OBJECTIVES: The aim of this study was to establish the relative safety and balance of risks for antidepressant treatment in older people. The study objectives were to (1) determine relative and absolute risks of predefined adverse events in older people with depression, comparing classes of antidepressant drugs [tricyclic and related antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs), monoamine oxidase inhibitors (MAOIs) and other antidepressants] and commonly prescribed individual drugs with non-use of antidepressant drugs; (2) directly compare the risk of adverse events for SSRIs with TCAs; (3) determine associations with dose and duration of antidepressant medication; (4) describe patterns of antidepressant use in older people with depression; and (5) estimate costs of antidepressant medication and primary care visits. DESIGN: A cohort study of patients aged 65 years and over diagnosed with depression. SETTING: The study was based in 570 general practices in the UK supplying data to the QResearch database. PARTICIPANTS: Patients diagnosed with a new episode of depression between the ages of 65 and 100 years, from 1 January 1996 to 31 December 2007. Participants were followed up until 31 December 2008. INTERVENTIONS: The exposure of interest was treatment with antidepressant medication. Antidepressant drugs were grouped into the major classes and commonly prescribed individual drugs were identified. MAIN OUTCOME MEASURES: There were 13 predefined outcome measures: all-cause mortality, sudden cardiac death, suicide, attempted suicide/self-harm, myocardial infarction, stroke/transient ischaemic attack (TIA), falls, fractures, upper gastrointestinal bleeding, epilepsy/seizures, road traffic accidents, adverse drug reactions and hyponatraemia. RESULTS: In total, 60,746 patients were included in the study cohort. Of these, 54,038 (89.0%) received at least one prescription for an antidepressant during follow-up. The associations with the adverse outcomes were significantly different between the classes of antidepressant drugs for seven outcomes. SSRIs were associated with the highest adjusted hazard ratios (HRs) for falls [1.66, 95% confidence interval (CI) 1.58 to 1.73] and hyponatraemia (1.52, 95% CI 1.33 to 1.75), and the group of other antidepressants was associated with the highest HRs for all-cause mortality (1.66, 95% CI 1.56 to 1.77), attempted suicide/self-harm (5.16, 95% CI 3.90 to 6.83), stroke/TIA (1.37, 95% CI 1.22 to 1.55), fracture (1.63, 95% CI 1.45 to 1.83) and epilepsy/seizures (2.24, 95% CI 1.60 to 3.15) compared with when antidepressants were not being used. TCAs did not have the highest HR for any of the outcomes. There were also significantly different associations between the individual drugs for seven outcomes, with trazodone, mirtazapine and venlafaxine associated with the highest rates for several of these outcomes. The mean incremental cost (for all antidepressant prescriptions) ranged between pound51.58 (amitriptyline) and pound641.18 (venlafaxine) over the 5-year post-diagnosis period. CONCLUSIONS: This study found associations between use of antidepressant drugs and a number of adverse events in older people. There was no evidence that SSRIs or drugs in the group of other antidepressants were associated with a reduced risk of any of the adverse outcomes compared with TCAs; however, they may be associated with an increased risk for certain outcomes. Among individual drugs trazodone, mirtazapine and venlafaxine were associated with the highest rates for some outcomes. Indication bias and residual confounding may explain some of the study findings. The risks of prescribing antidepressants need to be weighed against the potential benefits of these drugs. FUNDING: The National Institute for Health Research Health Technology Assessment programme.