Real-World Treatment Patterns and Overall Survival of Patients with Metastatic Castration-Resistant Prostate Cancer in the US Prior to PARP Inhibitors

This article is published with digital features, including a summary slide, to facilitate understanding of the article. To view digital features for this article go to https://​doi.​org/​10.​6084/​m9.​figshare.​14744073.

Prostate cancer (PC) is one of the most commonly diagnosed cancers and a leading cause of death among men in the United States (US) [1, 2]. Within 5 years of the PC diagnosis, an estimated 10–20% of patients develop castration-resistant prostate cancer (CRPC) [3, 4]. Metastatic CRPC (mCRPC) represents the most aggressive form of PC disease and is associated with a poor prognosis [5]. Further, mCRPC is a molecularly heterogeneous disease, with approximately 20–30% of patients harboring defects in DNA repair genes, including homologous recombination repair (HRR) gene alterations [6].

Despite the availability of next-generation hormonal agents (NHA; e.g., abiraterone and enzalutamide) for the treatment of mCRPC, as well as other therapies [e.g., sipuleucel-T (Sip-T) immuno-oncology, radium-223, cabazitaxel] [7], treatment response is short-lived due to tumor resistance [8], and patient median overall survival (OS) remains ≤ 2 years [9‐11]. However, the therapeutic landscape for mCRPC is poised to change with the 2020 approval of polyadenosine diphosphate–ribose polymerase inhibitors (PARPi), including olaparib and rucaparib, for patients harboring HRR gene alterations who have progressed on prior treatments [12, 13].

In the Phase 3 PROfound trial [12], 387 patients with mCRPC who progressed on NHA and had alterations in at least 1 of 15 HRR genes (e.g., BRCA1/2, ATM, CDK12, FANCA, PALB2) were randomized to either olaparib or a physician’s choice of NHA (i.e., abiraterone or enzalutamide). Among patients with alterations in BRCA1/2 or ATM, olaparib improved the primary endpoint of imaging-based progression-free survival (PFS) compared to physician’s choice of NHA (median PFS: 7.4 vs. 3.6 months) [12]. In the single arm TRITON2 clinical trial (n = 209 patients with HRR-mutated mCRPC) [14, 15], an objective response rate of 44% was obtained among 115 patients with BRCA1/2-mutated mCRPC treated with rucaparib after progression on androgen receptor-directed therapy and taxane chemotherapy.

Based on the results of these two trials, olaparib and rucaparib were the first PARPi to be approved by the US Food and Drug Administration (FDA) in May 2020: olaparib was approved for use in mCRPC patients with at least 1 of 14 HRR gene alterations, including BRCA1/2 alterations, who progressed on prior therapy with abiraterone or enzalutamide; rucaparib was approved for use in mCRPC patients with BRCA1/2 alterations who have progressed on prior NHA and taxane chemotherapy. Both agents were also included in the v2.2020 NCCN Clinical Practice Guidelines In Oncology (NCCN Guidelines^®): olaparib as a Category 1 recommended therapy post-abiraterone or enzalutamide for mCRPC patients with alterations in 14 HRR genes, and rucaparib as a Category 2A recommended therapy for mCRPC patients with BRCA1/2 alterations post-androgen receptor-directed therapy and taxane-based therapy [16].

While PARPi are expected to change the treatment landscape among eligible mCRPC patients harboring HRR gene alterations, there is limited real-world data on the treatment patterns of these patients in the pre-PARPi era to benchmark the changes. Although real-world treatment patterns after mCRPC diagnosis have been evaluated in a few recent studies in the US [10, 11, 17‐19], multiple knowledge gaps remain [10, 11]. First, line of therapy (LOT) sequences described in these prior studies do not go beyond the third line (3L) in the mCRPC setting. Second, treatment patterns were described at agent level only [10, 11], while analyses at treatment class level could provide a broader perspective, particularly as novel PARPi are available to patients who have responded inadequately to specific classes of treatments (e.g., NHA). Third, information regarding the duration of treatments is lacking for specific LOT sequences, including the time on and off treatments. Finally, there is a need for updated real-world OS estimates in this population, including the impact of the number of LOTs on patients’ OS. The present study provides a contemporary account of real-world treatment patterns among mCRPC patients in the US that addresses these limitations, with the aim of identifying current treatment gaps that newly emerging therapies could fill. The study focuses on treatment patterns among all patients with mCRPC (given that no targeted treatment options were approved for patients harboring HRR gene alterations during the 2013–2019 study period, the treatment patterns in the overall mCRPC population will serve as proxy for the treatment patterns of patients with mCRPC harboring HRR gene alterations in the pre-PARPi era).

The findings of this real-world study indicate that consecutive LOTs with NHA were the preferred 1L and 2L treatment options among mCRPC patients during the 2013–2019 study period, with increasing use of cytotoxic chemotherapy beyond 2L. Approximately half of the patients had ≥ 2 LOTs observed, and the median treatment duration was typically < 6 months per LOT. Median OS became shorter with each successive LOT (median OS: ranged from 19.4 months from start of 1L to death to 11.1 months from 3L to death), and patients who moved rapidly through multiple post-mCRPC LOTs had poorer OS compared to those who stayed longer on 1L. Taken together, these findings highlight the aggressive nature of mCRPC and the need for more effective novel treatments that prolong survival among patients with mCRPC.

The present study described treatment across the continuum of the patients’ PC disease. Pre-mCRPC diagnosis, chemotherapy was more common in metastatic states (i.e., mHSPC) than non-metastatic states (i.e., nmHSPC and nmCRPC), which is consistent with the NCCN Guidelines^® for recommended treatments in the pre-mCRPC setting during the study period and thereafter [16, 27‐31]. As of 2018, the FDA has expanded its approval of NHA to the pre-mCRPC setting, including the use of abiraterone and enzalutamide for mHSPC [32‐34] and apalutamide and enzalutamide for nmCRPC [33, 35]. Consistent with this, about one in five patients in the present study received NHA in the pre-mCRPC setting overall, ranging from < 5% in mHSPC clinical states to 34.1% in the nmCRPC clinical state. Given these high rates observed over the study period spanning from 2013 to 2019, it is also possible that NHA were being used off-label prior to their FDA approval for the pre-mCRPC setting. The emerging use of NHA in the pre-mCRPC setting may limit the benefits of NHA for the treatment of mCRPC, primarily due to the risk of acquired resistance and cross-resistance associated with repeated NHA use [36, 37]. However, this is less likely to be a concern among patients harboring HRR gene alterations, since they may benefit from a wider range of novel treatment options post-NHA (i.e., PARPi) compared to patients without HRR mutations.

Post-mCRPC diagnosis, the most common regimens by class were NHA, chemotherapy (i.e., docetaxel), and Sip-T, which is generally consistent with NCCN Guidelines for recommended treatments in the post-mCRPC setting during the study period and thereafter [16, 27‐31, 38]. In the present study, the predominant treatment class in 1L and 2L was NHA (e.g., abiraterone and enzalutamide), followed by chemotherapy (i.e., docetaxel). This pattern appears more consistent with later NCCN Guidelines (2017 onward) that include the use of NHA as category 1 recommended treatments irrespective of prior docetaxel therapy [16, 28‐31], whereas earlier guidelines primarily recommended NHA as treatment options for patients in the post-docetaxel setting or for those who are not candidates for docetaxel [27, 38]. Further, the present findings align with those of prior real-world studies by George et al. [10] and Higano et al. [11] that used the same data source to assess patients during an earlier time period (2013–2017). Whereas these two prior studies did not describe treatment patterns beyond 3L [10, 11], the present study shows an increasing reliance on chemotherapy agents in 3L and beyond. Specifically, docetaxel was the most common regimen in 3L, whereas cabazitaxel became the agent of choice beyond 3L.

In the present study, approximately 50% of patients did not receive a subsequent LOT after 1L, which is similar to the results of prior retrospective studies based on EHR data [10, 11]. Among patients with multiple LOTs, the most common LOT sequences post-mCRPC diagnosis were NHA → NHA (in patients with ≥ 2 LOTs) and NHA → NHA → chemotherapy (in patients with ≥ 3 LOTs), which reflects the increasingly common use of consecutive NHA in real-world settings [10, 11]. Interestingly, the duration of 1L NHA was longer among patients treated with NHA → NHA than patients treated with NHA → chemotherapy. These findings suggest that physicians may be seeking to maximize the benefits of NHA treatment, specifically among patients who responded to initial NHA therapy. However, given the increasingly common use of NHA in the pre-mCRPC setting, these post-mCRPC treatment patterns may change in the future.

The median duration of treatment in the present study was typically < 6 months (e.g., median duration of 1L: ranging from 2.8 months with chemotherapy to 5.1 months with NHA), which is consistent with previous findings reported by George et al. [10]; in particular, George et al. observed a median time on treatment of 5.4 months for IL abiraterone and 5.8 months for 1L enzalutamide, which further shortened with each successive LOT [10]. These median durations of NHA treatment are shorter than those observed in pivotal RCTs of enzalutamide and abiraterone, which ranged between 8.3 and 16.6 months for enzalutamide and between 8.0 and 13.8 months for abiraterone [39‐42]. Several factors may explain the shorter duration of NHA treatment in the current study compared with the RCTs. First, a relatively high proportion of patients included in this real-world study used NHA in pre-mCRPC settings. As more patients receive NHA in the earlier settings, the treatment patterns and subsequent time on treatment will likely evolve. Second, longer treatment durations in RCTs may also reflect the inclusion of patients with less severe disease compared to real-world populations [43, 44]. Consistent with this, some RCTs of the safety and efficacy of NHA had selected patients that were treatment-naïve and asymptomatic or minimally symptomatic with lower ECOG performance status (e.g., scores ≤ 1) [39, 42], while other RTCs had excluded patients with serious nonmalignant comorbidities [41]. Further, Gleason scores tended to be lower in the RCTs compared to the present study population (i.e., Gleason scores ≥ 8: ~ 50 vs. 66% of patients) [40, 41]. Finally, patients in RCTs generally show better adherence, persistence, and compliance to therapy [43, 44], which may have also contributed to the longer duration of NHA treatment compared to real-world studies. Future studies should further elucidate the shorter duration of NHA treatment in real-world studies compared with RCTs.

The present study builds upon prior research by assessing treatment duration by therapeutic class rather than at the agent level, and by accounting for the impact of LOT sequencing on outcomes. In the analysis of treatment durations across specific LOT sequences, the present study observed considerable variation in time on treatment across patients, with 1L NHA showing higher individual variability compared to 1L chemotherapy (IQR for time on treatment: 2.4–10.6 vs. 1.4–4.0 months, respectively). The time on 1L and 2L therapy combined was longer for NHA → NHA compared to the NHA → chemotherapy, although the time on treatment within LOT sequences tended to become shorter overall post-1L. While censoring may have been a concern in the present study, we note that treatment durations were short across all LOTs, including those that were observed to completion.

To our knowledge, the present study is the only analysis of the impact of LOTs on OS among mCRPC patients that accounts for changes in LOTs as a time-dependent covariate. This type of analysis clearly demonstrates that patients who moved more rapidly through multiple LOTs had a higher hazard of death at a given time point post-mCRPC diagnosis compared to those who had fewer LOTs at that time point. Prior studies have mainly focused of the impact of specific treatment sequences on OS among patients with mCRPC and were subject to several limitations [45‐48]. First, they primarily focused on patients who had been treated with NHA and cabazitaxel in different sequences following progression on docetaxel, a treatment sequence that is no longer representative of current mCRPC treatment approaches [45‐49]. The present study in turn captures a contemporary sample of patients with higher utilization of NHA. Second, studies comparing specific regimen sequences may be subject to confounding due to unmeasured clinical factors that influence regimen choices. An analysis of the impact of LOT number on OS, irrespective of the regimens used in each LOT, avoids this type of bias. Finally, prior studies did not account for censorship and the time-dependent nature of LOT changes post-mCRPC diagnosis. While one prior study did use a Cox regression model to assess the impact of LOT on OS, the LOTs were included as fixed variables in the model [47].

One of the main challenges in the treatment of mCRPC is the molecular heterogeneity of the disease [50, 51]. With the approval of PARPi, more treatment options have become available in 2020 for patients with HRR-mutated mCRPC who have progressed on conventional therapy [12, 13, 52]. Although consecutive NHA are currently the most common treatment sequence in 1L and 2L, the PROfound trial has shown a clear survival advantage with NHA → olaparib among patients with alterations in at least 1 of 14 HRR genes, including BRCA1/2, following progression on NHA [12]. Moreover, patients with BRCA1/2 gene alterations may benefit from rucaparib following progression on androgen receptor-directed therapy and taxane chemotherapy [53]. The availability of novel targeted therapies such as PARPi now provide a rationale for routine testing of appropriate patients for HRR mutations per clinical guidelines [31].

The present study had certain limitations. First, treatment selection in 2L and later lines is typically based on pre-existing patient characteristics (e.g., fitness level, response to prior LOTs), which may have influenced treatment outcomes; further, certain clinically important patient characteristics (i.e., comorbidities) were underreported in EHR data and, therefore, could not be adjusted for in the multivariate regression model for OS. Second, some patients may have received care outside the Flatiron Health network during the study period; in particular, treatments early in the course of PC disease may not have been fully captured for patients who had received their initial PC diagnosis outside the Flatiron Health network. Third, for most patients who did not die during the study period, censoring due to end of data availability underestimated the time on treatment associated with the patient’s last observed LOT. Fourth, LOTs were inferred from patterns of treatment administrations/orders recorded in the EHR data, and therefore occasional misclassification may have occurred. Fifth, disease-specific mortality data was not available in the database. Finally, the data used in this study are mostly drawn from community oncology practices and may not be representative of academic research centers.

In this real-world study in the US, consecutive LOTs with NHA remained the preferred treatment option. Despite a large number of LOT sequences beyond 1L, most patients died after one or two LOTs, highlighting a significant unmet need. The therapeutic landscape for mCRPC is continuing to evolve and treatments aimed at addressing the molecularly heterogeneity of the disease have emerged. In particular, recently approved PARPi may improve clinical outcomes and prolong survival among eligible patients. Updated NCCN Guidelines recommending genomic testing for HRR gene alterations among appropriate patients [31], combined with growing physician awareness of the importance of genetic testing in guiding treatment choices, have the potential to further transform the treatment patterns in mCRPC.