635 KEYNOTE-495/KeyImPaCT: updated analysis of a biomarker-directed, randomized, phase 2 trial of pembrolizumab-based combination therapy for non-small cell lung cancer
Roy S. Herbst, Wei‐Sen Lam, Matthew D. Hellmann, Matthew A. Gubens, Charų C. Aggarwal, Daniel S.w. Tan, Enriqueta Felip, Joanne Wing Yan Chiu, Jong-Seok Lee, James C. Yang, Edward B. Garon, Giovanna Finocchiaro, Myung‐Ju Ahn, Alexander Luft, Gregory Landers, Andrea Basso, Hua Ma, Julie Kobie, John Palcza, Răzvan Cristescu, Lawrence Fong, Alexandra Snyder, Jianda Yuan, Martin GutierrezRegular2022 · DOI
While immunotherapy has revolutionized the treatment of non-small cell lung cancer (NSCLC), it is not a "one size fits all" solution. Currently, clinicians often rely on single biomarkers like PD-L1 expression, which are frequently insufficient for determining which patients will benefit from complex combination therapies. The KEYNOTE-495/KeyImPaCT trial (NCT03516981) addresses this by using a prospective dual-biomarker approach—T-cell–inflamed gene expression profile (Tcell_infGEP) and tumor mutation burden (TMB)—to assess the clinical activity of first-line pembrolizumab-based combination therapies in advanced NSCLC.
Methods
The researchers conducted an open-label, Phase 2 trial using a sophisticated adaptive randomization design. Patients with previously untreated stage IV NSCLC were categorized into four quadrants based on their dual biomarker status:
Tcell_infGEP low / TMB non-high
Tcell_infGEP low / TMB high
Tcell_infGEP non-low / TMB non-high
Tcell_infGEP non-low / TMB high
Patients were randomly assigned to receive pembrolizumab (200 mg IV Q3W) combined with either the multikinase inhibitor lenvatinib (20 mg oral QD), the CTLA-4 inhibitor quavonlimab (75 mg IV Q6W), or the LAG-3 inhibitor favezelimab (initially 200 mg and then 800 mg IV Q3W). Adaptive randomization based on objective response rate (ORR) was employed, with frequent interim analyses used to quantify efficacy and assess for futility.
Results
At the data cutoff of March 21, 2022, 243 patients had been treated across the various arms (pembrolizumab+lenvatinib, n=80; pembrolizumab+quavonlimab, n=82; pembrolizumab+favezelimab 200 mg, n=30; pembrolizumab+favezelimab 800 mg, n=51).
Across all treatment combinations, patients in the "Tcell_infGEP non-low / TMB high" group—representing the most biologically active tumors—showed the most promising antitumor activity and durable responses. Specifically, the combination of pembrolizumab and lenvatinib met its prespecified efficacy threshold in the "Tcell_infGEP non-low / TMB non-high" subgroup with an ORR of 41% (95% CI: 21–64). In the "Tcell_infGEP non-low / TMB high" group, this same arm reached an ORR of 57%.
Table 1: Efficacy outcomes by biomarker-defined subgroups across treatment arms.
In the pembrolizumab + favezelimab arms, although the strict efficacy thresholds were not met, the "Tcell_infGEP non-low / TMB high" subgroup showed a trend toward improved ORR and numerically longer median progression-free survival (PFS) and overall survival (OS) compared to the other three biomarker-defined subgroups. The safety profile of each treatment arm was consistent with the known safety profile of each combination, with no new safety signals identified.
Discussion and Conclusion
The KEYNOTE-495 trial demonstrates that prospectively using dual biomarkers to guide adaptive randomization in a clinical setting is feasible and provides a clearer picture of drug efficacy than traditional trial designs. The study highlights the "double positive" advantage, where tumors with high inflammatory signatures and high mutation burdens respond best across different immunotherapy pairings.
While the small subgroup sizes in this Phase 2 trial lead to wider confidence intervals, the results pave the way for more personalized immunotherapy. Future Phase 3 confirmatory trials will likely use this dual-biomarker strategy to refine first-line treatment decisions, moving beyond a single-biomarker mindset to identify patients most likely to benefit from specific pembrolizumab-based combinations.
Summary updated
Abstract
<h3>Background</h3> The group sequential, adaptively randomized, open-label, phase 2 KEYNOTE-495/KeyImPaCT trial (NCT03516981) demonstrated the feasibility of using prospective T-cell–inflamed gene expression profile (Tcell<sub>inf</sub>GEP) and tumor mutation burden (TMB) dual biomarker status to assess the clinical activity of first-line pembrolizumab-based combination therapies in advanced non−small-cell lung cancer (NSCLC). Here, we present updated efficacy data relative to prespecified biomarker-defined subgroups for each treatment arm. <h3>Methods</h3> Patients with previously untreated stage IV NSCLC were categorized by Tcell<sub>inf</sub>GEP and TMB dual biomarker status (Tcell<sub>inf</sub>GEP<sup>low</sup>TMB<sup>non-high</sup>, Tcell<sub>inf</sub>GEP<sup>low</sup>TMB<sup>high</sup>, Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>non-high</sup>, Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>high</sup>) then randomly assigned 1:1:1 to receive pembrolizumab (200 mg IV Q3W) plus the multikinase inhibitor (targeting VEGFRs 1-3, FGFRs 1-4, PDGFRα, RET, and KIT) lenvatinib (20 mg oral QD), CTLA-4 inhibitor quavonlimab (75 mg IV Q6W), or LAG-3 inhibitor favezelimab (initially 200 mg and then 800 mg IV Q3W). Adaptive randomization based on objective response rate (ORR) was employed, and frequent interim analyses were performed to quantify efficacy and assess for futility. The primary end point of investigator-assessed ORR per RECIST v1.1 was assessed using prespecified efficacy thresholds for each biomarker-defined subgroup: an ORR of >5% within the Tcell<sub>inf</sub>GEP<sup>low</sup>TMB<sup>non-high</sup> subgroup, >20% within the Tcell<sub>inf</sub>GEP<sup>low</sup>TMB<sup>high</sup> subgroup and Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>non-high</sup> subgroup, and >45% within the Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>high</sup> subgroup evaluated using a Bayesian posterior probability. Secondary end points included progression-free survival (PFS) per RECIST v1.1, overall survival (OS), and safety. <h3>Results</h3> At data cutoff (March 21, 2022), 243 patients had been treated (pembrolizumab+lenvatinib, n=80; pembrolizumab+quavonlimab, n=82; pembrolizumab+favezelimab 200 mg, n=30; pembrolizumab+favezelimab 800 mg, n=51). Efficacy data are presented in<b> table 1</b>. ORR with pembrolizumab+lenvatinib in the Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>non-high</sup> subgroup met the prespecified efficacy threshold. The safety profile of each treatment arm was consistent with the known safety profile of each combination. <h3>Conclusions</h3> Pembrolizumab-based combination therapy continued to show promising antitumor activity and durable response in the Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>high</sup> subgroup across all combinations. Response in the Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>non-high</sup> subgroup treated with pembrolizumab+lenvatinib met the prespecified efficacy threshold. Although response in the pembrolizumab+favezelimab arm did not reach the efficacy bar, there was a trend toward improved ORR in the Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>high</sup> subgroup versus the other 3 biomarker-defined subgroups; median PFS and OS were also numerically longer in the Tcell<sub>inf</sub>GEP<sup>non-low</sup>TMB<sup>high</sup> subgroup compared with the other 3 biomarker-defined subgroups. Prospective assessment of dual biomarkers, as performed in this study, may help identify patients with NSCLC most likely to respond to pembrolizumab-based combination therapies. <h3>Acknowledgements</h3> Medical writing and/or editorial assistance was provided by Mehak Aggarwal, PharmD, and Holly C. Cappelli, PhD, CMPP, of ApotheCom (Yardley, PA, USA). This assistance was funded by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA. <h3>Trial Registration</h3> NCT03516981 <h3>Consent</h3> The study protocol and all amendments were approved by the institutional review board or ethics committee at each institution. All patients provided written informed consent
BibTeX
@inproceedings{herbst2022keynote,
author = {Roy S. Herbst and Wei‐Sen Lam and Matthew D. Hellmann and Matthew A. Gubens and Charų C. Aggarwal and Daniel S.W. Tan and Enriqueta Felip and Joanne Wing Yan Chiu and Jong-Seok Lee and James C. Yang and Edward B. Garon and Giovanna Finocchiaro and Myung‐Ju Ahn and Alexander Luft and Gregory Landers and Andrea Basso and Hua Ma and Julie Kobie and John Palcza and Răzvan Cristescu and Lawrence Fong and Alexandra Snyder and Jianda Yuan and Martin Gutierrez},
title = {{635 KEYNOTE-495/KeyImPaCT: updated analysis of a biomarker-directed, randomized, phase 2 trial of pembrolizumab-based combination therapy for non-small cell lung cancer}},
booktitle = {Regular},
year = {2022},
doi = {10.1136/jitc-2022-sitc2022.0635},
}