The baseline characteristics of the antidepressant user and nonuser groups were distinctly different. Relative to the nonusers, those using antidepressants had greater mobility, greater use of psychotropic and other types of drugs, and roughly twice as many falls in the 90-day lead-in period, which suggested a greater baseline risk for falls. Considering the possibility that the depression or its correlates, rather than the antidepressants, might increase the rate of falls by increasing the level of disability, the investigators controlled for an extensive set of reasons for baseline impairment, but found the correlation with antidepressants persisted.

Using Poisson regression models to adjust rate ratios of falling for the subjects' characteristics (such as age, sex, body-mass index, ambulatory status and previous falls, incontinence, cognitive impairment, use of physical restraints and use of certain other medications including anticonvulsants, antipsychotics and sedatives) the investigators determined that the rate ratio of falls relative to nonusers was highest with TCAs at 2.0, followed by 1.8 with SSRIs and 1.2 with trazodone. The rate of falls with TCAs and SSRIs increased with increased daily dosage.

Recognizing that cardiovascular disease can increase the severity of orthostasis caused by TCAs, Thapa and colleagues also assessed the rate of falls in relation to the number of cardiovascular medications being taken. The 2.0 rate ratio of falls with TCAs in patients taking no cardiovascular medications rose to 3.3 in patients taking three or more. There was, however, no such increased rate of falling with an increased number of cardiovascular medication taken by users of either SSRIs or trazodone. Thus, the little difference between the rate of falls occurring with TCAs and SSRIs in patients who did not receive cardiovascular medications grew substantially to favor the safety of SSRIs in elderly patients on multiple cardiovascular medications.

Avorn pointed out some methodological difficulties inherent in such an observational study. Since the antidepressants were not randomly assigned, for example, the SSRIs might have been prescribed to those patients who were deemed at most risk for falling, under the assumption that the SSRIs contribute less to the risk. In addition, reliance on the documentation procedures in the nursing home is likely to be less accurate than outcome monitoring in a prospective study. "If an elderly person falls in a nursing home and no one records it," Avorn posed, "does it still make a sound in the database?"

Despite such methodological limitations, Avorn found the conclusions to be validly drawn. While the SSRIs pose less risk than TCAs for causing falls in the frail elderly who are receiving cardiovascular medications, both the SSRIs and TCAs are associated with a dose-dependent increased risk of falls in the elderly in general. Elderly patients receiving these antidepressants should, therefore, be appropriately monitored, assisted and advised to stand and ambulate with care.

The "fear of falling," however, should not preclude appropriate antidepressant treatment that can avert the morbidity and mortality of depressive illness.