KEYNOTE Trial Data Suggest Features Predicting Response to Pembrolizumab

Get Permission

Both mutational load and tumor T-cell inflammation are common features of anti–PD-1 responsive tumors.… These markers appear to have a high positive and negative predictive value and may have clinical utility.
— Tanguy Seiwert, MD

Tweet this quote

A high mutational load and the presence of a T-cell–“inflamed” environment may independently predict for treatment response to pembrolizumab (Keytruda) and progression-free survival, according to a study presented by Tanguy Seiwert, MD, of the University of Chicago, at the 2017 ASCO-SITC Clinical Immuno-Oncology Symposium.1

“Nonsynonymous mutational load and neoantigen load as well as an 18-gene immune-related gene-expression profiling were significantly associated with overall response and progression-free survival to pembrolizumab across multiple indications,” Dr. Seiwert revealed. “This suggests that tumor antigenicity and T-cell infiltration may provide complementary information for expected pembrolizumab activity” and may be useful in characterizing responses to immunotherapies, he said.

Tumor mutational load has been shown to correlate with benefit from drugs blocking cytotoxic T-lymphocyte–associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1) in multiple tumor types. Tumor T-cell inflammation, which correlates with T-cell infiltration, gamma interferon expression, and indirectly programmed cell death ligand 1 (PD-L1) and PD-L2 expression, has also been advanced as a potential biomarker, said Dr. Seiwert.

The current findings were based on data from two multicenter, multicohort studies: KEYNOTE-012 and KEYNOTE-028. The analyzed population included 153 patients with PD-L1–positive tumors across 21 different tumor types who had whole-exome sequencing data available. Association with progression-free survival was assessed in all 153 patients, and overall response was evaluated in 135 of those with measurable disease at baseline. The researchers used expression values for 18 immune-related genes to develop a gene-expression profile, which indicated a T-cell–inflamed tumor microenvironment.

Association With Outcomes

Objective response to treatment was significantly associated with mutational load (P = .0001), neoepitope load (P = .0001), and gene-expression profile (P = .0004). Under a mutational load cutoff of 102, response rates exceeded 30% in patients ≥ 102 but were only 7% in patients below this level. Patients with more inflamed tumors had a response rate of 25%, compared with only 2% for those with noninflamed tumors. In patients with both high mutational load and tumor inflammation, the response rate exceeded 38%, vs 0% for those with low mutational load and noninflamed tumors.

The areas under the receiver operating curve (AUROC) for mutational load, neoepitope load, and gene-expression profile were high (> 0.7) and similar, he noted. Mutational load (P = .0004) and gene-expression profile (P = .0010) remained significant predictors of response in a multivariate model adjusted for each measure. “These measures separated responders from nonresponders quite well,” he added. “There’s no single measure that seems to be superior.”

The same pattern was seen for an association with progression-free survival, both for mutational load (P = .0168) and, even more robustly, gene-expression profile (P = .0003), he reported. Median progression-free survival was 109 days for patients with a mutational load ≥ 102 vs 59 days for < 102 (hazard ratio [HR] = 0.60). For gene-expression profile ≥ 0.318 (a cutoff indicative of inflammation), median progression-free survival was 96 days, vs 57 for < 0.318 (HR = 0.50). For patients with both high mutational load and inflammation, median progression-free survival was 175 days, vs 59 days for those with fewer mutations and a noninflamed tumor (HR = 0.46).

“Both mutational load and tumor T-cell inflammation are common features of anti–PD-1 responsive tumors,” Dr. ­Seiwert concluded. “Whether used separately or in combination, these markers appear to have a high positive and negative predictive value and may have clinical utility.” ■

Disclosure: Dr. Seiwert has received honoraria from Amgen, AstraZeneca, Bayer/Onyx, Bristol-Myers Squibb, Merck, and Merck Serono and (institutional) research funding from Boehringer Ingelheim and Genentech/Roche.


1. Cristescu R, et al: 2017 ASCO-SITC Clinical Immuno-Oncology Symposium. Abstract 1. Presented February 24, 2017.

Related Articles

Expert Point of View: Shridar Ganesan, MD, PhD

Among tumors with a high mutational burden, there are differences in the nature or flavor of the mutations involved, reflecting differences in underlying mutagenic processes.
— Shridar Ganesan, MD, PhD

Tweet this quote

Shridar Ganesan, MD, PhD, of Rutgers Cancer...