Article type
Year
Abstract
Objective: There is epidemiological evidence that people with both physical and mental illness are more disabled than those with either one alone. Depression is the commonest mental disorder in the physically ill, but often goes unrecognised and untreated, partly because of doubts among patients and professionals about the effectiveness of the most available treatment, antidepressant medication. In an effort to resolve this doubt, we set out to review systematically, according to Cochrane Collaboration methodology, all relevant placebo-controlled randomised trials of antidepressants for depression in the physically ill. This method is accepted as providing the least biassed estimate of treatment effectiveness short of new megatrials.
Methods: As individual trials identified in each illness were small, we decided to combine all physical illnesses, all depression diagnoses, all antidepressants and all depression endpoints. This increases statistical power, and estimates treatment effect across various settings, but involves assuming that the above "lumpings together" are valid. We searched electronically using terms to identify RCTs of all antidepressant drugs, and then inspected the abstracts for likely material. Hand and reference-list searching was also used. As a first step, the endpoints used were numbers of l)dropouts and 2)patients improving by group, analysed intention-to-treat with assignment of bad outcome to any randomised patients not accounted for at the end of the trial. Odds ratios for these outcomes were combined using Mantel-Haenszel methods.
Results: 12 trials met inclusion criteria (cancer 2, MS 1, heart 1, stroke 3, HIV/AIDS 1, RA 1, DM 1, COAD 1, mixed 1 ), covering a total of 576 patients. The odds ratio for improvement was 3.1 ( 95%CI 2.2-4.6) on antidepressants versus placebo; the OR for dropout was 0.88 ( 95%CI 0.56-1.3 ).
Discussion: Identified trials are small and positive; publication bias may exaggerate the observed treatment effect in this preliminary analysis: further analyses will examine the individual rating scales and drug classes. Nevertheless, the evidence so far indicates that antidepressants appear effective for the short term relief of depression symptoms in the physically ill, and that they do not cause more dropouts than placebo.
Methods: As individual trials identified in each illness were small, we decided to combine all physical illnesses, all depression diagnoses, all antidepressants and all depression endpoints. This increases statistical power, and estimates treatment effect across various settings, but involves assuming that the above "lumpings together" are valid. We searched electronically using terms to identify RCTs of all antidepressant drugs, and then inspected the abstracts for likely material. Hand and reference-list searching was also used. As a first step, the endpoints used were numbers of l)dropouts and 2)patients improving by group, analysed intention-to-treat with assignment of bad outcome to any randomised patients not accounted for at the end of the trial. Odds ratios for these outcomes were combined using Mantel-Haenszel methods.
Results: 12 trials met inclusion criteria (cancer 2, MS 1, heart 1, stroke 3, HIV/AIDS 1, RA 1, DM 1, COAD 1, mixed 1 ), covering a total of 576 patients. The odds ratio for improvement was 3.1 ( 95%CI 2.2-4.6) on antidepressants versus placebo; the OR for dropout was 0.88 ( 95%CI 0.56-1.3 ).
Discussion: Identified trials are small and positive; publication bias may exaggerate the observed treatment effect in this preliminary analysis: further analyses will examine the individual rating scales and drug classes. Nevertheless, the evidence so far indicates that antidepressants appear effective for the short term relief of depression symptoms in the physically ill, and that they do not cause more dropouts than placebo.