Per Kragh Andersen: A multi-state model for bleeding
   episodes and mortality in liver cirrhosis.

In a clinical trial in patients with liver cirrhosis
and esophageal varices, 286 patients were randomized
in a 2 by 2 factorial design to either prophylactic
sclerotherapy, to prophylactic treatment with
propranolol, to both or to neither. The purpose was to
reduce mortality and incidence of bleeding episodes.
Thus, data consist of times of bleeding and death,
and various approaches to the modelling of these data
will be discussed, including a competing risks model with
the two endpoints 'bleeding' and 'death', a standard survival
analysis neglecting information on bleeding episodes,
and a multi-state model with three possible transitions:
'no bleeding' -> 'bleeding', 'no bleeding' -> 'death'
and 'bleeding' -> 'death'.

In particular, it will be studied how survival probabilities
for patients with given characteristics may be estimated
in the various models and the uncertainty of the estimators
will be compared.