Docetaxel vs. Novel Hormonal Agent for Upfront Management of De Novo Metastatic Hormone-Sensitive Prostate Cancer: A Comprehensive Report of a Single-Center Experience

Background: Upfront docetaxel or novel hormonal agents (NHA) such as abiraterone and enzalutamide have become the standard of care for metastatic hormone sensitive prostate cancer (mHSPC). However, data comparing the ecacy of docetaxel and NHAs in this setting are limited. Patients and Methods: This was a retrospective cohort study of patients with de novo mHSPC treated with upfront docetaxel or an NHA between January 1, 2014 and April 30, 2019 within the Mount Sinai Health System. Clinical data were extracted from the medical record. The primary outcome was failure-free survival (FFS), dened as the time to next treatment. The primary predictor was treatment with docetaxel or NHA. FFS was compared between the two groups using the Kaplan Meier method and multivariable Cox proportional hazards models. We additionally assessed the prognostic value of post-treatment PSA. Results: We identied 94 de novo mHSPC patients; 52 and 42 treated with upfront docetaxel and NHAs, respectively. NHAs were associated with signicantly longer FFS compared to docetaxel (20.7 vs. 10.1 months, p=0.023). In a multivariable model adjusting for demographics and clinical factors, docetaxel was independently associated with worse FFS compared to NHAs (HR 1.96, 95% CI 1.12−3.45, p=0.019). High metastasis burden patients had a signicantly longer FFS with NHAs than docetaxel (25.12 vs. 9.63 months, p=0.014), while there was no signicant difference in FFS among low metastasis burden patients (NHA 20.71 vs. Docetaxel 26.5 months, p=0.9). Regardless of treatment, lower post-treatment PSA levels were associated with improved FFS (58.95 vs. 11.57 vs. 9.4 months for PSA ≤ 0.2, 0.2-0.4, >0.4ng/ml, respectively; p<0.001) Conclusion: Comparative analysis of real-world data demonstrated longer FFS in de novo


Background
The standard of care for metastatic hormone sensitive prostate cancer (mHSPC) continues to quickly evolve. While the use of androgen deprivation therapy (ADT) remains foundational, the earlier incorporation of advanced therapeutic agents previously reserved for castration resistant disease has become routine in patients with mHSPC. The CHAARTED and LATITUDE trials led to FDA approvals of docetaxel and abiraterone for mHSPC in 2015 and 2017 respectively; each demonstrating signi cant overall survival (OS) advantages over ADT alone [1,2]. Enzalutamide and apalutamide approvals followed in 2019, based on positive data from ARCHES, ENZAMET, and TITAN trials [3][4][5].
Data to guide the optimal selection among these 1st line treatment regimens for mHSPC are limited.
Meta-analyses comparing clinical trials of upfront docetaxel versus novel hormonal agents (NHA) in mHSPC have not shown any overall or progression-free survival bene t for either class of medication over the other [6][7][8]. To ll this gap, we performed a retrospective analysis of de novo mHSPC patients treated with docetaxel or novel hormonal agents (NHA) within the Mount Sinai Health System. We developed an automated oncology data retrieval and curation platform, and comprehensively extracted clinical features, outcomes, toxicities, and treatment patterns from the electronic medical record (EMR) in order to characterize and compare the clinical outcomes of patients treated with either upfront docetaxel or NHA for de novo mHSPC.

Data Source
Patients with prostate cancer were identi ed from the Mount Sinai Hospital at New York City data warehouse. Original EMR data from the hospital data warehouse were processed using the Sema4 Centrellis™ platform, consisting of automated abstraction engine, curation platform, patient dashboard, and cohort builder. Along with information from structured data elds, data from unstructured clinical notes were automatically abstracted, curated and integrated into structured elds. Cancer diagnosis metastases, medication, treatment, PSA values and other related data elements were stored in the database Patient Cohort Identi cation We identi ed prostate cancer patients diagnosed between January 2014 and April 2019 to capture the time frame when docetaxel and NHA were approved for mHSPC. 11,358 patients were assigned with at least one prostate cancer related ICD codes (185 in ICD-9 code and C61 in ICD-10 code) between January 1st, 2014 and April 30th, 2019 in the EMR system. However, not all patients with prostate cancer ICD codes assigned have con rmed prostate cancer. In addition, for patients diagnosed prior to visiting Mount Sinai hospital, their rst time ICD code recorded in the EMR system didn't always match the date of diagnosis. To correctly identify patients with pathologically con rmed cancer diagnosis dates, we used the automated oncology data retrieval and curation platform developed by Sema4 as demonstrated in Supplemental Fig. 1A. Both structured information and extracted data elements from unstructured clinical note using heuristic rules were used to assign cancer diagnosis dates as demonstrated in Supplemental Fig. 1B. Patients with their extracted cancer diagnosis dates and other diagnosis related data elements, such as stage, histology, TNM, were classi ed into three groups using a strati cation algorithm. Selected patients' data were manually reviewed by domain experts to evaluate the quality of the results We excluded patients diagnosed after April 2019 to ensure at least 12 months of follow-up. We de ned de novo metastatic disease as clinical documentation of metastatic disease within 3 months of the diagnosis date. Upfront docetaxel and NHA use were determined by chart review. After the cohorts were identi ed, chart review was carried out to con rm duration of upfront docetaxel of NHA therapy, adverse events, identi cation of 2nd line therapy, and start date of 2nd line therapy. Additional clinical and demographic variables such as PSA values, age, sex, race, performance status, Gleason score, and metastasis burden at time of diagnosis were also evaluated. This study was approved by the Mount Sinai Institutional Review Board (Protocol number: STUDY-21-00442).

Failure-Free Survival Analysis
Failure-free survival (FFS) was the primary outcome measure for this study, de ned as time from initiation of either docetaxel or NHA to the next line of treatment. For the purposes of this metric, any change or addition of systemic therapy -including chemotherapy, NHA immunotherapy, and clinical trials -was considered a failure of therapy, regardless of the reason for the change (e.g. progression, toxicity, patient preference, physician discretion). Patients not known to receive a 2nd line treatment were censored at the time of their last follow-up. FFS was represented as Kaplan-Meier curves as detailed under Statistical Analysis.

PSA Progression Free Survival (PFS) Analysis
PSA PFS was de ned as the time from initiation of therapy to rst PSA increase ≥ 25% and ≥ 2 ng/mL above that patient's nadir, con rmed by a second value at least 3 weeks later, per PCWG3 criteria [9].
Patients who were started on 2nd line therapy before criteria for PSA progression were met were excluded from this analysis. Patients who did not have PSA progression and who had not started 2nd line therapy were censored at the time of their last follow-up. PSA PFS was represented as Kaplan-Meier curves as detailed under Statistical Analysis.

Metastasis Burden and Failure Free Survival
High metastasis burden was de ned as visceral metastases or ≥ 4 bone lesions with ≥ 1 beyond the vertebral bodies and pelvis, per the CHAARTED trial [2]. Patients were assigned: high, low, or uncertain metastatic burden based on chart review by an oncologist. Differences in FFS between patient with high and low burden disease were compared in the 1. overall mHSPC cohort; 2. docetaxel treated cohort; and 3. NHA treated cohort. We further compared FFS between patients treated with docetaxel or NHA, strati ed by metastasis burden. FFS were represented as Kaplan-Meier curves as detailed under Statistical Analysis.

PSA nadir and Failure-Free Survival
The post-treatment PSA in the docetaxel cohort was de ned as the PSA value following completion of at least 4 cycles of docetaxel therapy. Patients who started 2nd line treatment prior to nishing 4 cycles of docetaxel (e.g. due to toxicity or refractory disease) were excluded from this analysis. For the NHA cohort, post-treatment PSA value was recorded after patients had received at least 7 months of ADT and at least 3 months of NHA. Patient were categorized in the docetaxel and NHA cohorts into three separate groups based on post-treatment PSA values: ≤0.2, 0.2-4.0, or ≥ 4.0 ng/mL. We compared FFS according to posttreatment PSA within each treatment cohort.

Statistical Analysis
Failure-free survival and PSA progression free survival data were presented as Kaplan-Meier curves. Survival curves were created with R (version 3.5.0; http://www.R-project.org/.) using the survival package (version 2.44-1; https://cran.r-project.org/web/packages/survival/index.html) and report the log rank p-value according to default parameters. We examined potential confounding variables that may also be associated with clinical outcome using single variable and multivariable Cox proportional hazards regression, reporting the hazard ratios and the Wald test statistic p-value for each variable. PSA values were log transformed due to the heavy right tail, and 2 individuals with missing values were assigned to the median value.

Patient cohort
Total 5725 patients were identi ed by the platform with pathologically con rmed diagnosis dates for prostate cancer between January 1, 2014 to April 30, 2019. Of these, 240 were determined to have had de novo metastatic disease, based on clinical documentation of metastatic disease within 3 months of the diagnosis date. The diagnosis date, metastatic status, and date of metastasis detection were based on terms extracted from clinical notes and pathology reports via the abstraction engine described above. We then identi ed patients who had been treated with docetaxel (N = 67) or an NHA (abiraterone or enzalutamide) (N = 99) at any time. Finally, we identi ed those who had been treated with upfront docetaxel (N = 52) or an NHA (N = 42), de ned as treatment within 7 months of diagnosis in conjunction with ADT and in the absence of any prior systemic treatment for prostate cancer.
The two groups of de novo mHSPC patients were balanced with respect to baseline characteristics; there were no signi cant differences in age, race/ethnicity, PSA at diagnosis, Gleason score, ECOG performance status, and metastatic burden. Table 1
Forty-ve patients (86.5%) completed 6 cycles of treatment. Six patients (11.5%) terminated docetaxel early due to adverse effects, most commonly leg edema, fatigue, LFT abnormalities, anaphylactic infusion reaction. One patient (1.9%) had docetaxel terminated before 6 cycles and moved to 2nd line treatment due to refractory disease.

Failure-free survival
Patients who received upfront NHA had a signi cantly longer median FFS of 20.7 months (95% CI = 17.0-NE[not estimable]) compared to 10.1 months(95% CI 8.75-18.2) in docetaxel treated patients (p = 0.023) (Fig. 1A). Single and multivariable analyses are presented in Table 2. NHA, low metastasis burden, and lower baseline PSA levels are all associated with longer FFS. Variables with pvalues < 0.15 were retained for the multivariable regression model. In the multi-variable model adjusting for age, baseline PSA, and metastasis burden, docetaxel was associated with worse FFS compared to NHA (HR 1.96, 95% CI 1.12 − 3.45, p = 0.019) ( Table 2). None of the other variables remain statistically signi cant, but age below 65 at time of diagnosis, high baseline PSA, and high metastatic burden are still modestly associated with shorter FFS, irrespective of treatment received (  (Fig. 2B). However in a cox regression model predicting FFS, the interaction between upfront therapy and metastasis burden was not signi cant (p = 0.28) We further analyzed the prognostic implications of metastasis burden in the combined mHSPC cohort and in the individual docetaxel and NHA cohorts separately. In the combined cohort of docetaxel and NHA treated patients, a high metastatic burden of disease was associated with a shorter median FFS as compared to patients with low burden of disease, 11.57 [95% CI 9.24-20.12] vs. 20.71 [95% CI 12.92-NE] months (p = 0.034) (Fig. 2C). In the docetaxel group, high burden disease is associated with shorter median FFS as well, 9.63 [95% CI 8.38-14.53] vs. 26.5 [95% CI 8.65-NE] months (p = 0.026) (Fig. 2D)

PSA progression
Since time to next treatment may be in uenced by decisions to change therapy due to toxicity rather than disease progression, we also evaluated PSA progression between the two treatment groups. Of note, 7 of 52 docetaxel-treated patients were started on a subsequent line of therapy prior to completing 6 cycles of docetaxel treatment (due to toxicity [6 patients] or refractory disease [1 patient], as mentioned above); PSA progression after docetaxel alone could not be evaluated in these patients. Among evaluable patients, we found that those treated with docetaxel were more likely to start a subsequent therapy in the context of rising PSA levels prior to meeting PCWG criteria for PSA progression (14/45 vs. 4/42, p = 0.02). The median PSA difference between PSA nadir and PSA at the start of 2nd treatment was 0.68 vs 1.58 ng/ml among those treated with docetaxel vs NHAs (p = 0.55).
The co-variants were reasonably well matched between the remaining 30 docetaxel-treated patients and 38 NHA-treated patients with respect to age, race, log PSA at time of diagnosis, Gleason score, ECOG score, and metastasis burden (Supplemental Table 5). Among these patients, there was a trend towards longer PSA PFS with NHA compared to docetaxel, though the difference was not statistically signi cant (19.0 [95% CI 15.6-NE] vs. 13.2 months [95% CI 7.7-22.6], p = 0.069) (Fig. 1B). Supplemental Table 6 presents the results for cox regression for both the univariable and multivariable analyses. In multivariable analysis, treatment (docetaxel vs. NHA) was not independently associated with PSA PFS (HR 1.64, 95% CI 0.87-3.09; p = 0.13) (Supplemental Table 6. Higher log PSA at time of diagnosis was independently associated with worse PSA PFS (Supplemental Table 6).

Post-treatment PSA as Prognostic Marker in de novo mHSPC
Prior to the advent of combination therapies for de novo mHSPC, when such patients were treated with ADT alone, post-ADT PSA levels were shown to be independent predictors of overall survival [11]. It is not known whether post-treatment PSA also holds prognostic value in the context of modern combination treatments.
To evaluate the prognostic signi cance of post-treatment PSA, we examined FFS strati ed by posttreatment PSA in the overall study cohort, as well as in the individual docetaxel and NHA cohorts. Patients were strati ed into 3 groups based on post-treatment PSA: ≤0.  (Fig. 3A). Similar outcomes were observed in the individual docetaxel and NHA cohorts ( Fig. 3B and C)).
In a multivariable analysis of the combined cohort, higher post-treatment PSA was independently associated with worse FFS (0.  Table 7).

Discussion
The treatment paradigm for mHSPC has rapidly shifted over the past 5 years from ADT monotherapy to combinations of ADT with docetaxel or NHAs such as abiraterone, enzalutamide, and apalutamide. However, data comparing these treatment options is limited. Network meta-analyses have addressed the question indirectly using cross-trial comparisons, nding no difference in overall survival between docetaxel and NHAs [6][7][8]. An opportunistic analysis from the STAMPEDE trial with limited power found improved failure-free survival and progression-free survival with abiraterone versus docetaxel, but no difference in overall survival [10].
Observational data may improve our understanding of the real-world use and outcomes of these treatments. We therefore conducted this study to compare the real-world use and outcomes of docetaxel vs. NHA in de novo mHSPC patients. To our knowledge, this is the rst observational analysis to directly address this question.

Failure Free Survival and Metastasis Burden
Upfront NHA was associated with longer median FFS (measured by time to next treatment) than docetaxel and remained an independent predictor of FFS after controlling for demographic and clinical factors. This corresponds with ndings from STAMPEDE, where the NHA cohort also showed a signi cantly longer FFS than the docetaxel cohort [10].
Time to next treatment was shorter with docetaxel than NHAs. Several factors may have contributed to this observation. First, more docetaxel-treated patients changed therapy due to toxicity (6 vs. 0, p = 0.022). Second, there appeared to be a lower clinical threshold for docetaxel-treated patients to be initiated on a 2nd line therapy compared to NHA-treated patients. For example, 14 docetaxel-treated patients were started on a 2nd line therapy before meeting formal criteria for PSA progression, compared to only 4 of NHA-treated patients (p = 0.03). In addition, the median increase in PSA from nadir at the time of 2nd line treatment initiation was 0.68 ng/ml among docetaxel-treated patients compared to 1.58 ng/ml among NHA treated patients; this difference was not statistically signi cant (p = 0.55), but suggests that clinicians were more inclined to initiate 2nd line therapy in docetaxel-treated patients than NHA-treated patients. The fact that docetaxel treatment is limited to 6 cycles whereas NHAs are taken on an ongoing basis may in uence clinicians' response to rising PSA levels.
High burden of metastatic disease has been associated with greater overall survival (OS) bene t from upfront docetaxel in post-hoc analyses of clinical trials. In our FFS analysis, NHAs were associated with median FFS of over 20 months regardless of disease burden (high: 25.12 vs. low: 20.71 months) while median FFS after docetaxel was signi cantly shorter in patients with high rather than low metastatic burden (9.63 vs. 26.5 months, p = 0.014). These observations raise the possibility of treatment effect heterogeneity between docetaxel and NHAs by metastatic burden. We assessed the interaction term, which was not signi cant (p = 0.28), but this analysis was limited by sample size. We were unable to assess OS in this cohort due to a lack of events; thus while NHAs were associated with longer FFS in high-burden patients, it is not clear whether this would translate into an OS bene t.

Second-Line Therapy
At the end of follow-up, 17.3%, of patients remained on ADT alone after completion of docetaxel, whereas 52.4% of the NHA cohort remained on 1st line therapy. This was due in part to earlier approval for docetaxel in this setting and, consequently, longer follow-up in the docetaxel cohort (maximum of 64 months) compared to the NHA cohort (maximum of 28 months). The apparent lower clinical threshold to initiate 2nd line treatment among docetaxel-treated patients, described above, may also have contributed.
Patterns of toxicity leading to dose reductions of treatment drug re ect otherwise well-described adverse effect pro les. Most patients, 86.5%, were able to successfully complete 6 cycles of docetaxel, with the main reasons for early termination being fatigue, peripheral edema, and LFT abnormalities. Dose reductions in docetaxel were recommended for fatigue, neuropathy, and febrile neutropenia. Meanwhile, no clear cases of early termination of an NHA was observed, and dose reductions occurred only in a minority of patients, primarily in response to abnormal LFT's or hot ashes.
The prevailing practice pattern following upfront docetaxel was to transition to an NHA, with a relatively even split between abiraterone and enzalutamide. Following upfront NHA, most practitioners opted to transition to a second NHA. However, understanding that sequencing two different AR-pathway targeting agents back to back is usually associated with relatively low rates of response [12], more recent studies have explored transitioning to chemotherapy [13], radium, or targeted therapy with PARP inhibitor [14].

PSA as a Prognostic Indicator
In the pre-CHAARTED era, where standard of care for mHSPC was ADT alone, post-ADT induction PSA level was shown to have strong prognostic signi cance, with lower PSA levels associated with better survival outcomes [11]. To our knowledge, there is no data examining the prognostic value of posttreatment PSA in the context of modern combination therapies for mHSPC. In the overall study cohort, we observed that patients with a post-treatment PSA of ≤ 0.2 enjoyed the longest FFS, whereas those with PSA 0.2-4 and ≥ 4 were associated with signi cantly shorter FFS. This suggests that the post-treatment PSA after combination treatment in mHSPC patients has prognostic value, though further evaluation of its association with overall survival is necessary.

Limitations
There are inherent limitations in our data given the retrospective nature of this study. Comparisons between the docetaxel and NHA-treated patients may be confounded by indication. We were unable to perform propensity-score matching due to limited sample size. However, baseline characteristics were not signi cantly different between the two groups.
Enzalutamide and apalutamide, which were recently approved for mHSPC, were not well-represented in our dataset. Future studies should address these newer agents as experience with them accrues.
We were unable to evaluate overall survival due to a limited number of events. We chose time to next treatment as a clinically relevant intermediate endpoint. Time to next treatment may be affected by several factors including frequency of follow-up, toxicity, and patient or provider-preference. We provided information on the decision to change treatment where available and found that most treatment changes were prompted by PSA progression. We conducted a separate PSA PFS analysis among a subset of evaluable patients which showed a trend towards longer PSA PFS with NHAs compared to docetaxel, though this association was not signi cant. Larger observational cohorts or prospective data are needed to compare long-term survival outcomes between these two treatment strategies for mHPSC.

Conclusion
In conclusion, this retrospective single-center study observed that upfront NHA treatment for de novo mHSPC was associated with longer FFS compared to docetaxel. Median FFS was longer among high-burden patients treated with NHAs rather than docetaxel, whereas median FFS was similar regardless of treatment in low-burden patients; the interaction was not signi cant. Finally, we showed that in the context of modern combination treatments for mHSPC, post-treatment PSA levels were associated with time to next treatment, suggesting prognostic value of this marker which should be evaluated in larger, prospective datasets. Abbreviations NHA-novel hormonal agents; mHSPC-metastatic hormone sensitive prostate cancer; ADT-androgen deprivation therapy; FFS-failure-free survival; EMR-electronic medical records; SDM-standardized mean difference; PFS-progression free survival; OS-overall survival Declarations Ethics approval and consent to participate This retrospective analysis was approved by the Mount Sinai Health System Institutional Review Board (Protocol number: STUDY-21-00442). All experimental protocol for involving human data is in accordance with the Mount Sinai Health System (institutional) and Declaration of Helsinki guidelines. Participant consent was obtained in writing before they were included in the database for analysis and the study was approved by the Mount Sinai Health System ethics committee as mentioned above.

Consent for publication
Not applicable Availability of data and materials The datasets generated and analyzed during the current study are not publicly available as it is part of Sema4 company database but are available from the corresponding author on reasonable request.

Competing Interests
The authors have declared no con ict of interest