Comparative Effectiveness of Immune Checkpoint Inhibitors vs Chemotherapy in Patients With Metastatic Colorectal Cancer With Measures of Microsatellite Instability, Mismatch Repair, or Tumor Mutational Burden

Julia C F Quintanilha; Ryon P Graf; Virginia A Fisher; Geoffrey R Oxnard; Haley Ellis; Nicole Panarelli; Douglas I Lin; Gerald Li; Richard S P Huang; Jeffrey S Ross; Parvathi A Myer; Samuel J Klempner

doi:10.1001/jamanetworkopen.2022.52244

. 2023 Jan 23;6(1):e2252244. doi: 10.1001/jamanetworkopen.2022.52244

Comparative Effectiveness of Immune Checkpoint Inhibitors vs Chemotherapy in Patients With Metastatic Colorectal Cancer With Measures of Microsatellite Instability, Mismatch Repair, or Tumor Mutational Burden

Julia C F Quintanilha ¹, Ryon P Graf ¹, Virginia A Fisher ¹, Geoffrey R Oxnard ¹, Haley Ellis ², Nicole Panarelli ³, Douglas I Lin ¹, Gerald Li ¹, Richard S P Huang ¹, Jeffrey S Ross ^1,⁴, Parvathi A Myer ³, Samuel J Klempner ^2,^✉

¹Foundation Medicine, Cambridge, Massachusetts

²Division of Hematology-Oncology, Department of Medicine, Massachusetts General Hospital, Boston

³Medicine/Gastroenterology, Montefiore Medical Center, Albert Einstein Cancer Center, New York, New York

⁴SUNY Upstate Medical University, Syracuse, New York

Accepted for Publication: December 2, 2022.

Published: January 23, 2023. doi:10.1001/jamanetworkopen.2022.52244

^✉

Corresponding Author: Samuel J. Klempner, MD, Division of Hematology-Oncology, Department of Medicine, Massachusetts General Hospital, 55 Fruit St, Boston, MA 02114 (skelmpner@partners.org).

Author Contributions: Drs Quintanilha and Graf had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Graf, Fisher, Ellis, Ross, Myer, Klempner.

Acquisition, analysis, or interpretation of data: Quintanilha, Graf, Fisher, Oxnard, Panarelli, Lin, Li, Huang, Ross, Myer.

Drafting of the manuscript: Quintanilha, Graf, Lin, Ross, Myer, Klempner.

Critical revision of the manuscript for important intellectual content: Graf, Fisher, Oxnard, Ellis, Panarelli, Li, Huang, Ross, Myer, Klempner.

Statistical analysis: Quintanilha, Graf, Fisher.

Supervision: Graf, Oxnard, Lin, Myer, Klempner.

Conflict of Interest Disclosures: Dr Quintanilha reported having equity interest in Roche and receiving grants from the American Heart Association outside the submitted work. Dr Graf reported receiving personal fees from Epic Sciences outside the submitted work. Dr Fisher reported receiving grants from and having equity interest in Roche outside the submitted work. Dr Oxnard reported receiving personal fees from Foundation Medicine and Roche during the conduct of the study. Dr Panarelli reported having a sponsored research agreement with Deciphera Pharmaceuticals outside the submitted work. Dr Lin reported having Roche stock options during the conduct of the study. Dr Li reported being a shareholder in Roche outside the submitted work. Dr Huang reported having equity interest in Roche during the conduct of the study. Dr Ross reported receiving personal fees from Foundation Medicine during the conduct of the study. Dr Klempner reported serving on the gastric cancer advisory board for and receiving personal fees from Eli Lilly, Sanofi-Aventis, Astellas, Merck, Bristol Myers Squibb, AstraZeneca, Daiichi Sankyo, Novartis, Mersana, Exact Sciences, and Natera outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported by Foundation Medicine, a wholly owned subsidiary of Roche and a for-profit company that produces molecular diagnostics regulated by the US Food and Drug Administration.

Role of the Funder/Sponsor: The funder 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.

Data Sharing Statement: See Supplement 2.

Additional Contributions: We thank the patients whose data made this research possible, the clinical and laboratory staff at Foundation Medicine, and the team at Flatiron Health.

^✉

Corresponding author.

PMCID: PMC9871803 PMID: 36689222

This comparative effectiveness research study compares outcomes of first-line therapy with immune checkpoint inhibitors vs chemotherapy among patients who have metastatic colorectal cancer with high microsatellite instability, mismatch repair deficiency, and/or high tumor mutational burden.

Key Points

Question

What is the comparative effectiveness of first-line immune checkpoint inhibitors (ICIs) vs chemotherapy in standard practice settings among patients with metastatic colorectal cancer (MCRC) with high microsatellite instability (MSI-H) determined by next-generation sequencing (NGS)?

Findings

In this comparative effectiveness research study of 138 patients with MCRC and MSI-H, patients receiving first-line ICIs vs chemotherapy had significantly more favorable time to next treatment, progression-free survival, and overall survival outcomes.

Meaning

These findings suggest that MSI determined by NGS is a robust biomarker of ICI outcomes in patients with MCRC.

Abstract

Importance

The KEYNOTE-177 trial demonstrated that patients with metastatic colorectal cancer (MCRC) with high microsatellite instability (MSI-H) and/or mismatch repair deficiency (DMMR) have better outcomes when receiving first-line immune checkpoint inhibitors (ICIs) compared with chemotherapy. Data on performance of ICIs in patients with MCRC in standard practice settings remain limited, and direct MMR vs MSI outcome association comparisons are lacking.

Objective

To validate MSI (determined by next-generation sequencing [NGS]) as a biomarker of ICI effectiveness among patients with MCRC in standard practice settings and examine the association of MSI assessed by NGS, DMMR by immunohistochemistry, and tumor mutational burden (cutoff, 10 mutations/megabase) with ICI outcomes.

Design, Setting, and Participants

This comparative effectiveness research study of outcomes in prospectively defined biomarker subgroups used data from a deidentified clinicogenomic database and included patients who received Foundation Medicine testing (FoundationOne or FoundationOne CDx) during routine clinical care at approximately 280 US academic or community-based cancer clinics between March 2014 and December 2021. The population included 1 cohort of patients with MSI-H MCRC who received first-line ICIs or chemotherapy and a second cohort who received ICIs in any line of therapy (LOT) for biomarker examination.

Exposures

ICI therapy or chemotherapy assigned at physician discretion without randomization.

Main Outcomes and Measures

The main outcomes were time to next treatment (TTNT), progression-free survival (PFS), and overall survival (OS). Hazard ratios were adjusted for known prognostic imbalances. Comparisons of explanatory power used the likelihood ratio test.

Results

A total of 138 patients (median age, 67.0 years [IQR, 56.2-74.0 years]; 73 [52.9%] female) with MSI-H MCRC received first-line ICIs or chemotherapy. A total of 182 patients (median age, 64.5 years [IQR, 55.2-72.0]; 98 [53.8%] female) received ICIs in any LOT. Patients receiving first-line ICIs vs chemotherapy had longer TTNT (median, not reached [NR] vs 7.23 months [IQR, 6.21-9.72 months]; adjusted hazard ratio [AHR], 0.17; 95% CI, 0.08-0.35; P < .001), PFS (median, 24.87 months [IQR, 19.10 months to NR] vs 5.65 months [IQR, 4.70-8.34 months]; AHR, 0.31; 95% CI, 0.18-0.52; P < .001), and OS (median, NR vs 24.1 months [IQR, 13.90 months to NR]; HR, 0.45; 95% CI, 0.23-0.88; P = .02). MSI added to DMMR better anticipated TTNT and PFS in patients receiving ICIs than DMMR alone. The same was not observed when DMMR evaluation was added to MSI.

Conclusions and Relevance

In this comparative effectiveness research study, MSI assessed by NGS robustly identified patients with favorable outcomes on first-line ICIs vs chemotherapy and appeared to better anticipate ICI outcomes compared with DMMR.

Introduction

Approximately 5% of metastatic colorectal cancers (MCRCs) are characterized by microsatellite instability (MSI) and/or deficiency in the mismatch repair pathway (DMMR). Microsatellite instable MCRCs generally harbor a high tumor mutational burden (TMB) and typically feature an enhanced response to treatment with immune checkpoint inhibitors (ICIs). Recently, National Comprehensive Cancer Network guidelines have incorporated ICIs as a first-line treatment option for patients with MCRC with high-MSI (MSI-H) and/or DMMR based on the results of the KEYNOTE-177 trial demonstrating the benefit of pembrolizumab in these patients.¹ The KEYNOTE-177 trial (N = 307) randomized patients with MSI-H and/or DMMR MCRC to first-line therapy with pembrolizumab or chemotherapy and showed a superior objective response rate and progression-free survival (PFS) with pembrolizumab compared with chemotherapy. Overall survival (OS) analysis did not reach statistical significance. Notably, there was a substantial (36%) crossover rate from chemotherapy to second-line ICIs after disease progression.¹

MSI in MCRC is due to sequence variations in 1 of the MMR genes (MSH2, MSH6, PMS2, or MLH1) or MLH1 promoter hypermethylation. Both mechanisms result in loss of protein expression by immunohistochemistry (IHC), which is an established method for detecting MMR deficiency. In CRC, BRAF V600E sequence variation has also been associated with MSI.² In addition to IHC for MMR assessment, other methods to directly detect MSI include polymerase chain reaction (PCR) of select unstable loci or next-generation sequencing (NGS)–based comprehensive genomic profiling. While there are advantages and disadvantages associated with the use of each method, there has been a limited number of studies directly comparing concordance and performance of each method with regards to ICI response in MCRC.

In addition to MSI, TMB has also emerged as a potential biomarker to select patients for ICI treatment. While the majority of MSI-H MCRCs harbor high TMB, a subset of MCRCs may be microsatellite stable (MSS) with high TMB, such as the POLE-variant MCRCs,³ which may be an additional indicator of ICI response independent of MSI. In clinical studies, TMB showed a nominally stronger association with objective response rate and PFS compared with MSI in a retrospective analysis of 22 patients with MSI-H treated with ICIs.⁴ In addition, a combination of MSI and TMB evaluation by NGS is suggested to be more precise than MSI and/or DMMR for the estimation of ICI response.⁵

Patients enrolled in clinical trials do not always represent the standard practice settings,^6,7,8 and the evaluation of drug and biomarker effectiveness in standard clinical practice is important for the validation of findings from clinical trials. We sought to evaluate MSI determined by NGS as a biomarker associated with outcomes of ICI treatment in a clinical context restricted to patients with MCRC with MSI-H tumors receiving first-line ICIs vs chemotherapy, similar to the KEYNOTE-177 trial paradigm. In addition, we sought to compare the performance of MSI determined by NGS, DMMR determined by IHC, and TMB in anticipating outcomes of ICI treatment.

Methods

Study Population

All patients included in this comparative effectiveness research study had a confirmed diagnosis of MCRC, underwent tissue genomic testing using Foundation Medicine comprehensive genomic profiling assays, and were included in the US-wide Flatiron Health and Foundation Medicine Clinico-Genomic MCRC database⁹ between March 2014 and December 2021. Retrospective deidentified longitudinal clinical data were derived from electronic health records from approximately 280 US cancer clinics (approximately 800 sites of care), comprising patient-level structured and unstructured data, curated via technology-enabled abstraction of clinical notes and radiology and pathology reports. These were linked to genomic data derived from Foundation Medicine testing by deidentified, deterministic matching.⁹ Clinical data included demographics, clinical and laboratory features, time of therapy exposure, and survival. Approval of the study protocol was obtained from the WCG Institutional Review Board prior to study conduct and included a waiver of informed consent based on the observational, noninterventional nature of the study. The study followed the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) guidelines.¹⁰

The study population was divided into 2 different cohorts for further analysis. The first cohort included patients who had MSI assessed by NGS and were treated in the first-line setting with ICIs (pembrolizumab monotherapy, nivolumab monotherapy, or the combination of ipilimumab plus nivolumab) or 5-fluorouracil (5-FU)–based chemotherapy regimens (folinic acid [leucovorin calcium], fluorouracil, and oxaliplatin [FOLFOX]; folinic acid, fluorouracil, and irinotecan hydrochloride [FOLFIRI]; folinic acid, 5-FU, oxaliplatin, and irinotecan [FOLFOXIRI] or folinic acid, fluorouracil, irinotecan hydrochloride, and oxaliplatin [FOLFIRINOX]; or capecitabine and oxaliplatin [CAPOX]). The second cohort included patients treated with ICIs in any line of therapy (LOT) that had MSI assessed by NGS, MMR assessed by IHC locally and included in electronic health records, and TMB assessed via a tissue specimen. Figure 1 shows the cohort selection and analysis overview for this study. Race and ethnicity was a patient-reported variable with categories modeled after the US Office of Management and Budget Standards for Race and Ethnicity, and it is reported in the Table to allow comparison with other existing cohorts.

Figure 1. — The 5-fluorouracil (5-FU) chemotherapy regimens included folinic acid (leucovorin calcium), fluorouracil, and oxaliplatin (FOLFOX); folinic acid, fluorouracil, and irinotecan hydrochloride (FOLFIRI); folinic acid, 5-FU, oxaliplatin, and irinotecan (FOLFOXIRI) or folinic acid, fluorouracil, irinotecan hydrochloride, and oxaliplatin (FOLFORINOX); or capecitabine and oxaliplatin (CAPOX). DMMR indicates deficient mismatch repair; EHR, electronic health record; ICI, immune checkpoint inhibitor; LOT, line of therapy; MCRC: metastatic colorectal cancer; MSI-H, high microsatellite instability; MSS, microsatellite stable; and TMB, tumor mutational burden.

Table. Baseline Clinical Characteristics of Patients With MCRC With MSI-H Tumors Who Received ICIs or 5-Fluorouracil–Based Chemotherapy in First-Line Therapy^a.

Characteristic	Patients, No. (%) (N = 138)		P value
Characteristic	ICIs (n = 49)	Chemotherapy (n = 89)
Age, median (IQR), y	72.0 (66.0-81.0)	64.0 (54.0-72.0)	<.001
Gender
Female	29 (59.2)	44 (49.4)	.27
Male	20 (40.8)	45 (50.6)	.27
ECOG performance status, No./available No. (%)^b
0	13/44 (29.5)	34/81 (42.0)	.005
1	20/44 (45.5)	34/81 (42.0)
2	5/44 (11.4)	13/81 (16.0)
≥3	6/44 (13.6)	0
Missing	5/49 (10.2)	8/89 (9.0)
Recurrent disease vs new diagnosis, No./available No. (%)
Recurrent disease	33/48 (68.8)	37/86 (43.0)	.004
New diagnosis	15/48 (31.2)	49/86 (57.0)
Missing	1/49 (2.0)	3/89 (3.4)
Race and ethnicity
Asian	3 (6.1)	0	.20
Black or African American	4 (8.2)	6 (6.7)
Hispanic/Latino	0	0
White	31 (63.3)	59 (66.3)
Other^c	8 (16.3)	19 (21.3)
Unknown or not documented	3 (6.1)	5 (5.6)
Tumor location
Right	25 (51.0)	48 (53.9)	.76
Left	8 (16.3)	17 (19.1)
Other or unknown	16 (32.7)	24 (27.0)
Next line of therapy, No./available No. (%)
Chemotherapy	3/8 (37.5)	12/56 (21.4)	<.001
ICIs	0	39/56 (69.6)
ICIs and chemotherapy	2/8 (25.0)	0
Other	3/8 (37.5)	5/56 (8.9)
Missing	41/49 (83.7)	33/89 (37.1)
BRAF, KRAS, and NRAS status, No./available No. (%)
All wildtype	13/49 (26.5)	27/83 (32.5)	.005, .007, and .44^d
KRAS or NRAS variant and BRAF wildtype	7/49 (14.3)	28/83 (33.7)
BRAF variant and KRAS or NRAS wildtype	29/49 (59.2)	27/83 (32.5)
BRAF variant and KRAS or NRAS variant	0	1/83 (1.1)
Missing	0	6/89 (6.7)

Open in a new tab

Abbreviations: ECOG, Eastern Cooperative Oncology Group; ICI, immune checkpoint inhibitor; MCRC, metastatic colorectal cancer; MSI-H, high microsatellite instability.

^{^a}

5-fluorouracil chemotherapy regimens included folinic acid (leucovorin calcium), fluorouracil, and oxaliplatin (FOLFOX); folinic acid, fluorouracil, and irinotecan hydrochloride (FOLFIRI); folinic acid, 5-FU, oxaliplatin, and irinotecan (FOLFOXIRI) or folinic acid, fluorouracil, irinotecan hydrochloride, and oxaliplatin (FOLFORINOX); or capecitabine and oxaliplatin (CAPOX).

^{^b}

Scores range from 0 to 4, with lower scores indicating better performance.

^{^c}

Smaller race and ethnicity categories were combined as “other” for deidentification prior to our receipt of the data but may include American Indian, Alaska Native, Hawaiian, Pacific Islander, and mixed race and ethnicity.

^{^d}

P values are for differences in BRAF, KRAS, and NRAS variants, respectively.

Comprehensive Genomic Profiling

Hybrid capture–based NGS assays (FoundationOne or FoundationOne CDx) were performed on patient tumor specimens in a Clinical Laboratory Improvement Amendments–certified, College of American Pathologists–accredited laboratory (Foundation Medicine, Inc). Samples were evaluated for alterations as previously described.¹¹ In the content of this research, MSI status was determined via NGS-based comprehensive genomic profiling assay, in which a principal-component-analysis MSI algorithm categorized a tumor as MSI-H or MSS, based on the analysis of 95 to 114 loci, as previously described,¹² and TMB was determined on up to 1.1 megabase (Mb) of sequenced DNA.¹³

Outcomes

Three validated outcome measurements were used as primary end points for this study, including time to next treatment (TTNT), PFS, and OS. Time to treatment discontinuation (TTD) was used as a secondary endpoint for this study. Details of the outcome calculations are in the eMethods in Supplement 1.

Statistical Analysis

The analyses performed in this study were prespecified in a prospectively declared statistical analysis plan and are summarized in eFigure 1 in Supplement 1. The statistical analysis plan followed the ISPOR guidelines,¹⁰ including the specification of the research question, prespecified analytical plans such as inclusion and exclusion criteria, potential biases, primary and secondary outcome measures, handling of missing data, and all methods herein described. The prespecified analyses included the comparison of the effectiveness of first-line treatment with ICIs vs chemotherapy in patients with MCRC who had MSI-H tumors (cohort 1) and the examination of the power of the association with ICI outcomes by biomarkers (MSI by NGS, DMMR by IHC, and TMB) in patients with MCRC treated with ICIs in any LOT (cohort 2).

χ² tests and Wilcoxon rank sum tests were used to assess differences between groups of categorical and continuous variables, respectively. Missing values were imputed with the expected values based on observed covariates using random forests (missForest package of R [R Project for Statistical Computing]). Differences in TTNT, PFS, OS, and TTD were evaluated with the log-rank test and Cox proportional hazards regression models. To adjust for potential confounders across all analyses, the baseline clinical risk score was estimated from known prognostic features as the linear estimate from a Cox proportional hazards regression model for OS in patients treated with any eligible line of chemotherapy (FOLFOX, FOLFIRI, FOLFOXIRI/FOLFIRINOX, or CAPOX). The prognostic features included LOT; age at treatment start; gender; race and ethnicity; recurrent disease vs new diagnosis; Eastern Cooperative Oncology Group (ECOG) performance status; clinical practice type (academic or community); primary tumor location; albumin, alkaline phosphatase, serum creatinine, hemoglobin, and lactate dehydrogenase levels; neutrophil-to-lymphocyte ratio; platelet count; use of opioids before therapy; and use of steroids before therapy. The risk scores were estimated as the mean of estimated values from 3 iterations of missing-forest imputation. Out-of-sample estimations were obtained by k-fold cross-validation from each imputed data set. eFigure 2 in Supplement 1 illustrates the cohort selection and method for prognostic risk score generation, and eFigure 3 in Supplement 1 shows the strong prognostic potential of the risk score generated and the multivariable model containing all patient baseline features included in the risk score generation. Three years of restricted mean survival times (RMSTs) were calculated as the area under the Kaplan-Meier curves at 1-month intervals up to 36 months. Inverse probability weighting was used for creating adjusted Kaplan-Meier curves for calculation of RMST estimates derived from those curves, as a way of adjusting the RMST analysis for the prognostic risk score. In this case, inverse probability weighting represents a pseudo-population of patients negative for biomarkers with the same distribution of clinical risk as the biomarker-positive group. Multiple-comparison adjustments were not performed, and the unadjusted and adjusted P values are reported to quantify the strength of association of the treatment group with each outcome (cohort 1) and of the biomarker with each outcome (cohort 2), not for null hypothesis significance testing. Two-sided tests were used throughout. The likelihood ratio test was performed to evaluate the explanatory power of MSI by NGS vs DMMR by IHC in anticipating outcomes in patients receiving ICI.¹⁴ This test evaluates the goodness of fit of 2 different models, the first model including DMMR by IHC and a second model adding MSI by NGS over the first model (and vice versa). The Cohen κ statistic was calculated for interrater reliability between MSI by NGS and DMMR by IHC.

The methods for additional exploratory analyses are in the eMethods in Supplement 1. R, version 4.1.3, was used for all statistical analyses. Significance was set at P < .05.

Results

Of a total of 138 patients (median age, 67.0 years [IQR, 56.2-74.0 years]; 73 females [52.9%] and 65 males [47.1%]), 49 (35.5%) received first-line ICIs (29 [59.2%] female, 20 [40.8%] male), and 89 (64.5%) received first-line chemotherapy (44 [49.4%] female, 45 [50.6%] male) (cohort 1) (Figure 1). Compared with patients receiving chemotherapy, patients receiving first-line ICIs were older (median age, 72.0 years [IQR, 66.0-81.0 years] vs 64.0 years [IQR, 54.0-72.0 years]; P < .001), were more likely to have worse ECOG performance status, had more recurrent disease, were more likely to not have started the next LOT yet, had a lower rate of KRAS variant and a higher rate of BRAF variant tumors, and were less likely to have received steroids before therapy. The majority of patients who received first-line ICIs started therapy in 2020 or later, while the majority of patients who received first-line chemotherapy started therapy before 2020. No significant differences were observed for other baseline features (Table and eTable 1 in Supplement 1).

A total of 182 patients (median age, 64.5 years [IQR, 55.2-72.0 years]; 98 [53.8%] female and 84 [46.2%] male) received ICIs in any LOT and had data available for all 3 biomarkers (cohort 2) (Figure 1). Of these 182 patients, 83 (45.6%) had MSI-H and 99 (54.4%) were MSS, 82 (45.1%) had DMMR and 100 (54.9%) had proficient MMR (PMMR), 101 (55.5%) had a TMB of 10 mt/Mb or greater, and 81 (44.5%) had a TMB of less than 10 mutations (mt)/Mb. Patients receiving first-line ICIs were older than patients receiving ICIs in the second or later LOT and were more likely to have worse ECOG performance status and BRAF sequence variation. Patients receiving ICIs in the first and second LOTs had more recurrent tumors and were more likely to have tumors located on the right side. Patients receiving ICIs in a later LOT were more likely to have alkaline phosphatase levels with values above the upper limit of normal and more likely to have received steroids before therapy. No significant differences were observed for other baseline features (eTable 2 in Supplement 1).

A total of 32 patients were included in both cohorts 1 and 2. This group represents patients with MSI-H MCRC who received first-line ICIs and had data available for all 3 biomarkers.

Outcomes Among Patients Receiving First-Line ICIs vs Chemotherapy

Among patients with MSI-H tumors, those receiving first-line ICIs had more favorable TTNT (median, not reached [NR] vs 7.23 months [IQR, 6.21-9.72 months]; adjusted hazard ratio [AHR], 0.17; 95% CI, 0.08-0.35; P < .001), PFS (median, 24.87 months [IQR, 19.10 months to NR] vs 5.65 months [IQR, 4.70-8.34 months]; AHR, 0.31; 95% CI, 0.18-0.52; P < .001), and OS (median, NR vs 24.1 months [IQR, 13.90 months to NR]; AHR, 0.45; 95% CI, 0.23-0.88; P = .02) (Figure 2). Taken at face value, RMST analysis showed that patients who received first-line ICIs had a mean benefit of 16.8 months (95% CI, 10.3-22.3 months) in the first 36 months from treatment initiation for TTNT, 12.1 months (95% CI, 6.4-17.5 months) for PFS, and 7.3 months (95% CI, 1.2-13.0 months) for OS (Figure 2). The weighted Kaplan-Meier plots for RMST analysis are shown in eFigure 4 in Supplement 1. No subgroups showed evidence of qualitatively different outcomes (eTable 3 in Supplement 1). The TTD analysis was concordant with the primary outcomes evaluated (eFigure 5 in Supplement 1).

Figure 2. — Kaplan-Meier plots show outcomes by therapy class. Overall survival (OS) estimates are left-truncated. Hazard ratios (HRs) were adjusted for clinical prognostic factors, namely line of treatment; age; gender; race and ethnicity; stage at diagnosis; Eastern Cooperative Oncology Group performance status; opioid prescription before therapy; steroid prescription before therapy; clinical practice type; albumin, alkaline phosphatase, serum creatinine, hemoglobin, and lactate dehydrogenase levels; neutrophil-to-lymphocyte ratio; primary tumor location; and platelet count. Chemotherapy included folinic acid (leucovorin calcium), fluorouracil, and oxaliplatin (FOLFOX); folinic acid, fluorouracil, and irinotecan hydrochloride (FOLFIRI); folinic acid, 5-fluorouracil, oxaliplatin, and irinotecan (FOLFOXIRI) or folinic acid, fluorouracil, irinotecan hydrochloride, and oxaliplatin (FOLFORINOX); or capecitabine and oxaliplatin (CAPOX). AHR indicates adjusted HR; PFS, progression-free survival and TTNT, time to next treatment.

Among patients with MSS tumors, differences in outcomes were not observed for those receiving first-line ICIs vs chemotherapy (eFigure 6 in Supplement 1). It is important to note that the low number of patients with MSS tumors receiving ICIs (n = 10) makes the interpretation of these results, as well as the interpretation of the interaction terms between therapy and each biomarker, more difficult (eFigures 7-9 in Supplement 1).

Association of MSI by NGS, MMR by IHC, and TMB With ICI Outcomes

Patients receiving ICIs in any LOT had more favorable TTNT for MSI-H vs MSS tumors (median, NR vs 6.64 months [IQR, 5.06 months to NR]; AHR, 0.30; 95% CI, 0.17-0.54; P < .001), DMMR vs PMMR (median, NR vs 8.64 months [IQR, 5.03 months to NR]; AHR, 0.37; 95% CI, 0.20-0.66; P = .008), and TMB of 10 mt/Mb or greater vs less than 10 mt/Mb (median, NR vs 5.29 months [IQR, 5.03-31.10 months]; AHR, 0.29; 95% CI, 0.16-0.51; P < .001) (Figure 3A). The same was observed for PFS, with more favorable PFS for MSI-H vs MSS (median, 12.19 months [IQR, 6.24-25.86 months] vs 2.46 months [IQR, 1.97-2.79 months]; AHR, 0.32; 95% CI, 0.21-0.47; P < .001), DMMR vs PMMR (median, 11.9 months [IQR, 6.24-24.87 months] vs 2.50 months [IQR, 2.04-2.79 months]; AHR, 0.38; 95% CI, 0.25-0.56; P < .001), and TMB of 10 mt/Mb or greater vs less than 10 mt/Mb (median, 6.90 months [IQR, 5.03-22.74 months] vs 2.17 months [IQR, 1.87-2.69 months]; AHR, 0.31; 95% CI, 0.21-0.45; P < .001) (Figure 3B). More favorable OS was observed for MSI-H vs MSS (median, NR vs 6.57 months [IQR, 4.14-8.02 months]; AHR, 0.25; 95% CI, 0.14-0.44; P < .001), DMMR vs PMMR (median, NR vs 6.6 months [IQR, 4.21-8.41 months]; AHR, 0.36; 95% CI, 0.11-0.61; P = .001), and TMB of 10 mt/Mb or greater vs less than 10 mt/Mb (median, NR vs 6.41 months [IQR, 4.04-7.95 months]; AHR, 0.30; 95% CI, 0.19-0.49; P < .001) (Figure 3C). The RMST analyses also showed a benefit in outcomes for MSI-H vs MSS and DMMR vs PMMR and with TMB of 10 mt/Mb or greater vs less than 10 mt/Mb (Figure 3D-F). The weighted Kaplan-Meier plots for RMST analysis are shown in eFigure 4 in Supplement 1. The TTD analysis was concordant with the primary outcomes evaluated (eFigure 10 in Supplement 1).

Figure 3. — P < .001 for all comparisons. Overall survival (OS) estimates are left-truncated. Hazard ratios (HRs) and restricted mean survival time (RMST) ratios were adjusted for clinical prognostic factors, namely line of treatment; age; gender; race and ethnicity; recurrent disease or new diagnosis; Eastern Cooperative Oncology Group performance status; opioid prescription before therapy; steroid prescription before therapy; clinical practice type; albumin, alkaline phosphatase, serum creatinine, hemoglobin, and lactate dehydrogenase levels; neutrophil-to-lymphocyte ratio; primary tumor location; and platelet count. DMMR indicates deficient mismatch repair; Mb, megabase; MSI-H, high microsatellite instability; MSS, microsatellite stable; mt, mutation; PFS, progression-free survival; PMMR, proficient mismatch repair; and TTNT, time to next treatment.

Concordance of MSI by NGS and DMMR by IHC

MSI determined by NGS and DMMR by IHC were highly concordant among patients who received ICIs in any LOT, with a Cohen κ statistic of 0.878 (Figure 4A and eFigure 11 in Supplement 1). However, the likelihood ratio test indicated that when MSI was added to the DMMR evaluation, there was an improvement in the explanatory power in anticipating TTNT, PFS, and OS, but when DMMR was added to the MSI evaluation, an improvement in the explanatory power in anticipating outcomes was not observed. Out of the 6 patients with MSI-H tumors determined by NGS and PMMR by IHC, 3 had sequence variations in 1 of the MMR genes (MLH1, MSH2, MSH6, or PMS2) and 5 had a BRAF V600E sequence variation, while none of the 5 patients with MSS by NGS and DMMR by IHC had one of these variations (Figure 4B). An examination of MSI by NGS and MMR by IHC can be observed in the swimmer plot, where 3 patients with MSI-H tumors determined by NGS but PMMR by IHC had favorable outcomes in ICI treatment (Figure 4C). All patients with MSI-H tumors also had a TMB of 10 mt/Mb or greater, and 2 patients with MSS and PMMR tumors with high TMB had a POLE sequence variation and favorable outcomes, consistent with published reports.

Among patients with MSS tumors receiving ICIs in any LOT, those with a TMB of 10 mt/Mb or greater (n = 18) tended to have nominally more favorable outcomes than those with a TMB less than 10 mt/Mb (n = 81) (eFigure 12 in Supplement 1). Among patients who received first-line chemotherapy, those with a TMB of 10 mt/Mb or greater (n = 185) and a TMB of less than 10 mt/Mb (n = 2597) had similar outcomes (eFigure 13 in Supplement 1).

Discussion

The intent of any tumor biomarker is to direct the care of patients, but routine biomarkers are often undervalued and understudied.^{15,16,17,18,19} The optimal biomarker test should be analytically valid and demonstrate clinical validity and utility.²⁰ In general, IHC- and/or NGS-based identification of DMMR and MSI-H MCRC have largely satisfied these criteria, but an onus remains on the field of oncology to continuously reevaluate biomarker performance, including in populations from standard practice settings. Recently, the clinical utility of using MMR or MSI detection to direct frontline MCRC management was directly tested in the KEYNOTE-177 trial, and pembrolizumab is now approved by the US Food and Drug Administration for this population.¹ Herein we have extended the study of MMR or MSI in MCRC and report, to our knowledge, the first comparison of outcomes between patients with MSI-H MCRC receiving first-line therapy with ICIs vs chemotherapy in standard practice settings. Importantly, our findings support MSI assessed by NGS as a biomarker of outcomes of first-line ICIs in patients with MCRC. Interestingly, our study showed that MSI by NGS, DMMR by IHC, and TMB have a similar explanatory power in anticipating outcomes in patients receiving ICI therapy but that MSI by NGS appears to better anticipate outcomes when compared with DMMR by IHC. Given the ability of NGS to simultaneously assess additional clinically relevant CRC biomarkers (BRAF, KRAS, NRAS, HER2, etc), our findings support NGS as the preferred method to identify patients who may benefit from ICIs.

To contextualize these clinical data, it is important to understand potential differences among this study’s population and the KEYNOTE-177 participants. In general, this study’s population was older and frailer (KEYNOTE-177 only included patients with ECOG performance status of 0 or 1, while our study also included patients with ECOG performance status of 2 or higher) and had more BRAF-variant, NRAS and/or KRAS wildtype tumors. Some of these differences may account for the median OS for chemotherapy in our study of approximately 24 months vs the roughly 36-month median OS for chemotherapy in KEYNOTE-177. Regarding ICIs in patients with MSI-H, we observed a more pronounced outcome of ICIs in patients with MSI-H in our study than was found in KEYNOTE-177 (HR of 0.45 in our study vs HR of 0.74 in KEYNOTE-177). In a post hoc analysis of non-CRC tumors comparing chemotherapy vs chemotherapy with ICIs or ICIs alone, the HR observed in our study was consistent.^21,22 Notably, patients in KEYNOTE-177 were locally determined to have DMMR by IHC or MSI-H by PCR and the rates of DMMR vs MSI-H were not reported, nor was central confirmation performed. It is provocative to consider that some patients in KEYNOTE-177 may have been misclassified as having DMMR and/or MSI-H, which would have a larger outcome on the pembrolizumab treatment arm.

One more unique aspect of this present study was the ability to assess MMR by IHC and MSI and TMB by NGS from the same samples. Both MMR and MSI had similar explanatory power for anticipating ICI outcomes and were highly concordant in our data. In formal comparison, the association of the likelihood ratio for superiority of MSI by NGS vs DMMR by IHC with patient outcomes was significant, but the converse did not hold. Together, this suggests that NGS-derived MSI assessments are likely superior to IHC, though we cannot account for factors that may have driven the ordering practitioners to obtain NGS testing. Recently, the College of American Pathologists published an expert consensus statement surrounding the optimal method for identifying defects in DNA MMR in solid tumors.²³ The authors note that IHC, PCR, and NGS are all widely acceptable methods for detection in CRC. We asked a question not about optimal detection method but rather about best performing biomarker in anticipating outcomes. While this may read as a nuanced difference, it is clinically relevant, and our data indicated that MSI by NGS was the strongest test (when compared with IHC) associated with ICI outcomes. Potential reasons for discrepant results between IHC and NGS include technical aspects of IHC in various laboratories, point sequence variations or small insertions or deletions that cause MSI and functionally deficient DMMR proteins without their expression loss as detected by the employed IHC antibodies,^24,25,26 and other DNA repair mechanisms that cause MSI not detectable by IHC. IHC has the advantage of being relatively less expensive and more readily available in pathology laboratories for diagnostic pathologists to directly score tumor cells, even in the setting of low tumor purity.¹²

Due to the interest in TMB as a biomarker in ICI selection, we incorporated TMB into our analyses. Patients treated with chemotherapy did not have different outcomes based on TMB status, suggesting TMB is not a prognostic marker independent of treatment. There were 2 patients with MSS and high TMB who had POLE sequence variations and favorable outcomes on ICIs. In general, excluding those 2 patients with POLE variations, patients with MSI-H had substantially higher TMB levels than patients with MSS. While evaluation of TMB and POLE variations could improve the detection of patients with MSS who may benefit from ICI therapy, these are rare patients, and we do not recommend selecting ICIs based on TMB without MSI in MCRC. Pathogenic POLE variations represent a unique subset of MSS and high-TMB tumors, often harboring very high TMB and responding well to ICI therapy.^27,28

Limitations

While we attempted to conduct a statistically rigorous analysis using validated methods, our work is not without limitations. We acknowledge that this is a rare patient population, and our relatively small sample size limits some biomarker comparisons. Similarly, our clinical data lack some potentially prognostic variables such as presence or absence of liver metastases, though based on KEYNOTE-177 patients,¹ we would not expect major differences in liver metastasis rates. Further, we developed a prognostic risk score to address differences among our cohorts.

Conclusion

In this comparative effectiveness research study, the findings demonstrated that patients with MCRC with MSI-H tumors in standard practice settings may have more favorable outcomes when receiving first-line ICIs vs chemotherapy. MSI assessed by NGS, MMR by IHC, and TMB had similar explanatory power in anticipating ICI benefit in patients with MCRC. However, while MSI by NGS and MMR by IHC were highly concordant, detection of MSI by NGS may be favored for evaluating patients with MCRC for ICI-based treatments when the biomarkers are discordant.

Supplement 1.

eMethods

eTable 1. Baseline Clinical Characteristics of Patients With MCRC With MSI-H Tumors That Received ICI or 5-Fu Based Chemotherapy in First Line of Therapy

eTable 2. Baseline Clinical Characteristics of Patients With MCRC That Received ICIs in Any Line of Therapy and Had MSI (NGS), DMMR (IHC), and TMB Biomarkers Available

eTable 3. TTNT, PFS, and OS for Subgroup of Patients With MCRC With MSI-H Tumors Receiving First-line ICIs vs Chemotherapy

eFigure 1. Statistical Analyses Plan

eFigure 2. Method for Prognostic Risk Score Generation

eFigure 3. Baseline Clinical Characteristics Available in the Electronic Health Record Used to Develop a Risk Score With Strong Prognostic Potential

eFigure 4. Weighted Kaplan-Meier Plots Used for the RMST Analysis

eFigure 5. Patients With MCRC With MSI-H Tumors Had More Favorable TTD When Receiving First-line ICIs vs Chemotherapy

eFigure 6. Patients With MCRC With MSS Tumors Receiving First-line ICIs vs Chemotherapy

eFigure 7. Treatment-MSI Interaction Models for Patients With MCRC That Received ICIs or 5-FU–Based Chemotherapy in First Line of Therapy

eFigure 8. Treatment–DMMR Interaction Models for Patients With MCRC That Received ICIs or 5-FU–Based Chemotherapy in First Line of Therapy

eFigure 9. Treatment–TMB Interaction Models for Patients With MCRC That Received ICIs or 5-FU–Based Chemotherapy in First Line of Therapy

eFigure 10. MSI (by NGS), MMR (by IHC), and TMB Have Similar Predictive Power for ICI TTD

eFigure 11. Venn Diagram Showing Concordance Between Biomarkers for MSI-H/DMMR/TMB ≥10 mt/Mb and MSS/PMMR/TMB <10 mt/Mb

eFigure 12. Among Patients With MSS Tumors Receiving ICIs, Those With TMB≥10 mt/Mb Tended to Have More Favorable Outcomes Than Those With TMB<10 mt/Mb

eFigure 13. Among Patients That Received First-line Chemotherapy, Those With TMB≥10 mt/Mb and TMB <10 mt/Mb Had Similar Outcomes

Click here for additional data file.^{(1.4MB, pdf)}

Supplement 2.

Data Sharing Statement

Click here for additional data file.^{(16KB, pdf)}

References

1.Diaz LA Jr, Shiu KK, Kim TW, et al. ; KEYNOTE-177 Investigators . Pembrolizumab versus chemotherapy for microsatellite instability-high or mismatch repair-deficient metastatic colorectal cancer (KEYNOTE-177): final analysis of a randomised, open-label, phase 3 study. Lancet Oncol. 2022;23(5):659-670. doi: 10.1016/S1470-2045(22)00197-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Tran B, Kopetz S, Tie J, et al. Impact of BRAF mutation and microsatellite instability on the pattern of metastatic spread and prognosis in metastatic colorectal cancer. Cancer. 2011;117(20):4623-4632. doi: 10.1002/cncr.26086 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Ahn SM, Ansari AA, Kim J, et al. The somatic POLE P286R mutation defines a unique subclass of colorectal cancer featuring hypermutation, representing a potential genomic biomarker for immunotherapy. Oncotarget. 2016;7(42):68638-68649. doi: 10.18632/oncotarget.11862 [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Schrock AB, Ouyang C, Sandhu J, et al. Tumor mutational burden is predictive of response to immune checkpoint inhibitors in MSI-high metastatic colorectal cancer. Ann Oncol. 2019;30(7):1096-1103. doi: 10.1093/annonc/mdz134 [DOI] [PubMed] [Google Scholar]
5.Xiao J, Li W, Huang Y, et al. A next-generation sequencing-based strategy combining microsatellite instability and tumor mutation burden for comprehensive molecular diagnosis of advanced colorectal cancer. BMC Cancer. 2021;21(1):282. doi: 10.1186/s12885-021-07942-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Yoo SH, Keam B, Kim M, Kim TM, Kim DW, Heo DS. Generalization and representativeness of phase III immune checkpoint blockade trials in non-small cell lung cancer. Thorac Cancer. 2018;9(6):736-744. doi: 10.1111/1759-7714.12641 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Graf RP, Fisher V, Weberpals J, et al. Comparative effectiveness of immune checkpoint inhibitors vs chemotherapy by tumor mutational burden in metastatic castration-resistant prostate cancer. JAMA Netw Open. 2022;5(3):e225394. doi: 10.1001/jamanetworkopen.2022.5394 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Graf RP, Fisher V, Creeden J, et al. Real-world validation of TMB and microsatellite instability as predictive biomarkers of immune checkpoint inhibitor effectiveness in advanced gastroesophageal cancer. Cancer Res Commun. 2022;2(9):1037-1048. doi: 10.1158/2767-9764.CRC-22-0161 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Singal G, Miller PG, Agarwala V, et al. Association of patient characteristics and tumor genomics with clinical outcomes among patients with non-small cell lung cancer using a clinicogenomic database. JAMA. 2019;321(14):1391-1399. doi: 10.1001/jama.2019.3241 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Berger ML, Mamdani M, Atkins D, Johnson ML. Good research practices for comparative effectiveness research: defining, reporting and interpreting nonrandomized studies of treatment effects using secondary data sources: the ISPOR Good Research Practices for Retrospective Database Analysis Task Force Report–Part I. Value Health. 2009;12(8):1044-1052. doi: 10.1111/j.1524-4733.2009.00600.x [DOI] [PubMed] [Google Scholar]
11.Frampton GM, Fichtenholtz A, Otto GA, et al. Development and validation of a clinical cancer genomic profiling test based on massively parallel DNA sequencing. Nat Biotechnol. 2013;31(11):1023-1031. doi: 10.1038/nbt.2696 [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Trabucco SE, Gowen K, Maund SL, et al. A novel next-generation sequencing approach to detecting microsatellite instability and pan-tumor characterization of 1000 microsatellite instability-high cases in 67,000 patient samples. J Mol Diagn. 2019;21(6):1053-1066. doi: 10.1016/j.jmoldx.2019.06.011 [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Chalmers ZR, Connelly CF, Fabrizio D, et al. Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden. Genome Med. 2017;9(1):34. doi: 10.1186/s13073-017-0424-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Vickers AJ, Cronin AM, Begg CB. One statistical test is sufficient for assessing new predictive markers. BMC Med Res Methodol. 2011;11:13. doi: 10.1186/1471-2288-11-13 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Dinan MA, Lyman GH, Schilsky RL, Hayes DF. Proposal for value-based, tiered reimbursement for tumor biomarker tests to promote innovation and evidence generation. JCO Precis Oncol. 2019;3:1-10. doi: 10.1200/PO.19.00210 [DOI] [PubMed] [Google Scholar]
16.Hayes DF, Allen J, Compton C, et al. Breaking a vicious cycle. Sci Transl Med. 2013;5(196):196cm6. doi: 10.1126/scitranslmed.3005950 [DOI] [PubMed] [Google Scholar]
17.Simon R, Altman DG. Statistical aspects of prognostic factor studies in oncology. Br J Cancer. 1994;69(6):979-985. doi: 10.1038/bjc.1994.192 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Altman DG, Lyman GH. Methodological challenges in the evaluation of prognostic factors in breast cancer. Breast Cancer Res Treat. 1998;52(1-3):289-303. doi: 10.1023/A:1006193704132 [DOI] [PubMed] [Google Scholar]
19.Teutsch SM, Bradley LA, Palomaki GE, et al. ; EGAPP Working Group . The Evaluation of Genomic Applications in Practice and Prevention (EGAPP) Initiative: methods of the EGAPP Working Group. Genet Med. 2009;11(1):3-14. doi: 10.1097/GIM.0b013e318184137c [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Hayes DF. Defining clinical utility of tumor biomarker tests: a clinician’s viewpoint. J Clin Oncol. 2021;39(3):238-248. doi: 10.1200/JCO.20.01572 [DOI] [PubMed] [Google Scholar]
21.Chao J, Fuchs CS, Shitara K, et al. Assessment of pembrolizumab therapy for the treatment of microsatellite instability-high gastric or gastroesophageal junction cancer among patients in the KEYNOTE-059, KEYNOTE-061, and KEYNOTE-062 clinical trials. JAMA Oncol. 2021;7(6):895-902. doi: 10.1001/jamaoncol.2021.0275 [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Shitara K, Ajani JA, Moehler M, et al. Nivolumab plus chemotherapy or ipilimumab in gastro-oesophageal cancer. Nature. 2022;603(7903):942-948. doi: 10.1038/s41586-022-04508-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Bartley AN, Mills AM, Konnick E, et al. Mismatch repair and microsatellite instability testing for immune checkpoint inhibitor therapy: guideline from the College of American Pathologists in collaboration with the Association for Molecular Pathology and Fight Colorectal Cancer. Arch Pathol Lab Med. 2022;146(10):1194-1210. doi: 10.5858/arpa.2021-0632-CP [DOI] [PubMed] [Google Scholar]
24.Hechtman JF, Rana S, Middha S, et al. Retained mismatch repair protein expression occurs in approximately 6% of microsatellite instability-high cancers and is associated with missense mutations in mismatch repair genes. Mod Pathol. 2020;33(5):871-879. doi: 10.1038/s41379-019-0414-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Rosty C, Clendenning M, Walsh MD, et al. ; Colon Cancer Family Registry Cohort . Germline mutations in PMS2 and MLH1 in individuals with solitary loss of PMS2 expression in colorectal carcinomas from the Colon Cancer Family Registry Cohort. BMJ Open. 2016;6(2):e010293. doi: 10.1136/bmjopen-2015-010293 [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Dudley B, Brand RE, Thull D, Bahary N, Nikiforova MN, Pai RK. Germline MLH1 mutations are frequently identified in Lynch syndrome patients with colorectal and endometrial carcinoma demonstrating isolated loss of PMS2 immunohistochemical expression. Am J Surg Pathol. 2015;39(8):1114-1120. doi: 10.1097/PAS.0000000000000425 [DOI] [PubMed] [Google Scholar]
27.Rousseau B, Bieche I, Pasmant E, et al. PD-1 blockade in solid tumors with defects in polymerase epsilon. Cancer Discov. 2022;12(6):1435-1448. doi: 10.1158/2159-8290.CD-21-0521 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Ma X, Riaz N, Samstein RM, et al. Functional landscapes of POLE and POLD1 mutations in checkpoint blockade-dependent antitumor immunity. Nat Genet. 2022;54(7):996-1012. doi: 10.1038/s41588-022-01108-w [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials