Illustration of association between change in prostate-specific antigen (PSA) values and time to tumor status after treatment for prostate cancer patients: A joint modelling approach

Abstract

Abstract Background Biomarkers’ information utilize for prediction of new patient’s suffering using statistical approaches to model outcomes. Simultaneous analysis of prostate-specific antigen (PSA) and time to tumor shrinkage after treatment allows it to make dynamic prediction adjustable over time. The motivation of this work is to elaborate the dynamic prediction about individual-specific trajectories of prostate cancer progression and tumor shrinkage, using baseline patients’ characteristics and longitudinal PSA data. Methods We developed joint models to predict low tumor status after treatment considering repeated PSA measurements, adjusted for baseline body-mass-index. We modeled 1,504 prostate cancer patients’ data jointly using mixed-effect sub-model for longitudinal PSA and an event time sub-model for tumor shrinkage. Results Baseline prognostic factors affected PSA level over time; also, an association of PSA on time with tumor status is evaluated. Patients who observed low tumor status had low PSA values during follow-up and also had exponentially decreasing PSA within one month after treatment. Factors included in this analysis were significant in mixed-effects and event time sub-models. Conclusions Medical science has evolved from subjective prediction to dynamic prediction, based on statistical knowledge with use of joint modelling of longitudinal data and event time outcomes. It provides individual event prediction with dynamic nature of time-accumulated evidence-based tools. We developed a joint model to predict prostate cancer patients’ tumor status after treatment with observing longitudinal PSA. It is evaluated that dynamically updated PSA information can improve prognosis, which is used to make follow-up and treatment decision-making easy and fruitful.