BMY/MRK: Why CM227 is very good news for BMY, and not such bad news for MRK

Print Friendly, PDF & Email


Richard Evans / Scott Hinds

203.901.1631 /.1632

revans@ /


February 8, 2018

BMY/MRK: Why CM227 is very good news for BMY, and not such bad news for MRK

  • Checkmate 227 (CM227) does not pit BMY’s Opdivo+Yervoy against MRK’s Keytruda; rather CM227 pits Opdivo+Yervoy against standard chemotherapy. Opdivo + Yervoy wins this comparison in patients with high tumor mutational burden (TMB), regardless of PD-L1 expression level. The result is far more likely to expand the use of immuno-oncology (IO) therapies in NSCLC than to result in dislocation of one IO regimen by another
  • TMB and PD-L1 are similarly indicative of response to (IO) treatment; i.e. patients with high TMB tend to respond IO just as well as patients with strong PD-L1 expression. However, these markers tend to select different patients, i.e. not all high TMB patients will have strong PD-L1 expression, and vice versa. As a final point, patients with both high TMB and strong PD-L1 expression tend to respond to IO better than patients scoring highly on either biomarker alone
  • Patients with high TMB but weak PD-L1 expression (about 20% of 1st-line NSCLC) will almost surely go to Opdivo + Yervoy, instead of traditional chemotherapy; this expands the role of IO in NSCLC, without taking any share away from Keytruda
  • Patients with strong PD-L1 expression (about 20% of 1st-line NSCLC) are likely to continue going to Keytruda. CM227 may result in the addition of Yervoy (to Keytruda) for the 7% of 1st-line NSCLC patients with both strong PD-L1 expression and high TMB
  • Under current treatment algorithms, Keytruda gets about 20% of NSCLC patients (strong PD-L1 expressors), plus some undetermined share of the 48% of patients who are weak PD-L1 expressors and whose tumors have non-squamous histology. CM227 does nothing to weaken Keytruda’s grip on the 20% of 1st-line NSCLC patients who are strong PD-L1 expressors. From Keytruda’s perspective the only effect of CM227 is to reduce the percentage of 1st-line NSCLC patients (from which Keytruda captures an undetermined share) who are weak PD-L1 expressors and whose tumors have non-squamous histology to 32% (from 48%)
  • Opdivo and Yervoy get no 1st-line NSCLC patients under the current treatment algorithm. Under CM227 both Opdivo and Yervoy should get 20% share of 1st-line NSCLC; specifically, those patients who have high TMB but weak PD-L1 expression. If detailed results from CM227 show that Yervoy adds efficacy in patients who have high TMB and strong PD-L1 expression, Yervoy should pick up an additional 7% patient share, for 27% total


Current treatment for 1st-line non-small-cell lung cancer (NSCLC) follows the algorithm in Exhibit 1. Patients whose tumors carry a driver mutation conferring susceptibility to a specific therapy receive that targeted therapy, and account for roughly one-fifth of patients. The remaining 80 percent of patients are then sorted by PD-L1 expression level. Those with expression levels above 50% account for roughly 25% of patients without driver mutations and are given Keytruda. Those with PD-L1 expression below 50% are assessed for histology (squamous v. non-squamous); patients whose tumors exhibit squamous histology are treated with platinum-based chemotherapy, and those whose tumors exhibit non-squamous histology are treated either with platinum-based chemotherapy (with or without Avastin), or carboplatin + pemetrexed + Keytruda. The net share of patients to Keytruda is the 20% of total NSCLC patients with PD-L1 expression levels above 50%, plus some undetermined share of the 48% of patients with PD-L1 expression levels below 50% and non-squamous histology

BMY’s Checkmate-227 (CM227) trial has shown that Opdivo + Yervoy is superior (in terms of progression free survival or ‘PFS’) to platinum-based chemotherapy (selected for tumor histology) in patients with high tumor mutational burden (high TMB), regardless of PD-L1 expression level, with high TMB being defined as 10 or more mutations per million nucleotides (mut/Mb)

Momentarily setting aside the question of whether BMY can get Opdivo / Yervoy approved for this subset of patients, the CM227 result begs the question of the relative roles of PD-L1 expression level and TMB in the NSCLC treatment algorithm

As a biomarker TMB:

1) appears to be similarly powerful as compared to PD-L1 (patients selected by high TMB have responses similar to those in patients selected by strong PD-L1);

2) appears to select a patient population that differs from but overlaps with the patient population selected by PD-L1 alone; and

3) when combined with PD-L1 identifies a subset of patients with even greater responses (patients with both strong PD-L1 and high TMB appear to have better responses than patients that measure high on one marker but not the other)

These characteristics of TMB as a biomarker are reflected in the following trials involving the approved PD1 inhibitors Keytruda, Opdivo, and Tecentriq, respectively

In a sub-analysis[1] of the Keytruda trial Keynote-001, authors disclosed patient-specific PD-L1, TMB, and outcomes data for 34 1st-line NSCLC patients. For the trial, patients were sorted into PD-L1 expression levels of negative (<1%), weak (1% to 49%), and high (>= 50%); using the patient specific data we further sorted patients into tumor mutational burden (TMB) tertiles with cut points of 254 and 425. Progression free survival (PFS) improved at higher PD-L1 levels, and at higher TMB levels, with duration of PFS being similar in the high PD-L1 and high TMB groups (Exhibit 2). 10/34 patients had strong PD-L1 as compared to 11/34 with high TMB; 5 patients had both strong PD-L1 and high TMB (Exhibit 3). Patients with high levels of both PD-L1 and TMB had the longest average PFS duration (Exhibit 4)

In the Opdivo trial Checkmate-026[2], authors disclosed objective response rates by combined PD-L1 / TMB status in 1st-line NSCLC patients receiving Opdivo (Exhibit 5). Patients with high levels on one biomarker but not the other had similar objective response rates (ORR’s; 32% for high TMB, weak PD-L1; 34% for lo/med TMB, strong PD-L1), and patients with high levels of both biomarkers had the highest ORR’s (75%)

In a sub-analysis[3] of the Tecentriq trials POPLAR and OAK (2nd-line NSCLC), authors showed that patients with high TMB[4] or strong PD-L1[5] had similar hazard ratios for PFS and overall survival (OS), and that patients with high levels of both TMB and PD-L1 had even lower hazard ratios than patients selected on the basis of either biomarker alone (Exhibit 6). Of the 229 biomarker evaluable patients, 126 (55%) had high TMB, 73 (32%) had strong PD-L1, and 30 (13%) had both high TMB and strong PD-L1

As a worst case, most conservative scenario we would expect TMB to fit into the current NSCLC algorithm just below PD-L1. If we structure the decision as a binary choice between PD-L1 and TMB, the clinical evidence in favor of PD-L1 expression level currently is stronger, being based on OS rather than PFS, and having been established as a labelled (and thus insured) indication. However TMB should be prioritized over histology (i.e. squamous v. non-squamous), since CM227 included patients with both histologies, and the strong signal from BMY’s releases and commentary is that efficacy (relative the current platinum-based standards of care) has been found in both subtypes

Placing TMB below PD-L1 expression and before tumor histology results in the algorithm shown in Exhibit 7. As before, patients with targetable driver mutations go to their appropriate targeted therapies, and the remainder are screened for PD-L1. Patients with PD-L1 expression of 50% or greater go to Keytruda; patients with PD-L1 expression below 50% are screened for TMB. High TMB patients are treated with Opdivo / Yervoy; low TMB patients are assessed for tumor histology, and treated with the appropriate platinum-based regimen for their histology

We assume that one-third of patients (with no driver mutation and PD-L1 below 50%) screened for TMB will be grouped as high TMB. There is as yet no standard cutoff for TMB; however we do know that BMY is using a cutoff of 10 mut/Mb for CM227, that about 17% of NSCLC patients will have TMB >= 20 mut / Mb[6], and that the top tertile of TMB expression has been an effective definition of ‘high’ in both the Keynote-001 and CM026 analyses mentioned earlier (Exhibits 2 thru 5)

Note that by prioritizing PD-L1 expression over TMB level, this algorithm automatically gives Keytruda any high TMB patients that have PD-L1 expression levels at or above 50% — i.e. Keytruda wins the jump ball for the strong PDL1 / high TMB patients. This algorithm results in Opdivo / Yervoy capturing 20% of newly diagnosed NSCLC patients; Keytruda still captures the 20% of newly diagnosed patients without driver mutations and with strong PD-L1, plus an undetermined share of the (now 32%, down from 48%) of patients with weak PD-L1, low TMB, and non-squamous histology

A further scenario is feasible if CM227 shows that Opdivo/Yervoy is superior to Opdivo alone in patients with high levels of both biomarkers. This outcome is possible based on the design of the trial, but we can’t handicap this scenario without the detailed results from CM227. Of the 1,700 patients in Part 1 of CM227, 1,200 are in Part 1a, in which patients are randomized 1:1:1 to Opdivo/Yervoy, Opdivo, or chemo. Part 1a sets the stage for a comparison of Opdivo/Yervoy v. Opdivo in all patients, including those with both strong PD-L1 and high TMB. If CM227 shows that Opdivo/Yervoy is more effective than Opdivo alone in patients with high levels of both biomarkers, this would strongly imply that adding a CTLA4 (Yervoy) to a PD1 (in this case Opdivo) adds efficacy for these patients. This would force clinicians to choose between PD1 only (presumably Keytruda) or PD1/Yervoy for patients with high levels of both biomarkers – and we believe many would want to add Yervoy for these patients. A CM227 finding that Opdivo/Yervoy is superior to Opdivo alone in patients with both strong PD-L1 and high TMB should result in the treatment algorithm shown as Exhibit 8, which differs from Exhibit 7 only in that patients with high levels of both biomarkers wind up on Keytruda + Yervoy

If Opdivo/Yervoy proves superior to Opdivo alone in combined strong PD-L1 / high TMB patients, we would expect many patients with strong PD-L1 / high TMB to get Keytruda + Yervoy rather than Opdivo + Yervoy, simply because Keytruda has the stronger (OS) evidence and labelled indication for strong PD-L1. It’s at least possible that BMY could encourage the use of Opdivo + Yervoy over Keytruda + Yervoy in these patients by modifying the packaging and pricing of Yervoy. Yervoy is currently available in 5 mg/ml strength in either 10ml (50mg) or 40ml (200mg) single-use vials. Dosing for Yervoy’s current indications is either 3 mg/kg (unresectable or metastatic melanoma) or 10 mg/kg (adjuvant melanoma); however dosing for Yervoy in CM227 was only 1mg/kg. BMY could offer a new NDC (or series of NDC’s) that co-packages Opdivo and Yervoy in the unique 3:1 Opdivo:Yervoy milligram strength ratio under development for NSCLC. If these co-packaged NDC’s were discounted relative to the combined prices of the two medications bought separately (and of Keytruda + Yervoy bought separately), this might encourage greater use of Opdivo + Yervoy, rather than Keytruda + Yervoy, in patients with high levels of both PD-L1 and TMB. Exhibit 9 summarizes share of 1st-line NSCLC patients by brand under the scenarios we’ve outlined. If we set aside any share impacts from co-packaging and discounting of Opdivo + Yervoy, then it’s clear from Exhibit 9 that CM227 has no impact on Keytruda’s share of strong PD-L1 patients. Keytruda is affected in that the share of patients who are candidates for either platinum-based chemo +/- Avastin or carboplatin + pemetrexed + Keytruda falls from 48% to 32% of 1st-line NSCLC patients – however having no clear sense of how many of these patients are going to Keytruda, it’s unclear just how significant of an impact this is on Keytruda’s overall NSCLC volume. The main impact of CM227 is to expand the use of immuno-oncology (IO) agents, namely Opdivo and Yervoy, in 1st-line NSCLC

Now to the regulatory controversy: the late inclusion of TMB as a biomarker in CM227 has raised concerns regarding the approvability of Opdivo + Yervoy in high TMB patients. If we assume, as BMY states, that the late inclusion was undertaken with FDA’s involvement, and that the inclusion preceded the un-blinding of outcomes data, then we can reasonably assume that the comparison of Opdivo + Yervoy v. chemo is valid from both statistical and regulatory viewpoints. This leaves the questions of the strength of the resulting evidence, and where that evidence fits into current knowledge and treatment patterns. Outcomes in CM227 were measured in terms of PFS, which can be cleaner (e.g. no crossover effect) but is certainly less definitive than OS. However PFS clearly is used by FDA as a basis for approval (e.g. Keytruda’s approval in combination with pemetrexed and carboplatin for weak PD-L1 NSCLC patients with non-squamous histology); and, PFS is clearly more substantive than other measures on which FDA has based approvals in the past, such as ORR (e.g. the MSI-H approval for Keytruda)

We believe the degree of skepticism regarding the impact of CM227 may flow from a mis-framing of the regulatory questions that flow from the trial’s result. In the context of CM227 the regulatory hurdle for Opdivo + Yervoy is not whether the combination’s PFS outcomes are more powerful than Keytruda’s OS-based efficacy in strong PD-L1 patients – CM227 is completely silent on this. Rather the question is whether the Opdivo + Yervoy combination managed to beat histology appropriate platinum-based chemo – which it did. Reasonably assuming the trial was modified in a manner that is scientifically / statistically appropriate and acceptable to FDA, there’s no reason to assume the combination won’t be approved for 1st-line NSCLC patients with high TMB. The clinical question of which high TMB patients get Opdivo + Yervoy is straightforward in the case of lo PD-L1 / high TMB, where Opdivo + Yervoy is clearly indicated over the current standard of histology-based chemotherapy. In the case of strong PD-L1 / high TMB, we suspect the path of least resistance is to stick with Keytruda, based on the strength of Keytruda’s evidence in strong PD-L1 patients. This reduces the strong PD-L1 question to one of whether or not to add Yervoy to Keytruda in strong PD-L1; we believe this is likely if Yervoy is proven to add efficacy in the strong PD-L1 / high TMB subset of CM227 (Part 1a)


  1. Rizvi et al., “Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer”, SCIENCE, 3 April 2015, Vol 348, Issue 6230, pp 124-128 
  2. Carbone et al., “First-Line Nivolumab in Stage IV or Recurrent Non-Small-Cell Lung Cancer”, NEJM; June 22, 2017; Vol 376; No.25; pp. 2415-2426 
  3. Gandara DR et al., Ann Oncol 2017;28(suppl 5);Abstr 12950 
  4. Defined in this instance as >= 16 mutations / Mb 
  5. Defined in this instance as being in the highest tertile of PD-L1 expression for both tumor cells (TC) and tumor-infiltrating immune (IC) cells 
  6. Chalmers et al., “Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden”, Genome Medicine (2017) 9:34 


©2018, SSR, LLC, 225 High Ridge Rd, 2nd Floor, Stamford, CT 06905. All rights reserved. The information contained in this report has been obtained from sources believed to be reliable, and its accuracy and completeness is not guaranteed. No representation or warranty, express or implied, is made as to the fairness, accuracy, completeness or correctness of the information and opinions contained herein. The views and other information provided are subject to change without notice. This report is issued without regard to the specific investment objectives, financial situation or particular needs of any specific recipient and is not construed as a solicitation or an offer to buy or sell any securities or related financial instruments. Past performance is not necessarily a guide to future results. In the past 12 months, through a wholly-owned subsidiary SSR Health LLC has provided paid advisory services to BioPharmX (BPMX), Pfizer Inc (PFE), Gilead Sciences (GILD), Bristol-Myers Squibb (BMY) and Sanofi (SNY) on both securities-related and non-securities-related topics. One or more of SSR Health’s analysts owns long positions in the following stocks: ACOR, AGEN, AGIO, AKAO, ALKS, ALNY, ANAB, ARRY, BMY, DERM, DOVA, DPLO, ESALY, FOLD, GWPH, INCY, IONS, KALA, LOXO, NSTG, PFSCF, PGNX, PRTK, PTLA, RARE, RHHBY, RIGL, TBPH, THERF, TRVN, TSRO, TTPH, TXMD, VRTX, VSTM

Print Friendly, PDF & Email