Development and validation of a model of PSA kinetics predicting prostate cancer aggressiveness during screening
Le Saux O. (1, 2), You B. (1, 2), Henin E. (1), Pierrillas P. (1), Perrin P. (3), Ruffion A. (3), Boniol M. (4), Freyer G. (1, 2), Tod M. (1), Colomban O. (1)
(1) EMR 3738 CTO, UCBL - HCL Faculté de Médecine Lyon-Sud, Université Lyon 1, Oullins, France. (2) Medical Oncology Unit, Centre Hospitalier Lyon Sud, Pierre-Bénite, France. (3) Urology Department, Centre Hospitalier Lyon Sud, Pierre-Bénite, France. (4) Strathclyde Institute of Global Public Health, Biostatistics, International Prevention Research, Lyon, France.
Objectives: Prostate cancer (PC) is the most common cancer among men [1]. A growing number of cancers is detected by screening. Tools to distinguish aggressive and indolent tumors are necessary to avoid over-diagnosis and overtreatment. We developed a kinetic model of PSA (Prostate-Specific Antigen), a protein produced by the prostate gland, to differentiate the aggressive PCs among screened PCs.
Methods: We used the data from the American PLCO trial [2], developed to evaluate PC screening. Between 1993 and 2011, 76693 men, enrolled at 10 study centers across the United States, were randomized to annual PSA screening for 6 years (n=38343) or to usual care. Patients, aged 55-74, were excluded in the context of prior PC or current cancer treatment. Individual pre-operative PSA data were analyzed with a semi-mechanistic non-linear mixed effect model, using NONMEM 7.3. Selection and evaluation of the best model were achieved using likelihood, goodness-of-fit plots and simulation-based diagnostics. Data from the screening group was randomly split in a learning (2/3) and a validation set (1/3). Aggressiveness of PC was defined as: biopsy Gleason score ≥7, and/or clinical stage ≥III, and/or fatal.
Results: On the 38343 screened men, 4250 developed a PC, 1643 developed an aggressive PC and 152 died from this cancer. Various models were evaluated. PSA kinetics was best described by a non-steady-state one compartment model with first and zero orders production by malignant and non-malignant prostate cells respectively and first order elimination. According to goodness-of-fit plots, PSA kinetics in healthy and cancer patients were properly fit over the 6-year period, and Visual Predictive Check showed good agreement between observed and simulated values. Relative Standard Errors of typical mean parameters and inter-individual variability were all less than 5%. Over the study period, median PSA production by cancer cells was 2 times more important than non-malignant cells’ production.
Agreement of the proposed model in the validation set, along with the relationships between kinetic parameters and PC aggressiveness will be presented.
Conclusions: Our semi-mechanistic model describes PSA kinetics in healthy and cancer patients. If relationships between kinetic parameters and PC aggressiveness were demonstrated, it would provide an interesting tool for distinguishing the most aggressive tumors among screened PC and for adjusting treatment delivered to patients.
References:
[1]. Howlader N, M.A., Krapcho M, et al., SEER Cancer Statistics Review (CSR) 1975–2010. National Cancer Institute Website. (http://seer.cancer.gov/csr/1975_2010/); Updated June 14, 2013 Accessed January 8, 2014.
[2]. Andriole, G.L., et al., Mortality results from a randomized prostate-cancer screening trial. N Engl J Med, 2009. 360(13): p. 1310-9.
