Use of artificial neural networks in evaluating prognostic factors determining the response to dendritic cells pulsed with PSMA peptides in prostate cancer patients

BACKGROUND: Our purpose was to compare the importance of over 22 measurements used in evaluating the clinical responses of patients with metastatic or locally recurrent prostate cancer, treated by dendritic cell (DC) infusions with prostate-specific membrane antigen (PSMA) peptides. METHODS: Artificial neural networks (ANNs) were employed for assessment, as well as the traditional methods of logistic regression. RESULTS: Twenty-six patients with metastatic disease and 37 patients with local recurrence were available for evaluation and comparison. ANN evaluation ranked the collective effects of DC infusion, immune responses (CD3+ cells, CD16+ cells, zeta chain+ cells), and cytokines, e.g., IL-6 and PSMA levels, very highly. Logistic regression identified all of these parameters to some degree, but in a different rank order. Patients with metastases showed a sharp rate of response secondary to the level of DC infusion, in contrast to those patients with local recurrence, in which it was more gradual. CONCLUSIONS: ANN analysis emphasizes the importance of level of DC infusion, immune parameters, cytokines, and markers such as PSMA in determining the response to PSMA peptide immunotherapy. The criteria of response were judged to be correct in 86% of metastatic patients and 83% of locally recurrent patients evaluated in this study.