Generalized anxiety disorder (GAD) is a prevalent and impairing disorder, associated with extensive psychiatric and medical comorbidity and usually characterized by a chronic course. Different drugs have been investigated in GAD; among them are the following: 1) SSRIs: paroxetine, sertraline, fluvoxamine and escitalopram; 2) SNRI1s: venlafaxine; 3) benzodiazepines (BZs): alprazolam, diazepam and lorazepam; 4) azapirones (AZAs): buspirone; 5) antihistamines (AHs): hydroxyzine; 6) pregabalin (PGB); and 7) complementary/alternative medicine (CAM): kava-kava and homeopathic preparation. We conducted an effect size (ES) analysis of 21 double-blind placebo-controlled trials of medications treating DSM-III-R, DSM-IV or ICD-10 GAD using HAM-A change in score from baseline or endpoint score as the main efficacy measure. Literature search was performed using MEDLINE and PsycINFO databases including articles published between 1987 and 2003 and personal communications with investigators and sponsors. comparing all drugs versus placebo, the ES was 0.39. Mean ESs, excluding children, were PGB: 0.50, AH: 0.45, SNRI: 0.42, BZ: 0.38, SSRI: 0.36, AZA: 0.17 and CAM: -0.31. Comparing ES for adults versus children/adolescents (excluding CAM) and conventional drugs versus CAM (excluding children/adolescents) we found significantly higher ES for children/adolescents and for conventional drugs (p < 0.001 and p < 0.01, respectively). No significant differences were found when comparing date of publication, location of site (i.e. US versus other), fixed versus flexible dosing, number of study arms, or number of outcome measures used. Medications varied in the magnitude of their ES, ranging from moderate to poor. Adolescents and children showed a much greater ES compared with adults. Subjects taking CAM had worse outcomes than placebo.