This double-blind, multicenter trial, carried out in general practice in Denmark, comprised 221 women and 70 men, aged 58 to 97 years, with major depression (with or without mild cognitive dysfunction) or dysthymia (DSM-III-R). Patients had a total score > or =13 on the 17-item Hamilton Depression Rating Scale (HDRS) and a score > or =20 on the Mini Mental State Examination scale. The efficacy and tolerability of citalopram (20-40 mg daily) and amitriptyline (50-100 mg daily) were compared over 12 weeks. The participating general practitioners were trained at corating sessions in the use of the HDRS and Melancholia Scale (MES) prior to and during the study. The inter-observer reliability was assessed to investigate if general practitioners were able to use scales that measure the severity of depression. The two treatments were considered equally effective; the 90% confidence interval for the difference between the treatment groups in change from baseline to end-point in HDRS total score (-0.84 to +1.23) was within the predefined interval (-4 to +4). Significantly more patients on citalopram (50%) than on amitriptyline (31%) reported no adverse events at all (P = .001). Moreover, patients on amitriptyline reported adverse events significantly earlier and more frequently than patients on citalopram. The inter-observer reliability was highly satisfactory, with intra-class correlation coefficients (ICC-U) of .83 for the HDRS and .82 for the MES; however, the ICC-U for the Clinical Global Impressions was .54, indicating a poorer consensus in the investigators clinical judgment. Training in the use of the HDRS and MES scales improved the inter-observer reliability.