In total, 146 patients were included in this study. Of them, 91 patients (62%) had de novo metastasis at prostate cancer diagnosis, 22 patients (15%) received radical prostatectomy and 37 patients (25%) received local radiation therapy. All the patients received ADT, the median (IQR) time to CRPC was 11 (6–27) months and 101 patients (69%) received previous ARAT therapy. The patients’ backgrounds at the start of DOC and CBZ treatment are presented in Table 1. The median duration of follow-up was 16 months from the start of DOC treatment. The percent PSA decrease exceeded 50% in 52 patients (36%) during DOC treatment and 35 (24%) during CBZ treatment (Fig. 1A). The changes of PSA levels during DOC and CBZ treatment were not correlated (R2 = 0.0005, Fig. 1B, C). The median (IQR) number of cycles of DOC and CBZ treatment were 6 (4–10) and 4 (3–8), respectively. The median (95% CI) PFS for DOC and CBZ were 4 (3–4) and 3 (2–3) months, respectively (Fig. 1D, E). The time to progression after DOC and CBZ treatment was not correlated (R2 = 0.0003, Fig. 1F).
In univariate Cox proportional hazard analyses, the associations of PFS from the start of DOC and CBZ treatment with the following variables were examined: age > 70 years, PS ≥ 1, GS ≥ 9, previous ARAT use, PSA > 20 ng/mL at DOC or > 40 ng/mL at CBZ, Hb < 13 g/dL, LDH > 220 U/L, ALP > 280 U/L, and ≥ 4 bone metastases or visceral metastases. The existence of ≥ 4 bone metastases or visceral metastases was associated with shorter PFS of DOC, and LDH > 220 U/L was associated with shorter PFS of CBZ (Table 2). The median (95% CI) combined PFS for DOC-CBZ was 11 (10–13) months. Cox proportional hazard analyses were performed for the combined PFS of DOC-CBZ using the parameter levels at the start of DOC treatment. PSA > 20 ng/mL at the start of DOC treatment was significantly associated with shorter PFS in univariate and multivariate analyses (hazard ratio [HR] = 1.81, 95% CI = 1.20–2.72, P = 0.004, Table 3). The combined PFS of DOC-CBZ was compared between patients with PSA > 20 ng/mL and PSA ≤ 20 ng/mL at the start of DOC treatment. The combined PFS of DOC-CBZ was significantly longer in patients with PSA ≤ 20 ng/mL (median PFS = 12 months, 95% CI = 11–16) than in patients with PSA > 20 ng/mL (median PFS = 10 months, 95% CI = 8–11) at the start of DOC treatment (P = 0.003, Fig. 2A) and was significantly longer in patients with PSA ≤ 40 ng/mL (median PFS = 13 months, 95% CI = 11–16) than in patients with PSA > 40 ng/mL (median PFS = 9 months, 95% CI = 8–11) at the start of CBZ treatment (P = 0.017, Fig. 2B).
The median (95% CI) OS from the initiation of DOC was 22 (18–26) months. Cox proportional hazard analyses were performed for OS using the same parameters used for PFS. In univariate analyses, PSA > 20 ng/mL, Hb < 13 g/dL, and ALP > 280 U/L were significantly associated with shorter OS. In multivariate analyses, PSA > 20 ng/mL remained significantly associated with shorter OS (HR = 2.55, 95% CI = 1.45–4.48, P = 0.001, Table 4), together with Hb < 13 g/dL and LDH > 220 U/L. OS from the initiation of DOC treatment was significantly longer in patients with PSA ≤ 20 ng/mL (median OS = 31 months, 95% CI = 22–44) than in those with PSA > 20 ng/mL (median OS = 18 months, 95% CI = 15–25) at the start of DOC treatment (P = 0.002, Fig. 3A). OS from the initiation of DOC treatment was also significantly longer in patients with PSA ≤ 40 ng/mL (median OS = 34 months, 95% CI = 22–42) than in those with PSA > 40 ng/mL (median OS = 18 months, 95% CI = 15–22) at the start of CBZ treatment (P = 0.003, Fig. 3B). In patients with PSA ≤ 20 ng/mL and PSA > 20 ng/mL at the start of DOC treatment, the median PSA levels at the start of CBZ were 15 ng/ml and 100 ng/ml, respectively. In patients with PSA ≤ 20 ng/mL at the start of DOC treatment, OS from the initiation of DOC treatment was not significantly different between patients with PSA ≤ 15 ng/mL (median OS = 31 months, 95% CI = 16–not reached) and PSA > 15 ng/mL (median OS = 25 months, 95% CI = 22–47) at the start of CBZ treatment (P = 0.962). In patients with PSA > 20 ng/mL at the start of DOC treatment, OS from the start of DOC treatment was significantly longer in patients with PSA ≤ 100 ng/mL (median OS = 26 months, 95% CI = 13–37) than in those with PSA > 100 ng/mL (median OS = 16 months, 95% CI = 13–20) at the start of CBZ treatment (P = 0.048, Fig. 3C).
The patients starting DOC at lower PSA levels might include patients for whom primary ADT was more effective. However, time to CRPC was not significantly different between high PSA (median: 10 months) and low PSA (median: 12 months) at the start of DOC (p = 0.29). Starting DOC at lower PSA levels means that DOC was started earlier in disease progression. OS from the initiation of DOC seemed to be longer when DOC treatment was started earlier. To resolve the bias, the OS from the development of CRPC was examined. OS from CRPC was compared between patients with PSA > 70 ng/mL and PSA ≤ 70 ng/mL at the start of ADT, between patients with PSA > 20 ng/mL and PSA ≤ 20 ng/mL at the start of DOC treatment, and between patients with PSA > 40 ng/mL and PSA ≤ 40 ng/mL at the start of CBZ treatment. OS did not differ according to PSA levels at the start of ADT (median OS: 44 months vs. 48 months, P = 0.166, Fig. 4A), but it was significantly longer in patients with low PSA levels at the start of DOC (median OS: 55 months vs. 41 months, P = 0.030, Fig. 4B). In addition, OS tended to be longer in patients with low PSA levels at the start of CBZ (median OS: 70 months vs. 44 months, P = 0.064, Fig. 4C).