Login to Access Video or Poster Abstract: MP70-07
Sources of Funding: None

Introduction

Whole-gland salvage cryosurgery (SCS) is a potential curative treatment for organ-confined radiorecurrent prostate cancer (PCa). To enhance patient selection and optimize follow-up, a prediction model was created for biochemical failure (BF).

Methods

Data on patients (pts) treated (1995-2004) with salvage SCS at one centre was prospectively collected. Recurrences were biopsy proven and metastatic disease was excluded with pelvic/abdominal CT and radionuclide bone scan. Cox regression was adopted to assess the influence of clinical characteristics on BF. Missing data was imputed 20 times. Factors with a p-valueÿ0.25 were left in the model. The model was internally validated using bootstrap resampling (500 times) after which the C-statistic and hazard ratios could be adjusted for optimism. Calibration at different time points was performed and a risk score were created to assess different prognostic groups.

Results

152 pts had follow-up data. A total of 89 pts experienced BF according to the Phoenix-definition (PSA-nadir+2 ng/ml). Median follow-up was 117 months (interquartile range 56-154). Five and ten year biochemical disease free survival (BDFS) was 45% (95%-CI 37-54%) and 35% (95%-CI 27-45%). Age at SCS, pre-salvage PSA, Gleason score and PSA-nadir after treatment were associated with BF after multivariable regression (table 1), adjusted C-statistic 0.76. The model was well calibrated up to 10 years. Four risk groups were created (score <22, 22-25, 25-30 and >30, see table 1). BDFS estimates at ten years were 68%, 41%, 27% and 12%, respectively (log-rank p<0.0001)(fig.1).

Conclusions

Selection of salvage pts is challenging, since guidance in the literature is scarce as to prognostic value of clinical characteristics. The presented model can guide patient selection and individualize follow-up. The model is not externally validation. Applicability might therefore be limited for other centres performing SCS.

Funding

None