Naji Tabet

Background Agitation is common and impacts negatively on people with dementia and carers. Non-drug patient-centred care is first-line treatment, but we need other treatment when this fails. Current evidence is sparse on safer and effective alternatives to antipsychotics. Objectives To assess clinical and cost-effectiveness and safety of mirtazapine and carbamazepine in treating agitation in dementia. Design Pragmatic, phase III, multicentre, double-blind, superiority, randomised, placebo-controlled trial of the clinical effectiveness of mirtazapine over 12 weeks (carbamazepine arm discontinued). Setting Twenty-six UK secondary care centres. Participants Eligibility: probable or possible Alzheimer’s disease, agitation unresponsive to non-drug treatment, Cohen-Mansfield Agitation Inventory score ≥ 45. Interventions Mirtazapine (target 45 mg), carbamazepine (target 300 mg) and placebo. Outcome measures Primary: Cohen-Mansfield Agitation Inventory score 12 weeks post randomisation. Main economic outcome evaluation: incremental cost per six-point difference in Cohen-Mansfield Agitation Inventory score at 12 weeks, from health and social care system perspective. Data from participants and informants at baseline, 6 and 12 weeks. Long-term follow-up Cohen-Mansfield Agitation Inventory data collected by telephone from informants at 6 and 12 months. Randomisation and blinding Participants allocated 1 : 1 : 1 ratio (to discontinuation of the carbamazepine arm, 1 : 1 thereafter) to receive placebo or carbamazepine or mirtazapine, with treatment as usual. Random allocation was block stratified by centre and residence type with random block lengths of three or six (after discontinuation of carbamazepine, two or four). Double-blind, with drug and placebo identically encapsulated. Referring clinicians, participants, trial management team and research workers who did assessments were masked to group allocation. Results Two hundred and forty-four participants recruited and randomised (102 mirtazapine, 102 placebo, 40 carbamazepine). The carbamazepine arm was discontinued due to slow overall recruitment; carbamazepine/placebo analyses are therefore statistically underpowered and not detailed in the abstract. Mean difference placebo-mirtazapine (−1.74, 95% confidence interval −7.17 to 3.69; p = 0.53). Harms: The number of controls with adverse events (65/102, 64%) was similar to the mirtazapine group (67/102, 66%). However, there were more deaths in the mirtazapine group ( n = 7) by week 16 than in the control group ( n = 1). Post hoc analysis suggests this was of marginal statistical significance ( p = 0.065); this difference did not persist at 6- and 12-month assessments. At 12 weeks, the costs of unpaid care by the dyadic carer were significantly higher in the mirtazapine than placebo group [difference: £1120 (95% confidence interval £56 to £2184)]. In the cost-effectiveness analyses, mean raw and adjusted outcome scores and costs of the complete cases samples showed no differences between groups. Limit