OBJECTIVES: Compare direct (medical and drug) costs between privately insured U.S. employees with major depressive disorder (MDD) who had treatment-resistant depression (TRD) and matched antidepressant-treated MDD controls without TRD. METHODS: Employees with >1 inpatient or >2 outpatient/other MDD diagnoses (ICD-9-CM: 296.2, 296.3) during 2004-2006, ages 18-64 years, were selected from a claims database. Employees who initiated a third antidepressant following 2 antidepressant treatments of adequate dose and duration were classified as TRD-likely (n=2,534). The index date was the date of first antidepressant. Control group was an age and gender matched cohort of randomly chosen antidepressant-treated employees with MDD without TRD. All were required to have continuous health coverage during the 6-month pre-index (baseline) and 12-month post-index (study) period. McNemar tests were used to compare baseline comorbidities. Wilcoxon signed-rank tests were used to compare annual per patient direct costs from third party payer perspective during the study period. Mental health (MH) related costs were identified by claims with MH disorder diagnosis (ICD-9-CM: 290-319) or MH-related drug costs. RESULTS: TRD-likely employees with MDD were on average 49.2 years old and 60.7% were women. Compared with controls, TRD-likely employees had significantly higher rates of MH disorders, chronic pain, fibromyalgia, but few differences in comorbidities included in the Charlson Comorbidity Index. Average direct annual costs were significantly higher for TRD-likely employees ($10,136) compared with controls ($7,793), $2,343 difference, p<.0001. Average MH-related costs were higher among TRD-likely employees ($2,714) compared with controls ($1,256), p<.0001; the MH-related cost difference of $1,458 accounted for 62.2% of the direct cost difference. MH-related cost differences were attributable to differences in drug (52.9%), outpatient (29.3%), and inpatient (12.2%) costs. CONCLUSIONS: TRD-likely employees with MDD had higher all-cause and MH-related direct costs compared with matched MDD controls. Excess costs of TRD-likely patients are underestimated when looking only at MH-related costs.