Loss of noradrenergic (NE) neurones in the locus ceruleus and compensatory changes in NE activity have been described in Alzheimer's disease (AD), but have never been linked to treatment. The hypothesis of this study was that central NE responsivity would predict aggression response to treatment with a NE medication, pindolol. Fifteen institutionalized AD subjects [Mini-Mental State Examination (MMSE), mean 3.3 +/- 4.6] with significant behavioural disturbances (Neuropsychiatric Inventory Score, mean 30.6 +/- 14.6) were studied. Growth hormone (GH) response to clonidine challenge (5 microg/kg) was used as a measure of central NE responsivity. Subjects were then randomized to 7 weeks of treatment with pindolol, maximum dose 20 mg b.i.d., or an identical placebo capsule in a cross-over design. The primary outcome measure was change on the retrospective Overt Aggression Scale (r-OAS). Five of 11 completers (45%) had decreased total r-OAS scores. There was significant improvement noted on the r-OAS verbal aggression subscale (paired t = -2.5, p = 0.03) compared to placebo, but not r-OAS total. Higher baseline aggression, higher MMSE and lower GH response predicted improvement in aggression, accounting for 82% of the variance (r = 0.91, F = 10.5, p = 0.006). Changes in NE responsivity, as reflected by a blunted GH response to clonidine challenge and more severe aggression, were associated with better response to the NE agent pindolol. Individual patient characteristics, including underlying neurotransmitter changes, may be useful for predicting response to therapy.