Across 26 trials the sleep effect peaked near 4 mg a day rather than climbing with dose, and moving it to about 3 hours before bed improved falling asleep and sleep length.
That pairing is the peak of a curve drawn across many different trials, not a regimen any single trial actually ran. Nobody has yet put it head to head against the usual 30-minutes-before dose, which is the comparison that would settle it. The pooled trials span 1987 to 2020 and differ widely in whom they enrolled.
