Well, Eric USES the shotgun approach.
The very dense data sets that result from shotgun approaches on small numbers of patients carry a significant risk of model overfitting, leading to spurious associations.