Tumor mutational burden stratified into high and low using 10 mutations per megabase as cutoff for all cancer types. Cancer types with fewer than 5 cases in either the low TMB (TMB-L) or high TMB (TMB-H) group are not shown (see the Table for information on response rates for these cancer types). NSCLC indicates non–small cell lung cancer; SCLC, small cell lung cancer.
A, TMB categorized using universal numerical cutoffs, with increments containing a minimum of 5% of patients in the cohort. B, TMB categorized using cancer type-specific cutoffs, based on the percentile rank of a tumor’s TMB within the corresponding cancer type.
OR indicates odds ratio.
eFigure 1. Flow Diagram
eFigure 2. Progression-Free Survival (PFS) Based on Tumor Mutational Burden (TMB) by Cancer Type
eFigure 3. Overall Survival (OS) Based on Tumor Mutational Burden (TMB) by Cancer Type
eFigure 4. Response Rates Based on Tumor Mutational Burden (TMB) by Cancer Type Using Tissue-Specific Cutoffs
eTable 1. Characteristics of Patients in the Study
eTable 2. Cancer Type–Specific Cutoffs for High Tumor Mutational Burden (TMB)
eTable 3. Odds Ratios (ORs) of Response Based on Tumor Mutational Burden (TMB) by Cancer Type
eTable 4. Outcomes for Cancer Types Previously Lacking a United States Food and Drug Administration (FDA) Approval for Treatment With Immune Checkpoint Inhibitors (ICIs)
Customize your JAMA Network experience by selecting one or more topics from the list below.
Valero C, Lee M, Hoen D, et al. Response Rates to Anti–PD-1 Immunotherapy in Microsatellite-Stable Solid Tumors With 10 or More Mutations per Megabase. JAMA Oncol. 2021;7(5):739–743. doi:10.1001/jamaoncol.2020.7684
Is a universal cutoff value to define high tumor mutational burden (TMB) associated with tumor response to cancer immunotherapy?
In this cohort study of 1678 patients with tumors representing 16 cancer types who were treated with anti–programmed cell death 1 or programmed cell death ligand-1 immunotherapy, response rates were generally higher with high TMB (≥10 mutations per megabase). However, the proportion of tumors with high TMB and the association, if any, between high TMB and response rates varied widely across cancer types.
These data validate the finding of higher response rates in tumors with high TMB (≥10 mutations per megabase), but the predictive value of a single cutoff value for TMB is limited and would likely be improved with cancer type–specific cutoffs.
In June 2020, the US Food and Drug Administration approved the anti–programmed cell death 1 drug pembrolizumab for patients with malignant solid tumors of any histologic type with high tumor mutational burden (TMB; ≥10 mutations per megabase). The predictive value of this universal cutoff for high TMB is not well understood.
To examine the performance of a universal definition of high TMB in an independent cohort of patients with solid tumors treated with immune checkpoint inhibitors.
Design, Setting, and Participants
This retrospective cohort study included 1678 patients at a single cancer referral center treated with immune checkpoint inhibitors from January 1, 2015, to December 31, 2018. Patients had 16 different cancer types and were treated with anti–programmed cell death 1 or programmed cell death ligand-1 immunotherapy. Tumors underwent next-generation sequencing.
At least 1 dose of immune checkpoint inhibitors.
Main Outcomes and Measures
Best overall response to immune checkpoint inhibitor therapy. The hypothesis tested was formulated after data collection and prior to analysis.
Of 1678 patients, 924 (55%) were male, and the median age was 64 years (interquartile range, 55-71 years). Using the universal cutoff of 10 mutations per megabase, 416 tumors (25%) were categorized as having high TMB. Across cancer types, the proportion of TMB-high tumors ranged from 0% of kidney cancers to 53% of melanomas (113 of 214). Tumors categorized as TMB-high had higher response rates compared with TMB-low tumors in only 11 of 16 cancer types. In the entire cohort, response rates increased with higher cutoffs for TMB-high categorization, reaching 41% (169 of 416) for TMB more than 10 and 56% (90 of 161) for TMB more than 18, the highest TMB decile. Response rates also increased with TMB percentile within cancer type. Using cancer-specific cutoffs, 457 tumors (27%) were categorized as TMB-high. Response rates within cancer type ranged from 4% for pancreatic cancer (1 of 26) to 70% for melanoma (46 of 66). Cancer-specific cutoffs were associated with numerically higher response rates for TMB-high compared with TMB-low tumors in 14 of 16 cancer types.
Conclusions and Relevance
The data from this cohort study validate the finding of generally higher response rates following immune checkpoint inhibitor therapy for tumors with TMB of 10 or more mutations per megabase, across multiple cancer types. However, the predictive value of a universal numerical threshold for TMB-high was limited, owing to variability across cancer types and unclear associations with survival outcomes. Further investigation will help define cancer type–specific TMB cutoffs to guide decision-making.
Immune checkpoint inhibitors (ICIs) targeting programmed cell death 1 (PD-1) or its ligand (programmed cell death ligand-1 [PD-L1]) achieve tumor responses in a subset of patients with many cancer types. However, most patients do not experience tumor response. One biomarker that has been associated with higher response rates and longer survival after ICI therapy for many cancer types is high tumor mutational burden (TMB), a measure of the number of somatic mutations in a tumor.1,2
In 2017, the US Food and Drug Administration (FDA) issued its first cancer drug approval agnostic to tissue or site: the use of anti–PD-1 therapy in tumors of any histologic type with microsatellite instability–high status. In 2020, the FDA approved anti–PD-1 therapy for solid tumors of any histologic type with TMB of 10 or more mutations per megabase.3 This approval was based on a finding of high response rates among 81 patients with TMB-high microsatellite-stable (MSS) tumors (determined using a commercial next-generation sequencing assay) representing 9 cancer types in the Keynote-158 trial.4,5
To our knowledge, response rates to ICI therapy using a universal TMB cutoff of 10 have not been analyzed across MSS cancers of all types in any other cohort or with other next-generation sequencing platforms. The aim of this study was to examine the performance of this universal definition of TMB-high in an independent cohort of patients with MSS solid tumors treated with ICI therapy.
We identified 2834 patients with solid tumors diagnosed from January 1, 2015, to December 31, 2018, who received anti–PD-1 or PD-L1 monotherapy or combination therapy at our center. Tumors and matched normal DNA from blood samples were profiled using MSK-IMPACT (Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets), an FDA-approved targeted next-generation sequencing platform.6 We excluded patients with a history of more than 1 cancer, enrolled in blinded trials, treated in neoadjuvant or adjuvant settings, and with unevaluable response, as well as cancer types with fewer than 25 cases and microsatellite-unstable tumors as determined with MSIsensor.7 The final cohort consisted of 1678 patients with 16 cancer types or cancers of unknown primary (eFigure 1 and eTable 1 in the Supplement). This study was approved by the Memorial Sloan Kettering Cancer Center Institutional Review Board. All patients provided written informed consent. This study was reported according to Reporting Recommendations for Tumor Marker Studies (REMARK) guidelines.
Tumor mutational burden is the total number of somatic, nonsynonymous mutations normalized to the exonic coverage of the MSK-IMPACT panel (mutations per megabase).1 The clinical variables analyzed were age and tumor stage (American Joint Committee on Cancer, 8th edition)8 at first ICI infusion, sex, cancer type, drug class, year of treatment start, and history of prior chemotherapy.
Best overall response to ICI was based on Response Evaluation Criteria in Solid Tumors, (RECIST), version 1.1 criteria, or manual review to categorize response based on change in target lesion size, if RECIST reads were unavailable.9 Progression-free survival was calculated from the time of the first ICI infusion to disease progression or death. Overall survival was calculated from the time of the first ICI infusion to death. Patients were censored at last contact (overall survival) or last appointment with a clinician (progression-free survival).
The previously reported TMB cutoff of 10 was first subjected to external validation by examining the association between this cutoff and response and survival rates in this independent cohort.4,5 Optimal cutoffs per cancer type were determined with the Youden statistic. For cancers with fewer than 50 patients, a minimum 20% of patients was required for the TMB-high group (eTable 2 in the Supplement).
For the predictive model, odds ratios (ORs) were calculated with logistic regression, and hazard ratios were calculated with Cox proportional hazards regression. A 2-sided P < .05 was considered statistically significant. Data were collected between June and September 2019 and analyzed between September 2019 and July 2020, and analyses were conducted using Stata, version 16 (StataCorp).
Of 1678 patients, 924 (55%) were male, and the median age was 64 years (interquartile range, 55-71 years). Using the universal cutoff of 10 mutations per megabase, 416 of 1678 MSS tumors (25%) were categorized as TMB-high. The proportion of TMB-high tumors ranged from 0% of kidney cancers to 53% of melanomas (113 of 214) (Table). Response rates for each cancer type are shown in Figure 1 and eTable 3 in the Supplement. In some cancer types, response rates for TMB-high tumors were low (0% in hepatobiliary cancer, pancreatic cancer, and mesothelioma and 14% in colorectal cancer [1 of 7]), although the numbers in these cohorts were small. Tumors categorized as TMB-high had numerically higher response rates compared with TMB-low tumors in 11 of 16 cancer types. Rates of progression-free survival and overall survival by TMB are shown in eFigure 2 and eFigure 3 in the Supplement. In the entire cohort, response rates increased with higher TMB cutoffs (OR, 1.04 per mutation per megabase; 95% CI, 1.02-1.05), reaching 41% (169 of 416) for TMB more than 10 and 56% (90 of 161) for TMB more than 18, the highest TMB decile. Response rates also increased with TMB percentile rank within cancer type (OR, 1.01 per percentile increase; 95% CI, 1.01-1.02), indicating that this association is not solely due to higher response rates in tumor types with higher TMB, such as melanoma and lung cancer (Figure 2).
The distribution of TMB and the definition of high TMB differ across cancer types.1 We examined response rates using tissue-specific cutoffs for high TMB. The cutoffs were determined for this exploratory analysis and should not be considered definitive optimized cutoffs until they are validated independently. Using cancer type–specific cutoffs, a similar percentage (27% ) of tumors were categorized as TMB-high. Response rates within cancer type ranged from 4% for pancreatic cancer (1 of 26) to 70% for melanoma (46 of 66) (eFigure 4 in the Supplement). Tissue-specific TMB cutoffs were associated with numerically higher response rates for TMB-high tumors in 14 of 16 cancer types. Progression-free survival and overall survival data are shown in eFigure 2 and eFigure 3 in the Supplement.
We analyzed cancer types that had not been approved by the FDA for ICI therapy until the tissue-agnostic approval for TMB-high tumors: sarcoma, colorectal cancer, pancreatic cancer, mesothelioma, ovarian cancer, cholangiocarcinoma, and cancer of unknown primary (n = 285). Using the universal cutoff of 10, the overall response rate in TMB-high tumors was higher compared with TMB-low tumors (27% [6 of 22] vs 11% [28 of 263]; OR, 3.15; 95% CI, 1.14-8.70; eTable 4 in the Supplement).
In this independent validation study of a universal cutoff (≥10 mutations per megabase) for high TMB in 1678 patients with MSS solid tumors treated with anti–PD-1 or PD-L1 therapy, we found that this universal definition of high TMB was associated with higher response rates, although associations were not present or consistent across all cancer types. This heterogeneity may result from applying a single universal TMB cutoff to cancer types with variable TMB distributions and differing mechanisms of neoantigen presentation and immune surveillance across different tissues and anatomical sites. Using cancer type–specific cutoffs, we observed higher response rates in TMB-high compared with TMB-low tumors across almost all cancer types.
Response rates below the cutoff of 10 mutations per megabase were still, on average, more than 20%, suggesting that a universal threshold may not be helpful in prioritizing specific therapies for an individual patient. Tumor mutational burden will likely need to be combined with other biomarkers to more robustly stratify responders and nonresponders.
This study has some limitations, including its retrospective nature and performance at a cancer referral center. Although the number of patients in this study was higher than the cohort used to justify FDA approval, the numbers were still low for some tumor types, limiting generalizability. Cancer type–specific cutoffs were optimized in this data set and require validation in independent data sets.
Our analysis validates the finding of higher response rates in tumors with TMB of 10 or more mutations per megabase, across multiple cancer types. However, the predictive value of a single universal threshold for TMB-high tumors is limited because of wide variability across cancer types. Although the recent FDA approval expands the pool of MSS tumors eligible for anti–PD-1 therapy, we recommend caution when considering this universal TMB cutoff for clinical decision-making, because response rates are not reliably higher with TMB above the cutoff value in all cancer types, and associations between this cutoff value and survival rates are unclear. Further investigation will help define the ideal TMB cutoffs to guide decision-making in individual cancer types and evaluate whether expanded use of anti–PD-1 and PD-L1 drugs based on TMB is a cost-effective strategy.
Accepted for Publication: October 22, 2020.
Published Online: February 18, 2021. doi:10.1001/jamaoncol.2020.7684
Corresponding Author: Luc G. T. Morris, MD, MSc, Department of Surgery, Memorial Sloan Kettering Cancer Center, 1275 York Ave, New York, NY 10065 (firstname.lastname@example.org).
Author Contributions: Drs Valero and Morris had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Chan and Morris jointly supervised the work and are co–senior authors.
Concept and design: Valero, Zehir, Chan, Morris.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Valero, Zehir, Chan, Morris.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Valero, Seshan, Chan, Morris.
Obtained funding: Chan, Morris.
Administrative, technical, or material support: Hoen, Chan, Morris.
Supervision: Chan, Morris.
Conflict of Interest Disclosures: Dr Hoen reported receiving funding from AstraZeneca outside the submitted work. Dr Zehir reported receiving honoraria from Illumina outside the submitted work. Dr Berger reported receiving personal fees from Roche outside the submitted work. Dr Chan reported receiving grant funding from Bristol Myers Squibb, AstraZeneca, Illumina, Pfizer, An2H, and Eisai; serving as an advisor for Bristol Myers Squibb, Illumina, Eisai, and An2H; holding equity in An2H; receiving royalties from PGDX; and being a cofounder of and holding equity in Gritstone Oncology during the conduct of the study. Dr Morris reported receiving laboratory research funding from AstraZeneca outside the submitted work. Drs Chan and Morris are inventors on a patent held by Memorial Sloan Kettering related to the use of tumor mutational burden in cancer immunotherapy. No other disclosures were reported.
Funding/Support: This study was supported in part by Fundación Alfonso Martín Escudero (Dr Valero); grants K08 DE024774 and R01 DE027738 from the National Institutes of Health, the Sebastian Nativo Fund, and the Jayme and Peter Flowers Fund (Dr Morris); grants R35 CA232097 and R01 CA205426 from the National Institutes of Health and the PaineWebber Chair (Dr Chan); and the National Institutes of Health/National Cancer Institute Cancer Center Support Grant P30 CA008748.
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: We are grateful to our patients and their families for their bravery and their support of cancer research. We thank members of the Molecular Diagnostics Service in the Department of Pathology, the Marie-Josée and Henry R. Kravis Center for Molecular Oncology, and the Chan and Morris Laboratories.
Additional Information: All deidentified clinical and tumor mutational burden data have been publicly deposited at Zenodo (https://doi.org/10.5281/zenodo.4419977).