Thierry Buclin (1), Hugo Teixeira Farinha (2), Patrice Jichlinski (3), Eric Grouzmann (1)
(1) Division of Clinical Pharmacology, (2) Faculty of Biology and Medicine, and (3) Urology Service; University Hospital (CHUV) and University of Lausanne, Switzerland
Objectives: PSA monitoring is used to detect cancer relapse after prostatectomy. PSA doubling time (PSADT) is a useful concept to interpret PSA results; however several modes of calculation compete. This retrospective observational study aimed to describe PSA trajectories and progression free survival after prostatectomy and to assess various PSADT calculations for their predictive performance regarding relapse.
Methods: 102 patients were drawn from a lab database over 10 years, having PSA concentration regularly monitored after prostatectomy (904 values). Their medical records were scanned for pTNM grade, Gleason score, capsular invasion, relapse-free survival, subsequent investigations and treatments. Relapse was defined by evidence of recurrence or metastases or by the initiation of secondary anticancer treatment. We modeled PSA trajectories according to Stein [1] using NONMEM 7, allowing prognostic factors to influence PSA regrowth rate. PSADT calculations used 1) the 2-point method, 2) the Log-slope method or 3) post-hoc individual predictions derived from population-based Bayesian estimates of PSA regrowth rate, from measurements limited to 1, 2, 3 and 5 years of follow up. We compared calculations including either all or only >0.1ng/mL PSA values, as sometimes recommended. The prognostic value of PSADT estimates (inversed) was assessed by survival analysis and Cox proportional hazard models.
Results: Cancer recurred in 52 patients (1-19 y follow up, median 5). Log-transformed PSA trajectories were fairly linear within patients, but markedly divergent between patients. T and N grades strongly predicted relapse (HR: 5.2 for T³3, 3.8 for N=1, p<0.001) and significantly influenced PSA regrowth rate (´5.14, resp. ´11.3). PSADT still improved outcome prediction, already at 1 year follow up, with Log-slope estimates predicting relapse (HR: 1.3 per y‑1 of PSADT‑1, p=0.04) better than Bayesian post-hoc regrowth rates (HR: 1.3, p=0.16) and 2-point estimates (HR: 1.0, p=0.97). Considering all PSA values was more efficient than selecting those >0.1ng/mL.
Conclusions: Known prognostic markers mainly account for highly variable PSA trajectories after prostatectomy for cancer. Yet the regular follow-up of PSA and calculation of PSADT remains warranted for relapse prediction and detection. A sophisticated population-based Bayesian approach does not improve the performance of the simple Log-slope method for PSADT calculation, while the 2-point method is worse.
References:
[1] Stein WD & al. Tumor growth rates derived from data for patients in a clinical trial correlate strongly with patient survival: a novel strategy for evaluation of clinical trial data. Oncologist. 2008;13:1046-54.
Reference: PAGE 21 () Abstr 2406 [www.page-meeting.org/?abstract=2406]
Poster: Other Drug/Disease Modelling