BACKGROUNDLimited evidence-base on long-term prognosis of treatment-resistant major depression (TRD) is a barrier to clinical decision-making. Therefore, the purpose of this study was to establish cause-specific mortality in TRD compared to non-TRD major depression.METHODWe identified all individuals with a diagnosis of major depression (MDD) who were treated with an antidepressant aged 15 to 65 years during 2004-2016 in Finland. Persons with over two treatment trials were defined to have TRD. Data were analysed with Cox proportional hazard models.RESULTS176,942 individuals with MDD (63 % women, median age at index diagnosis 40 years), of whom 11 % (n = 19,305) fulfilled the TRD criteria, were followed-up for 1,525,646 person-years (median 8.9 years). There were 959 deaths (6.1 deaths/1000 person-years) in TRD and 7662 deaths (5.6/1000 person-years) in non-TRD. All-cause mortality was 17 % higher (adjusted hazard ratio (aHR), 1.17; 95 % confidence interval (CI), 1.09-1.25) in TRD compared to non-TRD, when sex and age at index antidepressant prescription were controlled for. In TRD, increased mortality was observed for suicides (aHR, 1.90; 95%CI, 1.64-2.20) and for accidental poisonings (aHR, 1.81; 95%CI, 1.48-2.22), but not for natural causes (aHR, 0.98; 95%CI, 0.90-1.07). A higher proportion of accidental drug overdoses was observed in TRD than in non-TRD (62 % vs 42 %, respectively).LIMITATIONSDefinition of TRD lacks consensus. We used routine data to define TRD.CONCLUSIONSThe markedly increased mortality due to suicides and accidental overdoses suggests that persons with TRD may experience higher intensity symptoms and more severe suicidal ideation than persons with non-TRD major depression.