Biomarkers predict response to pembrolizumab in head and neck squamous cell carcinoma
CHICAGO — PD-L1 expression and T-cell activated gene expression profile independently predicted response, PFS and OS among a cohort of patients with head and neck squamous cell carcinoma treated with single-agent pembrolizumab, according to study results presented at American Association for Cancer Research Annual Meeting.
Tumor mutational burden also predicted best overall response and PFS but not OS.
The predictive ability of these measures appeared generally similar regardless of HPV status.
“When used alone or jointly, these biomarkers may have utility for characterizing responses to anti-PD-1 therapy in HNSCC,” Tanguy Seiwert, MD, assistant professor in the department of medicine and associate director of the head and neck cancer program at University of Chicago, said during his presentation. “However, they are not perfect. outlier responses can still occur, suggesting additional biology or dynamic changes may contribute and will be important to study.”
Two established biomarkers predict response to anti-PD-1 therapies. They are PD-L1 expression by immunohistochemistry — used to predict whether certain patients with lung cancer or gastric cancer will respond to first-line pembrolizumab (Keytruda, Merck) — as well as microsatellite instability-high status, which is used across multiple tumor types.
Other emerging biomarkers include tumor mutational burden and T-cell-activated gene expression profiles (GEP). Both have demonstrated potential to independently predict response to the anti-PD-1 monoclonal antibody pembrolizumab in multiple tumor types.
Seiwert and colleagues assessed the relationship between inflammatory biomarkers — such as PD-L1 expression and GEP — and tumor mutational burden with response to pembrolizumab among patients with HNSCC.
Researchers analyzed data from 415 patients enrolled in the KEYNOTE-012 trial (n = 261) or KN-55 trial (n = 154).
Of these, 258 patients (median age, 61 years; range, 25-90; 81% male) had tumors available for whole exome sequencing. The majority had ECOG performance status of 1 or 2 (69%), had metastatic staging (90%), and had undergone at least two prior treatments (71%).
Investigators assessed HPV status by p16 immunohistochemistry and whole exome sequencing. The two methods yielded “nearly identical” results, but Seiwert reported data based on whole exome sequencing, which identified 69% of patients as having HPV-negative disease.
Researchers used logistic regression to evaluate the relationship between the biomarkers and best overall response, and they used blinded Cox regression to assess PFS and OS. They adjusted for ECOG performance status and study cohort.
PD-L1 combined positive score (P < .0001) and GEP (P = .0012) appeared significantly associated with best overall response in the entire cohort, as well as for patients with HPV-positive disease. The associations among patients with HPV-negative disease appeared less robust, Seiwert said.
Tumor mutational burden appeared significantly associated with best overall response among all patients (P = .0006), with a stronger association observed among patients with HPV-negative disease (P = .0034) than HPV-positive disease (P = .0046).
“This is in contrast to the inflammation biomarkers, which seem to work slightly more in HPV-positive patients,” Seiwert said. “HPV-positive tumors have viral antigens. Maybe these are not measured by tumor mutational burden, and that could be one factor playing into the slightly differential association.”
Increasing neoantigen load (P = .0081)and clonally-weighted tumor mutational burden (P = .0032) appeared significantly related to response among all patients and those with HPV-negative disease, but not those with HPV-positive disease.
Receiver operating characteristic curves showed all three biomarkers “performed quite well” among all patients, with areas under the curve of 0.63 for tumor mutational burden by whole exome sequencing, 0.64 for PD-L1 combined positive score by immunohistochemistry, and 0.71 for T-cell-inflamed GEP.
When researchers analyzed how these biomarkers correlated with each other, they determined tumor mutational burden did not correlate with PD-L1 and GEP, and that PD-L1 combined positive score and GEP appeared moderately correlated (r = 0.473).
When they assessed tumor mutational burden in a multivariate model with either PD-L1 or GEP, tumor mutational burden and each inflammation marker remained significantly predictive for best overall response (P < .001 for all).
In addition, patients with high GEP and high tumor mutational burden appeared more likely to achieve response than those with low levels of both markers. Researchers observed this association in the overall population, as well as in both HPV subgroups.
Patients with higher levels of PD-L1 combined positive score (HR = 0.76; 955 CI, 0.54-1.09), GEP (HR = 0.57; 95% CI, 0.42-0.76) and tumor mutational burden (HR = 0.64; 95% CI, 0.46-0.89) achieved longer PFS than those with lower levels of all three markers. Results appeared similar in both HPV subgroups.
Researchers observed improved OS among patients with higher levels of PD-L1 combined positive score (HR = 0.67; 95% CI, 0.44-1.01) and GEP (HR = 0.5; 95% CI, 0.35-0.7) but not those with higher tumor mutational burden (HR = 0.98; 95% CI, 0.67-1.43). Results appeared comparable between HPV subgroups.
It remains unclear why tumor mutational burden is associated with improved PFS but not longer OS, Seiwert said. – by Mark Leiser
Seiwert TY, et al. Abstract LB-339. Presented at: American Association for Cancer Research Annual Meeting; April 14-18, 2018; Chicago.
Disclosures: Merck Sharp & Dohme supported the study. Seiwert reports grant or research support from Bristol-Myers Squibb, Jounce Therapeutics, Merck, and Merck Sharp & Dohme, as well as honoraria from AstraZeneca, Bristol-Myers Squibb, Innate Pharma, Merck, and Merck Sharp & Dohme. Please see the abstract for all other authors’ relevant financial disclosures.