INTRODUCTION — Deficient mismatch repair (dMMR) and its characteristic genetic signature, high levels of microsatellite instability (MSI-H) across the genome, define a unique biologic subset of cancers that are characterized by a high tumor mutational load and potential responsiveness to anti-programmed cell death 1 (PD-1)-based immune checkpoint inhibitor immunotherapy. This recognition led to the first tumor-agnostic anticancer therapy approval by the US Food and Drug Administration, in May of 2017, for the anti-PD-1 drug pembrolizumab, which covers both MSI-H and dMMR cancer subsets without specific regard to tumor type. Subsequently, in June 2020, pembrolizumab approval was extended to tumors with high levels of tumor mutational burden (TMB).
This topic provides an overview of the biology of dMMR and TMB, reviews methods to assess dMMR and TMB, the frequency of dMMR/MSI-H and TMB in a variety of tumor types, and summarizes data on the comparative clinical efficacy of immune checkpoint inhibitor immunotherapy in colorectal and noncolorectal cancers with dMMR and high levels of TMB in patients with advanced cancer. An overview of cancer immunotherapy is provided elsewhere. (See "Principles of cancer immunotherapy".)
BIOLOGY OF MISMATCH REPAIR AND TUMOR MUTATIONAL BURDEN — Mutations in one of the DNA mismatch repair (MMR) genes are found in the germline of individuals with Lynch syndrome (hereditary nonpolyposis colorectal cancer [HNPCC]) and in the minority of the 15 percent of sporadic colon cancers that have loss of DNA MMR (most of the defects in sporadic CRC are due to promoter hypermethylation of MLH1 rather than a mutation in a DNA MMR gene).
There are four relevant mismatch repair genes: mutL homolog 1 (MLH1), mutS homolog 2 (MSH2), mutS homolog 6 (MSH6), and postmeiotic segregation increased 2 (PMS2). Further, mutations in the 3' end of the epithelial cell adhesion molecule (EPCAM) gene can cause Lynch syndrome through promoter methylation-induced inactivation of the downstream MSH2 gene. (See "Lynch syndrome (hereditary nonpolyposis colorectal cancer): Clinical manifestations and diagnosis", section on 'Genetics' and "Molecular genetics of colorectal cancer", section on 'Mismatch repair genes'.)
Mismatch repair is one of a cell's mechanisms for repairing damage to DNA that primarily results from single base pair insertions or deletions (called indels) when slippage occurs during DNA replication by DNA polymerases. This type of DNA polymerase error tends to occur at areas of short, repetitive DNA sequences, termed microsatellites. Therefore, deficient mismatch repair (dMMR) can be discovered by looking at the variation in the length of a microsatellite in normal tissue compared with its length in the same patient's tumor tissue. When a high rate of variation in microsatellite length exists across the genome, a tumor is said to have high levels of microsatellite instability (MSI-H), which reflects underlying deficiency in mismatch repair capability. (See "Lynch syndrome (hereditary nonpolyposis colorectal cancer): Clinical manifestations and diagnosis", section on 'Genetics'.)
As expected, an inability to repair DNA damage results in the accumulation of mutations. Tumors that lack the mismatch repair mechanism harbor many more mutations (ie, they are hypermutated) than do tumors of the same type without such mismatch repair defects [1,2]. Invariably, almost all tumors with dMMR demonstrate a high tumor mutation burden (TMB) [3,4]. As an example, an analysis of 62,150 cancer samples analyzed by a large next-generation sequencing panel demonstrated 699 cases to have MSI-H and high TMB (defined in this study as ≥20 mutations per megabase [mut/Mb]) [3]. The vast majority of MSI-H cases (83 percent) also had high TMB, with 97 percent having TMB ≥10 mut/Mb (figure 1).
Although most MSI-H tumors are TMB-high, not all TMB-high cases are MSI-H. (See 'Tumor mutational burden' below.)
These mutations code for mutant proteins, which, like other cell proteins, are recycled via the immunoproteasome pathway. A minority of these mutations can give rise to neoantigens, mutation-derived antigens can be recognized by CD8+ T cells, and targeted by the immune system in vivo [5-8].
The greater immunogenicity of mutations generated by dMMR relates to the nature of single base pair insertions and deletions, which lead to frameshift mutations that may generate not one amino acid alteration but a sequence of multiple altered amino acids, creating an amino acid sequence "foreign" to the person. Some of these peptide sequences can then be presented by a person's human leukocyte antigen (HLA) class I molecules, generating novel antigens or "neoantigens." This is illustrated by a large dataset of 619 colorectal cancers (of all molecular subtypes) that underwent DNA exome sequencing in which the rate of mutations and the rate of high-affinity neoantigens predicted to bind to HLA class I were determined [9]. While frameshift indels represented 8.1 percent of the mutations, these same mutations represented 22.6 percent of the predicted neoantigens (figure 2).
Checkpoint inhibitors, deficient mismatch repair, and the immune response to cancer — Immunotherapeutic approaches to cancer therapy are based on the premise that the immune system plays a key role in the surveillance and eradication of malignancy, and that tumors evolve ways to elude the immune system. The same tolerance mechanisms that suppress the immune response to self-antigens to minimize autoimmune disease may also serve to blunt the immune response to these tumor antigens in vivo [10]. Based on data on the immunogenicity of mutated antigens in melanoma, it has been hypothesized that the "neoantigens" generated from tumor-specific mutations of self-antigens within certain cancers may be recognized by the immune system as foreign and could therefore trigger an antitumor immune response. (See "Principles of cancer immunotherapy".)
Several steps are required for the immune system to effectively attack tumor cells. These include tumor antigen uptake by antigen-presenting cells, such as dendritic cells, presentation of the tumor antigen to T cells, T cell activation and trafficking to the tumor, and direct attack of the tumor. Several immune checkpoints exist to dampen the immune response in order to protect against detrimental inflammation and autoimmunity. In the setting of malignancy, such immune checkpoints can result in immune tolerance to the tumor allowing escape from the immune response and progression of the malignancy. Inhibition of these checkpoints might be expected to halt/reverse disease progression. One well-characterized checkpoint being targeted in several tumor types is programmed cell death 1 (PD-1). PD-1 is upregulated on activated T cells, and upon recognition of tumor via the T cell receptor, PD-1 engagement by programmed cell death ligand 1 (PD-L1) expressed by tumor or other immune cells infiltrating the tumor tissue can lead to T cell inactivation and a "brake" on immune-mediated tumor eradication. (See "Principles of cancer immunotherapy".)
Proof of principle that cancers with dMMR might be particularly susceptible to inhibition of the PD-L1/PD-1 interaction was initially provided by a study of pembrolizumab in dMMR colorectal cancer, and subsequently shown in a variety of tumor types. These data are described in detail below. (See 'Clinical efficacy of anti-PD-1 therapy' below.)
Tumor mutational burden — As noted above, tumors with high mutational burden (particularly those arising in the setting of dMMR) are thought to be more immunogenic and responsive to immune checkpoint inhibitor immunotherapy. (See 'Checkpoint inhibitors, deficient mismatch repair, and the immune response to cancer' above.)
The vast majority of dMMR tumors have high TMB, with the median number of mutations often in the thousands. However, not all TMB-high tumors have dMMR (figure 1) [3], and in these cases, the number of mutations is much lower, approximately 200 nonsynonymous mutations (ie, coding variants that alter the amino acid composition of a protein) per exome (which is equivalent to approximately 10 mut/Mb on the FoundationOne CDx platform) [11-14]. (See "Genetics: Glossary of terms", section on 'Coding mutation or polymorphism'.)
Given these issues, TMB has been of increasing interest as a potential biomarker of benefit from immune checkpoint inhibitor immunotherapy beyond dMMR. In several retrospective studies, performed in different types of cancer, the hypothesis of correlation between high TMB and better response to immune checkpoint inhibitors seems to be validated [14-19]. However, the relationship between TMB and response to immune checkpoint inhibitors is imperfect across and within tumor types. As examples:
●Merkel cell carcinomas, renal cell cancers, and mesothelioma all have higher response rates to immune checkpoint inhibitors than would be anticipated from their TMBs [20-22]. (See 'Tumors with high mutational burden' below.)
●TMB has a poor predictive value for response to immune checkpoint inhibitors in some tumor types such as glioma and microsatellite stable colorectal and gastroesophageal cancer [23-25].
●Although a meta-analysis suggests that TMB predicts benefit for combined therapy that targets two different immune checkpoints (cytotoxic T-cell associated antigen-4 and PD-1/PD-L1) [26], individual studies have not always shown this association [14]. (See 'Tumors with high mutational burden' below.)
A major problem is that the approach for determination of TMB and the TMB thresholds predicting response to immune checkpoint blockade have been developed independently in each tumor type. Further, they are likely to differ across tumor types [27,28] and also across testing platforms (eg, blood versus tumor tissue [29,30]), even within a particular tumor; thus, no consistent pan-cancer testing approach has been validated. Efforts are underway to harmonize the assessment of TMB across different tumor types and platforms [29]. (See 'Approach to testing for high levels of TMB' below.)
Nevertheless, in June 2020, pembrolizumab was approved for unresectable or metastatic TMB-high (≥10 mut/Mb) solid tumors, as determined by a US Food and Drug Administration-approved test, although the FoundationOne CDx assay was the platform used in the supporting KEYNOTE-158 study. The approval was largely based on response rate and not benefits in more meaningful clinical endpoints. Subsequently, concerns have been raised as to whether this approval is too broad and whether immune checkpoint inhibitors should be considered in the context of the cause of the high TMB and not based solely on an absolute threshold [31].
The available prospective trials addressing the benefits of TMB as a predictor of benefit from immune checkpoint inhibitors are discussed below. (See 'Tumors with high mutational burden' below.)
ASSESSING MISMATCH REPAIR
Specific tests — Classically, deficient mismatch repair (dMMR) has been assessed by either direct immunohistochemical testing (IHC) for loss of the various mismatch repair proteins (mutL homolog 1 [MLH1], mutS homolog 2 [MSH2], mutS homolog 6 [MHS6], and postmeiotic segregation increased 2 [PMS2]) or by a comparison of the variation in length of a limited number of microsatellites between normal and tumor tissue by polymerase chain reaction (PCR). It can also be assessed by next-generation sequencing (NGS). For patients with colorectal cancer, assessment by any of these methods is acceptable. For cancers other than colorectal cancer, we recommend the use of IHC or NGS panels rather than PCR for microsatellite instability when evaluating for dMMR/high levels of microsatellite instability (MSI-H) due to the potentially lower sensitivity of PCR in other tumor types.
●Immunohistochemistry – Mutations in the mismatch repair genes that cause Lynch syndrome, and biallelic mutations in mismatch repair genes or hypermethylation of the MLH1 promotor in sporadic dMMR colon cancers typically result in a truncated or lost mismatch repair protein that can be detected as a loss of staining of the protein on tumor IHC. (See "Lynch syndrome (hereditary nonpolyposis colorectal cancer): Clinical manifestations and diagnosis", section on 'Immunohistochemistry'.)
●PCR – Microsatellite instability testing is performed using PCR to amplify a standard panel of DNA sequences containing nucleotide repeats. In the most commonly used panel (which includes two mononucleotides [BAT25 and BAT26] and three dinucleotides [D2S123, D5S346, and D17S250] [32-34]), if 30 percent or more of the markers show expansion or contraction of the repetitive sequences in the tumor compared with the normal mucosa from the same patient, the tumor is reported to have MSI-H. These microsatellites were chosen based on the desire to identify the inherited cause of dMMR (Lynch syndrome) in patients with colorectal cancer. (See "Lynch syndrome (hereditary nonpolyposis colorectal cancer): Clinical manifestations and diagnosis", section on 'Immunohistochemistry' and "Lynch syndrome (hereditary nonpolyposis colorectal cancer): Clinical manifestations and diagnosis", section on 'Microsatellite instability testing'.)
●Next-generation sequencing (NGS) panels – As of more recently, microsatellite instability testing can be performed with NGS panels. As there are thousands of microsatellites through the genome, readily captured with NGS methods, laboratories have developed different methods of gauging the level of variation at these microsatellites. There are multiple commercially available NGS testing panels (such as Caris Life Sciences, Foundation Medicine, Guardant360, or Memorial Sloan Kettering Integrated Mutation Profiling of Actionable Cancer Targets [MSK-IMPACT]) that determine MSI-H status based on a comparison of the microsatellites sequenced [4,35-37]. In each platform, the specific microsatellites analyzed vary, and the statistical methodologies determining variation in length are unique; however, all panels have demonstrated excellent sensitivity and specificity when compared with either PCR-determined MSI-H status [38-40] or IHC-based dMMR determination.
In fact, the larger number of microsatellites evaluated using NGS may lead to improved detection of MSI-H across different tumor types, as data have demonstrated that the variation in length of microsatellites is tissue specific, with different tumor types demonstrating different predilections for altered microsatellites. As examples:
•In a large exome analysis of 617 gastric, colorectal, and endometrial cases, MSI-H using exome data was compared with that using PCR detection and demonstrated a sensitivity of 95.8 percent and a specificity of 97.6 percent [41]. However, in 7 of the 16 discrepant cases, additional data demonstrated the exome MSI-H determination to be correct and the PCR determination to be incorrect.
•In another series of 91 prostate cancer patients, MSI-H by an NGS panel was compared with MSI-H by PCR, using biallelic inactivating mutations in mismatch repair genes as the gold standard. Using this gold standard, 29 patients had dMMR, and 62 had proficient mismatch repair (pMMR). The sensitivity and specificity for detection of dMMR were 93.1 and 94 percent for NGS and 72.4 and 100 percent for PCR [42].
These findings suggest that a specific, limited PCR panel devised primarily for colorectal cancer would likely demonstrate suboptimal performance across other tumor types, and that the use of IHC or NGS panels is preferred when evaluating for dMMR/MSI-H in noncolorectal cancers.
Frequency of dMMR across tumor types — The rate of deficient mismatch repair (dMMR) cancers has been established from a large-scale analysis of 11,139 tumor-normal pairs, primarily derived from The Cancer Genome Atlas (TCGA), covering 39 different cancer types (figure 3) [43]. In this analysis, which was performed mainly on primary tumors, the prevalence of MSI-H extended from 31.4 percent in endometrial carcinomas to 0.25 percent in glioblastoma. Cases of MSI-H were detected in 27 cancer types, 12 of which had a prevalence greater than 1 percent. The top 10 cancers with the highest prevalence were uterine endometrial carcinoma, colon adenocarcinoma, stomach adenocarcinoma, rectal adenocarcinoma, adrenocortical carcinoma, uterine carcinosarcoma, cervical cancer, Wilms tumor, mesothelioma, and esophageal carcinoma.
As dMMR cancers tend to have a good prognosis as localized tumors with a lower rate of developing metastases, the rate of dMMR in stage I to III cancers tends to be higher than that seen in stage IV or metastatic cases. Given that anti-programmed cell death 1 (PD-1) therapy is used for the treatment of dMMR patients with metastatic disease, the prevalence of dMMR/MSI-H across metastatic cancers is of greatest relevance to clinical care. These rates have been best estimated from next-generation sequencing panels applied to primarily metastatic patients. The rates across various different tumor types from Memorial Sloan Kettering Integrated Mutation Profiling of Actionable Cancer Targets (MSK-IMPACT), Caris Life Sciences, and Foundation Medicine are shown in the table (table 1) [35,44].
Approach to testing dMMR as a predictive marker — The US Food and Drug Administration label for pembrolizumab for solid tumors that have deficient mismatch repair (dMMR)/MSI-H indicates that the use should be in patients who "have progressed following prior treatment and who have no satisfactory alternative treatment options." For advanced colorectal cancer, there are broader but more clearly delineated indications:
●Patients previously treated with a fluoropyrimidine, oxaliplatin, and irinotecan.
●Use in first-line therapy is also approved based upon results of KEYNOTE-177 (NCT02563002), a randomized phase III trial which demonstrated a progression free survival benefit for front-line pembrolizumab compared with standard-of-care chemotherapy in first-line dMMR metastatic colorectal cancer. (See "Systemic therapy for nonoperable metastatic colorectal cancer: Selecting the initial therapeutic approach", section on 'Patients with deficient DNA mismatch repair/microsatellite unstable tumors'.)
MSI-H/dMMR testing is appropriate for all patients with colorectal and endometrial cancer as per guideline-based recommendations related to the high rate of Lynch syndrome in these cancer types. In addition, testing other metastatic or advanced tumor types with high rates of MSI-H/dMMR who would be candidates for immunotherapy is appropriate. As an example given that the rate in metastatic colorectal cancer is approximately 4 percent, the testing of cancers with similar rates, such as small bowel adenocarcinoma, gastric cancer, prostate cancer, and biliary cancer, would be one approach for clinical practice (table 1) [35,44]. More recently, an ASCO provisional clinical opinion recommended the testing of dMMR/MSI-H in all patients with metastatic or advanced solid tumors who would be candidates for immunotherapy [45]. Although the provisional opinion also recommends considering the prevalence of dMMR and/or MSI-H status in individual tumor types before making this decision, given both the high level and durable activity of immunotherapy for treatment of advanced dMMR/MSI-H solid tumors, the authors favor an aggressive approach to testing all metastatic or advanced solid tumors.
An important point is that regardless of the tumor type, MSI-H or dMMR may indicate the presence of Lynch syndrome. All patients with an MSI-H/dMMR cancer, regardless of age or family history, should undergo workup to determine sporadic or germline etiology [4]. (See "Lynch syndrome (hereditary nonpolyposis colorectal cancer): Clinical manifestations and diagnosis", section on 'Microsatellite instability testing'.)
APPROACH TO TESTING FOR HIGH LEVELS OF TMB — We recommend the use of a large next-generation sequencing (NGS) platform to assess TMB. Interpretation of TMB should be done in the context of the NGS panel utilized and the tumor type tested. At present only the FoundationOne CDx assay has been approved as a companion diagnostic for the tumor agnostic pembrolizumab approval for patients with a TMB of ≥10 mutations per megabase (mut/Mb). If another assay for TMB is used, clinicians should review the manufacturer's documentation to confirm the appropriate cutoff to define TMB-high tumors.
As noted above, a major problem with using TMB as a predictive factor for response to immune checkpoint inhibitors is that the method of determination and thresholds for TMB have been developed independently in each tumor type, and they are likely to differ across tumor types [27] and across platforms [29]; as a result, no consistent pan-cancer testing approach has been validated. The US Food and Drug Administration approval of pembrolizumab for tumors with high levels of TMB specified the use of the FoundationOne CDx assay, and the cutpoint of ≥10 mut/Mb, because this was the assay and threshold used in the seminal KEYNOTE-158 trial. This cutpoint was preselected based upon analyses of data linking TMB to responsiveness to immune checkpoint inhibitor blockade, mainly in lung and urothelial cancer [11-13], and the data employing higher thresholds as a method of improving antitumor efficacy, at least in lung cancer, are disparate [12,13,46]. As a result, TMB alone is not routinely used in NSCLC to make treatment decisions regarding immunotherapy [47]. (See "Management of advanced non-small cell lung cancer lacking a driver mutation: Immunotherapy".)
However, whether this is the optimal threshold for all other cancers is uncertain, as TMB is a continuous variable. Furthermore, as noted above, there appears to be a tumor tissue-type interaction between TMB and response to immune checkpoint inhibitors that has not been addressed in efficacy studies such as KEYNOTE-158.
There is a need for harmonization of TMB reporting across different NGS assays currently in use and across various tumor types. In one study investigating TMB estimation from 11 NGS panels compared with known whole exome TMB, certain NGS panels consistently overestimated or underestimated TMB [27]. In part this reflects the differences in NGS panel size, composition, and bioinformatics algorithms. To address this variability the Friends of Cancer Research TMB Harmonization Consortium has recommended best practices for NGS panel developers that include (1) TMB reporting in mutations/megabase, (2) standardized analytical validation studies, and (3) the alignment of NGS panel TMB to universal whole exome sequencing standards [29]. The second phase of their harmonization effort focused on variability across different panels and best practices for panel TMB alignment [48]. Clinicians should review the documentation provided by the vendor in interpreting individual TMB results.
Frequency of high TMB across tumor types — The highest levels of mutational load are found in melanomas, non-small cell lung carcinoma, and other squamous carcinomas, and these are also the cancers with the highest percentage of cases with high TMB [3,6]. Intermediate levels and a lower frequency of cases with high TMB are found in breast, uterine ovarian and kidney cancers, prostate and bladder cancers, head and neck cancers, and some subtypes of soft tissue sarcoma. High levels of TMB (variably defined) are found in approximately 12 percent of small bowel adenocarcinomas, and 5 percent of colorectal and gastric cancers [49,50]. The average across all solid tumor types according to data from FoundationOne is 13.3 percent [6].
In our view, tumoral testing for high levels of TMB is appropriate for any patient with a solid tumor for which there is no longer an appropriate treatment option, and who might be eligible for treatment with pembrolizumab.
CLINICAL EFFICACY OF ANTI-PD-1 THERAPY — Anti-programmed cell death 1 (PD-1) therapy with pembrolizumab was approved for use in solid tumors with high levels of microsatellite instability (MSI-H) or deficient mismatch repair (dMMR) by the US Food and Drug Administration (FDA) in May of 2017 based on the results of one prospective trial [1] and retrospective analyses of five uncontrolled single-arm clinical trials [51]. In some trials, patients were required to have MSI-H or dMMR cancers, while in others, a subgroup of patients was identified as having MSI-H or dMMR cancers by testing tumor samples after treatment had begun. A total of 15 cancer types were identified among 149 patients enrolled across these five clinical trials; the most common were colorectal, endometrial, and other gastrointestinal cancers. Of the 149 patients who received pembrolizumab in the trials, 40 percent had a complete or partial response, and the response lasted for six months or more in 78 percent of those patients [51].
Tumors with deficient mismatch repair — Immune checkpoint inhibitors are associated with a high level of activity independent of tumor type and drug used [52]. The following section will review separately the efficacy in colorectal cancer and other tumors with deficiency in mismatch repair.
Metastatic colorectal cancer
Pembrolizumab — The hypothesis that cancers with dMMR might be particularly susceptible to inhibition of the programmed cell death ligand 1 (PD-L1)/PD-1 interaction was initially provided in a phase II trial in which pembrolizumab, an immunoglobulin G4 (IgG4) monoclonal antagonist antibody to PD-1, was administered intravenously at a dose of 10 mg/kg every 14 days to 11 patients with dMMR metastatic colorectal cancer, 21 patients with proficient mismatch repair (pMMR) metastatic colorectal cancer, and 9 patients with noncolorectal dMMR metastatic cancers; all had been heavily pretreated [1]. In a later analysis of an expanded cohort of 54 patients with dMMR or pMMR metastatic colorectal cancer, patients with dMMR metastatic colorectal cancer had a 50 percent objective response rate (ORR) and an 89 percent disease control rate (objective response or stable disease) [44]. By contrast, the ORR was 0 percent, and the disease control rate was 16 percent in patients with pMMR metastatic colorectal cancer. After a median treatment duration of 5.9 months, no patients in the dMMR group who had responded had progressed. Overall survival and progression-free survival (PFS) were not reached in the dMMR group, compared with a median PFS of 2.3 months and an overall survival of 7.6 months in the pMMR group.
Patients with dMMR noncolorectal cancers had responses similar to those of patients with dMMR metastatic colorectal cancers (immune-related ORR 71 percent [5 of 7 patients], immune-related PFS rate 67 percent [4 of 6 patients]) [1].
Largely based on these data, in May of 2017, the FDA granted accelerated approval to pembrolizumab for the treatment of patients with advanced MSI-H or dMMR metastatic colorectal cancer that had progressed following conventional chemotherapy, and the approval was extended to a variety of advanced solid tumors other than colorectal cancer (eg, endometrial, other gastrointestinal, breast, prostate, bladder, thyroid, and other sites) that had MSI-H or dMMR, that had progressed following prior treatment, and for which there were no satisfactory alternative treatment options [53].
High levels of durable antitumor efficacy for pembrolizumab have now been confirmed in other cohorts and in a multicenter phase II trial of patients with previously treated dMMR metastatic colorectal cancer [44,54].
Nivolumab with or without ipilimumab — Benefit for a second anti-PD-1 monoclonal antibody, both alone and in combination with ipilimumab, in patients with dMMR metastatic colorectal cancer was suggested in a second trial (CheckMate 142) in which patients with dMMR (n = 59) or pMMR (n = 23) metastatic colorectal cancer received nivolumab (a fully human anti-PD-L1 monoclonal antibody) with or without ipilimumab (a monoclonal antibody directed against cytotoxic T-lymphocyte antigen 4 [CTLA-4]) [55]. The following results are available:
●In a preliminary report presented at the 2016 American Society of Clinical Oncology (ASCO) annual meeting, immunotherapy with nivolumab, with or without ipilimumab, benefited those with dMMR tumors (13 confirmed partial responses, median PFS 5.3 months). By contrast, there were no objective responses among those with pMMR tumors, and median PFS was 1.4 months [55].
●In a later analysis of 74 patients with dMMR metastatic colorectal cancer treated with nivolumab alone, 23 had an objective response (31 percent), and the median duration of response had not been reached; eight had responses lasting 12 months or longer [56]. Largely based on these data, in August of 2017, the FDA extended the approval of nivolumab to MSI-H or dMMR metastatic colorectal cancer that progressed following treatment with a fluoropyrimidine, oxaliplatin, and irinotecan [51]; the approval was not tissue agnostic.
●A subsequent report suggested an even greater degree of benefit for targeting both PD-1 and CTLA-4 as compared with nivolumab alone [57] in a nonrandomized comparison. Largely based on these data, in July of 2018, the FDA approved the combination of nivolumab plus ipilimumab for patients with previously treated MSI-H or dMMR metastatic colorectal cancer. Once again, this approval was not tissue agnostic. Preliminary data also support benefit for first-line nivolumab plus ipilimumab in dMMR metastatic colorectal cancer [58].
It is currently not known in which patients with MSI-H metastatic colorectal cancer to use combined nivolumab plus ipilimumab, or whether this combination is active in patients who relapse or progress on single-agent checkpoint inhibitor immunotherapy; data are extremely limited [59].
The use of checkpoint inhibitors in earlier lines of therapy was addressed in the KEYNOTE-177 trial, a randomized phase III trial of pembrolizumab compared with standard-of-care chemotherapy in first-line dMMR metastatic colorectal cancer [60]. Pembrolizumab was superior in efficacy (PFS) and better tolerated than conventional front-line for individuals with dMMR tumors, and as a result of these data, pembrolizumab was approved for first-line treatment of dMMR/MSI-H advanced metastatic colorectal cancer. (See "Systemic therapy for nonoperable metastatic colorectal cancer: Selecting the initial therapeutic approach", section on 'Patients with deficient DNA mismatch repair/microsatellite unstable tumors'.)
Other tumors with MSI-H/dMMR — Of the three anti-PD-1 monoclonal antibodies approved in the United States for treatment of dMMR/MSI-H cancer (pembrolizumab, nivolumab, and dostarlimab), two (pembrolizumab and dostarlimab) have received a tissue agnostic approval for treatment of any advanced dMMR/MSI-H tumor. Nivolumab is approved for treatment of dMMR/MSI-H advanced colorectal cancer.
Treatment results for anti-PD-1 immunotherapy in a variety of malignancies, and the integration of immune checkpoint inhibitor immunotherapy in the context of the other available therapeutic options for advanced disease of individual histologies are discussed elsewhere:
●(See "Treatment of metastatic endometrial cancer", section on 'Immune checkpoint inhibitors'.)
●(See "Treatment of advanced, unresectable gallbladder cancer", section on 'Immunotherapy'.)
●(See "Treatment of small bowel neoplasms", section on 'Indications for immunotherapy'.)
●(See "Systemic therapy for advanced cholangiocarcinoma", section on 'Immunotherapy'.)
●(See "Immunotherapy for castration-resistant prostate cancer", section on 'PD-1 pathway inhibition'.)
Pembrolizumab — A wide variety of tumor types other than metastatic colorectal cancer that have high levels of microsatellite instability (MSI-H)/deficient mismatch repair (dMMR) may benefit from pembrolizumab, and the broad FDA approval for this agent supports the use of this agent across a wide spectrum of histologies. As examples:
●In an analysis of 86 dMMR tumors receiving single-agent pembrolizumab, ORRs, durations of disease control, and overall survivals were similar in the 40 patients with metastatic colorectal cancer when compared with the 46 patients with noncolorectal cancers [44]:
•ORR of 52 versus 54 percent
•Disease control rate (objective response plus stable disease) of 82 versus 72 percent
•Two-year PFS of 59 versus 46 percent
•Two-year overall survival of 72 versus 57 percent
A summary of the treatment response according to primary tumor site in this series is provided in the table (table 2).
●A second larger cohort of 233 patients with noncolorectal dMMR tumors (27 different tumor types) receiving pembrolizumab monotherapy in the phase II KEYNOTE-158 study reported antitumor activity for the following tumor types with the greatest enrollment: endometrial cancer, gastric cancer, pancreatic cancer, small intestine cancer, ovarian cancer, and cholangiocarcinoma (table 3) [61]. Notably, approximately one-third of the patients with an objective antitumor response had a complete response.
Dostarlimab — Dostarlimab was initially approved for MMR-deficient advanced endometrial cancer. (See "Treatment of metastatic endometrial cancer", section on 'Tumors that are dMMR, MSI-H, or have high TMB'.)
However, in August 2021, the US FDA granted a tissue-agnostic approval to dostarlimab for any relapsed/refractory dMMR solid tumor [62]. Approval was based on a preliminary report of 106 patients with dMMR nonendometrial solid tumors (93 percent of gastrointestinal [GI] tract origin) treated on the GARNET study with dostarlimab 500 mg intravenously (IV) every three weeks for four courses, then 1000 mg every six weeks until treatment discontinuation [63]. In a preliminary report presented at the 2021 GI ASCO meeting, at a median follow-up of 12.4 months, the confirmed objective response rate was 38.7 percent, with a complete response rate of 7.5 percent. The estimated probability of maintaining a response at 18 months was 81 percent. Severe treatment-related adverse events were experienced by only 5.5 percent, most commonly lipase elevations.
Nivolumab — Information on the benefits of nivolumab monotherapy for dMMR noncolorectal cancer is available from a subprotocol arm of the National Cancer Institute Molecular Analysis for Therapy Choice (NCI-MATCH) trial [64]. Of the 42 enrolled patients, the most common histologies were endometrial (n = 13) and prostate (n = 5) adenocarcinoma and uterine carcinosarcoma (n = 4). The ORR was 36 percent, and 3 of 15 responses were complete. The estimated 6-, 12-, and 18-month PFS rates were 51, 46, and 31 percent, respectively.
The combination of nivolumab plus ipilimumab has not yet been tested in metastatic MSI-H noncolorectal cancers.
Response assessment — Individuals treated with immune checkpoint inhibitors for any cancer, can have pseudoprogression within the first several months of treatment, although the incidence is variable, in part related to differences in diseases, treatments, timing of restaging exams, as well as how the concept is defined. Response criteria specifically geared toward these agents (ie, immune-modified RECIST (table 4)) should be used. These issues are addressed in more detail elsewhere. (See "Principles of cancer immunotherapy", section on 'Immunotherapy response criteria'.)
Tumors with high mutational burden — As noted above, TMB has been of increasing interest as a potential biomarker of benefit from immune checkpoint inhibitor immunotherapy, and several reports now support a link between high levels of TMB and response to anti-PD-1 therapy in a variety of primary tumor types [11,14,20,65-72]. (See 'Tumor mutational burden' above.)
The most compelling data on the predictive capacity of TMB as an independent predictor of the response to immune checkpoint inhibitor immunotherapy come from the multicenter open-label phase II KEYNOTE-158 study, which established a link between TMB-high status (as determined by the FoundationOne CDx assay) and overall response rate with pembrolizumab [72]. The trial accrued patients with anal, biliary, cervical, endometrial, salivary, thyroid, or vulvar carcinoma, mesothelioma, a neuroendocrine tumor (NET), or small cell lung cancer (SCLC), who had an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 or 1, and had progressed or were intolerant of at least one prior line of standard therapy. Pembrolizumab was administered at 200 mg IV every three weeks.
There were 805 patients with evaluable TMB: 105 were TMB-high (defined as TMB ≥10 mutations per megabase [mut/Mb]), 14 of whom were also MSI-H. For TMB-high patients, the ORR (the primary endpoint) was 29 percent, including 28 percent in non-MSI-H patients, while the ORR for TMB-low patients was only 6 percent. Among the TMB-low patients, responses were seen in 33 of 383 (9 percent) of those with PD-L1-positive tumors, and 9 of the 274 (3 percent) who had PD-L1-negative tumors. Some responses were complete. While median PFS was similar for the TMB-high and TMB-low groups (median 2.1 months in each group), the tail of the PFS curve favored TMB-high patients (three-year PFS 32 versus 22 percent of those with TMB-low tumors). Among those with TMB-high tumors, responses were seen in SCLC (10 of 34, 29 percent), cervical cancer (5 of 16, 31 percent), endometrial cancer (7 of 15, 47 percent), NETs (2 of 5, 40 percent), vulvar cancer (2 of 12, 17 percent), anal cancer (1 of 14, 7 percent), thyroid cancer (2 of 2, both complete), and salivary gland cancer (1 of 3, 33 percent). None of the 63 enrolled patients with biliary tract cancer had TMB-high disease.
Largely based on these results, in June 2020, the FDA approved pembrolizumab for the treatment of adult and pediatric patients with unresectable or metastatic solid tumors that are tissue TMB-high (≥10 mut/Mb) by an FDA-approved assay (although the FoundationOne CDx assay was used in the supporting KEYNOTE-158 clinical trial), who have progressed following prior therapy, and who have no satisfactory alternative treatment options.
The FDA label does not explicitly include common tumors (eg, breast, colorectal, gastroesophageal, prostate) as they were not included in the KEYNOTE-158 study. At least some data in advanced colorectal cancer (using a different assay and different TMB threshold) suggest a very low response rate (4 percent) in TMB-high but microsatellite stable disease; similar findings are reported for advanced esophagogastric cancers [24,25,73].
On the other hand, the available data in breast cancer are conflicting:
●One report from the TAPUR study that included a small number of patients with metastatic breast cancer and high TMB treated with pembrolizumab noted a 37 percent objective response rate [74].
●In another analysis of data from the ARTEMIS study in which women with triple negative breast cancer who did not respond to four courses of doxorubicin/cyclophosphamide were stratified to molecularly targeted therapy or clinicians' choice of treatment, high levels of TMB did not predict for response to immune checkpoint blockade (objective response rate <20 percent), a similar finding to other cancers where the neoantigen load was not positively correlated with CD8 T-cell levels [75].
Doubts remain as to whether absolute TMB levels alone are a useful indicator of benefit from immune checkpoint inhibitors for all solid tumor types. A secondary analysis of a series of 1661 patients with various tumors treated with a checkpoint inhibitor at a single institution [66] showed that a TMB cutoff of 10 or more did not predict for benefit from checkpoint inhibitor immunotherapy in most of the limited subgroup of MMR-proficient tumors when stratified according to tumor type [31]. Only patients with metastatic head and neck cancer, non-small cell lung cancer, and melanoma with MMR-proficient high TMB tumors had improved overall survival with immunotherapy.
These data underscore the high level of tumor type dependency of high levels of TMB; caution is urged in extrapolating the FDA labeling to all tumor types. (See 'Approach to testing for high levels of TMB' above.)
The benefits of dual checkpoint inhibitor blockade is under study, but the data are so far less than definitive. Although a meta-analysis suggests that TMB predicts benefit for combined therapy that targets two different immune checkpoints (CTLA-4 and PD-1/PD-L1) [26], individual studies have not always shown this association [14].
SUMMARY AND RECOMMENDATIONS
●Deficient mismatch repair – Deficient mismatch repair (dMMR) and its characteristic genetic signature, high levels of microsatellite instability (MSI-H), define a unique biologic subset of cancers that have a high tumor mutational load and responsiveness to immune checkpoint inhibitor immunotherapy.
•Tissue-agnostic approvals – This recognition led to the first tumor-agnostic anticancer therapy approval by the US Food and Drug Administration (FDA), in May of 2017, for the anti-programmed cell death-1 (PD-1) monoclonal antibody pembrolizumab. Subsequently, a second anti-PD-1 monoclonal antibody, dostarlimab, was granted a tissue-agnostic approval. Both drugs are approved for solid tumors that have dMMR/MSI-H and have progressed following prior treatment and who have no satisfactory alternative treatment options. Notably, the safety and effectiveness of pembrolizumab in children with MSI-H central nervous system tumors has not been established. (See 'Other tumors with MSI-H/dMMR' above.)
•Assessment of dMMR – dMMR can be assessed by either direct immunohistochemical testing (IHC) for loss of the various mismatch repair proteins (mutL homolog 1 [MLH1], mutS homolog 2 [MSH2], mutS homolog 6 [MSH6], and postmeiotic segregation increased 2 [PMS2]) or by a comparison of the variation in length of a limited number of microsatellites between normal and tumor tissue by polymerase chain reaction (PCR). It can also be assessed by next-generation sequencing (NGS). (See 'Assessing mismatch repair' above.)
-For patients with colorectal cancer, assessment by any of these methods is acceptable.
-For cancers other than colorectal cancer, we recommend the use of IHC or NGS panels rather than PCR for microsatellite instability when evaluating for dMMR/MSI-H.
•Indications for testing – Given that anti-PD-1 therapy is used for the treatment of metastatic disease, the prevalence of dMMR/MSI-H across metastatic cancers is of greatest relevance to clinical care. An estimate of the frequency of dMMR across several tumor types, as derived from NGS panels applied to primarily metastatic patients, is presented in the table (table 1). (See 'Frequency of dMMR across tumor types' above.)
-Given the approval of pembrolizumab for front-line therapy, testing for dMMR/MSI-H should be undertaken for all patients with advanced colorectal cancer prior to initiating front-line therapy for metastatic disease. (See 'Metastatic colorectal cancer' above.)
-However, given the broad indication for pembrolizumab and dostarlimab in a variety of dMMR solid tumors, and the high level and durability of responses to immunotherapy in dMMR tumors, we favor a more universal testing approach that includes for all metastatic or advanced tumors if the individual would be a candidate for immunotherapy. (See 'Approach to testing dMMR as a predictive marker' above.)
●High levels of tumor mutational burden (TMB) – Tumors with high mutational burden (particularly those arising in the setting of dMMR) are thought to be more immunogenic and responsive to immune checkpoint inhibitor immunotherapy. (See 'Tumor mutational burden' above.)
•Tissue-agnostic approvals – Based upon an early report of the KEYNOTE-158 trial, pembrolizumab is now approved for the treatment of adult and pediatric patients with unresectable or metastatic solid tumors that are tissue TMB-high (≥10 mutations per megabase [mut/Mb]) by an FDA-approved assay, and who have progressed following prior therapy and who have no satisfactory alternative treatment options. (See 'Tumors with high mutational burden' above.)
•Assessment of TMB – We suggest the use of a threshold TMB of ≥10 mut/Mb if the NGS platform FoundationOne CDx assay is used, to select patients for immune checkpoint inhibitor immunotherapy, as was done in KEYNOTE-158. As other assays and platforms are now reporting TMB, clinicians should refer to the manufacturer's information to identify the specific cutoff for that assay to define TMB-high tumors. (See 'Approach to testing for high levels of TMB' above.)
•Indications for testing – In our view, tumoral testing for high levels of TMB is appropriate for any patient with a solid tumor for which there is no longer an appropriate treatment option, and who might be eligible for treatment with pembrolizumab. (See 'Frequency of high TMB across tumor types' above.)
