as monotherapy or in combination with chemotherapy, as well as some with combinations of immunotherapy agents, have been launched and completed enrollment. The first few presented and published since late 2016 have dramatically changed our approach to first-line treatment of advanced NSCLC, particularly among patients with adenocarcinoma that does not harbor a targetable driver, such as an EGFR mutation or ALK or ROS1 rearrangement.
The KEYNOTE-024 trial compared pembrolizumab monotherapy with any of several platinum-based chemotherapy doublets in the selected population (approximately 30% of the broad NSCLC population). Patients were selected for a higher probability of benefit from immunotherapy by having a tumor with high-level (>50%) expression of PD-L1 by the 22C3 antibody routinely employed in trials with pembrolizumab.8 Enrolling 305 patients with either previously untreated squamous or nonsquamous NSCLC, the trial was designed with a primary endpoint of progression-free survival (PFS). The results revealed a statistically significant efficacy with pembrolizumab across multiple endpoints, including PFS (median 10.3 mo vs 6 mo; HR, 0.50; P <.001), objective response rate (ORR) (44.8% vs 27.8%), and OS (medians not reached; HR, 0.60; P = .005), alongside an overall more favorable toxicity profile with pembrolizumab. This led to the rapid FDA approval of pembrolizumab in the first-line setting for this specific, biomarker-defined subset. Based on these impressive results, there was a clear and compelling incentive to test tumor PD-L1 expression during the initial workup of advanced NSCLC, with the subset of patients with high PD-L1 clearly favored to pursue pembrolizumab monotherapy.
Interestingly and somewhat surprisingly, a similarly designed trial, CheckMate-026, comparing nivolumab (Keytruda) to any of several chemotherapy combinations, was conducted in a less-selected population (PD-L1 cutoff, 1%).9 This trial not only failed to demonstrate any efficacy benefit with nivolumab compared with chemotherapy in the broader population but also gave no indication of benefit in terms of PFS or OS, even in the more-selected subsets with a higher threshold for PD-L1 expression. The differences in outcomes between KEYNOTE-024 and CheckMate-026 remain essentially unexplained, but the take-home message was that immune checkpoint inhibitors appeared to be less clearly interchangeable, which many of us had anticipated, and only pembrolizumab earned a place in the first-line setting for advanced NSCLC.
Alongside the larger phase III trials of pembrolizumab or nivolumab monotherapy versus chemotherapy, the randomized phase II cohort G of the KEYNOTE-021 trial compared carboplatin/pemetrexed plus concurrent pembrolizumab with the same chemotherapy alone in a population of 123 patients with advanced nonsquamous NSCLC and with no selection by PD-L1 expression level.10 This trial demonstrated a significant improvement in the primary endpoint of PFS (median 13 mo vs 8.9 mo; HR, 0.53; P = .010), as well as a higher ORR (55% vs 19%; P = .0016) with the chemotherapy/immunotherapy combination, but the trend toward more favorable OS, with a median follow-up of only 10.6 months in the population, was not statistically significant (HR, 0.90; P = .39). Notably, patients assigned to chemotherapy alone as first-line therapy were permitted to cross over to receive pembrolizumab upon progression, and 75% with progression after initial chemotherapy did, which could confound the analysis of OS. On the other hand, one might argue that if potential differences in OS conferred by first-line concurrent chemotherapy/immunotherapy are offset by patients’ ability to receive immunotherapy in a sequential manner, as is widely available as FDA-approved second-line treatment, there is no clear rationale to administer these treatment strategies concurrently in the first-line setting.
The clinical relevance of the KEYNOTE-021g trial has been the subject of significant debate and remains controversial. In addition to the concern that patients may do just as well in terms of OS if all or nearly all patients progressing on first-line chemotherapy actually receive second-line immunotherapy in a timely way, critics of the KEYNOTE-021g trial, including myself, have noted that it is unclear whether the trial results would have been favorable if the 30% of patients with high PD-L1 expression had been excluded, as the optimal first-line treatment approach for these patients is arguably pembrolizumab monotherapy, or at least not chemotherapy alone. Should we favor chemo/immunotherapy based on a randomized phase II trial of only 123 patients, particularly if nearly one-third with high PD-L1 expression arguably should be removed from the analysis due to the optimal comparison with chemo/immunotherapy now being pembrolizumab monotherapy, rather than chemotherapy alone, in this population? Moreover, should we make clinical decisions based on a randomized phase II trial when phase III trials that are asking essentially the same question but offering a more definitive answer have completed enrollment and are expected to report findings in the coming months?
Despite our musings about the quality of the data, in April 2017, the FDA approved the combination of carboplatin/pemetrexed with pembrolizumab as first-line therapy for advanced lung adenocarcinoma patients regardless of PD-L1 expression.11 This decision has been criticized by many thoracic oncology specialists, who up to now have not widely adopted this option (Table).12 Updated results with longer follow-up have demonstrated a more pronounced difference in OS over time for chemo/ immunotherapy over chemotherapy alone (Figure 2), now reaching statistical significance (median not reached vs 20.9 mo; HR, 0.59; P = .03).13 These data may well lead to a greater acceptance of this approach, although critics may rightly argue that it is not statistically valid to continually analyze the data until the P value crosses a threshold of .05 (“If you torture the data enough, they will inevitably tell you what you want”), and the longer follow-up does not change the size of the phase II trial or the fact that phase III data are imminent.
It is worth underscoring that these first-line trials of immunotherapy excluded patients with a known EGFR mutation or ALK rearrangement. These patients have established therapies with a response rate that exceeds that of immunotherapy alone or even with chemotherapy, and they can demonstrate prolonged responses with targeted therapy. Moreover, the limited available evidence suggests that patients with NSCLC that harbors a driver mutation are among those least likely to demonstrate a significant response to immune checkpoint inhibitors.14 All things considered, my approach to a patient with advanced lung adenocarcinoma who does not harbor a driver mutation is to test for PD-L1, and then pursue treatment depending on that result. For patients with high PD-L1 expression, I favor pembrolizumab monotherapy, given the significant potential for a prolonged response that spares a patient chemotherapy-related adverse effects. For those whose tumors have low or no PD-L1 expression, I am most inclined to favor the chemo/immunotherapy combination of carboplatin/pemetrexed/pembrolizumab in cases where I cannot confidently predict that the patient will have a good performance status and be a strong candidate for sequential chemotherapy followed by immunotherapy. This may be due to a large tumor burden, evidence of rapidly progressing disease leading into first-line treatment, significant comorbidities, marginal performance status, or even social issues that could complicate delivery of treatment over time.
Importantly, there is no evidence that concurrent chemo/immunotherapy is an inferior approach, even if it may not prove to be significantly superior. Any disinclination I have toward this strategy is largely grounded in a concern that patients are prone to receive ongoing pemetrexed/pembrolizumab maintenance as an extremely expensive combination, in which it is quite likely that only 1 agent is providing significant clinical benefit and the other, only ongoing expense and risk of cumulative toxicity. We have struggled to have our practice of oncology move toward precision medicine, and this combination blinds us, at least relatively, to which component provides benefit and which creates adverse effects.
Fortunately, we will learn far more in the coming months to years that will help clarify the ambiguities. The KEYNOTE-189 trial asks essentially the same question as KEYNOTE-021g but in a larger, phase III trial setting,15 and many other trials of chemotherapy compared with chemotherapy with a PD-1/PD-L1 checkpoint inhibitor, potentially also compared with a checkpoint inhibitor alone or a CTLA-4 inhibitor, will be reported very soon. The field and my interpretation of the optimal approach have changed rapidly over the past year. There is a range of fair strategies here, and I think the only clear mistake is to become entrenched in any perspective while we are learning so much from new trials and maturing data every few months.
H. Jack West, MD, is medical director of the Thoracic Oncology Program at Swedish Cancer Institute in Seattle, Washington. He is also the founder and president of GRACE: Global Resource for Advancing Cancer Education. At the 35th annual CFS®, West discussed the treatment of patients with stage IV lung adenocarcinoma and no actionable oncogenic drivers.