I wonder if such a small effect size is really statistically significant, especially if you correct for the multiple hypotheses that are implicit in that specific a choice of subgroup...