The FDA’s recent approvals of novel anti-PARP agents as maintenance therapy for patients with previously treated advanced ovarian cancer highlight an issue that has received inadequate attention in the peer-reviewed oncology literature.
Maurie Markman, MD
The FDA’s recent approvals of novel anti-PARP agents as maintenance therapy for patients with previously treated advanced ovarian cancer highlight an issue that has received inadequate attention in the peer-reviewed oncology literature: What is the significance of a biomarker when a drug benefits not only patients with that characteristic but also those without it?
Based on the existing preclinical and clinical literature, it was widely believed that the greatest benefits associated with PARP inhibition would be observed in individuals possessing a germline BRCA mutation or, in the absence of that event, a somatic mutation in that gene.1 Further, it was hypothesized that in the absence of a BRCA mutation, other germline or somatic mutational events might also be associated with deficiencies in DNA repair mechanisms, and that discovery of an effective biomarker would help define the specific patient populations who would be more likely to benefit from this approach.
Conversely, it was reasonably assumed by many that patients whose cancers failed to demonstrate the presence of an effective biomarker revealing DNA repair deficiency would be unlikely to achieve clinical value from an agent designed to inhibit PARP activity.
In fact, in several phase III randomized trials designed to address these questions, the general hypothesis was confirmed.2-4 That is, the population of patients with ovarian cancer who possessed either a germline or somatic BRCA mutation exhibited the greatest statistically defined impact of therapy, as revealed by relative hazard ratios (HRs), on the subsequent time to disease progression.
Further, and as expected, patients whose cancers possessed a prospectively defined molecular defect suggesting an inability to efficiently repair DNA damage (homologous recombination deficiency [HRD]) also benefited from this approach to disease management, although the impact, as measured by HRs, was not quite as robust as with the BRCA mutation—positive population.
However, perhaps surprising to many in the ovarian cancer research community, patients whose cancers did not possess either a BRCA mutation or the presence of a biomarker suggesting DNA repair deficiency also exhibited a statistically defined improvement in the time to subsequent disease progression.2-4
To be clear, the HRs documenting improvement in outcome associated with therapy of this final group were not as robustly positive as observed in the presence of a BRCA mutation or an HRD biomarker, but clinical benefit for this patient population was objectively observed. In fact, the reported data led to the FDA’s broad approval of several anti-PARP agents as maintenance treatment for women with advanced ovarian cancer who had achieved either a clinically defined complete or partial response to second-line or later platinum-based chemotherapy, independent of the presence of a molecular biomarker.
Therefore, the issue to be addressed is as follows: How does one interpret the results of these studies regarding the significance of a biomarker in the clinical decision to employ a PARP inhibitor as a maintenance strategy in ovarian cancer? Is knowledge of the BRCA mutation status or the presence/absence of a measurable finding suggesting the impaired ability to repair double-strand DNA damage (an HRD) irrelevant based on the study findings, including statistical analyses, or on the FDA’s decision not to mandate such an analysis in this clinical setting?
This question highlights the increasing complexity of clinical medicine in general, and oncology in particular, and the very realistic potential that, for varying reasons, investigators, regulators, payers, and treating physicians will seek to oversimplify decisions when the actual data make such simplifications highly problematic.
Based on robust trial information, it was quite reasonable for regulators to approve the use of the agents for all patients responding to second-line or later platinum-based treatment, independent of their biomarker status. This decision also permitted the drug manufacturers to advertise the medicines for an expanded patient population compared with only the biomarker-positive groups.
However, despite this rational regulatory decision, it would be inappropriate to conclude that the biomarker status was either biologically or clinically irrelevant in this specific disease management setting.
In fact, based on the impressive study results, the opposite conclusion would be far more reasonable because the data reveal a greater statistical likelihood that the population with a BRCA mutation or, to a lesser extent, an HRD-positive cancer (in the absence of a BRCA mutation) will experience clinical benefit from this management approach compared with the group without these molecular findings. The probability that an individual patient will experience clinical benefit will be defined by the population to which she belongs. Further, although population-based trial data, no matter how robust, can never be used to inform patients on their exact future clinical course for efficacy or toxicity, such data can be highly useful in the decision-making process.
Ultimately, this is perhaps the most appropriate way to explain to patients and their families the utility of a clinical or molecular biomarker with proven efficacy. The test results cannot provide absolute answers regarding the best course of action or predict with certainty the benefits or harms, but they can be very helpful to the patient, family, and the clinical team in the selection of therapy and decisions regarding future management of the cancer.