Marieke van Son

29 RADICAL VS. FOCAL TREATMENT FOR LOCALIZED PROSTATE CANCER from 2017 onwards). All patients underwent MRI either for staging prior to focal therapy and radiotherapy, or to guide surgical technique regarding nerve sparing prostatectomy. Outcome assessment The primary outcome was failure-free survival (FFS), a composite endpoint of (1) need for local salvage treatment, (2) development of metastatic disease, (3) use of systemic treatment (ADT or chemotherapy) or (4) progression to a watchful waiting (WW) strat- egy. Secondary outcome was overall survival (OS). Prostate cancer-specific survival could not be assessed, as causality of death was often difficult to gauge. Salvage treatment was defined as any secondary treatment after EBRT, prostate bed radiother- apy for rising PSA after LRP if there were no adverse pathologic findings and >1 focal re-do or any whole-gland treatment after FT. WW was defined as no intention to treat despite biochemical recurrence after EBRT (PSA nadir+2 ng/mL) or LRP (PSA>0.2 ng/ mL) or histologically proven recurrence after focal (ISUP ≥2 of any length). Prostate biopsies were mostly taken after two consecutive PSA rises and suspected recurrence on mp-MRI, with a small proportion of patients undergoing standard prostate biopsies as part of the FT protocol. Statistical analysis All analyses were performed using R version 3.5.0. To compare treatments, a PSW-anal- ysis was performed using the matching weights approach 25,26 . Missing data was con- sidered to be missing at random and was imputed upfront with single imputation (mice package). Each patient was assigned a propensity score based on age, PSA, ISUP grade, MCCL, T-stage and year of treatment (VGAM package). Patients were then weighted to correct for imbalances between treatment groups, with more weights applied to pa- tients with equal probabilities of assignment to either treatment group. After weighting, covariates with a standardized mean difference (SMD) <0.1 were considered sufficiently balanced between treatment groups. Next, a weighted Cox regression analysis was performed to estimate the average treatment effect on hazard of failure and mortality (survey package). To visualize survival over time, PSW-adjusted Kaplan Meier survival curves were fitted, using a weighted log-rank test to detect differences in FFS and OS (survey package). All analyses were also performed in a three-way setting (EBRT versus LRP versus FT), comparing multiple pairs at once. For all three-way analyses, the signif- icance level was set at p <0.017 (Bonferroni correction). For all two-way comparisons, significance was set at p <0.05. RESULTS Overall, 440 EBRT, 390 LRP and 530 FT patients were eligible. Treatment details are summarised in Table 1. Although patients may have had different types of treatment failure, the total number of failures represents each patient’s first event. Local sal- vage treatment after EBRT consisted of focal HIFU (n=2). LRP patients received either 2