A total of 163 patients fulfilled the inclusion criteria for this study and had long-term PSA monitoring follow-up records available. The age range of patients was 53–89 years, with a median age of 72 years. There were 12 patients at the T1-2N0M0 stage (limited early stage with PSA > 50 ng/mL), 32 at T3-4M0 (focal progressive), and 119 at M1 (metastatic). All patients had prostate cancer confirmed by prostate puncture biopsy. Nine (5.5%) patients underwent surgical castration and the remaining 94.5% were treated with subcutaneous goserelin acetate (3.6 mg once/28 days) as androgen deprivation therapy. All patients received oral bicalutamide (50 mg every day) concomitantly with castration treatment. The median follow-up was 43.5 months. Thirteen (8.0%) patients were lost to follow-up. Basic characteristics of the patients are summarized in Table 1.
Screening for prognostic risk factors
Risk factors associated with biochemical progression of PSA were analyzed using univariate and multivariate Cox regression. As shown in Table 2, in terms of PSA biochemical progression, only whether PSA declined to ≤ 2 ng/mL (hazard ratio [HR] 0.462, P = 0.001), and PSA time to nadir less than 9 months (HR 1.736, P = 0.021) were independent risk factors for disease progression. PSA half time more than 3 months tended to increase the risk of disease progress (HR 1.346), however, it was not statistically significant (P = 0.338). PSA falling to < 0.2 ng/ml were associated with a good prognosis, while the risk for disease progression was lower when the time to PSA nadir was longer.
Dynamic changes in PSA levels and survival with biochemical progression-free
Absolute levels of PSA after treatment: As shown in Table 3, the lower the PSA nadir value, the lower the risk for bPFS. After controlling for other variables (Gleason score, stage, initial PSA level, age, body mass index, and tumor load), lower PSA nadir value still predicted lower bPFS.
Rate of PSA change after treatment: As shown in Table 4, for patients in whom PSA level decreased to < 4 ng/ml (“normalization”) after treatment, the risk for PFS was lower when the PSA “normalization time” was shorter, but a statistical correlation with castration resistance could not be demonstrated. In this case, when grouped of 8 ~12, 12~24, and >24 weeks for intergroup comparison, shorter time to nadir was associated with better patient prognosis, and PSA “normalization time” more than 24 weeks could be used as an independent predictor (HR 0.695, P = 0.001). This suggests that prognosis is better when PSA level falls into the reference ranges faster after androgen deprivation therapy.
Analysis of PSA time to nadir—in contrast to PSA “normalization time”—revealed that longer PSA time to nadir was associated with a decreased risk for bPFS and remained a statistically significant predictor after controlling for other risk factors. In this study, the same conclusions were obtained by choosing 6, 10, and 12 months as the cut-off values.
Finally, in terms of PSA halving time, group comparisons using 0.5, 1, 1.5, and 3 months as cut-off values revealed that the shorter the PSA halving time, the worse the prognosis. The faster the PSA decreased to the lowest value, the worse the prognosis was, regardless of the adjustment for PSA nadir value. However, none of the group comparisons demonstrated statistically significant.
Optimization of PSA nadir values and time cut-offs for reaching the nadir: Results showed (Figure 1) that the area under the curve (AUC) for PSA nadir and PSA time to nadir were 0.804 and 0.833, respectively, which were > 0.7 but both were < 0.9, indicating that both had a certain degree of accuracy for prognostic judgments. In particular, the sensitivity of PSA nadir value was 65.7% and the specificity was 73.6% when 0.2 ng/ml was used as the best cut-off value. Moreover, PSA nadir > 0.2 ng/ml suggested a poor prognosis, while PSA nadir ≤ 0.2 ng/ml suggested a good prognosis. Similarly, for PSA time to nadir, 9 months was the most appropriate threshold, with a sensitivity of 71.6% and a specificity of 73.9%. Patients with a time to nadir > 9 months had a better prognosis.
In addition, the AUC for PSA halving time was 0.563, which was < 0.7, thus indicating a lack of accuracy of this indicator as a prognostic criterion for bPFS.
As shown in Figure 2, the median bPFS was 27.6 months and 13.5 for cases in two groups with PSA nadir < 0.2 ng/ml and ≥ 0.2 ng/ml, respectively, and there was significant difference between groups (P < 0.001 [log-rank]). The median bPFS was 27.8 months and 13.5 months for cases in two groups with TTN ≥ 9 months and < 9 months, respectively, and there was significant difference between the two groups (P < 0.001 [log rank]).