iv) In the analysis of multiple occurrence of distinct infections, assuming equality of the regression coefficients using a global test of significance gives non-meaningful results (mainly when a significant test of equality of the coefficient has been performed before). With this idea, in the studied example, it seems more interesting to consider death as a fourth kind of event, keeping different regression coefficients, rather than considering it as all kinds of events. Moreover, we cannot really justify discarding from consideration events which are not the first of each kind.
The different results, in terms of significance, show that it is important to define very cautiously the main endpoint in the planning of clinical trials. Nevertheless, none of these results can be interpreted in terms of a multivariate generalization of the event-free survival. The main issue is the weight associated with the mortality comparative to the infections; this weight is not determined by the clinician but by the frequency of the events. The WLW model, in its present form, does not allow to take into account accurately both the recurrence and the multiplicity of the events.