Prognostic significance of the logistic regression model for assessing the risk of recurrence in patients with prostate cancer after radical prostatectomy

Abstract

Purpose of the study. The study’s objective is to investigate the influence of risk factors for recurrence of prostate cancer (PCa) after radical surgical treatment on the unfavourable course of recurrence of the disease with the formation of a set of the most significant factors of a model that reflects the likelihood of relapse.Materials and methods. A retrospective analysis of clinical, pathomorphological and perioperative parameters of 803 patients with PCa after radical prostatectomy was carried out. By means of logistic regression, a model for assessing the risk of recurrence for patients with prostate cancer was built, which included 7 indicators, one of which was measured by quantity (time between biopsy and start of treatment) and six categorical ones (ISUP grade group, cT, cN+, positive surgical margin, PSA level after surgery, pN+). The construction of a logistic regression model consisted in obtaining a characteristic of the logistic function Ψ for the standard equation y = exp(ψ) / (1 + exp(ψ)).Results. Substituting the coefficients obtained for each of the indicators, we obtain Ψ to assess the risk of relapse Ψ = 0.485 × X1+ 1.937 × X2 + 0.789 × X3 + 3.229 × X4 + 0.443 × X5 + 0.880 × X6 + 0.015 × X7–6.65. In the resulting formula, each of the regression coefficients describes the size of the contribution of the corresponding factor. In our case, all regression coefficients were positive, which means that this factor increases the overall risk of relapse. The quality of the resulting model is determined by the chi-square = 284.3; p < 0.001; OR = 28.45. The sensitivity of this model was 86.6 %, specificity 81.5 %, diagnostic accuracy 82.7 %.Conclusion. This model makes it possible to obtain the probability of recurrence after radical prostatectomy depending on the severity of a specific set of predictive signs (a positive effect is predicted for y > 0.5, a negative one for y ≤ 0.5) and the degree of influence of one or a group of predictive signs on the likelihood of relapse, such as the ISUP grade group, locally advanced disease, clinically detectable lymph node lesion, positive surgical margin, PSA level of more than 0.09 ng/ml 1 month after surgery, the presence of regional metastases and the time between biopsy and the start of treatment.