A Comparative Analysis of Accelerated Failure Time Models on the Survival of Breast Cancer Patients

A comparative analysis of three parametric survival models was carried out. In particular, the Weibull, Lognormal, and Log-logistic models were used to estimate the survival rates of 105 patients diagnosed and treated of breast cancer at the Federal Medical Centre, Makurdi from June, 2014 to May, 2019. The Log-likelihood, Bayesian Information Criteria (BIC) and Akaike Information Criteria (AIC) were used to determine the best model. Results revealed that the Weibull AFT model was the best model compared to the Lognormal AFT and Log – logistic AFT models. Further results showed sharp downward slope for both survival and hazard plots. These are indication of a decrease in survival rate over time as well as a high rate of hazard in the early days after diagnosis. Using the Weibull AFT model, it was observed that stage at diagnosis and occupation were the significant risk factors at 0.05 level of significance. The study also revealed that 96.20% death from breast cancer occurred at stage 4 with 0.12 probability of survival. Only 3.8% of patients who were diagnosed at stage 3 died and with no death recorded of patients diagnosed at stage 1 and 2. The immediate implication is that early diagnosis is vital to the survival of breast cancer patients.