It is shown that the next-to-leading order (NLO) corrections to the QCD Pom
eron intercept obtained from the Balitsky-Fadin-Kuraev-Lipatov (BFKL) equat
ion, when evaluated in non-Abelian physical renormalization schemes with Br
odsky-Lepage-Mackenzie (BLM) optimal scale setting, do not exhibit the seri
ous problems encountered in the <(MS)over bar> scheme. A striking feature o
f the NLO BFKL Pomeron intercept in the BLM approach is that it yields an i
mportant approximate conformal invariance. (C) 1999 American Institute of P
hysics. [S0021-3640(99)00115-2].