Yes, there is evidence of this, albeit not subjected to rigorous statistical & probabilistic methods to root out the effects of random chance & surmise, from janoshik's impressions of dimer results. He cannot conduct such rigorous statistical & probabilistic methods on data that is subject to stringent privacy interests, legally.
There is also theoretical support for this. Dimer reflects GH polypeptide that has aggregated, think of it as "more bound up" or "sticky polypeptides." These "sticky, bound up GH" structures cannot activate the GH receptor as "cleanly" to fully activate or bind with the "torsional force" to the GH receptor versus unbound, more free GH, and so these dimerized GH structures don't augment IGF-I to the same degree, which acts in a complementary fashion to GH. That is, e.g., where GH is hyperglycemic, IGF-I is hypoglycemic; where GH promotes efflux, IGF-I promotes influx, etc. So in many ways you can get more "negative" GH effects that are not countervailed by more "positive" IGF-I effects with high dimer (% high molecular weight proteins).