December 17, 2015: Why immuno-oncology drug developers need a coherent payer-facing strategy for patient stratification
Current market context: The last 15 months have seen the emergence of a compelling class of immuno-oncology therapies targeting the PD(L)-1checkpoints. Clinical data have been compelling, suggesting potentially durable effects in a subset of patients with one of several highly immunogenic tumour types. Clinical success has been mirrored by a tsunami of investment, amidst suggestions from some analysts that immune therapies may come to represent 60% on oncology background therapy within 10 years. While early enthusiasm is certainly justified, it is important to note that up to 80% of patients with the most immunogenic tumour types do not respond in many trials and that the currently approved immunogenic settings (lung and melanoma) may represent the low hanging fruit in terms of monotherapy options. The next wave of success will almost certainly be associated with carefully timed therapy combinations informed by biomarker-defined patient selection. We are already starting to see some early examples of associated operational challenges associated with patient selection and some highly informative recent payer reactions, both of which will be discussed in the remainder of this update.
Clinical data – mixed signals on need for stratification: Pivotal clinical trial data generated in checkpoint inhibitor studies have resulted in a heterogeneous role for PD-L1 based patient selection in the clinic, despite the common drug target for the two currently approved therapies. BMS’ Opdivo therapy was first approved by the FDA in a broad (all-comers) advanced melanoma population in late 2014 and subsequently approved in a broad squamous NSCLC population in early 2015. A further approval followed in October 2015 for an unstratified NSCLC population based on an OS increase of nearly 3 months, but FDA noted an enriched response in PD-L1 positive patients and proposed a potential role for PD-L1 testing as a complementary (non-mandated) enrichment tool. Additional Opdivo approvals in renal cell carcinoma and in melanoma (with Yervoy) have followed, also in unstratified populations. Meanwhile, Merck’s Keytruda, which was first approved by FDA in a broad (all-comers) advanced melanoma population in late 2014, was subsequently approved in a stratified NSCLC population (with a PD-L1 companion diagnostic) in October 2015. Accordingly, as of time of writing (December 17, 2015), two approved PD-L1 targeting therapies are available, with very mixed regulatory messages on the benefits of patient stratification. Additional PD(L)-1targeting therapies from Roche and Astra Zeneca are in late development, but associated patient selection requirements have yet to be clarified by regulators. It should be noted, however, that both companies have a sophisticated approach to patient stratification across their oncology portfolio. As a consequence of the foregoing, as noted by others, the current regulatory and evidentiary landscape is rather heterogeneous in terms of patient stratification for immuno-oncology therapy. It is important to note, however, that the response of regulators will not necessarily be universally reflected in corresponding payer coverage in the same populations, as evidenced by events of the last 24 hours, highlighted later in this feature.
Stratification – the companion/complementary diagnostic assay landscape:
One key source of the variability in the clinical landscape is associated with assay and protocol variability. In particular, assays for the PD(L)-1 target are highly heterogeneous in terms of cells targetted, antibody reagents used, and cut-off thresholds for positivity. A recent report (November, 2015) in JAMA Oncology describes a resulting 25% discordance across assay types in the lung setting. This ambiguity renders clinical decision making highly challenging. Several initiatives are underway to address such disparities. Within the US, these challenges are being addressed by FDA and 4 sponsors via the “blueprint” initiative. NCCN has also launched a clinical study with BMS to address assay comparability, while the International Association for the Study of Lung Cancer has launched a project to develop a PD-L1 atlas with similar objectives. Reported PD(L)-1 status may also be misleading, even in a world where assay protocols are ultimately harmonized. Specifically, it is important to remember that the level of this adaptive immunological resistance biomarker is known to evolve under treatment and that an assay result based on archived tissue from a previous line setting may lack contemporary relevance. For example, in the pivotal BMS Checkmate-017 trial which led to approval for Opdivo in squamous NSCLC, the sponsor noted the absence of a significant subgroup effect. However, the PD-L1 biomarker status recorded in this trial was based on a mix of fresh and archived tissue and therefore any inference relating to status at time of intervention should be viewed with caution. It is plausible that a stratified response would be seen in this setting (as in non-squamous NSCLC), but that the effect is hidden by the trial design. Hence, we should perhaps be careful about concluding that these results, and perhaps also data from other all-comers indications, suggest limited benefits from stratification. Hence, there may be a broader role for PD(L)-1 testing than is currently apparent. Even if this does prove to be the case, however, such imperfect single marker testing will only be the beginning for this drug class. Indeed, as I argued in a recent editorial in the Personalized Medicine journal, we are likely to see the emergence of a host of more informative multi-parametric dynamic predictive and monitoring biomarkers based on a plethora of technologies including NGS, PCR, flow cytometry and Elispot. Such multi-parametric assessment offers the opportunity for broader decision support and perhaps higher Positive Predictive Value testing. Within this broader group of technologies, as with the current IHC-based approaches, there will likely also be a need for standardization initiatives such as the new PrecisionFDA initiative to develop NGS standards. It is rapidly becoming clear within the immuno-oncology field, as with targeted therapy before it, that developers need a strategy to dynamically chase the genome to track adaptive therapy resistance and immunogenicity using validated test platforms.
Payors perspective and future adoption:
There is increasing evidence that international payers may reflect the above complexity in their coverage decisions. However, within the US, broad adoption continues in all-comers NSCLC and melanoma populations, with only a minority of NSCLC patients (reportedly 22%, according to a leading analyst) being tested, although a higher proportion are tested in melanoma (where there are no related label requirements). As also noted by the same analyst, PD-L1 testing does not appear to be a treatment-gating factor for the current on-label indications in the US. Within the UK, NICE has already issued favourable coverage for Keytruda in an unstratified melanoma population, and is currently preparing guidance for Opdivo in a similar population. In the squamous NSCLC subset, however, NICE has just (December 16, 2015) released an appraisal consultation document with a preliminary recommendation that nivolumab not be used in the UK NHS in the “all-comers” squamous NSCLC population previously cleared for marketing authorization by FDA and EMA. In its analysis, the NICE Technology Appraisal Committee deemed that the drug, while clinically effective, did not meet cost effectiveness thresholds and also commented that, as noted earlier, PD-L1 biomarker data should be considered with caution since much of the data was obtained from archived tissue and therefore not contemporaneous with treatment. This preliminary decision may portend a limit to the all-comers adoption scenario in some settings and may suggest an increasing payer focus on patient stratification in immuno-oncology, especially in the ever-growing settings where cost-effectiveness criteria are important. In this scenario, diagnostics such as PD(L)-1 and NGS-based testing will play a key determinative role as new monotherapies and combinations emerge in diverse clinical settings. The author firmly believes that, current ambiguity notwithstanding, diagnostic stratification competency will emerge as a core differentiator for successful developers of immuno-oncology medicines.
Iain D. Miller, Ph.D.
Healthcare Strategies Group