Objectives: To evaluate critically the evidence linking psychotropic drugs with falls in older people.
Design: Fixed-effects meta-analysis.
Data sources: English-language articles in MEDLINE (1966 - March 1996) indexed under accidents or accidental falls and aged or age factors; bibliographies of retrieved papers.
Study selection: Systematic evaluation of sedative/hypnotic, antidepressant, or neuroleptic use with falling in people aged 60 and older.
Data extraction: Study design, inclusion and exclusion criteria, setting, sample size, response rate, mean age, method of medication verification and fall assessment, fall definition, and the number of fallers and non-fallers taking specific classes of psychotropic drugs.
Results: Forty studies, none randomized controlled trials, met eligibility criteria. For one or more falls, the pooled odds ratio (95% confidence interval) was 1.73 (95%CI, 1.52-1.97) for any psychotropic use; 1.50 (95%CI, 1.25-1.79) for neuroleptic use; 1.54 (95%CI, 1.40-1.70) for sedative/hypnotic use; 1.66 (95%CI, 1.4-1.95) for any antidepressant use (mainly TCAs); 1.51 (95%CI, 1.14-2.00) for only TCA use; and 1.48 (95%CI, 1.23-1.77) for benzodiazepine use, with no difference between short and long acting benzodiazepines. For neuroleptics in psychiatric inpatients, the pooled OR was 0.41 (95%CI, 0.21-.82); for all other patients, the pooled OR was 1.66 (95%CI, 1.38-2.00). Comparing > or =1 with > or = 2 falls, mean subject age <75 versus > or =75 years old, communities with <35% versus > or =35% fallers, or subject place of residence did not affect the pooled OR. Increased falls occurred in patients taking more than one psychotropic drug.
Conclusion: There is a small, but consistent, association between the use of most classes of psychotropic drugs and falls. The evidence to date, however, is based solely on observational data, with minimal adjustment for confounders, dosage, or duration of therapy. The incidence of falls and their consequences in this population necessitate that future large randomized controlled trials of any medication in older persons should measure falls prospectively as an adverse outcome event.