Objective: Two-month head-to-head clinical trials of escitalopram and venlafaxine demonstrated similar efficacy and better tolerability for escitalopram. As routine practice may differ from controlled trial, policy makers wonder how clinical trial findings translate into real life outcomes in community practice. This work compares early treatment discontinuation (ETD) rates at 1 and 2 months and its economic consequences at 6 months, for patients with depression treated with venlafaxine and escitalopram. Methods: Using US denominator-based claims database PharMetrics (includes data from 86 managed care health plans covering 45 million patients), we included adult patients diagnosed with depression who started venlafaxine or escitalopram between January 1, and December 31, 2004. We compared ETD at 1 and 2 months using Cox proportional hazard models and health care costs at 6 months, using log-linear regression. Propensity scoring was used to account for channelling by indication. Results: A total of 13,227 patients started escitalopram; 5,922 patients started venlafaxine. ETD at 2 months was 47% for venlafaxine, 45% for escitalopram. At 1 month, venlafaxine patients had a 50% greater risk of ETD than escitalopram patients (Hazard Ratio=0.493 [95%CI 0.432-0.564]); this difference decreased at 2 months (Hazard Ratio=0.955 [95%CI 0.912-0.999]). Six-month health care costs were higher with venlafaxine (+USD626, p<0,01). Patients continuing treatment at 2 months had 36% chance of still being on treatment at 6 months. Patients (all treatments) with ETD at 2 months incurred more costs over 6 months (+USD350) compared to patients continuing treatments. Conclusion: Early treatment discontinuation rate was higher with venlafaxine than escitalopram, possibly due to intolerance to venlafaxine. Absence of ETD was associated with long term persistence and lower total treatment costs.