The use of antidepressants (AD) in children and adolescents has become controversial due to a possible increased risk of suicide, revised evidence about effectiveness, and growing usage. Several studies have been conducted through healthcare databases, mainly restricted to hospitalized cases.

To evaluate the rate of self-harm among children (10–14 years) and adolescents (15–19 years) who use AD and to compare rates between classes of AD, using a common data model.

A retrospective cohort study was conducted in children and adolescents, age 10–19, members of the Quebec (QC) public drug plan or residents of British Columbia (BC), and who received a new AD treatment in 1997–2008 (QC) and 1997–2006 (BC). Self-harm (fatal or non-fatal) was the main outcome. Data sources consisted of prescription, medical services, and hospitalization databases. AD were categorized into: fluoxetine (only AD approved for paediatric use for depression in Canada), non-fluoxetine SSRIs, TCAs, and others. Multivariate logistic regression analyses were conducted using high dimensional propensity scores.

Were included 51,868 and 28,200 AD users from BC and QC, respectively: 70% were females, and 60% had received a diagnosis of depression. Risk of hospitalization for self-harm was 38.15 and 19.23/1000 person-years in BC and QC, respectively. Rates were higher among adolescents (BC: 47.52; QC: 21.36) than children (BC: 17.00, QC: 12.99). For both hospitalized and non-hospitalized cases, while there was no statistically significant difference in risk associated with fluoxetine relative to non-fluoxetine SSRIs, TCAs were associated with a lower risk in BC (OR = 0.47; 95% CI: 0.31–0.72) and Quebec (OR = 0.41; 95% CI: 0.30–0.46).

In two independent large cohorts, there was no apparent differences in risk across ADs. Lower risk of TCAs may be due to residual confounding by indication.