[Analysis of response to treatment. Considerations on changing prognostic variables (time-dependent)]

In the prognostic evaluations the initial variables, obtained at the time of diagnosis, and the evolutive or time-dependent variables, that appear during the course of the disease, should be clearly differentiated. If this separation is not strictly maintained inaccurate results may be obtained. The errors resulting from the use of an inappropriate methodology are illustrated in the present article on the basis of two series of patients. On the other hand, the advantages and shortcomings of alternative methods, which are valid for prognostic studies with evolutive or time-dependent variables, are discussed. Finally, recommendations are given for the particular instance where the response to therapy in a therapeutic trial is being assessed.