THOMAS J. HERZOG, MD
Professor of Obstetrics and Gynecology
University of Cincinnati Cancer Institute
Cost-effectiveness analysis comparing “PARP inhibitors-for-all” to the biomarker-directed use of PARP inhibitor maintenance therapy for newly diagnosed advanced stage ovarian cancer
Gyn Oncol. Published online August 27, 2020. doi:10.1016/j.ygyno.2020.08.003.
Analyzing the Assessment of Cost Effectiveness for PARP Inhibitors in Ovarian Cancer
Background: Poly(adenosine diphosphate [ADP]-ribose) polymerase (PARP) inhibitors have been transformative in the contemporary treatment of ovarian cancer whereby progression-free survival (PFS) medians and hazard ratios (HR) have been improved dramatically. These agents leverage the concept of synthetic lethality whereby a combination of deficiencies in a gene or pathway leads to cellular death, whereas a defect in just one allows the cell to survive. These deficts can be mutations, epigenetic alterations, or inhibitors, as observed with PARP inhibitors.
Three phase III trials in ovarian cancer were presented at the European Society of Medical Oncology(ESMO 2019) and published (New England Journal of Medicine) in late 2019. PARP inhibitors to front-line regimens in ovarian cancer. These trials compared maintenance PARPi and bevacizumab to bevacizumab maintenance alone (PAOLA-1), PARPi to placebo (PRIMA), and the addition of a PARPi to chemotherapy and continued PARPi as maintenance compared to placebo (VELIA) (1-3). These three papers, of which two led to FDA approval (PAOLA-1 and PRIMA) are summarized comparatively in terms of eligibility, trial design, and clinical impact (4). The theory behind the concept of utilizing maintenance is that since the majority of women with advanced ovarian cancer will recur, why not attempt to prevent recurrence by using additional treatment or exposure to novel agents rather than waiting for the recurrence to occur.
PAOLA-1 trial was an investigator-initiated phase III trial in patients who received platinum and taxane based chemotherapy with at least 3 cycles of bevacizumab . Patients were randomized (2:1) after completing chemotherapy to either bevacizumab alone for 15 months with 24 months of placebo, or bevacizumab for 15 months with 24 months of oral olaparib for 24 months. The primary endpoint of this trial was investigator assessed PFS, and it was powered to show a HR of 0.75 at 458 events with the median PFS increasing from 15.8 to 21.1 months. A total of 762 patients were enrolled and the HR for PFS in the intention to treat (ITT) cohort was 0.59 (95% CI 0.49–0.72; P < 0.001), corresponding to a median PFS of 22.1 months in the bevacizumab & olaparib arm compared to 16.6 months in the control group (bevacizumab alone). Thus the trial was positive for the primary endpoint. Exploratory post-hoc subgroup analysis was performed and the efficacy data was positive in the BRCAmut group (HR = 0.31; 95%CI 0.20–0.47), and the HRD cohort PFS median was extended from 16.6 months to 28.1 months. (HR = 0.43; 95% CI 0.28–0.66), with HRD being identified by the ESMO Myriad My-Choice® test using a cutoff of ≥42 to define HRD. However in the HR proficient (HRP) subgroup, median PFS was similar (16.9 months vs 16.0 months; HR = 0.92; 95% CI 0.71–1.17). (1, 4)
PRIMA was a phase III trial designed to study the effect of niraparib maintenance in patients with not only germ-line or somatic mutations, but also those with HRD, as well as those who are HRP. The primary endpoint for this study was PFS as assessed by blinded independent central review (BICR) of radiologic images. A total of 733 patients were randomized (2:1) after completing chemotherapy to either niraparib or placebo for 36 months. The statistical design used a hierarchal step down analysis starting with the HRD cohort analyzed initially followed by the ITT population. This trial met its primary endpoint as well with a median PFS of 21.9 months for those who received niraparib vs. 10.4 for those receiving placebo (HR = 0.43; 95% CI 0.31-0.59). For the ITT population, the median PFS was 13.8 months for those who received niraparib vs. 8.2 for those receiving placebo (HR = 0.62; 95% CI 0.50-0.76). The exploratory secondary endpoints were met for PFS including for those who had HRD without BRCAmut, thus HRD and BRCA-wild type, (HR = 0.50; 95% CI 0.31-0.83). Interestingly, the HRP tumors also had an impressive HR with a HR of 0.68, although the magnitude of effect was less with a median PFS difference of only 2.1 months.
The VELIA trial was a third trial presented at the Presidential Session at ESMO. This phase III trial studied veliparib in the front-line setting in a three arm trial that randomized patients to chemotherapy plus placebo followed by placebo maintenance (control), chemotherapy plus veliparib followed by placebo maintenance (veliparib-combination), or chemotherapy plus veliparib followed by veliparib maintenance (veliparib-throughout). The VELIA trial demonstrated a statistically positive improvement in investigator assessed PFS in the overall population thus meeting its primary endpoint. Additionally in exploratory analyses, statistically significant improvements in PFS were observed in the BRCAmut and HRD/BRCAmut cohorts, but not in the HRD/BRCAwt or HRP groups. There are no approvals for veliparib for ovarian cancer, but it is discussed in the study below.
Hence clinicians can now choose from two new PARP inhibitors in the maintenance setting for front-line ovarian cancer treatment. The FDA regulatory labels are different from each of the studies with the PAOLA-1 study with bevacizumab and olaparib having an indication in those with HRD as assessed by the Myriad My Choice companion diagnostic test. The regulatory approval from the PRIMA trial permits niraparib to be prescribed to all patients regardless of subsets. Thus biomarker testing is not necessary for use as all patients with advanced epithelial ovarian cancer qualify.
Study: The treatment paradigm for front-line ovarian cancer and specifically the use of maintenance with PARP inhibition has changed dramatically in the past year. So how are clinicians supposed to process these new data and implement them most effectively clinically? One of the concerns besides short and long-term toxicity is the cost of therapies in terms of financial toxicity for patients and their families as well as for the healthcare system. A study recently published by Gonzalez et. al. examined the cost-effectiveness of PARP inhibitors in the front-line ovarian cancer maintenance setting. These authors specifically examined the approaches of PARP inhibitors “for all”- across all biomarker subgroups versus a biomarker driven strategy, whereby only those who have a positive biomarker such as HRD would receive PARP inhibitors. The authors examined the cost of these two aforementioned frontline PARPi maintenance prescribing strategies. The authors utilized modified Markov decision models to compare the PARPi for all patients irrespective of biomarker status strategy versus the biomarker-directed maintenance strategy. Comparisons incorporated the study designs of the PAOLA-1, PRIMA, and VELIA trials. The authors presented outcomes that included overall costs and incremental cost-effectiveness ratios (ICERs) reported in United States dollars per quality adjusted progression-free life-year (QA-PFY) gained thus using PFS not overall survival (OS) data. Sensitivity analyses were performed to adjust for uncertainties of inputs for the models.
Results: When modeling was conducted, the authors concluded that PARPi for all was much more costly than the biomarker driven strategy for each trial. The mean cost per patient for the PARPi for all strategy was $366,506, $166,269, and, $286,715, for the PAOLA-1, PRIMA, VELIA trials, respectively. For the biomarker driven strategy, the mean cost per patient was $260,671, $98,188, and $167,334 for the PRIMA, VELIA, and PAOLA-1 trials. ICERs of PARPi-for-all compared to biomarker-directed maintenance were: $3,347,915/QA-PFY (PAOLA-1), $593,250/QA-PFY (PRIMA), $1,512,495/QA-PFY (VELIA). Modeling that inputed contemporary drug pricing, demonstrated no significant PFS improvement in a biomarker negative cohort that would make PARPi for all cost-effective versus a biomarker driven maintenance strategy. The authors concluded that the biomarker driven maintenance therapy strategy in the front-line setting is preferred unless there are dramatic reductions in drug costs.
Clinical impact: These results help to provide context for clinicians in implementing these new phase III data for novel precision drugs such as PARP inhibitors and anti-vascular endothelial growth factors, specifically bevacizumab. The results are not surprising considering that one would only need to look at the benefit scene in the HRP groups from each trial to conclude that the greatest value would be in those two had a biomarker. It is likely the biomarker positive subgroups that drove the overall efficacy seen in the ITT populations.
There are multiple issues and potential shortcomings in this analysis. Firstly, these types of modeling exercises are fraught with use of input values that are not completely known. Sensitivity analyses help to adjust for uncertainty, but these methods are not always accurate either. Overall the authors did a laudable job of conducting the modeling, but the inherent nature of these types of studies needs to be acknowledged. Another issue with this particular analysis was the use of these phase III trials whereby the outputs were compared to one another. Although the authors will contend that these were compared internally, readers will likely compare cost approach of each trial to one another. These trials are each unique as there were significant differences in statistical design, eligibility criteria, length of study drug treatment, time of randomization, and means of assessing primary endpoint. Importantly, the most expensive outputs seen in the analysis were for the PAOLA-1 trial that utilize an active control arm with bevacizumab and thus the experimental arm used both bevacizumab and olaparib. Obviously this would be the most expensive approach; however, the authors are measuring the delta between the combination in the experimental arm versus the single agent active FDA approved therapy in the control arm. When comparing the values between the three studies, this is not taken into account as again the comparator is different in PAOLA-1.
Another issue is the use of quality adjusted PFS rather than OS. This was used to do to the lack of OS data at this time. The details of how the PFS date it was inputted is needed. It appears that median PFS values were used. If so, this ignores the experimental effects across the entire study which would be better captured with hazard ratios. This relationship is seen in the PRIMA trial, where a very modest gain was seen for median PFS of under three months versus a more substantial gain in the hazard ratio, which was below 0.7. In addition, the model did not appear to adjust for individualized dosing of niraparib which would lower costs.
Overall this was a well-done study that is subject to the limits inherent in such a modeling design. While it does appear that the biomarker driven strategy is preferred, we need to continue to improve the accuracy of HRD determination. The counter argument to the strategy suggested by the authors is that there are a number of patients who still benefit in the HRP cohort. While this number is much smaller, it is still significant especially when an individual patient is being counseled versus considering the more global view from the perspective of the greatest economic value for the healthcare system. Improvements in HRD testing that a more functional should help close the gap between these two strategies of biomarker driven versus PARPi for all.