Whether by luck or design Bristol-Myers Squibb had the advantage of a broad Opdivo label in lung cancer, but today this was seemingly swept away by Asco data suggesting that the drug might, after all, not work in PD-L1-negative tumours, narrowing its target population.
Bristol lost some $7bn of market cap, although patients’ PD-L1 status is still viewed as a poor biomarker. The quest for predictors of response to immunotherapy is one of Asco’s hottest themes, and will only grow in importance in light of possible payer pressure. Ironically, it was a tiny investigator-sponsored study that today marked a striking advance in finding a more reliable biomarker.
This trial, presented as an Asco late-breaker, looked specifically at the predictive power of defects in mismatch repair (MR), a process that corrects mismatches generated during DNA replication but which in some tumours is impaired. It concerned Merck & Co’s rival PD-1 agent, Keytruda.
The most impressive result was from a comparison of two cohorts of colorectal cancer patients: those with MR-deficient cancer had 62% objective response and 92% disease control rates, versus zero and 16% respectively for those with MR-proficient cancer.
Dr Dung Le, from Johns Hopkins University, said tumours that harbour MR deficiency had far higher rates of genetic mutations than those without, equating to an antigenic environment that could be hit efficiently with immunotherapy. Moreover, MR status can be assayed cheaply and easily, she said, and testing is already part of standard practice for this cancer.
The study could thus show a way in to treating colorectal cancer – so far one of the tougher tumour types to target with immunotherapy – with anti-PD-1/PD-L1 therapy, though it is surely too early to say how relevant this biomarker is for others.
New biomarkers have to be found because the transient nature of PD-L1 expression on tumours makes PD-L1 itself a relatively poor marker. That said, with full disclosure at Asco today of Bristol’s Checkmate 057 data, testing Opdivo in second-line non-squamous lung cancer, a patient’s PD-L1 status suddenly became relevant again.
This is because ‘057 suggested a survival benefit for Opdivo versus docetaxel in tumours expressing PD-L1 at over 1%, 5% and 10%, but not if the data are cut by expression below these cutoffs. The possibility is that Opdivo’s target market could be limited to PD-L1-positives in non-squamous disease.
This is in stark contrast to the Checkmate 017 squamous NSCLC trial, on which Opdivo got lung cancer approval and which at an Asco update showed a continued effect in PD-L1-negatives, hazard ratios for survival favouring Opdivo regardless of PD-L1 expression.
While Bristol stock crashed in afternoon trading Merck and Roche were both up 2%. Evercore ISI’s Mark Schoenebaum said ‘057 opened the door to competitors who are a little further behind in this setting but are also demonstrating activity in tumours that highly express PD-L1.
Lung, of course, is the far bigger and more lucrative indication for these novel checkpoint inhibitors, after the fast approvals of Merck & Co’s Keytruda and Bristol’s Opdivo in melanoma. Also disappointing for Bristol investors was that in ‘057 Opdivo’s survival advantage was a meagre 2.8 months, with 27% reduction in risk of death (p=0.0015).
Until now Bristol had the upper hand, since Keytruda has been filed for lung cancer on the strength of Keynote-001, a study that had enrolled only PD-L1-positives.
Roche also was in something of a pickle: it had recruited all-comers into its all-important 287-patient Poplar study of MPDL3280A, but according to its Asco abstract this showed a clear preference for PD-L1-positives, with a 0.47 hazard ratio for high PD-L1 expressers but a highly unimpressive 1.22 in non-expressers.
The overall survival was 11.4 months for MPDL3280A versus 9.5 months for docetaxel, perhaps also not hugely impressive. Some inconsistency within the PD-L1 biomarker data, as well as against the earlier Fir trial, has perplexed analysts, though most agree that based on Poplar a filing is warranted, albeit with a restrictive label for PD-L1-positives.
The relevance of PD-L1 status also haunts earlier-stage competitors. Pfizer, presenting its first data on the Merck KGaA-licensed avelumab, showed a 14.4% response rate in NSCLC, including 10.0% response in PD-L1-negatives. A monotherapy NSCLC trial of Astra’s MEDI4736 did not fully split out PD-L1 status, but 23% response in PD-L1-positives seemed to drive the 14% overall rate.
A vital fact to remember is that assays across the different studies as far from standard, and definitions of PD-L1 positivity vary.
And beyond Dr Le’s fascinating MR-focused trial, work on other new biomarkers continues; researchers presenting a separate study of Keytruda in head and neck cancer suggested the possibility of using an interferon-gamma signature method as a promising predictor of response. Merck yesterday struck a deal with Nanostring Technologies to work on assays of gene expression signatures.
This suggests that the issue of PD-L1 positivity might not remain relevant for too long, and combination therapies will in any case contribute to dialling up response by producing an immunogenic tumour environment much more susceptible to immunotherapy.
For now, however, an incremental shift in perception has allowed Merck to turn the tables on Bristol.