Unfavourable overall survival data for PARP inhibitors in certain ovarian cancers prompts calls for caution in interpretation of the analyses and stimulates the search for surrogate measures of clinical benefit
2022 saw a wave of US FDA regulatory approval removals for PARP inhibitors (PARPi) in the treatment of some patients with ovarian cancer, including BRCA-mutated and homologous recombination deficiency (HRD)-positive disease. These removals – which were made voluntarily by sponsors, prompted by long-term overall survival (OS) data analyses – have not only limited options for treatment in the US, but also fuelled uncertainties at a global level about how and when to use these agents. This has left many ovarian cancer experts wondering, is OS the right endpoint for PARPi trials?
Interpreting OS data with PARPi is the subject of a Special Symposium at the ESMO Gynaecological Cancers Congress 2023 (Barcelona, 23–24 February). Symposium Co-Chair, Dr Robert L. Coleman from The Woodlands, Houston, TX, USA, thinks it is important to look at how these FDA regulatory decisions came about. “The ARIEL4 (Lancet Oncol 2022;23:465–478) and SOLO3 (J Clin Oncol 2020;38:1164–1174) trials – investigating single-agent PARPi therapy for relapsed BRCA-mutated ovarian cancer – both met their primary endpoints of progression-free survival (PFS),” he says. “It was the post-marketing, long-term follow-up of OS, requested by the US FDA, that revealed hazard ratio (HR) point estimates in favour of comparator treatment for some patients. Voluntary removal of FDA regulatory approval for rucaparib and olaparib in the treatment of heavily pre-treated BRCA-mutated ovarian cancer, was followed by knock-on effects of FDA regulatory approval removals for niraparib in heavily pre-treated HRD-positive ovarian cancer and, further, for niraparib and rucaparib as maintenance therapy in patients with non-BRCA-mutated, recurrent platinum-sensitive disease.”
OS remains a critical component of risk–benefit analysis for PARPi use, but it needs to be carefully considered in the context of the trial design, its patient population and, crucially, post-progression treatments. Coleman explains: “Firstly, data interpretation should be informed by the statistics. The fact that an HR point estimate and its 95% confidence intervals span 1.0 implies only that the null hypothesis cannot be rejected; it does not imply harm from the investigational treatment. In addition, caution is urged regarding subjective interpretation of the degree HR extends over 1.0 and the suggestion of harm from PARPi treatment. It is certainly worth noting that none of the trials were powered for OS. We also need to consider that the trials in question employed nested cohort analyses. Today, it is a given that patients considered for PARPi will have their tumours tested genomically for BRCA and HRD and will fall into one of three categories: HRD/BRCA-mutated, HRD/non-BRCA-mutated or non-HRD. In a bid to assign patients to these three groups, the recent OS analyses involved un-nesting of the trial data, confounding interpretation of the results.” Coleman also points towards the well-recognised problem that OS is influenced by a range of post-progression treatment factors, including receipt of surgery or radiotherapy and crossover to the alternative study agent, and by the disease status at progression. “One particular issue is the considerable bias towards re-treating patients with platinum following progression on PARPi,” he says. “The mechanisms involved in the development of resistance to PARPi can also lead to resistance to platinum, but not non-platinum agents; so patients receiving post-PARPi platinum can be expected to show less survival benefit than those receiving non-platinum agents.”
A number of important lessons have been learned from this experience, according to Coleman. “It is confirmation that nested cohort analyses are not an appropriate design for trials involving binary biomarkers. Future trials should be designed specifically for the patient population under investigation, for example HRD/non-BRCA-mutated ovarian cancer,” he says. “In addition, we know that there are difficulties with statistical power in a disease like ovarian cancer that has long post-progression survivorship and that it is not feasible to address these in terms of defining strict treatment strategies for all patients involved or by enrolling very large numbers of patients. Instead, we need better definition of post-PARPi progression therapies and identification of biomarkers that will help us to understand tumour status at progression and so inform the best treatment approach at that time point.” In terms of alternative or additional endpoints to OS, Coleman thinks that the time to second objective progression (PFS2) merits particular attention as a way of providing some confidence that an intervention is persisting through an initial line of therapy and, as such, can provide a useful surrogate of benefit.
Concluding, Coleman says: “It will be interesting to see how the results from phase III trials involving maintenance olaparib, immunotherapy and bevacizumab in ovarian and endometrial cancer – DUO-O (NCT03737643), DUO-E (NCT04269200) and KEYLYNK-001/ENGOT-ov43/GOG-3036 (NCT03740165), all of which are due to report later this year – are received in the current climate.”
How to interpret overall survival data with PARP inhibitors. ESMO Gynaecological Cancers Congress 2023
Special Symposium, 23.02.2023, h. 10:15 – 11:35, Auditorium 113