OBJECTIVES: To compare the impact of various measures of adherence to antidepressant (AD) therapy on the estimation of the rate of depressive relapse using claims database. METHODS: Using the Korean Health Insurance Review & Assessment Service (HIRA) claims database (2006-2008), patients aged 18-84 without medical visit due to depression (ICD 10=F06.3, F31.3, F31.4, F32, F33, F34.1, F38.1, F41.2) within 6 months before the first observed prescription of ADs between July 2006 to Jun 2007 (18 months index period), and with at least 3 psychiatric visits within 3 months from index date were identified. ADs were categorized as tricyclic antidepressants (TCAs), selective serotonin reuptake inhibitors (SSRIs), and new antidepressants (NADs). The adherence measures of AD therapy include 90-days medication possession ratio (MPR, <75% as nonadherent), persistence (duration of uninterrupted therapy > 90 days with 14-day permissible gap), and switching of AD class within 180-days were used. Relapse was defined as hospitalization or emergency department visit due to depression, suicide attempts, or reinitiation of AD therapy after at least 6 months from previous AD treatment. Cox proportional hazards model was used to estimate risk ratios of relapse with 95% confidence intervals for each adherence measure. RESULTS: A total of 88,079 patients satisfied the selection criteria, among which mean age of 45.2 years and 67.3% of women. Overall relapse rate was 29.5%. Adherence (MPR≥75%) or persistence (duration of uninterrupted therapy>90 days) showed non-significant estimates of relapse risk (aRR=0.99[0.97-1.02] and 1.01[0.97-1.04], respectively). Patients who switched AD class within 180 days showed increased risk of relapse (aRR=1.18[1.15-1.21]). CONCLUSIONS: Various definitions of adherence led to different estimates of relapse rate. Diverse aspects of adherence should be considered when studying the association between the medication adherence and clinical outcomes.