Approximately 2 out of every 3 depressed diabetic patients seen in the primary care setting receive no specific antidepressant treatment (1). This may be because their depressive disorders are not diagnosed and thus their physicians accept depression as an inevitable outcome of chronic illness, or because too little is known about the efficacy of treatments for depression in diabetes. In our 10-week clinical trial, half of the patients received supportive care for depression provided in the context of diabetes education, and half received supportive treatment plus CBT. Earlier controlled studies had found that supportive care provided in the context of health education was effective for postnatal depression (2). Thus, our participants were given more treatment than that often provided depressed diabetic patients, and each received a treatment with some evidence of efficacy. Several safeguards were used to protect participants during the trial, regardless of treatment assignment.

Concluding that ethical standards demand conventional antidepressant treatments for all participants at this point is premature because the efficacy of these treatments in diabetic patients has not been convincingly established. Generalization from the psychiatric treatment literature may be unsafe. Depression in diabetes appears in some ways different from depression in otherwise medically well patients. The cause may be different, with greater relative contributions from organic and psychosocial sources. The course is also decidedly different and is influenced by medical factors (for example, glycemic status). Findings from animal studies suggest the possibility of a resistance to conventional antidepressants in depressed diabetic patients (3). Diabetes frequently results in lifestyle restrictions, financial strain, pain, and disability, realities that may influence the course and limit the effectiveness of treatment. Consequently, for a variety of reasons, neither pharmacologic nor nonpharmacologic approaches can be considered effective at face value in the diabetic patient, and treatment versus no-treatment comparisons remain the scientific gold standard for establishing their utility.

Furthermore, whether random assignment to a no-treatment condition is ethically defensible depends, in part, on the efficacy of existing treatment; the more effective the treatment, the more inappropriate is its withholding (4). The potency of antidepressant treatment in nondiabetic samples is modest at best, the probability of depression remission being just 25% greater in those who receive active treatment over those who do not (5). In 40% of treated patients, depression does not remit. With such small margins of benefit and high rates of nonresponse, improper conclusions from trials lacking no-treatment control conditions are easy to envision. Active treatments might be considered comparable when neither or only one truly exceeds the expectations of placebo. It is the responsibility of the scientific community to determine the efficacy of treatment before issuing recommendations for such a prevalent problem.