Though the United States Food and Drug Administration (FDA) has approved the combination, should the rest of the world have followed suit without an appropriately powered phase III comparison measuring the primary outcome of survival?
Study design
One of the challenges of KEYNOTE-021 was that it was designed prior to the establishment of first-line single agent pembrolizumab in patients with PD-L1 tumor proportion score (TPS) ≥ 50% based on the KEYNOTE-024 randomized phase III trial. This demonstrated improved survival and symptom control, as well as superior response rate, PFS and a more favourable toxicity profile compared to first-line platinum doublet chemotherapy. Thus, KEYNOTE-021G does not reflect our current challenge, namely how should we incorporate immunotherapy for those with PD-L1 TPS ≤ 50%, and for those with TPS ≥ 50%. Indeed, the small sample size has led to conflicting results within subgroups. The addition of pembrolizumab to chemotherapy markedly improved response rates in 21 patients without PD-L1-negative tumours compared to chemotherapy alone in 23 patients (62% versus 17%). Similarly, in those with PD-L1 TPS ≥ 50% tumours, the combination improved response rates compared to chemotherapy alone (80% or 16/20 patients versus 41% or 7/17 patients). However, counter to our expectations, those with intermediate PD-L1 tumour staining (TPS 1-49%) appeared to benefit from chemotherapy rather than the combination (39% or 9/23 patients versus 26% or 5/19 patients receiving the combination). Can we truly rely on these results, based on a few dozen patients, to define a new standard of care for all our patients?
Choosing an endpoint
Is response rate the best endpoint for practice-changing trials in lung cancer? It is an efficient measure to screen for treatment activity, and avoids potential confounding from such variables as post-progression therapy or treatment cross-over. But our patients need to live longer and feel better, and deserve more than temporary relief from “scan-xiety”. One could argue that a better endpoint is the time to starting docetaxel when comparing upfront combination therapy versus the standard sequential approach of chemotherapy followed by a checkpoint inhibitor or vice-versa. Response rate alone may underestimate the impact of immune checkpoint inhibitors on long-term survival, and may not be an appropriate endpoint to screen for activity of this class of agents. As a surrogate endpoint, neither response rate nor progression-free survival in advanced non-small cell lung cancer has not been shown to have strong correlation with survival [1,2]. For example, what if the FDA had approved selumetinib in patients with advanced KRAS mutant lung adenocarcinoma based on its highly positive phase II randomized trial results? This study demonstrated markedly better response rates (37% versus 0), and a doubling of median progression-free and overall survival when selumetinib was added to second-line docetaxel [3]. However, the larger phase III study of this combination did not confirm these results [4].
Translating promise into practice change
Randomized phase II trials allow us to explore treatment activity using time-dependent endpoints such as overall and progression-free survival, and even quality of life (when blinded). We often use these trials to provide a “go-no go” signal for phase III development, as they can decrease the risk of proceeding to a phase III study if the new treatment is not truly effective. But it is important to recall that randomized phase II studies do not necessarily predict the outcome of phase III trials. It has been suggested that the probability of a positive randomized phase II trial result predicting a positive phase III result is approximately 1 in 20, similar to a chance finding using a p value of 0.05 [5]. While these trials increase the level of comfort investing in phase III development, they are not designed to define practice change.
To establish the superiority of immunotherapy plus chemotherapy over chemotherapy alone, an estimated 570 patients are needed in the ongoing phase III KEYNOTE-189 study [6,7]. An initial press release reports that the combination improved both PFS and overall survival significantly [8]. While final data have not been released, it is our hope that these results will significantly improve patient outcomes, and change the standard of care so that more may benefit from immune checkpoint inhibitors.
Unanswered questions remain
As neither study restricted enrolment by PD-L1 tumour staining, we hope that we can interpret the impact of adding immunotherapy to chemotherapy in those with PD-L1 TPS ≤ 50%. If benefit is statistically significant but clinically marginal, we will need to know the impact on patient quality of life and symptoms, in order to balance the increased toxicity of the combination. For patients with PD-L1 TPS ≥ 50%, the appropriate first-line comparator is no longer chemotherapy but pembrolizumab. Survival, quality of life and toxicity will be key endpoints when comparing pembrolizumab plus or minus platinum doublet chemotherapy in this group. And can we translate these findings to our patients with squamous carcinoma, who were not eligible for KEYNOTE-189? Other phase III trials, such as KEYNOTE-489, have been designed to answer that question and will hopefully be reported soon. Finally, will there still be a role for biomarkers such as PD-L1 and tumour mutation burden to help us select our patients mostly likely to benefit from immunotherapy?
Immunotherapy has had a profound impact on the treatment and outcome of patients with advanced lung cancer. We look forward to expanding the population that benefit from this exciting class of agents – high level benefit based on high level data, which we can use to change practice and improve outcomes worldwide.