T-Cell-Inflamed Gene-Expression Profile, Programmed Death Ligand 1 Expression, and Tumor Mutational Burden Predict Efficacy in Patients Treated With Pembrolizumab Across 20 Cancers: KEYNOTE-028.

PURPOSE: Biomarkers that can predict response to anti-programmed cell death 1 (PD-1) therapy across multiple tumor types include a T-cell-inflamed gene-expression profile (GEP), programmed death ligand 1 (PD-L1) expression, and tumor mutational burden (TMB). Associations between these biomarkers and the clinical efficacy of pembrolizumab were evaluated in a clinical trial that encompassed 20 cohorts of patients with advanced solid tumors. METHODS: KEYNOTE-028 ( ClinicalTrials.gov identifier: NCT02054806) is a nonrandomized, phase Ib trial that enrolled 475 patients with PD-L1-positive advanced solid tumors who were treated with pembrolizumab 10 mg/kg every 2 weeks for 2 years or until confirmed disease progression or unacceptable toxicity occurred. The primary end point was objective response rate (ORR; by RECIST v1.1, investigator review). Secondary end points included safety, progression-free survival (PFS), and overall survival (OS). Relationships between T-cell-inflamed GEP, PD-L1 expression, and TMB and antitumor activity were exploratory end points. RESULTS: ORRs (with 95% CIs) ranged from 0% (0.0% to 14.2%) in pancreatic cancer to 33% (15.6% to 55.3%) in small-cell lung cancer. Across cohorts, median (95% CI) PFS ranged from 1.7 months (1.5 to 2.9 months) to 6.8 months (1.9 to 14.1 months) in pancreatic and thyroid cancers, respectively, and median OS from 3.9 months (2.8 to 5.5 months) to 21.1 months (9.1 to 22.4 months) in vulvar and carcinoid tumors, respectively. Higher response rates and longer PFS were demonstrated in tumors with higher T-cell-inflamed GEP, PD-L1 expression, and/or TMB. Correlations of TMB with GEP and PD-L1 were low. Response patterns indicate that patients with tumors that had high levels of both TMB and inflammatory markers (GEP or PD-L1) represent a population with the highest likelihood of response. Safety was similar and consistent with prior pembrolizumab reports. CONCLUSION: A T-cell--inflamed GEP, PD-L1 expression, and TMB predicted response to pembrolizumab in multiple tumor types. These biomarkers (alone/in combination) may help identify patients who have a higher likelihood of response to anti-PD-1 therapies across a broad spectrum of cancers.

Pan-tumor genomic biomarkers for PD-1 checkpoint blockade-based immunotherapy.

Programmed cell death protein-1 (PD-1) and programmed cell death ligand-1 (PD-L1) checkpoint blockade immunotherapy elicits durable antitumor effects in multiple cancers, yet not all patients respond. We report the evaluation of >300 patient samples across 22 tumor types from four KEYNOTE clinical trials. Tumor mutational burden (TMB) and a T cell-inflamed gene expression profile (GEP) exhibited joint predictive utility in identifying responders and nonresponders to the PD-1 antibody pembrolizumab. TMB and GEP were independently predictive of response and demonstrated low correlation, suggesting that they capture distinct features of neoantigenicity and T cell activation. Analysis of The Cancer Genome Atlas database showed TMB and GEP to have a low correlation, and analysis by joint stratification revealed biomarker-defined patterns of targetable-resistance biology. These biomarkers may have utility in clinical trial design by guiding rational selection of anti-PD-1 monotherapy and combination immunotherapy regimens.

Programmed Death-Ligand 1 Expression and Response to the Anti-Programmed Death 1 Antibody Pembrolizumab in Melanoma.

Purpose Expression of programmed death-ligand 1 (PD-L1) is a potential predictive marker for response and outcome after treatment with anti-programmed death 1 (PD-1). This study explored the relationship between anti-PD-1 activity and PD-L1 expression in patients with advanced melanoma who were treated with pembrolizumab in the phase Ib KEYNOTE-001 study (clinical trial information: NCT01295827). Patients and Methods Six hundred fifty-five patients received pembrolizumab10 mg/kg once every 2 weeks or once every 3 weeks, or 2 mg/kg once every 3 weeks. Tumor response was assessed every 12 weeks per Response Evaluation Criteria in Solid Tumors (RECIST) v1.1 by independent central review. Primary outcome was objective response rate. Secondary outcomes included progression-free survival (PFS) and overall survival (OS). Membranous PD-L1 expression in tumor and tumor-associated immune cells was assessed by a clinical trial immunohistochemistry assay (22C3 antibody) and scored on a unique melanoma (MEL) scale of 0 to 5 by one of three pathologists who were blinded to clinical outcome; a score ≥ 2 (membranous staining in ≥ 1% of cells) was considered positive. Results Of 451 patients with evaluable PD-L1 expression, 344 (76%) had PD-L1-positive tumors. Demographic and staging variables were equally distributed among PD-L1-positive and -negative patients. An association between higher MEL score and higher response rate and longer PFS (hazard ratio, 0.76; 95% CI, 0.71 to 0.82) and OS (hazard ratio, 0.76; 95% CI, 0.69 to 0.83) was observed ( P < .001 for each). Objective response rate was 8%, 12%, 22%, 43%, 57%, and 53% for MEL 0, 1, 2, 3, 4, and 5, respectively. Conclusion PD-L1 expression in pretreatment tumor biopsy samples was correlated with response rate, PFS, and OS; however, patients with PD-L1-negative tumors may also achieve durable responses.

Pembrolizumab for the treatment of non-small-cell lung cancer.

BACKGROUND: We assessed the efficacy and safety of programmed cell death 1 (PD-1) inhibition with pembrolizumab in patients with advanced non-small-cell lung cancer enrolled in a phase 1 study. We also sought to define and validate an expression level of the PD-1 ligand 1 (PD-L1) that is associated with the likelihood of clinical benefit. METHODS: We assigned 495 patients receiving pembrolizumab (at a dose of either 2 mg or 10 mg per kilogram of body weight every 3 weeks or 10 mg per kilogram every 2 weeks) to either a training group (182 patients) or a validation group (313 patients). We assessed PD-L1 expression in tumor samples using immunohistochemical analysis, with results reported as the percentage of neoplastic cells with staining for membranous PD-L1 (proportion score). Response was assessed every 9 weeks by central review. RESULTS: Common side effects that were attributed to pembrolizumab were fatigue, pruritus, and decreased appetite, with no clear difference according to dose or schedule. Among all the patients, the objective response rate was 19.4%, and the median duration of response was 12.5 months. The median duration of progression-free survival was 3.7 months, and the median duration of overall survival was 12.0 months. PD-L1 expression in at least 50% of tumor cells was selected as the cutoff from the training group. Among patients with a proportion score of at least 50% in the validation group, the response rate was 45.2%. Among all the patients with a proportion score of at least 50%, median progression-free survival was 6.3 months; median overall survival was not reached. CONCLUSIONS: Pembrolizumab had an acceptable side-effect profile and showed antitumor activity in patients with advanced non-small-cell lung cancer. PD-L1 expression in at least 50% of tumor cells correlated with improved efficacy of pembrolizumab. (Funded by Merck; KEYNOTE-001 ClinicalTrials.gov number, NCT01295827.).

Clinical utility estimation for assay cutoffs in early phase oncology enrichment trials.

Predictive enrichment strategies use biomarkers to selectively enroll oncology patients into clinical trials to more efficiently demonstrate therapeutic benefit. Because the enriched population differs from the patient population eligible for screening with the biomarker assay, there is potential for bias when estimating clinical utility for the screening eligible population if the selection process is ignored. We write estimators of clinical utility as integrals averaging regression model predictions over the conditional distribution of the biomarker scores defined by the assay cutoff and discuss the conditions under which consistent estimation can be achieved while accounting for some nuances that may arise as the biomarker assay progresses toward a companion diagnostic. We outline and implement a Bayesian approach in estimating these clinical utility measures and use simulations to illustrate performance and the potential biases when estimation naively ignores enrichment. Results suggest that the proposed integral representation of clinical utility in combination with Bayesian methods provide a practical strategy to facilitate cutoff decision-making in this setting.

Evaluating surrogate variables for improving microarray multiple testing inference.

The use of surrogate variables has been proposed as a means to capture, for a given observed set of data, sources driving the dependency structure among high-dimensional sets of features and remove the effects of those sources and their potential negative impact on simultaneous inference. In this article we illustrate the potential effects of latent variables on testing dependence and the resulting impact on multiple inference, we briefly review the method of surrogate variable analysis proposed by Leek and Storey (PNAS 2008; 105:18718-18723), and assess that method via simulations intended to mimic the complexity of feature dependence observed in real-world microarray data. The method is also assessed via application to a recent Merck microarray data set. Both simulation and case study results indicate that surrogate variable analysis can offer a viable strategy for tackling the multiple testing dependence problem when the features follow a potentially complex correlation structure, yielding improvements in the variability of false positive rates and increases in power.

Statistical performance of cladistic strategies for haplotype grouping in pharmacogenetics.

Haplotypes comprising multiple single nucleotide polymorphisms (SNPs) are popular covariates for capturing the key genetic variation present over a region of interest in the DNA sequence. Although haplotypes can provide a clearer assessment of genetic variation in a region than their component SNPs considered individually, the multi-allelic nature of haplotypes increases the complexity of the statistical models intended to discover association with outcomes of interest. Cladistic methods cluster haplotypes according to the estimates of their genealogical closeness and have been proposed recently as strategies for reducing model complexity and increasing power. Two examples are methods based on a haplotype nesting algorithm described by Templeton et al. (Genetics 1987; 117:343-351) and hierarchical clustering of haplotypes as described by Durrant et al. (Am. J. Hum. Genet. 2004; 75:35-43). In the context of assessing the pharmacogenetic effects of candidate genes, for which high-density SNP data have been gathered, we have conducted a simulation-based case study of the testing and estimation properties of two strategies based on Templeton's algorithm (TA), one being that described by Seltman et al. (Am. J. Hum. Genet. 2001; 68:1250-1263; Genet. Epidemiol. 2003; 25:48-58), as well as the method of Durrant et al. using data from a diabetes clinical trial. Even after adjusting for multiplicity, improvements in power can be realized using cladistic approaches with treatment group sizes in the range expected for standard trials, although these gains may be sensitive to the cladistic structure used. Differences in the relative performance of the cladistic approaches examined were observed with the clustering approach of Durrant et al. showing statistical properties superior to the methods based on TA.

Effect of initial combination therapy with sitagliptin, a dipeptidyl peptidase-4 inhibitor, and metformin on glycemic control in patients with type 2 diabetes.

OBJECTIVE: To assess the efficacy and safety of initial combination therapy with sitagliptin and metformin in patients with type 2 diabetes and inadequate glycemic control on diet and exercise. RESEARCH DESIGN AND METHODS: In a 24-week, randomized, double-blind, placebo-controlled, parallel-group study, 1,091 patients with type 2 diabetes and A1C 7.5-11% were randomized to one of six daily treatments: sitagliptin 100 mg/metformin 1,000 mg (S100/M1000 group), sitagliptin 100 mg/metformin 2,000 mg (S100/M2000 group), metformin 1,000 mg (M1000 group), metformin 2,000 mg (M2000 group) (all as divided doses administered twice daily [b.i.d.]), sitagliptin 100 mg q.d. (S100 group), or placebo. Patients who had an A1C >11% or a fasting glucose value >280 mg/dl after the run-in period were not eligible to be randomized; these patients could participate in an open-label substudy and were treated with S100/M2000 for 24 weeks. RESULTS: The mean baseline A1C was 8.8% in the randomized patients. The placebo-subtracted A1C change from baseline was -2.07% (S100/M2000), -1.57% (S100/M1000), -1.30% (M2000), -0.99% (M1000), and -0.83% (S100) (P < 0.001 for comparisons versus placebo and for coadministration versus respective monotherapies). The proportion of patients achieving an A1C <7% and <6.5% was 66 and 44%, respectively, in the S100/M2000 group (P < 0.001 vs. S100 or M2000). For the open-label cohort (n = 117; baseline A1C 11.2%) treated with S100/M2000, the within-group mean A1C change from baseline was -2.9%. The incidence of hypoglycemia was low (0.5-2.2%) across active treatment groups and not significantly different from that in the placebo group (0.6%). The incidence of gastrointestinal adverse experiences was similar for coadministration therapies compared with their respective metformin monotherapy. CONCLUSIONS: The initial combination of sitagliptin and metformin provided substantial and additive glycemic improvement and was generally well tolerated in patients with type 2 diabetes.

Effect of the dipeptidyl peptidase-4 inhibitor sitagliptin as monotherapy on glycemic control in patients with type 2 diabetes.

OBJECTIVE: To examine the efficacy and safety of once-daily oral sitagliptin as monotherapy in patients with type 2 diabetes. RESEARCH DESIGN AND METHODS: In a randomized, double-blind, placebo-controlled study, 741 patients (baseline HbA(1c) [A1C] 8.0%) were randomized to sitagliptin 100 or 200 mg or placebo for 24 weeks. RESULTS: Sitagliptin 100 and 200 mg produced significant (P < 0.001) placebo-subtracted reductions in A1C (-0.79 and -0.94%, respectively) and fasting plasma glucose (-1.0 mmol/l [-17.1 mg/dl] and -1.2 mmol/l [-21.3 mg/dl], respectively). Patients with baseline A1C >or=9% had greater reductions in placebo-subtracted A1C with sitagliptin 100 and 200 mg (-1.52 and -1.50%, respectively) than those with baseline A1C <8% (-0.57 and -0.65%) or >or=8 to <9.0% (-0.80 and -1.13%, respectively). In a meal tolerance test, sitagliptin 100 and 200 mg significantly decreased 2-h postprandial glucose (PPG) (placebo-subtracted PPG -2.6 mmol/l [-46.7 mg/dl] and -3.0 mmol/l [-54.1 mg/dl], respectively). Results for the above key efficacy parameters were not significantly different between sitagliptin doses. Homeostasis model assessment of beta-cell function and proinsulin-to-insulin ratio improved with sitagliptin. The incidence of hypoglycemia was similar, and overall gastrointestinal adverse experiences were slightly higher with sitagliptin. No meaningful body weight changes from baseline were observed with sitagliptin 100 (-0.2 kg) or 200 mg (-0.1 kg). The body weight change with placebo (-1.1 kg) was significantly (P < 0.01) different from that observed with sitagliptin. CONCLUSIONS: In this 24-week study, once-daily sitagliptin monotherapy improved glycemic control in the fasting and postprandial states, improved measures of beta-cell function, and was well tolerated in patients with type 2 diabetes.

Stratification and weighting via the propensity score in estimation of causal treatment effects: a comparative study.

Estimation of treatment effects with causal interpretation from observational data is complicated because exposure to treatment may be confounded with subject characteristics. The propensity score, the probability of treatment exposure conditional on covariates, is the basis for two approaches to adjusting for confounding: methods based on stratification of observations by quantiles of estimated propensity scores and methods based on weighting observations by the inverse of estimated propensity scores. We review popular versions of these approaches and related methods offering improved precision, describe theoretical properties and highlight their implications for practice, and present extensive comparisons of performance that provide guidance for practical use.

Estimation of survival distributions of treatment policies in two-stage randomization designs in clinical trials.

Some clinical trials follow a design where patients are randomized to a primary therapy at entry followed by another randomization to maintenance therapy contingent upon disease remission. Ideally, analysis would allow different treatment policies, i.e., combinations of primary and maintenance therapy if specified up-front, to be compared. Standard practice is to conduct separate analyses for the primary and follow-up treatments, which does not address this issue directly. We propose consistent estimators for the survival distribution and mean restricted survival time for each treatment policy in such two-stage studies and derive large-sample properties. The methods are demonstrated on a leukemia clinical trial data set and through simulation.