Leveraging Diagnostics to Realize the Full Potential of Precision Oncology

As we approach the twentieth anniversary of the approval of Herceptin, the field of precision oncology has never been more exciting or dynamic, and utility seems to be broadly recognized. Precision medicines have accounted for over 20% of new molecular entities approved by FDA in each of the last 3 years, per the Personalized Medicine Coalition, and biomarker-based stratification is currently utilized in 30% of late-phase trials. Indeed, a key 2016 report sponsored by BIO quantified an increase from 8.4% to 25.9% in likelihood of successful transition from Phase I to ultimate FDA approval resulting from a targeted development approach. New tools such as ctDNA (“liquid biopsy”) seem set to enable a new era of therapy prediction and monitoring.  The recent $1B financing of liquid biopsy leader Grail, including investments by J&J, Merck and Celgene, is just the latest example of the mainstreaming of precision drug development and early detection. 2016 also witnessed a compelling supportive message from the investment community, as BMS saw its stock price plummet by 20% in response to the failure of its Opdivo Checkmate 026 trial to meet its endpoints in 1st line NSCLC, a failure attributed by many to be associated with lack of adequate patient stratification. Indeed, competitor Merck, which adopted a more aggressive stratification approach for its Keytruda therapy in the same setting, saw a stock price gain of approximately 16% on the same day, in August 2016, based on its previously announced Keytruda success in the Keynote 24 trial. Clearly, financial markets are now awake to the benefits of precision targeting, and no longer expecting “all-comers” marketing strategies from pharmaceutical companies.

However, the precision medicine field is in many ways less advanced than we might expect nearly 20 years after the heady early days of Herceptin. Notwithstanding a few notable exceptions in the infectious disease, rare genetic disease and autoimmune fields, precision medicine in 2017 is still largely confined to about 10 single gene “rule-out” tests for a select set of targeted medicines in the field of oncology, together with a small number of multi-gene prognostic tools such as Oncotype. Under typical current oncology scenarios, simple diagnostic technologies are used to interrogate the expression or mutation status of the drug target, while not accounting for the broader biology which informs durable response. Absence of the driving target mutation or associated expression for such “rule-out” tests is generally associated with non-response, while presence is associated with a degree of response enrichment, typically around 25%. The result is often a strong but short-lived response in the biomarker-positive patient subset, as new resistance clones and other biological effects come to dominate. The current generation of immuno-oncology therapies appear to offer the potential for more durable response, but, as with the early wave of targeted therapy, patient selection in the clinic has yet to move much beyond limited target-expression based testing in select clinical settings. Utility of precision medicine in the clinic also lags significantly behind its potential. For example, in recent next generation sequencing-based precision programs described by City of Hope and MD Anderson, potentially actionable results were found in 39-95% of cancer patients but only 5-16% of said patients were ultimately matched to genomically guided therapies. Several (non-biological) factors were cited to account for this significant gap, including patient compliance, provider education, data capture, and lack of decision support. Further, as noted by one recent commentator, only about 30% of patients with diverse, relapsed cancers who receive therapy respond, with a rather modest median progression-free survival of 5.7 months, reflecting a broader unmet need associated with understanding of tumour biology. Accordingly, the median PFS benefit in the 5-16% of cancer patients ultimately getting precision therapy may be as low as 2 months.

As noted in the foregoing, the gap between potential and actual patient benefit can largely be attributed to a mix of operational and biological factors. The remainder of this article will consider the biology of patient selection, with a focus on immuno-oncology.

It is evident that we need to leverage a better understanding of the underlying biology to evolve the stratified medicine we have today into a truly more personalized medicine. As a group of leading MD Anderson researchers recently noted in relation to immuno-oncology, “checkpoint responder/non-responder identification is a complex picture likely involving the interplay of tumor genomic characteristics, tumor modulation of the local microenvironment, and the extent of immune surveillance in the tumor milieu at the time of initiation of therapy”. Current testing of PD-L1 status, itself contentious and operationally challenged, is only the beginning of the diagnostic odyssey. Ultimately, we will need to leverage tools addressing each of the genomic and immunological components to select an optimal basket of therapies for cancer patients to address biology much more comprehensively than at the target-specific or checkpoint induction level. It is illustrative here to consider a subset of some of the emerging pre-eminent approaches currently under study. These include profiling markers of genomic instability such as tumour mutational burden (TMB), DNA repair and microsatellite instability (MSI), together with immunological parameters relating to cell clonality and infiltration.

Tumour mutational burden (TMB), a biomarker of genomic instability, has now been shown in several studies to be associated with checkpoint inhibitor response. Mechanistically, this appears to be primarily associated with antigenicity, via the generation of a diverse set of immune-stimulating neoantigens in high-TMB patients. Various authors have noted the generally higher TMB at the group level in frontline immuno-oncology settings such as NSCLC and melanoma vs less immunogenic tumour types, but TMB assays may find greatest utility in the identification of subgroups within tumour types with generally lower immunogenicity. Ultimately, they may manifest as part of the informational output from large gene panels conducted in tissue (e.g. FoundationONE) or blood (e.g. Guardant360, FoundationACT), or smaller gene panels assayed in the ctDNA setting, such as those reportedly being developed by PGDx and others. Foundation Medicine, for example, defines TMB-high patients as being those with greater than 20 mutations/Mb. Other markers of genomic instability include homologous recombination and mismatch repair/microsatellite instability deficiencies.

Immunological parameters such as human leukocyte antigen (HLA) typing will be an important tool for immuno-oncology response prediction. The HLA region, which encodes for the major histocompatibility complex, represents the most polymorphic region of the genome, which underpins the fine tuning of the adaptive immune system. Previously best known for its utility in stem cell transplantation and abacavir HIV therapy hypersensitivity testing, HLA testing has yet to demonstrate significant utility in immuno-oncology. However, the important role of HLA-encoded Major Histocompability Complex  class I proteins in the presentation of neoantigens to immune cells has the potential to play an important role in both checkpoint and T cell receptor therapy solution.

The inherent diversity of B and T cells will also likely prove to be an important determinant of response. In this context, T-cell clonality assays are being used to tease apart the diversity and antigenic specificity of T-cells by assessing the amino acid sequence of the V, D and J segments of the hypervariable complementarity-determining region 3 (CDR3) of the T-cell receptor. One leader in this space is Adaptive Biotechnologies, although multiple other labs are offering related services. Potential applications include both tracking of minimal residual disease clones and monitoring of specific infused therapeutic T cell populations in the context of Chimeric Antigen Receptor T cell immunotherapy, such as CD19-recognizing CART T cell populations in B cell malignancies.

A further immunological consideration and potential biomarker is the intra-tumour population of Tumour-Infiltrating Lymphocytes (TILs).  The density of cytotoxic CD8+ TILs can potentially serve as a biomarker of immunological activation, as shown in several checkpoint inhibitor trials. Indeed, TIL profiling is a core component of the conceptual “Cancer Immunogram” proposed testing framework recently introduced by leading researchers from the Netherlands Cancer Institute and UCLA to assess T cell recognition, activation and infiltration. As the Cancer Immunogram hypothesis posits, it will likely be necessary to combine some or all the above genomic instability and immunological interrogation techniques to profile the therapeutic options for oncology patients.

Emergence and adoption of the foregoing complex testing strategies will present a significant “big data” challenge, both in early development and in the clinic. Biomedical data analysis providers are already helping pharmaceutical companies and clinicians extract key learnings from massive parallel genomic datasets, and are increasingly layering in clinical data, often in a cloud-based collaborative environment. Some early innovators in this space include Seven Bridges Genomics and DNAnexus.

Regulatory and product development complexity will accompany biological complexity. For example, many of the testing strategies considered above may emerge as complementary tests, whereby regulators such as FDA effectively decline to mandate stratification on the label as required (companion) tests. Such a scenario will allow pharmaceutical developers and payers the flexibility to develop marketing strategies and reward scenarios reflective of the latest evidence base. A portfolio of complementary tests, likely combined as predictive panels, will then emerge as part of the initial tumour board workup, allowing clinicians to optimize therapeutic strategy from a portfolio of diverse immuno, targeted and chemotherapies while also, ideally, providing for the dynamic monitoring of tumour evolution over time. Such panel tests would evolve the field beyond current high Negative Predictive Value testing which leaves late-stage biomarker-negative patients with few options. As more broadly actionable baseline and monitoring testing options and associated therapeutic strategies evolve, stakeholders should find themselves at more of a decision node than a binary yes/no treatment scenario. Under this likely multi-parametric profiling scenario, which would represent a step change in patient benefit, the current generation of target-specific companion tests will represent merely one dimension in a multi-dimensional molecular workup supporting the emergence of the next generation of precision medicine. The growing complexity of patient selection, which will present both a challenge and an opportunity to pharmaceutical developers, will be a key success driver in the years to come.


By Iain D. Miller, Ph.D., July 2017

Dr. Miller consults regularly on precision medicine strategy, both as founder of Healthcare Strategies Group and as an Associate of Alacrita. The original version of this article first appeared in the Journal of Precision Medicine in May, 2017.

Share on FacebookShare on Google+Tweet about this on TwitterShare on LinkedIn

Stay in touch