Prognostic factors for patients treated with abiraterone

Aim: To evaluate prostate-specific antigen response (PSAr) defined as a ≥50% decrease in PSA concentration from the pretreatment value, as a prognostic factor in patients with metastatic castration-resistant prostate cancer (mCRPC) treated with abiraterone acetate (AA). Methods: Retrospective evaluation of patients with mCRPC treated with AA. Results: 124 patients were identified. Median overall survival and progression-free survival for patients achieving PSAr versus patients without PSAr were 29.3 versus 9.7 months and 17.0 versus 5.2 months, respectively. Multivariate analysis confirmed that PSAr correlated with better overall survival (hazard ratio: 0.19; 95% CI: 0.10−0.38; p < 0.001) and progression-free survival (hazard ratio: 0.24; 95% CI: 0.14−0.41; p < 0.001). Conclusion: PSAr can be utilized as prognostic and predictive factors in mCRPC patients treated with AA.

of resistance, emphasizing the need for biomarkers for patients who are candidates for new-generation hormonal agents as AA [7,8]. Prostate-specific antigen (PSA) is widely used to monitor prostate cancer and its decline after chemotherapy has been acknowledged as a valid surrogate for OS and progression-free survival (PFS) at 3 months [9][10][11][12][13]. Retrospective studies confirmed that patients with mCRPC with a 50% decline in PSA from baseline have a survival benefit compared with patients who do not achieve such magnitude of reduction. However, the role of PSA as a surrogate predictor for OS in the course of treatment with new-generation hormonal agents and after chemotherapy remains uncertain [9][10][11]14]. The aim of the present retrospective analysis was to evaluate PSA response as a prognostic factor in patients treated with AA. The proposed hypothesis is that PSA response is able to identify patients more likely to benefit from AA treatment and who will survive longer.

Study population
Patients with mCRPC treated with AA, both pre-and post-docetaxel, at Hospital de Santa Maria − Centro Hospitalar Universitário Lisboa Norte between January 2013 and December 2017, were consecutively included and retrospectively evaluated. In the predocetaxel setting, patients were asymptomatic or minimally symptomatic, with no need for opiate analgesia. In the postdocetaxel setting, patients had confirmed progression or intolerable toxicity under chemotherapy treatment. All patients with confirmed bone metastases were under antiresorptive therapy, either with denosumab or zoledronate. The study's primary end point was the correlation of PSA response to AA treatment − defined as a ≥50% decrease in PSA concentration from the pretreatment baseline value (which was confirmed in a second PSA evaluation) − with OS and PFS. Secondary end points included the association of OS and PFS with other clinical and laboratory baseline characteristics retrieved from patients' clinical records whenever available: age, Gleason score, disease sites, previous docetaxel therapy, primary tumor treated, performance status (Eastern Cooperative Oncology Group [ECOG]), hemoglobin, LDH, ALP and total PSA. Additionally, the following data were retrieved: time of PSA response to AA, time of disease progression, time with androgen blockade (androgen-deprivation therapy) to mCRPC, time of AA discontinuation and time of death or last follow-up visit. During AA treatment, patients were monthly evaluated for PSA values. Radiographic assessment with computed tomography or Tc 99 bone scan was performed whenever biochemical or clinical progression was suspected. Progression and treatment response were defined according to the Prostate Cancer Clinical Trials Working Group 2 (PCWG2) criteria.

Statistical analysis
Sample size was not preplanned, as this was a convenience sample, only determined by the predefined inclusion criteria. Patients were stratified according to each prognostic factor included in the analysis: Gleason score (≥8 and <8), performance status (0 or 1 vs ≥2), visceral metastases (yes/no), hemoglobin levels (≥12 and <12 g/dl), LDH (≥median and <median), ALP (≥upper limit of normal and <upper limit of normal), total PSA (≥median and <median), PSA response (yes/no), primary tumor treated (yes/no) and previous docetaxel treatment (yes/no). Median and interquartile range (IQR) were reported for continuous variables, and frequencies for categorical variables. To compare categorical variables among groups, χ 2 and Fisher exact tests were used whenever appropriate. The Mann-Whitney test was used to compare medians. Outcome measures were PFS and OS. PFS was defined as time between AA therapy's start and time of progression, as established by PCWG2 criteria. OS was defined as time between AA therapy's start and death, irrespective of its cause. Median OS and PFS were obtained using the Kaplan-Meier method and log-rank test was used to compare group outcomes. To assess the effect of analyzed parameters in the duration of response and prognosis, univariate and multivariate analyses were conducted using Cox regression model. All statistical tests were two tailed and statistical significance was assumed when p > 0.05. IBM's SPSS statistics v.24 was used for statistical analysis. . Baseline characteristics of groups with and without PSA response were well balanced, with the exception of ECOG performance score, hemoglobin and PSA levels, as significantly more patients in the no PSA response group had hemoglobin levels <12 g/dl, higher PSA values and worse ECOG performance scores (p = 0.022, 0.019 and 0.010, respectively). Characteristics of both the groups are detailed in Table 1.
The overall median follow-up was 11.5 months. A total of 73 patients (58.9%) had died by the last follow-up, 24 of which in the group with PSA response and 49 in the group without PSA response. 27 patients remain on AA treatment, with progression being the primary reason for treatment discontinuation.
Duration of response in patients treated with AA, stratified by PSA response Duration of response in terms of PFS is presented in Table 2. Median PFS was significantly longer in the group with PSA response: 17.0 versus 5.3 months in the group without PSA response. This indicates that PSA response resulted in a 77% reduction in the risk of progression compared with having no PSA response (hazard ratio [HR]: 0.23; 95% CI: 0.14−0.39; p < 0.001). PFS Kaplan-Meier curves are shown in Figure 1. In univariate analysis, patients with PSA response, baseline ALP and total PSA below the median, not previously treated with docetaxel and better ECOG performance score were associated with a better PFS (Table 2).
In univariate analysis, PSA response, treatment for the primary tumor, absence of visceral metastases, absence of previous chemotherapy, better ECOG performance score, baseline hemoglobin ≥12 g/dl and baseline PSA and ALP below the median were associated with better OS (Table 3). After adjusting for significant factors at univariate analysis, multivariate analysis revealed that a ≥50% reduction in PSA remained predictive of better survival prognosis (HR: 0.19; 95% CI: 0.10-0.38; p < 0.001), as well as baseline PSA below the median (HR: 0.35; 95% CI: 0.20-0.59; p < 0.001) and previous treatment for the primary tumor (HR: 0.47; 95% CI: 0.27-0.80; p = 0.006; Table 3).

Discussion
With the growing number of therapeutic options capable of extending survival in mCRPC patients, there is a need for biomarkers that can simultaneously guide treatment decisions and predict which patients will benefit the most from new treatments as AA. Although this is true for all tumors, it is particularly relevant for prostate cancer, due to factors related with disease heterogeneity and difficulty in assessing disease in bone and interpreting the clinical significance of post-therapy PSA changes in different settings. In this cohort of mCRPC patients treated with AA, PSA response was associated with both OS and PFS benefit. After adjusting for other clinically relevant and statistically significant baseline characteristics, this factor remained prognostic for increased survival and predictive of treatment response. The development of new surrogate markers for clinical outcomes is becoming increasingly important with the emergence of multiple lines of treatment with survival benefit for mCRPC patients. With OS being the gold-standard end point in Phase III clinical trials, the multiple sequencing treatment options that are becoming available will predictably disturb the analysis of these end points. Therefore, surrogates for OS can provide clinicians with data that better inform them throughout patient treatment, allowing for more robust go/no-go decisions. Defining a new surrogate end point requires meeting several criteria defined by Prentice criteria or by the proportion of treatment effect, as previously explained [11]. Table 3. Prognostic role of baseline characteristics and prostate-specific antigen response: univariate and multivariate analysis for progression-free survival.  PSA remains a questionable surrogate for survival in late-stage prostate cancer. PCWG2 states that the clinical significance of a post-therapy PSA decline remains controversial and advises against its early use. In fact, an increase in serum PSA, or 'flare', may occur in some patients before they experience a subsequent and sometimes significant PSA decline. This is why PCWG2 recommends securing a sufficiently large drug exposure window and avoid relying on serum PSA decline as a surrogate for clinical benefit. Perhaps even more importantly, the group advises not to interpret a rise in serum PSA as progression and early withdrawing a therapy from which the patient may benefit [12,13]. Drugs targeting androgen receptor (AR) signaling, such as AA, may have a different association to an early PSA decrease, since PSA is a pharmacodynamic biomarker of androgen receptor signaling in absence of aberrations on the PSA promoter or key regulators of PSA production and secretion [15]. Some studies reported that patients with mCRPC who experienced ≥50% PSA decline from baseline had improved survival compared with those who did not. Additional studies, like the retrospective analysis of PSA decline in the SWOG 99-16 trial, which compared docetaxel plus estramustine to mitoxantrone plus prednisone, indicated that a PSA reduction of at least 20−40%, a 2-month PSA decline of 30% and PSA velocity of decline at 2 and 3 months after therapy satisfy the stringent requirements of statistical surrogacy for prediction of OS. The TAX327 trial is another such example, showing that a ≥30% PSA decline after docetaxel treatment fulfilled the Prentice criteria for surrogacy. It is worth mentioning that most of these results were observed in patients treated with first-line chemotherapy, and the evidence supporting PSA decline as a surrogate for OS across multiple drug classes is scarce and conflicting [11,12,15]. Although this study showed a potential role for PSA response as a predictive and prognostic factor for response in mCRPC, it has some limitations, the primary being its small sample size and retrospective nature. Further validation of its results is, therefore, required to assist in implementation of the best clinical practice and allow patients unlikely to benefit from AA to switch to another treatment option. Additionally, validation of PSA as a biomarker in mCRPC has the potential to help identify patient subgroups that may benefit the most from AA, maximizing cost-effectiveness of this treatment [16]. Finally, it should be noted that new molecules, such as Radium-223, with proven therapeutic effect in mCRPC patients, do not usually induce a PSA response, therefore other biomarkers should be additionally investigated.

Conclusion
In the present study, PSA response was an independent predictive and prognostic factor for survival outcomes, with PSA response being associated with response duration and survival benefit in patients treated with AA. Biochemical markers are becoming increasingly important as response and prognostic surrogates, especially for diseases difficult to evaluate radiographically (>90% bone metastases). Further prospective studies in larger patient populations are required to confirm these findings.

Future perspective
Further advances are expected to occur in this area in upcoming years, with new and more effective treatment options. Associated with that, more than prognostic also predictive biomarkers will be important to investigate, as they can potentially predict response before treatment's start, avoiding toxicities and maximizing treatment cost-effectiveness.