Covariate level stratification assessment of mortalities due to competing risks and breast cancer in Ghana

Abstract

Ascertaining the extent of competing risk in survival data that has the capacity to set the basis for selecting appropriate models for analysing and predicting breast cancer mortality rates has been of interest. In this study, the estimates of two cumulative incidence functions were compared with the complement of the Kaplan-Meier to assess the presence of competing events. The results revealed that the complement of the product-limit estimator (1-KM) overestimated the survival probability at periods beyond the 32nd month of follow-up, an indication of competing risk. Older patients (>50 years) were found to be more likely to die from competing events (P<0.01,x2=52.26)than those in the age category of ≤50 years. Further analysis was conducted when the estimates of the Fine-Grey and Cox regression models were compared. The estimates were largely not identical, coupled with the overestimation of hazards in favour of the Cox model, a further sign of the presence of competing risk. Breast cancer deaths and deaths due to competing events were examined through stratification by each level of each covariate. The triple-negative (TRN) breast cancer molecular subtype was the most associated with the greatest absolute risk, which peaked at about 80 % by the 60th month. Patients who experienced a recurrence of BC and experienced competing events had about 35 % higher absolute risk by the 60th month. Again, BC patients who were hospitalized have about 25 % higher absolute risk than their counterparts who were not hospitalized. Poorly differentiated BC tumors (Grade 3) were associated with a 40 % increase in absolute risk in comparison with moderately differentiated tumors (Grade 2). Pre-menopausal patients experienced about 7 % increased absolute risk compared with post-menopausal patients. Older patients were solely the category that experienced the risk of dying from competing events, accounting for an absolute risk of 22 % by the 60th month. It is recommended for Ghanaian clinicians to integrate these findings into BC management in order to optimize treatment.

Keywords

• Competing risk
• Breast cancer
• Ghana
• Kaplan-Meier
• Cox model
• Fine-grey model