Percentage of free to total PSA as a biomarker of survival in metastatic castration‐resistant prostate cancer

Abstract

Objective

To analyse whether the percentage of free to total prostate‐specific antigen (%fPSA) is a prognostic biomarker in metastatic castration‐resistant prostate cancer (mCRPC), as novel studies suggest an elevated %fPSA is associated with adverse oncological outcomes for men with biochemical recurrence of prostate cancer.

Patients and Methods

A biobank prospectively collated at mCRPC diagnosis was analysed for %fPSA. Clinicopathological characteristics, systemic therapies and survival outcomes were recorded. Patients were stratified by a %fPSA cut‐off of 15%. Cox proportional hazard models evaluated whether %fPSA was associated with overall survival (OS) and cancer‐specific survival (CSS) across the cohort and by treatment.

Results

A total of 254 patients analysed with newly diagnosed mCRPC: 161 (63%) men having a %fPSA ≥15%. The median follow‐up was 25.6 months. The median cohort OS and CSS was 39.6 and 43.8 months, respectively. Patients with a %fPSA ≥15% had lower median PSA level (31.30 vs 50.80 ng/mL; P = 0.007) and otherwise comparable clinicopathological and treatment profiles to men with a %fPSA <15%. Adjusting for PSA and on multivariable analysis, a %fPSA ≥15% was associated with shorter OS (multivariable hazard ratio [HR] 1.56, 95% confidence interval [CI] 1.02–2.40; P = 0.039). Among men treated with docetaxel, a %fPSA ≥15% was associated with worse OS (HR 1.84, 95% CI 1.03–3.26; P = 0.038) and CSS. Conversely, %fPSA was not associated with outcomes for men receiving androgen receptor pathway inhibitors (abiraterone acetate or enzalutamide).

Conclusion

An elevated %fPSA appears to be an adverse prognostic biomarker. Findings are consistent with biochemical recurrence studies, suggesting a biological basis. Validation and mechanistic studies are warranted.

Abbreviations

ADT

androgen‐deprivation therapy

ARPI

androgen receptor pathway inhibitors

CCI

Charlson Comorbidity Index

(m)CRPC

(metastatic) castration‐resistant prostate cancer

CSS

cancer‐specific survival

HR

hazard ratio

IQR

interquartile range

4Kscore

four‐kallikrein panel score

LDH

lactate dehydrogenase

OS

overall survival

%fPSA

percentage of free to total PSA

Introduction

Prostate cancer represents a significant healthcare burden, comprising 29% of all malignancies diagnosed in men globally [1]. Since the advent of PSA screening, most men with prostate cancer (94.4%) are diagnosed with localised disease only and have an excellent prognosis [2]. Men with metastatic hormone‐sensitive prostate cancer, either diagnosed de novo or metachronously, are treated with a combination of androgen‐deprivation therapy (ADT), androgen receptor pathway inhibitors (ARPI) and/or chemotherapy [3, 4, 5, 6, 7]. However, despite significant treatment advances over the last decade, progression to castration‐resistant prostate cancer (CRPC) remains inevitable for most and is the second leading cause of all cancer‐related deaths in men [2].

Prostate‐specific antigen is an androgen‐regulated protease produced by benign and malignant prostate epithelial cells [8]. In seminal fluid, a portion of PSA is inactivated by proteolysis, before passing through the basement membrane into peripheral blood as free PSA. The degree of PSA inactivation is reduced in malignant cells, such that the percentage of free to total PSA (%fPSA) is lower in prostate cancer than in BPH. Therefore, in the setting of a PSA level of 4–10 ng/mL, a lower %fPSA favours a diagnosis of prostate cancer [9].

While total PSA levels are monitored throughout a patient's prostate cancer journey, the %fPSA is only routinely calculated when diagnosing prostate cancer in patients with a total PSA between 4 and 10 ng/mL [9]. However, several recent studies have suggested that the %fPSA may be a useful biomarker for patients with metastatic prostate cancer [10, 11, 12]. Contrary to localised studies where a decreased %fPSA was associated with an increased risk of a prostate cancer diagnosis, in the metastatic setting, these studies demonstrated that an increased %fPSA at the time of biochemical recurrence is associated with earlier commencement of ADT, development of metastases [10], as well as progression to CRPC and cancer‐related death [11]. Therefore, we aimed to evaluate whether an elevated %fPSA in men with mCRPC is associated with adverse oncological outcomes.

Patients and Methods

Following Institutional Review Board approval, a prospective biobank of patients diagnosed with metastatic CRPC (mCRPC) was reviewed. Serum samples of these patients were collected prospectively, as per protocol at various states throughout the patient journey and stored at the time of mCRPC diagnosis at the Princess Margaret Cancer Centre. The %fPSA, was then calculated on the stored pre‐treatment mCRPC samples by dividing the free PSA by total PSA.

Chart review was performed to collect baseline clinicopathological characteristics at the time of CRPC diagnosis including age, Charlson Comorbidity Index (CCI), family history of prostate cancer, previous radical prostatectomy, Gleason score (if available) and the location of metastases, classified as bone, lymph node or visceral disease. Until the diagnosis of mCRPC, all patients had only received ADT. After diagnosis of mCRPC, first‐line systemic therapies beyond ADT were recorded and patients were followed for all‐cause and prostate cancer‐related deaths. First‐line systemic therapies after diagnosis of mCRPC were grouped into ADT plus chemotherapy (docetaxel) and ADT plus ARPI (abiraterone acetate or enzalutamide). Patients who did not receive docetaxel, abiraterone acetate or enzalutamide were excluded from the survival analysis.

Clinicopathological characteristics were summarised with the median and interquartile range (IQR) for continuous variables and the frequency and percentage for categorical variables. Patients were then stratified by a %fPSA cut‐off of 15%. The cut‐off approach was employed instead of a continuous variable as, from a statistical perspective, the continuous calculation would require a logarithmic transformation and, from a clinical perspective, the cut‐off of 15% is consistent with %fPSA studies published in the biochemical recurrence setting and is more readily interpretable [10, 13]. To account for the difference in total PSA, univariate analyses subdivided %fPSA ≥15% by the median total PSA across the entire cohort (39 ng/mL). Covariates associated with higher %fPSA were evaluated using Fisher's exact test for categorical variables and the Mann–Whitney U test for continuous variables.

The primary study endpoint was overall survival (OS), defined as the time from diagnosis of mCRPC to death of any cause, and the secondary outcome was cancer‐specific survival (CSS), defined as the time from diagnosis of mCRPC to death from cancer. The Kaplan–Meier method was used to estimate OS with survivors censored at last follow‐up. Cox proportional hazard models were used to evaluate whether the %fPSA, tested with the cut‐off of 15%, was associated with CSS and OS across the entire cohort and by treatment subgroup. Univariate Cox regression analysis was performed to identify clinicopathological features and therapies associated with OS. A multivariate analysis of %fPSA and significant findings from the univariate model was performed to identify independent predictors of outcome. A P < 0.05 was considered statistically significant. All statistical analyses were conducted using R version 4.3.1 (R Core Development Team, Vienna, Austria).

Results

A total of 254 patients in the biobank were treated for mCRPC. The median (IQR) age at mCRPC diagnosis was 71 (64–78) years. The median (range) PSA level was 35.4 (9.3–119.4) ng/mL and 66 patients had had a prior prostatectomy (Table 1). The median (IQR) %fPSA was 19.8 (12–25)%. Of these, 161 (63%) men had a %fPSA ≥15%, with 93 (37%) men having a %fPSA <15%. The median (IQR) follow‐up was 25.6 (16.0–37.7) months.

Table 1.

Cohorts’ characteristics.

Characteristic	%fPSA <15%	%fPSA ≥15%	P
	n = 93	n = 161
PSA, ng/mL, median (IQR)	50.8 (13.7–255.3)	31.3 (6.5–93.7)	0.007
Age, years, median (IQR)	69 (62–77)	72 (64–78)	0.128
Positive family history, n (%)	14 (15.1)	29 (18.0)	0.605
CCI score, median (IQR)	9 (8–10)	9 (8–10)	0.456
Radical prostatectomy, n (%)	30 (32.3)	36 (22.5)	0.103
Bone metastases, n (%)	72 (77.4)	127 (78.9)	0.875
Visceral metastases, n (%)	17 (18.3)	33 (20.5)	0.744
Lymph node metastases, n (%)	39 (41.9)	88 (54.7)	0.068
First‐line systemic therapy, n (%)			0.568
Abiraterone	19 (20.4)	24 (14.9)
Docetaxel	42 (45.2)	79 (49.1)
Enzalutamide	22 (23.7)	45 (28.0)
Radium	1 (1.1)	4 (2.5)
No treatment	6 (6.5)	7 (4.3)
Others	3 (3.2)	2 (1.2)

Of the 254 patients, 241 received systemic therapy after the diagnosis of mCRPC: 121 (48%) received docetaxel, 43 (17%) received abiraterone acetate, 67 (26%) received enzalutamide, five received radium‐223 and five received other therapies. The abiraterone acetate and enzalutamide groups were merged to a combined 110 (43%) patient cohort (ARPI).

Patients with a %fPSA ≥15% had a lower total PSA level (median [IQR] 31.30 [6.45–93.70] vs 50.80 [13.67–255.30] ng/mL; P = 0.007) and otherwise comparable clinicopathological and treatment profiles to men with a %fPSA <15%. There was no association between %fPSA and the location of metastases (lymph node, bone or visceral).

Among the 231/254 patients treated with either docetaxel or ARPI, there were 108 deaths, of which 97 were prostate cancer related. The estimated median (IQR) OS and CSS were 39.6 (33.2–46) and 43.8 (35.7–50.6) months, respectively. On bivariate analysis, the %fPSA dichotomised by the median PSA level, demonstrated that a %fPSA >15% and a high PSA (PSA level >39 ng/mL) was associated with shorter OS and CSS (P < 0.001) (Fig. 1). Other significant associations with OS on univariable analysis were age (hazard ratio [HR] 1.02, 95% CI 1.00–1.04; P = 0.04), CCI (HR 1.21, 95% CI 1.05–1.40; P = 0.10), prior radical prostatectomy (HR 0.50, 95% CI 0.31–0.80; P = 0.004) and the presence of bone metastases (HR 1.68, 95% CI 1.03–2.73; P = 0.039).

Fig. 1.

Panel of Kaplan–Meier analyses.

On multivariable analysis, a %fPSA ≥15% was associated with reduced OS (HR 1.56, 95% CI 1.02–2.40; P = 0.039). Other significant variables were treatment with docetaxel (HR 1.72, 95% CI 1.13–2.65; P = 0.011), CCI (HR 1.26, 95% CI 1.00–1.58; P = 0.051) and PSA level (HR 1.00, 95% CI 1.00–1.00; P < 0.001), which were all associated with shorter OS, while prior prostatectomy was a favourable OS factor (HR 0.53, 95% CI 0.32–0.86; P = 0.01) (Fig. 2). A %fPSA ≥15% demonstrated an increased HR for CSS on multivariable analysis but did not reach statistical significance (HR 1.42, 95% CI 0.93–2.18; P = 0.11).

Fig. 2.

Multivariable forest plot analysis for OS in CRPC. HR (95% CI) values given. AIC, Akaike information criterion; mets, metastases.

On treatment arm subanalysis, among men treated with docetaxel, a %fPSA ≥15% was associated with reduced OS (HR 1.84, 95% CI 1.03–3.26; P = 0.038) and CSS (HR 1.85, 95% CI 1.02–3.35; P = 0.042) (Fig. 3). On multivariate analysis, a %fPSA ≥15% for patients treated with docetaxel remained an adverse factor for OS (HR 2.61, 95% CI 1.38–4.95; P = 0.003) and CSS (HR 2.13, 95% CI 1.15–3.95; P = 0.016). In patients treated with ARPI, a %fPSA ≥15% was not associated with OS or CSS on univariate or multivariate analysis.

Fig. 3.

Treatment subanalysis stratified by treatment arm.

Discussion

Outcomes among patients with mCRPC are variable and associated with known prognostic factors such as tumour volume, lactate dehydrogenase (LDH) levels and functional status. However, there is an ongoing clinical need to identify additional low‐cost biomarkers that can predict patient outcome and segregate patients for future clinical trials.

To our knowledge, this is the first study of patients with mCRPC that explores the association of the %fPSA and oncological outcomes. We found that a %fPSA ≥15%, adjusted for total PSA level, was associated with shorter OS and CSS in patients with newly diagnosed mCRPC receiving standard first‐line therapy.

While this is the first study of %fPSA in a mCRPC population, analyses have previously been done on %fPSA for patients with biochemical recurrence. In the 1990s, studies identified heterogeneity of %fPSA in biochemical recurrence, particularly for men harbouring aggressive disease where some had a markedly elevated %fPSA [14, 15]. However, without the longitudinal data, it was unclear whether this information was clinically meaningful and the use of %fPSA remained as an adjunct focussed on screening for new diagnoses [16].

In 2020, Woon et al. [10] published the first study identifying that %fPSA at the point of biochemical recurrence, in patients originally treated with a radical prostatectomy, was associated with earlier progression to ADT, CRPC and metastatic stage. Subsequently, in 2021, a similar study, but for patients who were treated with radiation therapy, found that an elevated %fPSA at the time of biochemical recurrence was associated with earlier progression to metastases, castration resistance and death [11]. In 2023, Pellegrino et al. [12] evaluated whether the four‐kallikrein panel score (4Kscore®) test (OPKO Lab, Nashville, TN, USA) a panel of kallikreins including free PSA, was able to predict disease progression in patients with biochemical recurrence. Pellegrino et al. [12] reported that the 4Kscore test was not associated with oncological outcomes; however, when individual composites were analysed, the %fPSA was the only component that was associated with developing metastases and improved the discrimination of a validated predictive model for disease progression.

These recent studies have hypothesised several mechanisms for the adverse outcomes seen with an elevated %fPSA. Goldberg et al. [11] postulated that the different composition of metastatic prostate cancer cells results in altered prostatic cytodifferentiation and PSA production compared with primary prostate tissue. Other theories raised included PSA being cleaved and processed differently in metastatic cells resulting in more inactivated forms or that somatic mutations may accumulate in the genes involved with free PSA expression. Woon et al. [10] proposed that PSA processing alters after prostatectomy with metastatic tissue producing a less complexed form of PSA. However, in our cohort, OS differences based on %fPSA ≥15% remained regardless of prior prostatectomy. Pellegrino et al. [12] highlighted that as free PSA undergoes renal clearance, a greater %fPSA may reflect a more dynamic disease state and could therefore correlate with a shorter PSA doubling time. Finally, Pellegrino et al. [12] suggested that an elevated %fPSA could reflect an increased proportion of enzymatically inactive PSA secretion, which may be associated with angiogenesis and invasion, thus promoting metastasis.

An interesting finding in our cohort was that patients with an elevated %fPSA had significantly reduced median total PSA. Notably, this was not seen in the aforementioned %fPSA studies of patients with biochemical recurrence. This finding led us to propose another mechanism: that an elevated %fPSA may reflect an increased concentration of dedifferentiated tumour, which would correlate with both the poorer prognosis and reduced PSA secretion [17]. This would also be consistent with studies reporting a greater incidence of neuroendocrine differentiation found in CRPC compared with earlier disease states [18]. However, this cohort was on the first line of systemic therapy making tumour transformation less likely. Histological correlation is required to further explore this hypothesis.

An unexpected result from our study was the stark difference in the treatment arm subanalysis. Patients treated with docetaxel who had a %fPSA ≥15% experienced markedly reduced OS and CSS. Whereas, there was no predictive capability for %fPSA in patients treated with an ARPI. Due to the non‐randomised nature of the biobank, clinician selection of treatment (docetaxel or ARPI) introduces an inherent bias, with patients exhibiting a more aggressive disease phenotype possibly being prescribed docetaxel first. Given this, direct comparison between these groups was not performed. Furthermore, this cohort accrued during an era prior to triplet therapy and where all patients with mCRPC were naïve to systemic therapy other than ADT, which does not align with contemporary approaches to treatment [19]. As the first study to explore the relationship between %fPSA and systemic treatment, our findings provide new avenues for future research on predictive biomarkers.

Our study is limited by the retrospective design. The mCRPC cohort is heterogeneous with patients with mCRPC diagnosed and followed from various stages along the prostate cancer journey. Disease sites were recorded with conventional imaging, which may underestimate the extent of disease [20]. Furthermore, some prognostic factors such as disease volume and LDH were not obtainable for the multivariable analysis. Comparing the %fPSA against other established prognostic factors would help address whether clinicians should measure %fPSA for prognostication. Histological evaluation to assess the theory of tumour dedifferentiation was not uniformly available and original biopsies were often the much older primary prostatectomy specimens. Liquid biopsies or postmortem analysis may provide other opportunities to evaluate this question [21]. Nonetheless, our study provides novel insights towards exploring new applications for an established biomarker.

Conclusion

An elevated %fPSA appears to be an independent adverse prognostic biomarker for men with mCRPC. While the data requires validation, the %fPSA findings are consistent with all published studies of patients with biochemical recurrence, suggesting that there is an underlying biological basis. Furthermore, an elevated %fPSA appears to be associated with a reduced response to docetaxel. As %fPSA is a readily accessible, inexpensive, and reproducible test, with promising clinical applications both in prognostication and informing treatment, further validation and mechanistic studies are warranted.

Funding

None.

Disclosure of Interests

None.