Underutilization of Guideline-Recommended Mismatch Repair/Microsatellite Instability Biomarker Testing in Advanced Colorectal Cancer

In 2015, the KEYNOTE-016 trial first demonstrated the efficacy of PD-1 checkpoint inhibitors for patients with DNA mismatch repair–deficient (dMMR)/microsatellite instability high (MSI-H) advanced colorectal cancer, in which 40% of the patients with dMMR/MSI-H colorectal cancer experienced objective responses compared with none of the MMR-proficient cases (1). dMMR cancer arises as a consequence of the loss of function of MMR genes (typically MLH1, MSH2, MSH6, or PMS2)—occurring either following acquisition of a second pathogenic mutation in patients with inherited mutations in MMR genes (“Lynch syndrome”) or, more commonly, through somatic silencing of MLH1 via promoter hypermethylation. dMMR causes a high tumor mutational burden in microsatellite repeat regions (“MSI-H”), resulting in expression of tumor neoantigens and consequent lymphocytic infiltration within the tumor (2). Blockade of the T-cell inhibitory pathway bolsters native antitumoral responses in dMMR/MSI-H tumors.

The success of PD-1 checkpoint inhibitors for advanced dMMR/MSI-H cancers in KEYNOTE-016/-164/-012/-028/-158 and CheckMate-142 galvanized a change in National Comprehensive Cancer Network (NCCN) guidelines 1.2017 to recommend pembrolizumab or nivolumab as a second-line treatment option for advanced dMMR/MSI-H colorectal cancer (1, 3–5). Also, in early February of 2017, ASCP/CAP/AMP/ASCO jointly released a comprehensive practice guideline on molecular biomarker testing for patients with colorectal cancer, recommending that MMR testing—previously only recommended for patients at increased risk for Lynch syndrome—be conducted universally (6–8).

Those clinical trials’ successes across multiple cancer types culminated in the first cancer site-agnostic FDA accelerated approval of pembrolizumab in May, 2017 for advanced solid tumors with dMMR/MSI-H, including advanced colorectal cancers that had progressed following fluoropyrimidine, oxaliplatin, and irinotecan chemotherapy; followed by nivolumab in July, 2017 for the same indication (9–11). These milestones marked a paradigm shift in the management of advanced colorectal cancer and cemented the utility of MMR/MSI status as a predictive biomarker. Because those landmark advances, the NCCN guidelines have been further updated to recommend universal MMR/MSI testing for all newly diagnosed colorectal cancers and in 2020, the FDA expanded approval of pembrolizumab to the first-line setting for advanced dMMR/MSI-H patients with colorectal cancer (12).

A previous study of patients with colorectal cancer newly diagnosed in 2010 to 2012 found that only 21% of stage IV patients received MMR/MSI testing, including only 36% of young patients for whom testing had been indicated at the time (13). Given the recent expansion in guideline-recommended MMR screening criteria, we undertook an evaluation of the upfront MMR/MSI testing practice patterns in the U.S. for patients with advanced colorectal cancer diagnosed in 2017 to determine whether testing utilization has improved.

This IRB-approved retrospective cohort study examined the National Cancer Database (NCDB), a nationwide observational database containing 71% and 76% of newly diagnosed colon and rectal cancers, respectively, in the U.S (14). NCDB data are abstracted from medical records by certified tumor registrars from pathology reports, laboratory reports, and physician notes for patients managed at Commission on Cancer (CoC)–accredited hospitals (15). Adult patients (≥20-years-old) who presented between 2010 and 2017 with stage IV histologically confirmed colorectal invasive adenocarcinoma were identified using ICD-O-3 histological codes and primary colorectal site codes (C18.0, C18.2–18.9, C19.9, and C20.9). Patients with unknown MMR/MSI testing status (38.2% of patients herein) or previous cancer diagnoses were excluded. 2010 was the first year that MMR/MSI testing data were reported in the NCDB, and 2017 was the latest available year for which “test not done” and unknown whether tested were distinctly coded. The NCDB releases data approximately 3 years after a patient's diagnosis.

The outcome of interest was the receipt of MMR/MSI testing. A lack of testing was defined as “test not done (test not ordered and not performed)” from the site-specific collaborative stage variables for colorectal cancers, as previously reported (13). The NCDB does not specifically distinguish between MMR immunohistochemical and MSI molecular-testing modalities (16). In addition, the NCDB lacks details about testing strategy, sample details, or timing. MMR/MSI testing data are abstracted ≥4 months after initial diagnosis and registries continuously update the reported data for patients; however, it is possible that for some newly diagnosed patients with stage IV colorectal cancer that their MMR/MSI testing occurred after registry abstraction and so was not captured (16).

Independent variables of interest were patient age at diagnosis, sex, race/ethnicity, Charlson–Deyo comorbidity index, and tumor features, including primary site, histological subtype (adenocarcinoma, mucinous adenocarcinoma—including “mucin-producing adenocarcinoma,” or signet ring cell carcinoma), metastatic site(s) (liver, lung, bone, brain, distant lymph node, and/or other site, including carcinomatosis), surgery type at the primary site (none, diagnostic or excisional biopsy, or resection), and biopsy/resection of a regional or distant site (binary). Histologic types incompatible with a colorectal primary site were excluded (n < 10; sebaceous adenocarcinoma, serous cystadenocarcinoma). Socioeconomic variables were patient's insurance status, median household income for the patient's ZIP code of residence, and population size of the patient's county of residence. Care setting variables were the location (by U.S. Census Bureau-designated geographic divisions as follows: New England, Middle Atlantic, South Atlantic, East North Central, East South Central, West North Central, West South Central, Mountain, Pacific) and CoC category of the reporting facility.

To evaluate the practice patterns of testing following the updated guidelines in 2017, the analysis was restricted to those patients diagnosed in 2017. Univariable associations were assessed by the χ² test. Features independently associated with MMR/MSI testing utilization were identified with multivariable logistic regression. Average adjusted predicted probabilities (%_AAP) of MMR/MSI testing were estimated from the regression models (17). Analyses were conducted using Stata (SEv15.1, StataCorp). Two-sided P values <0.001 were stipulated as significant.

The data analyzed in this study were obtained from the NCDB.

From 2010 to 2017, 72,830 adults with newly diagnosed advanced colorectal adenocarcinoma were identified. MMR/MSI testing levels increased from 16.4% (n = 1,342/8,185) of patients in 2010 to 56.4% (n = 5,657/10,022) in 2017 (Fig. 1A). By contrast, MMR/MSI testing levels for all <50-year-old patients with colorectal cancer rose from 35.3% in 2010 to 69.2% in 2017.

To examine the MMR/MSI testing patterns following the updated national guidelines in early 2017, predictors of testing were evaluated using multivariable logistic regression for those patients diagnosed in 2017 (n = 10,022, of whom n = 7,408 had complete data for multivariable analysis) and described using average adjusted predicted probabilities (%_AAP). MMR/MSI testing in 2017 was associated with patient age at diagnosis (Supplementary Table S1), ranging from 64.4%_AAP of patients <50 years old [99.9% confidence interval (CI), 59.3–69.5] to 45.9%_AAP of those 80 years and older (99.9% CI, 40.2–51.5), with patients <50 years more likely to be tested than any other age group (all P < 0.001). Testing was also associated with the extent of primary site surgery, with 71.0%_AAP of resected patients receiving testing (99.9% CI, 68.4–73.7), compared with just 44.5%_AAP of biopsy-only (99.9% CI, 40.9–48.0) and 35.2%_AAP of no primary site surgery (99.9% CI, 30.4–40.0) cases (both P < 0.001). MMR/MSI testing was not associated with patient sex, race/ethnicity, Charlson–Deyo comorbidity index, tumor histology, or metastatic sites.

The relationships between testing and patient socioeconomic status were evaluated. Patient insurance status was not associated with MMR/MSI testing following multivariable adjustment. In unadjusted analysis, 63.7% of privately insured patients received MMR/MSI testing, compared with 50.3% of uninsured and 58.8% of Medicaid-insured patients (P < 0.001). Socioeconomic status (reported by the NCDB as the median household income of the ZIP code of patient's residence) was independently associated with testing in adjusted multivariate analysis: Patients from the poorest quartile of households received less testing (49.6%_AAP, 99.9% CI, 45.5–53.7) than patients from the 3rd (56.9%_AAP, 99.9% CI, 53.3–60.6; P < 0.001) or 4th quartiles (57.6%_AAP, 99.9% CI, 54.3–60.9; P < 0.001).

Although testing levels improved the most over time at community cancer programs (rising from 9.3% in 2010 to 44.7% in 2017; Fig. 1B), they remained significantly lower in 2017 (46.6%_AAP, 99.9% CI, 41.0–52.1) than those at integrated network programs (55.3%_AAP, 99.9% CI, 50.5–60.2; P < 0.001) and academic/NCI-designated comprehensive cancer centers (62.8%_AAP, 99.9% CI, 59.7–65.8; P < 0.001; Supplementary Table S1). Patients managed at academic/NCI-designated comprehensive cancer centers were independently more likely to receive MMR/MSI testing than those at any other cancer program type (all P < 0.001). In 2017, MMR/MSI testing was reported as unknown (code 999) for 24.2% of stage IV patients with colorectal cancer—a rate that did not significantly vary by hospital type (Supplementary Table S2).

Regarding geographic practice patterns in 2017, whereas the fewest patients (n = 443) presented to hospitals in New England (i.e., CT, MA, ME, NH, RI, VT), they were most likely to receive MMR/MSI testing (67.4%_AAP, 99.9% CI, 59.7–75.3) compared with patients diagnosed elsewhere in the U.S. (all P ≤ 0.001) except the Pacific division (P = 0.05; i.e., AK, CA, HI, OR, WA; Supplementary Table S1). Patients managed at hospitals in East North Central (49.3%_AAP, 99.9% CI, 44.9–53.7; i.e., IL, IN, MI, OH, WI) and East South Central (52.4%_AAP, 99.9% CI, 45.5–59.3; i.e., AL, KY, MS, TN) regions were among the least likely to be tested. This was despite New England being tied for having the fewest NCI-designated Comprehensive Cancer Centers (n = 3), whereas East North Central is tied for second most (n = 8). In addition, although New England and East North Central had the highest proportions of patients with advanced colorectal cancer who were managed at community cancer programs (14.2% and 14.9%, respectively; with the lowest being 3.5% in Mountain [i.e., AZ, CO, ID, MT, NM, NV, UT, WY) and 5.6% in Middle Atlantic (i.e., NJ, NY, PA) divisions], 57.0% of patients at New England cancer programs came from the richest quartile of ZIP codes, compared with 25.4% in East North Central and 12.8% in East South Central (P < 0.001).

With the approval of checkpoint blockade immunotherapy for the treatment of dMMR/MSI-H advanced colorectal cancer, universal screening became the recommended standard of care in early 2017 (5, 6, 9, 10). Although testing levels more than tripled between 2010 and 2017—with a ∼20% increase from 2016 to 2017—after the guidelines were updated in early 2017, >43% of patients with advanced colorectal cancer in the U.S. were still not receiving MMR/MSI testing within 4 months of presenting with stage IV disease. Because the NCDB reports data from initial presentation, it is possible that some patients received MMR/MSI testing more than 4 months after their initial diagnosis and were not captured by continuing review by registries. Our results are corroborated by those from a recent analysis of 23 oncology practices from across the eastern U.S., in which 43% of their metastatic colon cancer cases (both synchronous and metasynchronous) in 2017 had not received MMR/MSI testing (18). We found that patients with advanced colorectal cancer who were older, from the poorest quartile of households, or managed at community cancer programs were independently less likely to receive guideline-aligned MMR/MSI testing.

Before the approval of checkpoint inhibitor therapy for dMMR/MSI-H–advanced colorectal cancer, the underutilization of MMR testing for patients with colorectal cancer at increased risk for Lynch syndrome—recommended by the NCCN in 2011—was well documented: Just 48% of U.S. patients with colorectal cancer younger than 50 (all stages) received MMR testing in 2012 (13, 19, 20). We found that testing levels increased to only 69% in 2017 for this high-risk population. Furthermore, for stage IV disease in particular, 69%_AAP of <50-year-olds received upfront MMR/MSI testing, compared with 46%_AAP of ≥80-year-olds. This age discrepancy might reflect persistent adherence to outdated Lynch syndrome screening criteria, which rarely triaged elderly patients to receive screening (7, 8). However, universal testing provides considerable clinical benefit in older patients because approximately 80%–85% of dMMR/MSI-H colorectal cancers are sporadic, with an average age at presentation of ≥75 years (21).

A 2018 survey of 151 oncologists/pathologists/surgeons in the U.S. found that although 84% were aware of MMR/MSI testing guidelines (67% of NCCN, 13% of ASCP/CAP/AMP/ASCO); only 70% of physicians agreed that it is standard practice to perform universal testing for all patients with colorectal cancer (22). The largest barrier to testing endorsed by the physicians was insufficient tissue sample to run the test (48% of physicians); In our analyses, primary tumors that underwent diagnostic biopsy were substantially less likely to have upfront MMR/MSI testing than those that were resected, potentially reflecting insufficient tissue availability for molecular testing. Although molecular testing may require more tissue than might be available in some biopsies, dMMR IHC is more permissive in this regard, because only four 4-μm tissue sections are generally necessary. Furthermore, advances in plasma-based genotyping of circulating-tumor DNA may provide an alternative for MSI-H detection in cases of insufficient tissue (23, 24).

Another barrier to testing cited by surveyed physicians was worry regarding insurance coverage (31% of physicians; ref. 22)—a concern that was shared by 27% of surveyed patients with advanced colorectal cancer (25). In our assessment, however, lack of insurance was not independently associated with upfront MMR/MSI testing in multivariable analysis when patients’ ZIP code-level household income was also included. By contrast, lower ZIP-code level household income was independently associated with lower testing levels, suggesting that broader socioeconomic considerations might affect access to biomarker testing in underappreciated ways (26–31). Geographic discrepancies in upfront MMR/MSI testing utilization in the U.S. appeared, at least in part, to be associated with differences in the socioeconomic statuses of patient populations. Of note, although racial disparities have been previously identified in colorectal cancer patients’ outcomes, in our analyses, race/ethnicity was not independently associated with MMR/MSI testing utilization (32–34). To assess for potential collinearity between race/ethnicity and socioeconomic status, we performed a sensitivity analysis, in which even after excluding ZIP code-level household income status or insurance status from our multivariable model, a significant association between race/ethnicity and MMR/MSI testing was not found.

Although upfront MMR/MSI testing levels improved across all cancer program types, in 2017 patients managed at community programs were still a third less likely to be tested than their counterparts at academic or NCI-designated comprehensive cancer programs. Approaches to MMR/MSI testing vary by institution, with some institutions reflexively performing testing on colorectal cancers as initiated by pathologists and other institutions regarding MMR/MSI testing for Lynch syndrome screening as a genetic test, which require the treating physician to initiate testing with patient consent. It is conceivable that systematic differences in the approach to MMR/MSI testing could have contributed to the difference in testing rates in academic and community settings. Given the crucial role of MMR/MSI testing as a prognostic and predictive biomarker for patients with colorectal cancer, one potential approach for encouraging more widespread adoption among community programs would be to use MMR/MSI testing as a performance metric for gauging the quality of an institution's colorectal cancer care. Reflexive MMR/MSI testing at the time of initial diagnosis could join the National Quality Forum (NQF)’s currently endorsed quality measures for patients with colorectal cancer. In addition, MMR/MSI biomarker test reporting could be made a required part of the CAP Colon and Rectum Protocol and Laboratory Accreditation Program to encourage integration into pathology laboratory workflows nationwide (35–37). An important element that the NCDB did not capture was the reason for a lack of testing and whether patients refused testing. Previous studies found that 36% of surveyed physicians cited patient refusal as a barrier to testing, with 29% of surveyed patients with colorectal cancer reporting personal reservations about the benefits of MMR/MSI testing—together suggesting that improved patient education across institution types may be helpful in expanding testing (22, 25).

Underutilization of guideline-recommended biomarker testing for patients with advanced colorectal cancer also extends beyond MMR/MSI: Among patients with metastatic colon cancer from 23 oncology practices across the U.S., overall, guideline-aligned biomarker (i.e., MMR/MSI, KRAS, NRAS, and BRAF) testing was used in only 40% of patients between 2013 and 2017 (18). Adoption of multigene next-generation sequencing assays may improve multi-biomarker testing rates, because these panels can also be used to infer MSI status by mutational signature; indeed, NGS is already supplanting MSI PCR testing in some academic institutions (18, 38). Overall, our results parallel those from other cancer contexts, in which utilization of biomarker testing lagged for patients from lower socioeconomic statuses or managed at community programs, indicating that more general barriers to molecular testing may exist (39–42).

Owing to NCDB design, this study has multiple limitations. NCDB data likely include coding errors and incomplete records (15, 43). 31% of patients herein were coded as MMR/MSI testing status unknown, and because a lack of testing could erroneously be coded as unknown testing status (code 999), we conducted a sensitivity analysis in which both unknown status and “test not done” (code 998) were considered to be a lack of testing. Using that definition, 43% of advanced colorectal cancers in 2017 were tested, and our multivariable findings remained unchanged (Supplementary Table S3). In addition, the rate of MMR/MSI testing status reported as unknown did not significantly vary by institution type in 2017, suggesting comparable cancer registry reporting rates across hospital settings.

Notably, although cancer registries report MMR/MSI testing status from at least 4 months after initial presentation and continuously update data for included patients, it is possible that some patients had subsequent testing following data abstraction for that patient that was not captured—thereby underestimating the MMR/MSI testing rates. Implementation of practice changes following the issuance of new guidelines takes time and because complete MMR/MSI test utilization data were not available in the NCDB for patients diagnosed after 2017, we were unable to measure how testing levels changed over time after 2017. On the basis of results from similar molecular testing contexts, practice changes can be slow to manifest nationally in the U.S. For instance, despite NCCN guideline recommendations in 2009 for KRAS testing to guide the use of anti-EGFR therapy in patients with metastatic colorectal cancer, just 25% of patients were tested in 2010—rising to only 35% by 2013 (42, 44).

MMR/MSI testing utilization for patients with stage IV colorectal cancer has improved in the last decade; however, we found that after consensus guidelines recommended universal testing in early 2017, 44% of newly diagnosed patients with stage IV colorectal cancer did not receive upfront testing. Testing utilization disproportionately lagged for patients who were older, from the poorest quartile of households, or managed at community cancer programs—less than half of these patients received upfront testing. Now that pembrolizumab has been approved for first-line management of stage IV colorectal cancer, upfront MMR/MSI testing should be incorporated into NQF performance metrics and CAP accreditation requirements.

D. Schrag reports personal fees from Pfizer, non-financial support from Grail, grants from the AACR, and personal fees from JAMA outside the submitted work. No disclosures were reported by the other authors.

D.J. Papke Jr.: Software, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. N.I. Lindeman: Writing–original draft, writing–review and editing. D. Schrag: Writing–original draft, writing–review and editing. J.B. Iorgulescu: Conceptualization, data curation, software, formal analysis, supervision, investigation, visualization, methodology, writing–original draft, writing–review and editing.

J.B. Iorgulescu gratefully acknowledges funding support from the National Cancer Institute (K12CA090354) and Conquer Cancer Foundation. The National Cancer Data Base (NCDB) is a joint project of the Commission on Cancer (CoC) of the American College of Surgeons and the American Cancer Society. The CoC's NCDB and the hospitals participating in the CoC NCDB are the source of the de-identified data used herein; they have not verified and are not responsible for the statistical validity of the data analysis or the conclusions derived by the authors.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.