It's been some time since I read this work, but at one point there was a situation where medical schools in a state were temporarily (like a year or two) prevented from using MCATs in admission decisions due to a judge's court order. Most applicants still had MCATs due to other schools, though.

Some researchers collected data before, during, and after this time, and followed-up on the physicians later.

My fuzzy memory (which might be incorrect) is that MCATs were predictive of GPA in the first two years of medical school like 0.20 on a Pearson correlation metric, and it dropped from there as you went out increasingly further in time and included things like ratings on clinical rotations. So it was predictive, but fairly weakly so and seemed less predictive as time went forward.