Marieke van Son

150 CHAPTER 8 DISCUSSION This study provides two clinically useful multivariable prediction models for BF in pa- tients with radiorecurrent PCa treated with FS-HDR-BT. Model 1 can be used to support clinical decision making and patient guidance at baseline, while model 2 could be used during follow-up to counsel patients regarding their prognosis and potentially adapt follow-up intensity accordingly. The predictors in both models and the direction of their effects were mostly as ex- pected. Increased age was associated with a lower hazard of BF. Although causal inference is not applicable in prediction, this could be explained by the potentially longer disease-free survival interval (DFSI) between primary and salvage treatment indicating more indolent tumors. DSFI was longer in elderly patients (median 92 versus 108 months for <75 years versus ≥75 years, respectively). Data on pre-salvage Gleason score is mostly lacking in our cohort, which hinders assessing this relation. Both a higher pre-salvage PSA level and larger GTV were associated with an increased hazard. Both indicate higher tumor load and were therefore expected to be correlated with BF. For pre-salvage PSADT, which was non-linearly related to the outcome, hazard decreased with longer doubling times. This was expected given previous reports[15]. However, from approximately 32 months onward, the hazard increased slightly again, as displayed by a HR of 1.18 for PSADT’. PSADT was ≥32 months in only 19 patients (12.7%). Median post-primary PSA nadir, post-salvage PSA nadir, and pre-salvage PSA were higher in these patients compared to those with a PSADT of <32 months (1.1 vs 0.5 ng/ml, 0.9 vs 0.6 ng/ml, and 6.1 vs 4.6 ng/ml, respectively), but the percentage of patients classified as high-risk (NCCN) at primary treatment was comparable (39% vs 42%). Therefore, we have no clear explanation, and these findings might be caused by the limited sample size. Seminal vesicle involvement, which is a sign of extensive dis- ease, was associated with an increased hazard of BF. A longer post-salvage time to PSA nadir was associated with a lower hazard, potentially reflecting tumor biology (a faster response after radiotherapy could be a sign of more malignant/dedifferentiated PCa) as previously observed[27]. Finally, a larger reduction in PSA level was protective of BF. Several studies have identified predictors for BF in patients with radiorecurrent PCa treated with focal or whole-gland salvage high-intensity focused ultrasound (HIFU), low-dose rate brachytherapy (LDR-BT), and cryotherapy[15,28–30]. However, it is ques- tionable to what extent predictors from whole-gland salvage studies are applicable to focal salvage treatments. Spiess et al. reported a risk stratification model in a whole- gland salvage cryotherapy cohort (n=132), using the Phoenix definition of BF[29]. Upon multivariable analysis, post-salvage PSA nadir and pre-salvage Gleason score were identified as predictors for BF. PSA nadir was also identified as a predictor of BF after salvage whole-gland HIFU in a small cohort of 50 patients[30]. Peters et al. showed that DFSI between primary and salvage treatment, T-stage before salvage, prostate volume (cm3), PSA, and PSADT were predictors of BF in patients treated with focal salvage HIFU[15]. This model shows overlap with our model, indicating that pre-salvage PSA