Patients background in ARSI or vintage treatment groups

From the original cohort including 304 patients for ARSI and 295 patients for vintage treatment (Table S1), we selected 426 mHSPC patients (213 patients from each group) using PSM to align their background. Characteristics of patients are summarized in Table 1. Median age and iPSA level were 74 years old and 209.5 ng/ml in the ARSI group, and 75 years old and 169.98 ng/ml in the vintage group, respectively. The percentage of ISUP GG5 was 66.2% in the ARSI group and 67.1% in the vintage group. Visceral metastasis was observed in 19.3% of the ARSI group and 16.5% of the vintage group. High-volume and high-risk disease were seen in 63.4% and 69% of the ARSI group and 62% and 65.3% of the vintage group. No significant differences were observed between the ARSI and vintage groups except for cN stage (P = 0.0048) and serum ALP level (P = 0.0375). In the ARSI group, abiraterone acetate was the most commonly used upfront treatment (55.4%) followed by apalutamide (28.2%) and enzalutamide (16.4%). In the vintage group, all patients in this study received bicalutamide administration. Median follow-up period was 24 months, and 93 patients died during this study.

Table 1 Characteristics of patients after Propensity Score MatchingPSA kinetics in the ARSI and vintage treatment group

Kaplan–Meier analysis showed that patients who received an ARSI as the initial treatment was associated with prolonged PSA PFS as compared to those who received a vintage agent (P = 0.0063) (Fig. 1A). Median PSA PFS was 38.5 months in the ARSI group and 28.5 months in the vintage group, respectively. Regarding OS, no significant difference was seen between two groups (P = 0.3134) (Fig. 1B).

Fig. 1

Comparison of survival between ARSI and Vintage treatment. A PSA PFS, B OS

In terms of PSA response in the early stage of treatment, medial PSA levels at 3 months were 0.52 ng/ml in the ARSI group and 1.47 ng/ml in the vintage groups (P = 0.3042). On the other hand, ARSI treatment group achieved a lower PSA level at 3 months than vintage treatment group when the cut-off PSA value was set at ≥ 0.5 ng/ml (Table 2). A PSA level < 0.5 ng/ml at 3 months was seen in 49.1% of patients in the ARSI group and 36.9% of patients in the vintage group (P = 0.0205) at 3 months. When looking at the deeper PSA response (PSA nadir < 0.2 ng/ml), no significant differences were found between two groups.

Table 2 PSA dynamics in each treatment group

In terms of PSA reduction rate at 3 months, no significant difference was observed between the two groups (Table 2). PSA reduction ≥ 90% was seen in 86.5% of the ARSI group and 88.3% of the vintage group (P = 0.8726).

Furthermore, no significant difference was observed in nadir PSA level with median PSA nadir of 0.1345 ng/ml in the ARSI group and 0.263 ng/ml in the vintage group (P = 0.1228) (Table 2). No significant differences were seen with all cut-off values of PSA nadir (Table 2). A nadir PSA level < 0.2 ng/ml was seen in 52.8% of the ARSI group and 46.6% of the vintage group (P = 0.1052).

Time to PSA nadir was shorter in the ARSI treatment group (213.5 days) than in the vintage group (243 days; P = 0.0003) among progression-free cases, although no significant difference was observed in patients with progression (185.5 days vs 175.5 days, respectively; P = 0.7509).

Additional analysis using iPSA values (> 200 ng/ml or ≤ 200 ng/ml) was performed. The median PSA values at 3 months were 0.708 ng/ml for the ARSI and 4.413 ng/ml for the vintage in patients with iPSA > 200 ng/ml, whereas they were 0.4365 ng/ml for the ARSI and 0.448 ng/ml for the vintage in those with iPSA ≤ 200 ng/ml (Table S2 and S3). With respect to nadir PSA level, the rate of achieving PSA < 0.2 ng/ml was higher in the ARSI group (49.5%) than in the vintage group (32.2%) in patients with iPSA > 200 ng/ml (P = 0.0151) although no significant difference was seen in those with iPSA ≤ 200 ng/ml, indicating the ARSI treatment is more likely to have an evident PSA-suppressing effect when iPSA levels are high (Table S2 and S3).

Cut-off values for nadir PSA level and time to nadir to predict OS

We investigated the cut-off values for nadir PSA and time to nadir to best stratify OS in each treatment group using survival tree analysis (Fig. 2). As a result, the optimal cut-off nadir PSA level was 1.5 ng/ml for the vintage group and 0.45 ng/ml for the ARSI group (Fig. 2A). The optimal cut-off time to nadir was 70 days for the vintage group and 145 days for the ARSI group.

Fig. 2

Methodology of survival tree analysis

Kaplan–Meier analysis showed that patients who achieved a PSA nadir ≤ 1.5 ng/ml had prolonged OS compared to those who did not in the vintage group (P < 0.0001) (Fig. 3A). In the ARSI group, patients who achieved a PSA nadir ≤ 0.45 ng/ml had better OS than those who did not (P < 0.0001) (Fig. 3B). Regarding analysis of time to PSA nadir, men with time to nadir ≥ 145 days was associated with improved OS than time to nadir < 145 days in the vintage group (P < 0.0001) (Fig. 4A). Furthermore, a time to nadir of 70 days allowed significant stratification of prognosis in the ARSI group (P < 0.0001) (Fig. 4B). Notably, the prognosis of men who achieved PSA nadir ≤ 0.45 ng/ml in the ARSI group was similar to that of patients with PSA nadir ≤ 1.5 ng/ml in the vintage group (3-year survival was 85.1% and 87.7%, respectively: P = 0.8582) (Fig. S1). Furthermore, the prognosis of patients who did not have favorable response was similar between the ARSI and vintage groups (P = 0.8835) (Fig. S1). In addition, we validated the prognostic impact of the same PSA nadir level and time to nadir in both treatment groups (Fig. S2−5). No significant differences in OS were observed between the ARSI and vintage treatment when the same cut-off was used.

Fig. 3

Survival analysis divided by optimal cut-off PSA nadir level. A Vintage treatment group. B ARSI treatment group

Fig. 4

Survival analysis divided by optimal cut-off for time to PSA nadir. A Vintage treatment group. B ARSI treatment group

PSA kinetics as a prognostic factor for OS

To examine the prognostic impact of PSA kinetics for OS, we performed Cox proportional hazard modeling using other clinical factors. In the ARSI treatment group, elevated LDH (> 194U/L), PSA nadir (> 0.45 ng/ml), and time to PSA nadir (> 70 days) were identified as predictive factors for OS in univariate analysis (Table 3). After multivariate analysis, PSA nadir (> 0.45 ng/ml) was the only independent predictive factor for OS (hazard ratio (HR) = 4.66, 95% confidence interval (CI): 1.96–12.35, P = 0.0004) (Table 3).

Table 3 Uni- and multivariate cox proportional hazard models for OS in ARSI group

In the vintage treatment group, clinical T stage 4 (HR = 1.97, 95%CI: 1.12–3.45, P = 0.02), EOD score ≥ 2 (HR = 1.83, 95%CI: 1.01–3.31, P = 0.046), PSA nadir (> 1.5 ng/ml) (HR = 2.39, 95%CI: 1.26–4.53, P = 0.0075), and time to PSA nadir (> 145 days) (HR = 3.88, 95%CI: 2.1–7.16, P < 0.0001) were independent predictive factors for OS in multivariate analysis (Table 4).

Table 4 Uni- and multivariate cox proportional hazard models for OS in Vintage groupCharacteristics of patients with poor PSA response associated with worse survival

In addition, a novel risk classification was established using PSA nadir level and time to nadir period in each treatment group. We divided patients into Favorable (presence of both favorable factors), Intermediate (presence of only one favorable factor), and Unfavorable (presence of no favorable factors) groups. Kaplan–Meier analysis revealed that our risk classification allowed distinct stratification of survival outcomes for patients receiving vintage treatment (Fig. 5A). The Favorable group showed a significantly decreased risk of death as compared to the Intermediate (P = 0.003) and Unfavorable (P < 0.0001) response groups. Likewise in the ARSI treatment group, these risk classifications were able to stratify patient prognosis (Favorable vs Intermediate: P = 0.0028; Intermediate vs Unfavorable: P = 0.0237) (Fig. 5B).

Fig. 5

Survival analysis by risk classification using PSA nadir and time to PSA nadir. A Vintage treatment group. B ARSI treatment group

To identify the clinical factors correlated with poor PSA response and shorter survival, we further examined the characteristics of patients classified by our risk groups in both treatment groups. In the ARSI treatment group, men with Intermediate or Unfavorable responses showed higher rates of EOD score ≥ 2 as compared to men with Favorable response (P = 0.0226) (Table 5). The Favorable group had a higher frequency of M1c stage (P = 0.0044) due to lung metastasis (P = 0.0015), which may have resulted in favorable outcomes (Table 5). No significant differences in age, iPSA level, ISUP GG, baseline values of peripheral blood markers, disease volume/risk, or types of ARSI drugs were found between Favorable and Intermediate/Unfavorable groups (Table 5). On the other hand, ISUP GG and frequency of clinical T stage4 tended to be higher in the Intermediate/Unfavorable groups than Favorable group (P = 0.079 and P = 0.0778, respectively).

Table 5 Characteristics of patients regarding PSA response after ARSI treatment

In the vintage treatment group, ISUP GG5 (P < 0.0001), EOD score ≥ 2 (P = 0.0031), high-volume disease (P = 0.038), and high-risk disease (P = 0.0344) were more likely to be observed in the Intermediate/Unfavorable groups in comparison with the Favorable group (Table 6). No significant differences in age, iPSA, tumor stage, or baseline values of peripheral blood markers were seen (Table 6).

Table 6 Characteristics of patients regarding PSA response after Vintage treatment