Keytruda approval for TMB-high tumors: A double-edged sword?
By Martin Strebl-Bantillo and Dennis Chang
Immune checkpoint inhibitors (ICIs)—including anti-CTLA4, anti-PD1, and anti-PD-L1 antibodies—have transformed the landscape of cancer treatment, reshaped standards of care, and for some patients cured their disease. To date, the FDA has approved 7 ICI agents1 that together drove more than 24 billion USD in global sales in 20192. At the forefront of this class is Keytruda (pembrolizumab, or just “pembro”), Merck’s crown jewel; in 2019 alone, global sales exceeded 11 billion USD3. As of this writing, Keytruda is FDA-approved for more than two dozen distinct cancer indications across more than 15 tumor types4. One of the most recent approvals, granted on June 16, 2020, was for any cancer—regardless of site of origin or histology—with high tumor mutational burden (TMB) and refractory to standard of care.
TMB, defined as the frequency of mutations in the cancer cell genome, is relevant for immunotherapy as immune cells can detect cancer primarily because cancer cells harbor mutant proteins, generating neoantigens not normally found on healthy cells. Accordingly, TMB is an indirect measure of neoantigen load, and therefore an indirect measure of the immunogenicity of the tumor. The Keytruda approval marks the first approved indication using TMB as a biomarker, an achievement that is remarkable because BMS and others preceded Merck in examining TMB in clinical trials5.
However, the potential implications are yet unclear, and the approval may turn out to be a double-edged sword for Keytruda’s commercial success. While the indication may have tremendous upside, practical challenges could render its impact more incremental, and as we outline below, it may even present a potential vulnerability to be exploited in this fiercely competitive landscape.
The broad potential
This is not the first histology-agnostic indication granted to Keytruda: In 2017, it was approved for cancers with high microsatellite instability or mismatch repair deficiency (MSI-H/dMMR) and refractory to prior therapy. MSI-H/dMMR cancers account for ~1.5-3.8% of all cancers6. Genomic studies suggest that TMB at a level of ≥10 mutations per megabase of DNA sequence occurs in ~5–6 times as many patients as MSI-H, or ~10–20% of all cancers7. In Merck’s registrational trial, Keynote-158, ~13% of patients screened for the study were found to be TMB-H (per the ≥10 mutations/megabase threshold), consistent with these estimates. The current forecast of $24B worldwide sales for Keytruda in 20262—as astronomical as that seems, being higher than the annual sales of any other drug in history—may actually be an underestimate if physicians broadly use Keytruda for patients with TMB-H cancers.
However, a few factors need to be considered. The requirement to have exhausted standard therapies shrinks the opportunity considerably and first-line approvals (similar to Keytruda’s first-line approval in MSI-H/MMRd colorectal cancer) will be required to maximize the potential. However, some indications (e.g., lung and skin cancers) are already treated with checkpoint inhibitors in earlier lines of therapy, and it is unclear how much benefit Keytruda will have in those patients, regardless of TMB status. Furthermore, now that Keytruda has blazed the trail for a TMB-H indication, others are likely to follow and the competition may temper Keytruda’s commercial returns.
Constraints of TMB testing
It is broadly accepted that TMB correlates with response to ICIs, but the specifics of how to measure and apply this relationship are less than clear. The situation is further complicated by inconsistencies in how TMB is measured, and what mutation cutoff is chosen to qualify patients as “high” TMB in different tumor types8.
Nonetheless, the FDA approved the use of Keytruda in TMB-H tumors, using the ≥10 mutations/ megabase threshold broadly, the first acknowledgement by the FDA that the biomarker is good enough to support patient selection for specific treatments. Foundation Medicine will provide the FDA-cleared companion diagnostic (CDx) to provide the TMB scores for treatment decisions, which should gain a certain reference quality in measuring TMB as a consequence.
Harmonization among panels will become no less important. Foundation has the pole position, but NGS competitors like Caris and Tempus also have TMB offerings. More TMB tests will continue to emerge from other NGS developers (e.g., Illumina) and lab-developed tests at large cancer centers will remain the standard there (e.g., MSK-IMPACT at Memorial Sloan-Kettering). In anticipation of the evolving complexity, Friends of Cancer Research is already leading an effort to harmonize performance standards across the various panels9, but this is a nontrivial undertaking.
Furthermore, the cost of NGS is still high and associated with significant reimbursement hurdles10. This may lead to preferred contracts at certain hospitals and practices with specific NGS tests, but given the complexity of TMB, physicians in those settings may be more hesitant to substitute other NGS tests for Foundation’s CDx, potentially limiting the uptake of testing. In addition, the turnaround time of approximately 2+ weeks for obtaining test results11 makes it impractical or undesirable for many patients.
A more fundamental concern is not unique to TMB: all available I-O biomarkers leave a lot to be desired in their predictive power. As seen before for PD-L1 and MSI, there are TMB-high patients who do not respond to ICI, and TMB-low patients who do. A combinatorial approach to biomarker testing may be required to optimize patient selection for ICI (more on that below), but that may take considerable time to develop, and until then, test uptake may only be half-hearted.
Given these caveats, the TMB-H approval could therefore lead to merely incremental growth for Keytruda, at least in the near term.
A potential vulnerability
The tumor type with the highest global incidence and mortality is NSCLC, and this is also the market driving the largest portion of Keytruda sales (about 2/3 of total sales in 2019)2. In first-line treatment of NSCLC (with wild-type EGFR and ALK), Keytruda was the first ICI approved, and it had a series of successful pivotal trials that supported its approval as a monotherapy for patients with tumors with PD-L1 expression ≥1%, and in combination with chemotherapy for patients regardless of PD-L1 status (for both squamous and non-squamous NSCLC)4. Although other immunotherapy regimens are now also approved for first-line treatment of NSCLC, Keytruda is clearly the dominant ICI.
Competitors are naturally looking for patient segments “won” by Keytruda to reclaim. NSCLC with low TMB had previously been shown to respond poorly to ICI therapy12. Could Keytruda (and other members of its class) be excluded from TMB-low NSCLC? If so, the bar to beat for novel therapies (including novel immune mechanisms) would be lowered to that of older standards of care like platinum chemotherapy. Payers would likely support an exclusion of Keytruda from patient segments where it does not deliver meaningful value, given the high cost.
However, TMB alone seems unlikely to limit Keytruda use. At the World Conference on Lung Cancer 2019, Merck emphasized through multiple talks that in NSCLC, Keytruda + chemotherapy shows better efficacy than chemotherapy alone in both TMB-low patients and in TMB-high cohorts13.
TMB in conjunction with PD-L1 expression may be more predictive than either alone, as they are independent markers associated with ICI efficacy. Other biomarkers such as STK11 and KEAP1 mutations are also associated with poor ICI efficacy. Is it possible that there is a segment defined by an overlay of multiple immune biomarkers (potentially including low PD-L1 and low TMB and other markers) where Keytruda would be contraindicated due to lack of efficacy? If so, and if this segment is still sizable (e.g., ≥10% NSCLC), this may be an attractive segment for novel therapeutics development.
Of course, generating the evidence for this overlay of multiple biomarkers is non-trivial and will take years, and identifying novel mechanisms that will succeed in this ICI-primary resistant setting will also be challenging. However, Merck may have inadvertently contributed to this endeavor by setting a standard for both PD-L1 (in its multiple NSCLC approvals) and TMB measurement and threshold definition.
In conclusion, Keytruda’s approval for TMB-H cancers is an important milestone, establishing a regulatory path, a reference for diagnostics development, and proof of principle of clinical utility for TMB. It reaffirms Merck’s leadership in immuno-oncology, and in the precision medicine of I-O specifically. It remains a double-edged sword, however, as it may provide a starting point for competitors to exclude Keytruda from patient subsegments. In the end, will this approval be in Merck’s favor?
The likely answer is “yes”.
The upside is high, and the testing challenges are at least partially surmountable, while the hurdle for competitors to reclaim biomarker-defined patient segments back from Keytruda is demanding.
If the opportunity to seize on the potential vulnerability motivates companies to develop more sophisticated I-O biomarker panels, this advance will benefit the entire field. Thus, regarding whether this approval is of benefit to patients with cancer, the answer is unequivocally “yes”.
Perhaps Merck itself will develop this opportunity for novel I-O agents or combinations, though the path is unclear, and competition will be fierce. Merck’s leadership and innovation in pursuing the tissue-agnostic TMB-H indication is all the more commendable because they took on some risk to advance the field and do the right thing for patients.
1 FDA; the 7 agents are Yervoy, Keytruda, Opdivo, Tecentriq, Imfinzi, Bavencio, and Libtayo
2 Evaluate Pharma (accessed August 2020)
3 Merck & Co. 2019 Annual Report
4 Keytruda product label
5 E.g., CheckMate-227, which began enrolling in Aug. 2015 (Hellman 2018 NEJM. 378:2093)
11 Foundation Medicine Patient FAQs
12 ASCO 2018 Annual Meeting, abstracts 12000, 12001; Carbone 2017 NEJM. 376:2415
13 WCLC 2019 Meeting Abstracts (1) OA04.05; (2) OA04.06