Abundance gradients in spiral discs: is the gradient inversion at high redshift real?