*2.2. Clinical Outcome and Prognostic Variables*

In the 26,434 patients analyzed for OS, the median values for OS in Cohorts A, B, and C were 26 (95% confidence interval (CI) 25.0–27.0), 26 (95% CI 25.3–26.7), and 30 (95% CI 29.1–30.9) months (Figure 2A). In the 26,032 patients analyzed for CSS, the median values of CSS were 31 (95% CI 29.7–32.3), 31 (95% CI 30.1–31.9), and 35 months (95% CI 32.4–33.6) in Cohorts A, B, and C, respectively (Figure 2B). The detailed age-standardized 1- to 5-year OS and CSS are shown in Figure 3.


**Table 1.** Basal characteristics of patients.

**Figure 2.** Kaplan–Meier estimations of overall survival (OS) (**a**) and cancer-specific survival (CSS) (**b**) according to cohort allocation. *p*-value from log-rank test.

Age, race, and metastatic stage (the latter was only analyzed in Cohorts B and C) were identified as significant prognostic factors at univariate analysis (data not shown) and were included in the multivariable models.

#### *2.3. Multivariable Models*

The multivariable models for OS and CSS showed a substantially increased risk of death according to age, with the highest risk in patients ≥85 (Tables 2 and 3). Black patients showed a slightly higher risk of death compared to white, whereas Asians/Pacific Islanders showed better outcomes compared to white. A 9% decreased risk of death and an 8% decreased risk of cancer-specific death were found in Cohort C compared with Cohort A (hazard ratio (HR): 0.91 (95% CI 0.87–0.95), *p* < 0.001 for OS; HR: 0.92 (95% CI 0.88–0.96), *p* < 0.001 for CSS), whereas no statistically significant differences in OS and CSS were found between Cohorts A and B. Exploratory multivariable models were also performed in Cohorts B and C to include the metastatic stage classification (AJCC, 6th edition), which was found to be associated with distinct OS and CSS outcomes (Tables S2 and S3). In these multivariable models, significant OS and CSS advantages were reported in Cohort C compared with Cohort B (HR: 0.94 (95% CI 0.91–0.97), *p* = 0.001 for OS; HR: 0.89 (95% CI 0.85–0.92), *p* < 0.001 for CSS). In the exploratory subgroup analysis comparing the OS and CSS of Cohort C with Cohort B, a significant interaction was found among the subgroups of the AJCC metastatic classification. More pronounced OS and CSS advantages in Cohort C were shown in M1c patients compared with patients with metastases that were limited to nodes or bone (M1c HR: 0.87 (95% CI 0.81–0.94), interaction *p* = 0.014 for OS; M1c HR: 0.81 (0.75–0.88), interaction *p* = 0.015 for CSS) (Table 4).

**Figure 3.** Age-standardized 1- to 5-year OS (**A**) and CSS (**B**) of patients according to year of diagnosis.


**Table 2.** Multivariable analysis for OS.

**Table 3.** Multivariable analysis for CSS.



**Table 4.** Subgroup analysis of OS and CSS between Cohorts C and B.

Multivariable models including age and race were used to compute the hazard ratios (HR) and their 95% confidence intervals (CI) for OS and CSS in the metastatic subgroups of patients diagnosed in 2011–2014 vs. 2004–2010. \* *p*-value for interaction; <sup>1</sup> Multivariable model including age, race and metastatic stage for OS and CSS (Cohort C vs. Cohort B).

#### **3. Discussion**

Several randomized trials demonstrated that both chemotherapy and ARSi provided a significant survival benefit in mPCa [1–8]. However, the real-world survival outcomes of patients with de novo mPCa remain largely unexplored.

A recent analysis compared 590 patients with mCRPC, who were diagnosed and treated in two treatment eras (2004–2007 vs. 2010–2013) at the Dana–Farber Cancer Institute [11]. The authors demonstrated a 41% decreased risk of death in the newer treatment era, with a median OS gain of 6 months. In addition, the cumulative benefit from the newer therapies was more pronounced in longer-term survivors and de novo patients. Although this study provided useful information, all patients had castration-resistant disease, only 216 had de novo mPCa, and they were all managed in a top-level institution.

In another study, Helgstrand and colleagues analyzed the incidence and mortality data of patients with de novo mPCa included in the SEER database and in the Danish Prostate Cancer Registry [13]. In patients diagnosed between 2000 and 2009, the median OS was 22 months in SEER and 30 months in the Danish Registry. The five-year overall mortality was 80.0% in both registries in the period of 2000–2004, remained stable (80.5%) according to SEER in 2005–2008, and decreased to 73.2% according to the Danish Registry in 2005–2009.

Although the monocentric experience of the Dana–Farber Cancer Institute and the Danish data confirmed the potential survival gain offered by newer treatments, the SEER analysis by Helgstrand and colleagues did not show substantial survival changes after 2004.

In the present SEER-based analysis, we investigated whether the introduction of both chemotherapy and ARSi in mCRPC had substantially changed the real-world OS and CSS in the population of patients with de novo mPCa diagnosed in the United States of America in three different time periods (2000–2003—Cohort A, 2004–2010—Cohort B, 2011–2014—Cohort C). Although the patients were allocated to these cohorts regardless of having received a specific treatment, we highlight that docetaxel was approved by the FDA for the treatment of mCRPC in 2004, whereas ARSi was approved from 2011 onwards (Figure 1).

More than 26,000 patients diagnosed between 2000 and 2014 were included in our analysis; of these, 6047 were allocated to Cohort A, 11,815 to Cohort B, and 8572 to Cohort C (Table 1). We found that age had a significant impact on patients' OS and CSS (Tables 2 and 3). In the multivariable model, patients older than 85 showed a double risk of dying compared with patients between 15 and 54 years old, and the hazard ratio for death was also significantly unfavorable in patients aged 75–84. Although this figure might be at least in part attributable to the reduced expected survival, older patients may also be less likely to receive the same treatments as their younger counterparts, especially chemotherapy.

We did not find a significant difference in the OS and CSS between Cohort A and Cohort B (Figure 2). Conversely, we observed a statistically significant improvement in the OS and CSS of patients included in Cohort C, who showed a decreased risk of death of 9%, a decreased risk of cancer-specific death of 8%, and a median OS gain of 4 months compared with Cohort A. The comparison of Cohort C with Cohort B, adjusted for the metastatic stage, also demonstrated an OS improvement of 6% and a CSS improvement of 11%. When compared with the other metastatic stages, we found that patients with M1c disease showed the worst survival, but had a more pronounced OS and CSS improvement in the newer ARSi era compared with M1a or M1b patients (Table 4). Although the reason for this observation remains unknown, the presence of visceral metastases might lead to more aggressive pharmaceutical approaches and more adherence to treatment that could result in increased benefit compared with the other stages.

The median OS gain of chemotherapy and ARSi in randomized trials for mCRPC was 2–4 months in first-line [1,5,6] and 4–5 months in second-line [3,4]. Although our study was not designed to demonstrate the potential benefit of chemotherapy or ARSi, a more robust OS and CSS improvement would have been expected in patients diagnosed in 2011–2014, after the introduction of several agents in clinical practice (Figure 1). A median OS improvement of 4 months in Cohort C compared with Cohort A appears to be quite discouraging. Regardless of cohort analysis, the probability of survival after 3 years from diagnosis was 40.0% in 2000 and 46.8% in 2014 (Figure 3). Similarly, the five-year probability of survival was 24.0% in 2000 and 28.2% in 2012. Several reasons might explain these disappointing results.

First, the degree of benefit seen in clinical trials does not necessarily translate into the real-world setting. Screen failure rates on trials are relatively high and can easily affect the ultimate generalizability of trial results to the real-world population.

Second, our study was based on patients diagnosed with de novo mHSPC who were supposed to receive androgen-deprivation therapy (ADT) as a first-line treatment for metastatic disease, and subsequently docetaxel or ARSi as a first-line treatment for mCRPC. The number of patients who died without receiving a first-line treatment for mCRPC or refused therapies for mCRPC was unknown. The information on the number of lines of treatment, type of treatment, disease burden, number and site of metastases, body mass index, performance status, and comorbidities was not available in the SEER database, and these potential confounders were not included in our analysis. In addition, we acknowledge that some patients could have received chemotherapy or ARSi outside of the defined cohort allocation in the context of clinical trials or some years after mPCa diagnosis.

Third, the medical costs and the health insurance policies might have significantly reduced the extensive use of ARSi and chemotherapy in the general population of patients with de novo mPCa diagnosed and treated in the United States, affecting their survival outcomes. Ramsey and colleagues reported that the cumulative incidence of bankruptcy in the first 5 years after prostate cancer diagnosis is 38% (nearly 50% in metastatic stage), and the risk of mortality is almost twice as high among patients with prostate cancer who file for bankruptcy compared with those who do not [12]. Further studies should investigate whether insurance policies or limited access to healthcare services could contribute to such disappointing survival gains observed in the SEER registry after the introduction of chemotherapy and ARSi.

Fourth, patients with de novo mPCa showed worse time to castration and survival compared with those who relapsed after local therapy, irrespective of treatment received [14,15]. Therefore, the intrinsic aggressiveness of de novo mPCa could have also led to decreased survival gains in this patient population. Although discouraged by international guidelines in recent years, possible premature discontinuation of ARSi and chemotherapy based on PSA progression without clinical or radiographic progression could have also affected the outcome data of patients diagnosed between 2004 and 2014 [16].

Finally, we acknowledge that our study excludes the possible benefit induced by docetaxel or ARSi in mHSPC, given their approval for this setting in the latest years (Figure 1). The earlier use of these agents provided OS gains that exceeded 12 months in randomized trials for mHSPC [8]. Future analyses could also detect additional survival benefits that might be provided by an increased knowledge in the sequencing of agents for mCRPC and by the biomarker-driven selection of patients suitable for specific drugs (i.e., poly (ADP-ribose) polymerase (PARP) inhibitors) [17–19].
