Agnostic pembrolizumab: MSI-H, dMMR and high TMB

A 47-year-old teacher has metastatic intrahepatic cholangiocarcinoma and progressed after first-line chemotherapy. The tumor genomic panel arrives with a highlighted line: tumor mutational burden of 12 mutations per megabase, classified as high. The same report states that the tumor is microsatellite stable. In the consultation, she asks whether she can have the immunotherapy she read was approved for any cancer with this result. The resident opens the package insert and finds the indication. The oncologist asks him to read the rest before answering. The scene is fictional and was written for teaching.

This text uses the case as a clinical reasoning scenario for students and health professionals. The topic is the use of pembrolizumab guided by biomarker, regardless of the organ of origin, based on the KEYNOTE-158 study and the American package insert, and the question Dr. Alexei proposed for the series: when to test for MSI, MMR and tumor mutational burden, and why a positive biomarker does not mean automatic indication in any scenario? The text does not replace your country's guideline or your service's protocol. If you play the case of the day on NEXO, you will recognize the challenge: a positive clue must be read together with the others.

Key points

• Mismatch repair deficiency, called dMMR, and high microsatellite instability, called MSI-H, were the first biomarker with organ-agnostic approval for immunotherapy.
• In the studies used for the package insert, about 33 out of every 100 patients with MSI-H or dMMR tumors responded to pembrolizumab, with wide variation between tumor types.
• High tumor mutational burden, from 10 mutations per megabase, led to accelerated approval with response in about 29 out of every 100 patients.
• The indication applies to unresectable or metastatic disease that progressed after treatment and with no satisfactory alternative.
• Local and central tests disagreed in an important share of samples, and the package insert recommends confirmation with an authorized test when possible.
• In some tumors, such as high-grade gliomas, activity was low even with the biomarker present.

Why this is a reasoning scenario

For decades, cancer treatment was chosen by organ of origin and histology. Agnostic approval reversed this logic for some biomarkers: what matters is a molecular feature of the tumor, whether it is from the intestine, the uterus or the stomach. A review on the development of agnostic drugs describes how basket studies allowed testing the same drug in several tumor types with the same alteration (Thein et al., 2024).

The change is real, but it creates a reasoning trap. A positive report seems to answer the treatment question on its own. Dr. Alexei's question asks the opposite: to understand when to test, how to confirm the result and when the biomarker is not enough. The post on pretest probability helps to think about how the same result carries different weight in different contexts.

A strong clue does not close the case

In the teacher's case, the positive biomarker is high mutational burden, not microsatellite instability. This difference matters, as the text will show. The post on heuristics in clinical reasoning describes premature closure, which happens when a convincing clue ends the investigation too early.

Other agnostic biomarkers

Pembrolizumab is not alone on this list. A review with data from the AACR GENIE genomic registry lists five biomarkers with agnostic approval in the United States: repair deficiency or high instability, high mutational burden, NTRK fusions, RET fusions and the BRAF V600E mutation outside colorectal cancer (Gouda and colleagues, 2023). The same review notes that pembrolizumab approval for dMMR and MSI-H came in 2017 and for high mutational burden in 2020. The list shows that the agnostic model applies to both immunotherapy and targeted therapies, but each approval has its own evidence base.

What MSI, dMMR and mutational burden are

The mismatch repair system corrects errors that arise when DNA is copied. It depends mainly on four genes, MLH1, MSH2, MSH6 and PMS2, and the proteins from these genes work in pairs (Zhao and colleagues, 2019). When the system fails, errors accumulate in repetitive regions of DNA called microsatellites, and the tumor starts to have many mutations. The failure can be inherited, as in Lynch syndrome, or arise only in the tumor, for example through silencing of MLH1 (Zhao and colleagues, 2019).

They are two ways of looking at the same phenomenon. Immunohistochemistry looks for loss of the repair proteins, and the result is dMMR. Polymerase chain reaction or sequencing looks for instability in the microsatellites, and the result is MSI-H. Tumor mutational burden, in turn, counts how many mutations the tumor has per megabase of DNA analyzed. Tumors with deficient repair almost always have high burden, but many tumors with high burden have normal repair.

Why the immune system starts to see the tumor

More mutations mean more abnormal proteins, and some of them become neoantigens that lymphocytes recognize. PD-1 blockade releases the brake on these lymphocytes. A review on mutational burden shows that high burden selects for benefit with immunotherapy in several tumors, but that disease-specific cutoffs are not yet well established for many cancer types (Chan and colleagues, 2019). The same review calls for calibration and harmonization between platforms before broader use.

KEYNOTE-158 and the agnostic approval

The US label for pembrolizumab describes the basis of the indication for MSI-H or dMMR tumors: 504 patients in three non-randomized studies, KEYNOTE-164 with colorectal cancer, KEYNOTE-158 with other solid tumors and KEYNOTE-051 with children (Keytruda label). KEYNOTE-158 included 373 patients with non-colorectal tumors that had progressed after prior treatment. All studies excluded patients with autoimmune disease or a condition requiring immunosuppression.

Objective response in the pooled set was about 33 in every 100 patients, with complete response in about 10 in every 100 (Keytruda label). The most striking finding is duration: among those who responded, 77 in every 100 maintained the response for at least 12 months, and 39 in every 100 for at least 36 months. Responses lasting years, in patients already treated, explain why the approval happened without a randomized study.

The colorectal arm

In KEYNOTE-164, with 124 patients with previously treated MSI-H or dMMR colorectal cancer, objective response was 33 in every 100 in both cohorts, and median duration of response was not reached (Le and colleagues, 2020). Median progression-free survival, however, was short, 2.3 and 4.1 months, which shows the other side: patients who do not respond progress quickly.

Tumor type MSI-H or dMMR	Patients	Responded
Small intestine	27	16
Endometrium	94	47
Biliary tract	22	9
Stomach or gastroesophageal junction	51	20
Colorectal	124	42
Ovary	25	8
Pancreas	22	4
Breast	13	1
Brain	27	1

Source of the table numbers: Keytruda label.

Response varies by tumor

The package insert table is the best argument against automatic reading of the biomarker. With the same molecular result, 16 of 27 patients with small bowel cancer responded and 47 of 94 with endometrial cancer, but only 4 of 22 with pancreatic cancer and 1 of 27 with brain tumors (Keytruda package insert). Agnostic means that approval does not require a specific organ, not that the effect is the same in all of them.

In the endometrium, the dedicated analysis of KEYNOTE-158 showed an objective response in 48 of every 100 patients, with a median progression-free survival of 13.1 months and median overall survival not reached (O'Malley and colleagues, 2022). In high-grade gliomas with loss of repair proteins, an observational study with 13 patients found no objective response, and the authors concluded that there was no apparent benefit (Lombardi and colleagues, 2020).

The case of gliomas

Gliomas show another subtlety. The package insert recommends testing mutational burden, MSI and MMR in the sample of the primary tumor obtained before temozolomide, because subclonal repair alterations can arise during that treatment (Keytruda package insert). A positive result after chemotherapy may reflect an acquired mutation in part of the cells, not the biology of the whole tumor.

High mutational burden: a more fragile indication

The indication for high mutational burden came from a planned retrospective analysis of ten KEYNOTE-158 cohorts. Among 790 patients with sufficient tissue, 102 had a burden of at least 10 mutations per megabase, and the objective response in that group was 29 of every 100 (Keytruda package insert). The package insert states that the approval is accelerated, based on response rate and duration, and that continuation may depend on confirmatory studies.

Two details help in the teacher's case. First, the package insert notes that no patient with high burden was identified in the cholangiocarcinoma cohort (Keytruda package insert). That is, there is no direct study data for her tumor. Second, the package insert says that safety and efficacy have not been established in children with central nervous system tumors with high burden, which is a reminder that the biomarker does not replace evidence by context.

What the pan-tumor analysis showed

A retrospective analysis of 2234 participants from several studies found a response of about 31 of every 100 with high burden, measured by exome, versus about 10 of every 100 with low burden (Cristescu and colleagues, 2022). The association appeared independently of PD-L1 expression. This reinforces the value of the marker, but it also shows that most patients with high burden do not respond.

When to test for the biomarkers

Testing for dMMR or MSI has two goals. One is therapeutic, to identify who may benefit from immunotherapy. The other is hereditary, to identify families with Lynch syndrome, which may have screening implications for relatives. In colorectal and endometrial cancer, the test is done routinely for these two reasons. In other advanced tumors, it usually enters the genomic panel when standard options run out or before that, depending on the tumor.

Situations in which the test is usually requested

• Colorectal and endometrial cancer, at diagnosis.
• Advanced solid tumors with a prospect of exhausting standard options.
• Clinical suspicion of Lynch syndrome from personal or family history.
• Metastatic prostate cancer, in which liquid biopsy can help when tissue is scarce.
• Gliomas, preferably in the sample taken before temozolomide.

In the prostate, a study with liquid biopsy in 460 patients with advanced disease identified high instability in 15, and among 9 treated with pembrolizumab, 4 had a drop in prostate-specific antigen of at least half (Barata and colleagues, 2020). The authors argue for widespread testing, precisely because the biomarker is rare but the response can be intense.

How rare the biomarker is

Frequency weighs on the decision to test. The GENIE review cites studies suggesting a prevalence of repair deficiency of around 3 in every 100 tumors across all cancers, with a large concentration in endometrium and bowel (Gouda and colleagues, 2023). In tumors where the biomarker is rare, most tests will be negative, and a single positive result deserves attention to sample and technique quality. In tumors where the biomarker is common, routine testing makes more sense and is already part of many guidelines.

The history of this indication began with small studies. The review by Zhao and colleagues describes the 2017 work by Le and colleagues, with 86 patients with dMMR tumors of 12 different types, in which the objective response to PD-1 blockade reached about 53 in every 100 (Zhao and colleagues, 2019). These early results opened the way for the larger studies that supported the label, in which the response rate was lower, as usually happens when the study population broadens.

Confirming the result

The label contains a fact every resident should know. In samples considered MSI-H or dMMR by local testing in cohort K of KEYNOTE-158, a central sequencing test confirmed MSI-H in 59 of 104, and a central immunohistochemistry panel confirmed dMMR in 105 of 169 (Keytruda label). That is why the label recommends confirming with an authorized test when possible, and accepts high mutational burden as an alternative when confirmation is not feasible.

Why a positive result is not an automatic indication

The answer to the second part of Dr. Alexei's question brings together what the data show. A positive biomarker indicates a higher chance of response, but the decision depends on the tumor type, the line of treatment, the available alternatives, the quality of the test and the patient's clinical condition. The label indication for both biomarkers is for unresectable or metastatic disease that progressed after treatment, and in the case of high burden, with no satisfactory alternative.

What to weigh before indicating

• Which biomarker it is: MSI-H or dMMR has more solid data than isolated high mutational burden.
• Which tumor it is, and whether there are response data in that specific type.
• Whether the test was done on a validated platform and on an adequate sample.
• Whether there is an alternative with better evidence for that tumor and that line.
• Whether the patient has autoimmune disease, a transplant or another condition excluded from the studies.
• Whether there is time to wait for the response, which can take weeks.

For the professor in the scene, high burden opens a possibility, but with less evidence than microsatellite instability, with no direct data in her tumor and with other options to consider, such as targeted therapies if the panel shows treatable alterations. An honest conversation includes these limits. The post on diagnostic uncertainty helps structure this kind of decision.

Dose and safety

For adults with MSI-H or dMMR tumors, the label indicates pembrolizumab 200 mg every 3 weeks or 400 mg every 6 weeks, and 2 mg/kg up to 200 mg every 3 weeks in children (Keytruda label). In the pivotal studies, treatment lasted up to 24 months. The label states that adverse events in patients with MSI-H or dMMR were similar to those in other solid tumors treated with pembrolizumab alone.

Immune-mediated events can affect the lung, bowel, liver, endocrine glands, kidney and skin, and require early recognition. In KEYNOTE-164, grade 3 and 4 treatment-related events occurred in about 13 to 16 of every 100 patients (Le and colleagues, 2020). The conversation with the patient must include these risks, even when the expected response is good.

In practice, the decision also involves logistics. The infusion is given in the hospital or in a clinic, every three or six weeks, and requires periodic tests of thyroid, liver and kidney function. The patient needs to know whom to call if diarrhea, shortness of breath or unusual fatigue appears, because early recognition changes the outcome of immune-mediated events.

Common reasoning errors

• Treating high mutational burden and MSI-H as if they carried the same weight of evidence.
• Indicate immunotherapy without confirming a discordant local test when confirmation is possible.
• Forget to investigate Lynch syndrome after a dMMR result.
• Assume the response will be the same in all tumor types.
• Use a glioma test done after temozolomide without considering acquired alterations.

How NEXO helps train this reasoning

In NEXO, the case of the day is fictional, with a real clinical presentation and up to six progressive clues, and you propose the diagnosis by looking for the code in ICD-10. Agnostic biomarkers are a good exercise because the report result seems to decide everything, when in fact it is one clue among several. The post on clinical decision-making helps you choose tests with a purpose, and the post on differential diagnosis helps you organize clues.

Three exercises with cases

Change the biomarker. Say what changes if the professor's report showed MSI-H instead of isolated high burden.

Change the organ. Say how you would think about a glioblastoma with MSH6 loss detected after temozolomide.

Change the family. Say what to do if a 38-year-old patient with dMMR endometrial cancer has a mother with bowel cancer.

The limits of use of the game are described on the NEXO economy page. If you research clinical reasoning teaching, the page for researchers explains how to get in touch. Professionals who write or review cases find information in curation and in creators, and institutions can learn about the proposal on the partnerships page.

Frequently asked questions

Is pembrolizumab approved for any MSI-H cancer?

In the United States, the label includes unresectable or metastatic MSI-H or dMMR solid tumors that progressed after treatment, with specific rules for some tumors (Keytruda label). Check the approval in your country.

What is the chance of response?

In the set of 504 patients with MSI-H or dMMR, about 33 in 100. It varies widely by tumor, from about half in endometrial cancer to almost none in brain tumors.

Is high mutational burden enough?

It is an indication approved on an accelerated basis, with response in about 29 in 100 patients, and with less data by tumor type (Keytruda label).

Do I need to repeat the test?

The label recommends confirming MSI-H or dMMR with an authorized test when possible, because there was significant discordance between local and central tests.

Does a dMMR result mean Lynch syndrome?

Not necessarily. The deficiency may be only in the tumor, for example through MLH1 silencing, and hereditary investigation follows its own path (Zhao and colleagues, 2019).

What other agnostic treatments exist?

Besides pembrolizumab, there are agnostic approvals for NTRK fusions, RET fusions and BRAF V600E mutation outside colorectal cancer, as well as dostarlimab for dMMR (Gouda and colleagues, 2023).

Is it worth testing if the tumor rarely has the biomarker?

It depends on the tumor and the moment. In advanced disease with options exhausted, a broad panel may find rare and treatable alterations, but the result must be interpreted with care.

Final checklist

• I identified which biomarker is positive: MSI-H, dMMR or high mutational burden.
• I checked the test platform and sample, and the need for confirmation.
• I looked for response data in the specific tumor type.
• I confirmed that the disease is unresectable or metastatic and progressed after treatment.
• I compared it with alternatives with better evidence for that tumor.
• I reviewed contraindications such as autoimmune disease and transplant.
• I considered investigation of Lynch syndrome when indicated.
• I explained the expected benefit, immune-mediated risks and time to response.

Want to practice reasoning with progressive clues? Play the case of the day.

References

1. Barata P, Agarwal N, Nussenzveig R, et al. Clinical activity of pembrolizumab in metastatic prostate cancer with microsatellite instability high (MSI-H) detected by circulating tumor DNA. J Immunother Cancer. 2020. PMID 32788235.
2. Chan TA, Yarchoan M, Jaffee E, et al. Development of tumor mutation burden as an immunotherapy biomarker: utility for the oncology clinic. Ann Oncol. 2019. PMID 30395155.
3. Cristescu R, Aurora-Garg D, Albright A, et al. Tumor mutational burden predicts the efficacy of pembrolizumab monotherapy: a pan-tumor retrospective analysis of participants with advanced solid tumors. J Immunother Cancer. 2022. PMID 35101941.
4. Gouda MA, Nelson BE, Buschhorn L, et al. Tumor-Agnostic Precision Medicine from the AACR GENIE Database: Clinical Implications. Clin Cancer Res. 2023. PMID 37061987.
5. Le DT, Kim TW, Van Cutsem E, et al. Phase II Open-Label Study of Pembrolizumab in Treatment-Refractory, Microsatellite Instability-High/Mismatch Repair-Deficient Metastatic Colorectal Cancer: KEYNOTE-164. J Clin Oncol. 2020. PMID 31725351.
6. Lombardi G, Barresi V, Indraccolo S, et al. Pembrolizumab Activity in Recurrent High-Grade Gliomas with Partial or Complete Loss of Mismatch Repair Protein Expression: A Monocentric, Observational and Prospective Pilot Study. Cancers (Basel). 2020. PMID 32823925.
7. Merck Sharp & Dohme LLC. KEYTRUDA (pembrolizumab) injection, for intravenous use. Prescribing information. DailyMed. 2026. DailyMed.
8. O'Malley DM, Bariani GM, Cassier PA, et al. Pembrolizumab in Patients With Microsatellite Instability-High Advanced Endometrial Cancer: Results From the KEYNOTE-158 Study. J Clin Oncol. 2022. PMID 34990208.
9. Thein KZ, Myat YM, Park BS, et al. Target-Driven Tissue-Agnostic Drug Approvals-A New Path of Drug Development. Cancers (Basel). 2024. PMID 39061168.
10. Zhao P, Li L, Jiang X, et al. Mismatch repair deficiency/microsatellite instability-high as a predictor for anti-PD-1/PD-L1 immunotherapy efficacy. J Hematol Oncol. 2019. PMID 31151482.