Prognostic role of docetaxel-induced suppression of free testosterone serum levels in metastatic prostate cancer patients

To date, only few data concerning the biologically active, free form of testosterone (FT) are available in metastatic prostate cancer (mPC) and the impact of FT on disease, therapy and outcome is largely unknown. We retrospectively studied the effect of docetaxel on FT and total testosterone (TT) serum levels in 67 mPC patients monitored between April 2008 and November 2020. FT and TT levels were measured before and weekly during therapy. The primary endpoint was overall survival (OS). Secondary endpoints were prostate-specific antigen response and radiographic response (PSAR, RR), progression-free survival (PFS), FT/TT levels and safety. Median FT and TT serum levels were completely suppressed to below the detection limit during docetaxel treatment (FT: from 0.32 to < 0.18 pg/mL and TT: from 0.12 to < 0.05 ng/mL, respectively). Multivariate Cox regression analyses identified requirement of non-narcotics, PSAR, complete FT suppression and FT nadir values < 0.18 pg/mL as independent parameters for PFS. Prior androgen-receptor targeted therapy (ART), soft tissue metastasis and complete FT suppression were independent prognostic factors for OS. FT was not predictive for treatment outcome in mPC patients with a history of ART.


Methods
Patients. For this translational biomarker study, data from patients with mPC treated with docetaxel at Hannover Medical School from April 2008 to November 2020 were retrospectively analyzed and followed up until death or until December 2020. Eligibility criteria were a histologically confirmed mPC and disease progression according to the PCWG2/3 criteria 4,16 . Patients who received at least two cycles of docetaxel were included. All data were collected following patient informed consent, in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice guidelines, and with the Hannover Medical School institutional review board approval (13th August 2008). Patients were stratified into three subgroups: (1) castration-naïve disease (mCNPC), (2) castration-resistant disease (mCRPC) and (3) mCRPC patients with a history of ART (mCRPC-ART).
Treatment plan. At baseline, medical history and physical examination were performed, including an initial staging. Patients received 75 mg/m 2 of docetaxel intravenously every 3 weeks (q3w), 50 mg/m 2 every 2 weeks (q2w) or 30-35 mg/m 2 weekly on days 1, 8 and 15 (q1w). Concomitant use of dexamethasone and oral prednisone (5 mg) twice a day were part of the regimen. ADT was continued throughout therapy. Morning FT and TT (8-11 am) were obtained before and weekly during treatment (on average 28 samples per patient) using an enzyme immunoassay (ELISA from IBL, International GmbH, Hamburg, Germany) and a direct, competitive, chemiluminescence immunoassay (CLIA) (LIAISON ® Testosterone Assay, Diasorin S.p.A., Saluggia, Italy). PSA levels, carcinoembryonic antigen (CEA) levels, neuroendocrine tumor markers (neuron-specific enolase (NSE), chromogranin A (CgA)), lactate dehydrogenase (LDH), alkaline phosphatase (AP), hemoglobin and analgesic requirement were parameters at baseline to possibly predict OS 16 . Patients were assessed according to PCWG2/3 criteria and RECIST 1.1 by CT and bone scan every three months or if tumor progression was suspected 4,17,18 . Pain and use of pain medications were monitored by clinician interview. Toxicity was graded according to the Common Terminology Criteria for Adverse Events v3.0 (CTCAE).
Data analysis. The primary study end point was OS, defined by the initiation of docetaxel therapy until death. Secondary endpoints were PFS, defined as the time between the start of docetaxel therapy until progression according to PCWG2/3 criteria 4,16 , PSA response (PSAR), which is defined by a decline of > 50% from baseline, and FT reduction of 100% from baseline and safety. Radiographic response (RR) was evaluated according to RECIST 1.1. 17 . Follow-up data were collected throughout December 2020.
Statistical analyses were performed using SPSS statistics v26.0. Categorical variables were summarized, numeric variables were analyzed in median and range. Logistic regression was used to estimate the prognostic significance of FT suppression in predicting ≥ 50% decline in PSA from baseline. Cox proportional hazards regression modeling was used to determine the prognostic significance of baseline characteristics on PFS and OS. Chi-square tests and t-tests were applied to estimate p values of variables at baseline. Uni-and multivariate Cox regression analyses were used to demonstrate the impact of covariates on PFS and OS. Non-proportionality was assessed by plotting the Kaplan-Meier survival distribution as a function of the survival time for each level of the covariate and plotting the function log(-log(survival probability)) as a function of the log survival time 19 . Additionally, extended Cox modelling with time-by-covariates and conditional landmark analyses were used to remove potential guarantee-time bias, specifically the time-window bias, which is introduced because of differential exposure opportunity time windows between subjects 20 . Only p values of < 0.05 were considered statistically significant for all comparisons.

Results
Patient characteristics. A total of 67 patients with a histologically confirmed metastatic adenocarcinoma of the prostate were analyzed. Seven patients were castration-naïve (mCNPC, group 1), 26 patients were castration-resistant after ADT (mCRPC, group 2) and 34 patients were castration resistant after ADT plus ART (mCRPC-ART, group 3). The median age at time of diagnosis was 69 years (Table 1).
At baseline, FT levels below the detection limit were more common in group 3 than in groups 1 and 2 (p < 0.001). PSA, FT and TT were higher in group 1 than in groups 2 and 3, whereas hemoglobin was lower. Overall, 95.5% patients had bone metastases and 62.7% had soft tissue metastases. A median of two organs were involved by metastatic disease. All other characteristics were well-balanced (Table 1).

Effects of docetaxel chemotherapy on TT and FT.
In the overall study population, serum levels of TT were reduced from a median of 0.12 ng/mL at baseline to non-detectable levels (< 0.05 ng/mL) at nadir (p = 0.014) and FT levels were reduced from 0.32 pg/mL at baseline to non-detectable levels at nadir (< 0.18 pg/ mL) during docetaxel chemotherapy (p = 0.006; Fig. 1). The rate of patients with TT levels under the detection limit (< 0.05 ng/mL) increased from 23/55 (41.8%) at baseline (all 23 receiving abiraterone at that time point) Table 1. Patient characteristics, prior treatments and laboratory values at baseline before start of docetaxel chemotherapy for 67 patients. ECOG Eastern cooperative oncology group, BMI body mass index, TUR-prostate transurethral resection of the prostate, RTX radiation therapy, PSMA prostate-specific membrane antigen, ADT androgen-depleting therapy, ART androgen receptor targeted therapy, PSA prostate-specific antigen, FT free testosterone, TT total testosterone, Hb hemoglobin, AP alkaline phosphatase, ULN upper limit of normal, LDH lactate dehydrogenase, NSE neuron-specific enolase, CgA chromogranin A, NLR neutrophil to lymphocyte ratio. p values were estimated by applying t-tests to evaluate significant differences between the groups at baseline.

Multivariate analyses.
Univariate analyses of baseline parameters associated with OS are summarized in Table 3. In multivariate analyses with stepwise regression, only prior ART and the presence of soft tissue metastases (lymphatic, hepatic, pulmonary, brain) remained independent predictors of a shorter OS (Table 3).
In univariate and time-dependent covariate analyses (T_; ln(T_); T_ ≥ 365 days; data not shown) of treatmentdependent parameters (e. g. FT-, PSA-response), only FT median < 0.3 pg/mL and complete FT suppression (reduction of 100%) during docetaxel therapy were associated with a longer OS. In multivariate analysis, complete FT reduction (100%) remained an independent predictor for better OS (Table 3).
In addition to extended time-dependent Cox modelling, log(-log(survival probability)) and conditional landmark analyses were performed and demonstrated that the FT reduction = 100% during docetaxel therapy had no association with time regarding OS (data not shown).
In multivariate analysis of the baseline parameters, only the requirement of non-narcotics remained an independent predictor of a better PFS (Table 3). In univariate analysis of treatment-dependent parameters, PSAR, FT suppression (100%), FT median values < 0.2 pg/mL and FT nadir values < 0.18 pg/mL during docetaxel therapy were associated with a better PFS. In multivariate analysis with stepwise regression, only PSAR, FT reduction = 100% and FT nadir values < 0.18 pg/mL remained independent predictors of a better PFS. Applying extended time-dependent Cox modelling (T_; ln(T_); T_ ≥ 365 days), log(-log(survival probability)) and conditional landmark analyses demonstrated that these parameters had no association with time, with the exception of ln(T_) for PSAR, which revealed a significant time-dependence (p = 0.015, Table 3) (data not shown).

Discussion
In this study, we demonstrate that TT and FT serum levels are reduced during docetaxel chemotherapy and that FT suppression under the detection limit (100%) resulted in better PFS and OS in mCNPC and mCRPC patients, but not in mCRPC patients with a history of ART. Interestingly, in contrast to TT, only FT was a significant predictor for PFS and OS, demonstrating a major biological role of FT for treatment outcome. mCRPC-ART patients had a significantly lower FT reduction rate due to low FT levels at baseline (8/27 vs. 6/7 and 11/20, respectively) and FT reduction was no longer a predictor for better PFS or OS (Table 2). mCRPC-ART patients experienced a lower PSAR and shorter PFS and OS. These results are consistent with several previous studies that showed decreased efficacy of docetaxel in PC patients with a history of ART 11,12,21,22 . Our data suggest that the worst clinical outcome of mCRPC-ART patients towards docetaxel is due to progressing castration resistance.
Multiple lines of evidence suggest that docetaxel and prednisone might directly interfere with testosterone biosynthesis and metabolism in mPC patients and contribute to this FT suppressing effect. Prednisone was demonstrated to lower serum TT, androstenedione, dehydroepiandrosterone (DHEA) and dehydroepiandrosterone sulfate (DHEAS) levels in some metastatic PC patients by suppressing the hypothalamic-pituitary-adrenal axis, but had no antitumor activity in mCRPC 23,24 . Consistent with this observation, a history of previous prednisone treatment had no effect in multivariate analyses on PFS or OS in our study.
The role of docetaxel in reducing testosterone levels is less clear, although serum androgens (TT, androstenedione and DHEA) decline during docetaxel treatment 5 . Docetaxel metabolism is largely catalyzed by CYP3A4 25 and docetaxel was shown to induce CYP3A4, which is responsible for the greatest portion of testosterone 6βand 16β-hydroxylation [26][27][28] . CYP3A4 induction may lower testosterone levels by inactivation through 6β-and 16β-hydroxylation. The effect of docetaxel on CYP17A1 is unclear [26][27][28] .
Franke et al. reported castration-dependent pharmacokinetics of docetaxel in PC patients. Docetaxel clearance was increased by approximately 100% in castrated men and was associated with a two-fold reduction in area under the curve, although hepatic activity of CYP3A4 was unchanged 29 . Conversely, castration-naïve patients were exposed to higher amounts of the drug, which was accompanied by more severe hematotoxicity 29 . This study also demonstrated that lower intracellular docetaxel levels caused by lower baseline levels of testosterone resulted in a lower response rate to treatment 29 . Consistent with these results, group 1 patients (mCNPC) had a significantly higher rate of grade 3 and 4 neutropenia compared to group 3 patients (mCRPC-ART) in our study (57.1% vs. 20.6%, p = 0.047; Table 2) and a significantly better clinical outcome (Figs. 2, 3, Table 2).
Ryan et al. showed that conversion from higher to lower androgen levels (e.g. above/below median) during docetaxel therapy contributed to superior survival as the reduction is the driving mechanism behind the clinical      In recent years, several large phase 3 trials in patients with mCNPC (e.g. CHAARTED, STAMPEDE, GETUG-3, LATTITUDE, TITAN, PREVAIL) demonstrated that the addition of docetaxel and ART (abiraterone, apalutamide and enzalutamide) to ADT is associated with significantly improved PFS and/or OS compared to ADT alone 30 . Our data demonstrate that docetaxel therapy is associated with similarly low testosterone levels (FT + TT) as achieved by Abiraterone + ADT (Fig. 1, Tables 1, 2).
As with many retrospective analyses, the retrospective design of our study also has some limitations. For example, the patient population size is small, and although high numbers of FT and TT measurements were accomplished, these were not always assessed on a regular basis (e.g. weekly). Furthermore, progression was mainly due to PSA progression (e.g. PSA progression or radiographic progression, whichever presented first). Despite all of our efforts to address possible lead-time bias (e.g. use of extended time-dependent Cox modelling (T_; ln(T_); T_ ≥ 365 days), log(-log(survival probability)) and conditional landmark analyses), there is still the risk that our analyses are subject to lead-time bias, as group 3 patients (mCRPC-ART) had more advanced Table 3. Baseline factors influencing OS and PFS: univariate and multivariate analysis. HR hazard ratio, CI confidence interval, ECOG Eastern Cooperative Oncology Group, ART androgen receptor targeted therapy, FT free testosterone, CRP C-reactive protein, NLR neutrophile to lymphocyte ratio, PD progressive disease, PSA prostate-specific antigen, FT free testosterone. ** = time-dependent covariate analysis revealed for FT reduction = 100% and FT median < 0.3 pg/mL p > 0.05 for overall survival applying T_, ln(T_) and T ≥ 365 days. ♦ = ln(T) revealed for PSA response p = 0.015, for the remaining time dependent parameters using T_,ln(T) and T ≥ 365 days there was no significant time-dependence. www.nature.com/scientificreports/ disease at baseline compared to groups 1 (mCNPC) and 2 (mCRPC). Group 1 was small due to a recent trend towards abiraterone treatment in this setting and patients had a very high-volume disease that required intensive treatment. In addition, scanning intervals were not always uniformly assessed and confirmatory scans were not conducted in general.

Conclusion
This study represents the strongest evidence to date that FT plays a fundamental role during docetaxel chemotherapy. In mCNPC and mCRPC patients, complete FT suppression (= 100%) during chemotherapy was an independent predictor of PSAR, RR, PFS and OS. However, in mCRPC patients with a history of ART, FT was not linked to the clinical outcome. Our data suggest that castration-dependent pharmacokinetics of docetaxel seem to reduce its clinical effectiveness in mCRPC-ART patients.