In 2010, only three therapies improved overall survival (OS) in metastatic prostate cancer, including docetaxel, cabazitaxel and sipuleucel-T1. The burgeoning research in angiogenesis inhibition and phase II trials had buoyed hopes that a combination of the VEGF inhibitor bevacizumab with docetaxel could further enhance survival. When the data were presented in 2010, a progression-free survival (PFS) advantage of bevacizumab and docetaxel versus placebo and docetaxel was observed (9.9 versus 7.5 months, HR 0.80, 95% CI 0.71–0.91, stratified log rank P < 0.001). The price of this PFS was increased toxic effects (75.4% versus 56.2%, P ≤ 0.001) and treatment-related mortality (4.0% versus 1.2%, P = 0.005)2. Despite these short-term benefits in PFS, no OS advantage was found (22.6 versus 21.5 months; HR 0.91, 95% CI 0.78–1.05; P = 0.181)3. Regardless of a relative paucity of options for treatments of prostate cancer at the time, regulatory bodies did not approve this combination for patients with terminal prostate cancer. The disappointment in the field was clear, but the recognition that the toxic therapy did not actually improve OS relative to the best possible control arm was accepted by most and the field moved on to a decade that would yield more than a dozen new therapeutic options across all stages of prostate cancer1. Yet, somehow nearly 15 years later with a multitude of therapeutic options, a drive to reduce the threshold for approving therapies in prostate cancer is apparent and it might be time to question whether that desire is appropriate.

Emerging PFS data in prostate cancer

A recent presentation of data from CONTACT-2 indicated that patients with metastatic castration-resistant prostate cancer (mCRPC) treated with second-line cabozantinib and atezolizumab had a superior PFS (6.3 versus 4.2 months; HR 0.65; 95% CI 0.50–0.84; P = 0.0007) relative to the control arm of androgen receptor pathway inhibitor (ARPI) switch — a clinical strategy of swapping one ARPI for a second one after progression on the initial ARPI4. Clinical data have shown the ARPI switch to be an inferior control arm because of overlapping mechanisms of resistance among all ARPIs5. Toxic effects were amplified in the experimental arm, but underpowered subsets suggested relative efficacy of the combination of cabozantinib and atezolizumab. The OS data were not mature; with 49% of the target number of events having occurred, no statistical difference was observed (HR 0.79; 95% CI 0.58–1.07; P = 0.13). The PFS difference was statistically significant, but a debate occurred at the podium regarding whether the findings were truly clinically meaningful and filled an “unmet need”. Astute questions were raised about censoring and how that could have influenced the PFS readout. Some hailed this study as the first time an immune checkpoint inhibitor had a ‘positive’ study in prostate cancer whereas others correctly questioned what the contribution was of the immunotherapy to this combination, given a history of negative immune checkpoint inhibitor trials in prostate cancer.

“a drive to reduce the threshold for approving therapies in prostate cancer is apparent”

This report is not the first time cabozantinib has had a preliminary PFS advantage without an OS signal. In COMET-1, in which cabozantinib was compared with prednisone, an improvement in radiographic PFS was reported as a secondary end point (median 5.6 versus 2.8 months, HR 0.48; 95% CI 0.40–0.57; stratified log-rank P < 0.001), but no meaningful improvement in OS was observed (11.0 versus 9.8 months; HR 0.90; 95% CI 0.76–1.06; stratified log-rank P = 0.213)6. In addition, this disconnect between PFS and OS was recognized and resulted in the demise of agents such as satraplatin for prostate cancer. Based on preliminary data that suggested a PFS benefit, satraplatin plus prednisone with placebo plus prednisone in mCRPC after docetaxel were compared in the SPARC study. Similar to CONTACT-2, a small, but statistically significant PFS benefit of 11.1 weeks versus 9.7 weeks was observed (HR 0.67; 95% CI 0.57–0.77; P < 0.001)7. This PFS benefit did not translate to an OS benefit (61.3 versus 61.4 weeks; HR 0.98; 95% CI 0.84–1.15; P = 0.80). Thus, upon review, prostate cancer studies historically have yielded PFS improvements that did not improve OS.

Recent PFS data leading to regulatory approval

Peer review of the complete CONTACT-2 dataset is required to understand the true merits of cabozantinib and atezolizumab in mCRPC, but whether the investigators are wrong to expect approval of this novel combination that met its primary end point of PFS, if only by 2 months versus a substandard control therapy, is a pertinent question. A review of recent history suggests that perhaps the expectations of a forthcoming regulatory approval are not actually out of line. In 2023, full regulatory approvals of three poly(ADP-ribose) polymerase (PARP) inhibitor and ARPI combinations in biomarker-driven subpopulations were made based solely on PFS data8,9,10. These approvals were made despite the fact that that single-agent PARP inhibitors were already approved in mCRPC, and inadequate study design of all three combination trials assessed only the value of PARP inhibitor and ARPI combinations in patients with homologous recombination repair mutations (HRR+) versus a control HRR+ group receiving only ARPI with no crossover11. These trial design deficiencies meant that across the three trials, subsequent treatment with PARP inhibitors was a paltry 30%, 17% and 1%8,9,10. This asymmetrical study design with lack of appropriate subsequent therapy virtually ensures numerical advantages in PFS and eventually OS. Most practitioners hailed the new option for patients with HRR+ mutations, but some questioned whether combination therapies given earlier and longer (with accompanying exposure to toxic effects) than the existing standard of care of sequential therapy of ARPI followed by PARP inhibitor are truly superior — a question that, unfortunately, can never be answered from these data, given that the study designs left the substantial majority of patients with HRR+ undertreated for the duration of their lives11.

In 2023, new data regarding the radioligand 177lutetium-prostate-specific membrane antigen (177Lu-PSMA) in the chemotherapy-naive setting were reported, after results of a previous trial demonstrated an OS advantage in patients with mCRPC who had already received chemotherapy compared with ARPI switch12. In the PSMAfore trial, patients were again randomized to the same treatment options of 177Lu-PSMA and ARPI switch. As was widely expected, a superior PFS advantage in patients treated with 177Lu-PSMA to ARPI switch was observed (12.0 versus 6.6 months; HR 0.43)13. However, this study should be commended because, unlike the three previous trials discussed involving PARP inhibitors, it had substantial crossover of 84.2%. Interestingly, this crossover probably yielded the most unintended yet important data from this trial: the OS of the two arms of the study were similar (19.3 versus 19.7 months) despite an initial and statistically significant PFS advantage for 177Lu-PSMA relative to ARPI. Earlier administration of 177Lu-PSMA in the treatment course for mCRPC did not improve survival based on this PSMAfore data. Many still expect regulatory approval, perhaps based on the reduced toxic effects compared with docetaxel, but some would question the true therapeutic value in both 177Lu-PSMA phase III trials given inferior control arms (ARPI switch), and whether further studies earlier in the natural history of prostate cancer (such as biochemical recurrence) are truly warranted.

Survival benefit can still be demonstrated despite crossover

The 2023 experience of 177Lu-PSMA (PSMAfore) and the ARPI and PARP inhibitor combinations could offer a cautionary tale about study design to investigators and industry alike. One myth in prostate cancer that should be clarified is that substantial crossover dooms a study to a negative OS readout. While we could cite TAX327 and the FDA approval of docetaxel in 2004, it might be best to look no further than the PROfound trial of olaparib versus ARPI switch in BRCA1 and BRCA2+ chemotherapy-naive mCRPC14. In PROfound, which yielded a positive PFS (7.4 versus 3.6 months; HR 0.34, P < 0.001), 66% of patients in the control group ultimately went on to receive olaparib15. Despite this substantial crossover, the OS benefit of earlier receipt of olaparib based on study randomization was maintained (18.5 versus 15.1 months; HR 0.64; P = 0.02). The therapeutic signal was not lost in this trial and should raise questions about PFS benefits in mCRPC trials that cannot translate into an OS benefit despite high rates of crossover.

Some might argue that with so many mCRPC therapies available, losing an OS advantage is easy despite having PFS benefit because of the numerous therapies available for prostate cancer. Such a concern could be accentuated the earlier you are in the disease process, like the first-line trials of ARPI–PARP inhibitor combinations in mCRPC8,9,10; but this concern should be dispelled by recent history as well. Since 2015, five therapeutic strategies have been developed in metastatic castration-sensitive prostate cancer (mCSPC). All of these trials had statistically significant PFS advantages, and therapies or combinations of therapies upon progression to later stages of disease (such as ARPI or docetaxel) were allowed in each. Yet the patients in these control arms (androgen deprivation alone or in combination with docetaxel) would have had to have had prolonged survival timelines (~4 years or longer) to get exposure to the treatments they were not randomized to in the respective mCSPC studies. Despite the high functional crossover that was inevitable, the sustained OS improvement in these trials indicates that the PFS signal observed in mCSPC was never lost, despite timelines that would eclipse survival in mCRPC studies.

“The unmet need is for prostate cancer therapies that can meaningfully enhance OS”

The increasing acceptance of PFS as an approvable end point by regulatory bodies in prostate cancer probably comes from a genuinely altruistic endeavour to get patients new therapies in expedited fashion. But whether patients are truly being served if PFS is gained using inferior control arms that do not truly represent acceptable treatment options for patients on trials or in practice is an important question to ask. If PFS is virtually ensured by an unbalanced study design and subsequent appropriate treatment is limited, as in the PARP and ARPI studies, OS data cannot reinforce PFS findings. PFS might be an acceptable intermediate end point, but OS data from these same studies must validate those PFS findings after ensuring subsequent appropriate care as much as possible. Both PROfound and PSMAfore had substantial crossover in the third-line setting, so ensuring appropriate subsequent therapies has been proven feasible in recent trials. It cannot be acceptable to win the sprint to start the race, but then ignore the rest of the marathon.

Defining a future path

Hearing that prostate cancer still has an unmet need for therapies is common at every scientific meeting. Patients would agree with that sentiment but perhaps with a slightly different perspective. The unmet need is for prostate cancer therapies that can meaningfully enhance OS, not treatments with potentially high toxic effects that can yield statistically significant PFS advantages relative to substandard control arms, yet do not translate into clinically meaningful OS benefits. With approximately ten therapeutic strategies available for mCRPC (including HRR+ options), patients cannot possibly receive all therapies, nor would it be advisable for them to do so. But as the recent results of studies on single-agent PARP inhibitors have shown, biomarker stratification enhances sustainable benefits in appropriately selected patients. Thus, efforts should be channelled into biomarker-driven studies that either identify new therapies for certain subpopulations or determine the best available options and/or sequences for existing treatment strategies. Investigators, industry and regulatory bodies must be fully transparent with practitioners, patients and their families about the practical implications of clinical trial data as it pertains to patient care and not just hazard ratios. Only with the correct focus on what is a need and what is truly unmet can the field move forward to provide new, optimal and sustained therapeutic benefits for patients with prostate cancer.