Abstract Part of the recent literature on the validation of biomarkers as surrogate endpoints proposes to undertake the validation exercise in a multi‐trial context which led to a definition of validity in terms of the quality of both trial level and individual level association between the surrogate and the true endpoints (Buyse et al., 2000). These authors concentrated on continuous univariate responses. However, in many randomized clinical studies, repeated measurements are encountered on either or both endpoints. When both the surrogate and true endpoints are measured repeatedly over time, one is confronted with the modelling of bivariate longitudinal data. In this work, we show how such a joint model can be implemented in the context of surrogate marker validation. In addition, another challenge in this setting is the formulation of a simple and meaningful concept of “surrogacy”. We propose the use of a new measure, the so‐called variance reduction factor, to evaluate surrogacy at the trial and individual level. On the other hand, most of the work published in this area assume that only one potential surrogate is going to be evaluated. We also show that this concept will let us evaluate surrogacy when more than one surrogate variable is available for the analysis. The methodology is illustrated on data from a meta‐analysis of five clinical trials comparing antipsychotic agents for the treatment of chronic schizophrenia.