ERASMUS

Survival Analysis

Exercises for Wednesday (Day 3)

Analysis of Competing Risks:

1. On the class web page is data from a study of 194 patients with squamous cell carcinoma. The study was from a clinical trial conducted by the Eastern Cooperative Oncology Group. Patients were followed until failure defined as: local spread of disease (cause 1); or metastatic spread of disease (cause 2) -- there were 67 subjects whose follow-up was censored. Three covariates are of interest:

performance status 0=ambulatory,

1=non-ambulatory;

treatment 0=treatment A,

1=treatment B;

age in years at start of the study.

(a) Provide appropriate statistical summaries that describe the overall time-until failure (disease spread). Such an analysis considers all failure causes combined.

(b) Plot Kaplan-Meier curves for each of the two treatment groups. What do these plots suggest about treatment?

(c) Use Cox regression to look at the association between the risk of either local or metastatic spread of disease and the three covariates (e.g. analysis of all-cause failure). Interpret your model estimates.

(d) Use Cox regression to evaluate the association between the three covariates and the ``cause-specific'' risk of disease spread for each of the two causes. Interpret your model estimates.

(e) The investigator that reported analysis of these data used an exponential model for each cause-specific hazard model. Provide summaries that either support this model assumption, or suggest that it was inadequate. (note this investigator is a former chair of biostat at a major east coast university).

(optional) Produce cause-specific cumulative incidence curves by treatment group, and by performance status. Interpret these curves. Specifically, interpret how these displays support the regression conclusions that you obtained in 1(d).