Accelerated failure time survival model for prostate cancer

Abstract

Background: Prostate cancer is the second most common cancer in men globally and ranks sixth among male cancer-related deaths. This study utilized survival analysis to assess post-diagnosis outcomes, aiming to enhance prostate cancer patient longevity. Insights were gained by determining median survival times, identifying prognostic factors, and applying an AFT model to prostate cancer data.

Methods: The study collected data from 502 prostate cancer patients featured in a 1980 paper by D. P. Byar and S. B. Green, focusing on stage 3 or 4 cases. Over six years, they monitored these patients' survival under varying estrogen doses, analyzing 15 variables. Data import was done using excel, R and SPSS were used for the analysis. Utilizing the Kaplan-Meier method and AFT models, the study identified predictive factors and determined the most suitable model for prostate cancer analysis.

Results: Among the 502 patients, 70.7% died, and 29.3% were right-censored, with a median survival of 35 months. Factors enhancing survival included a 1.0 mg estrogen dose, higher body weight, and elevated serum hemoglobin levels. In contrast, age, tumor size, cardiovascular disease history, and a composite index of stage and histology grade were associated with reduced survival odds in prostate cancer patients.

Conclusions: Prostate cancer patients had a median survival of 35 months, with survival rates at 60 months (31.7%) and 75 months (23.2%). The 1.0 mg estrogen dose proved most effective among tested doses. The Weibull AFT model, with an AIC score of 3469.032, was the best-fit model.