In Reply We welcome the opportunity Petersen and Nazareth provide with their letter to emphasize the importance of assessing the likely direction and quantifying the likely magnitude of biases in our study,1 including confounding by indication.
As we stated originally, although it is possible that unmeasured confounding accounts for the dose-response relationship we observed, it is not obvious what unmeasured factors might have led to meaningful confounding of our results. For example, our sensitivity analyses indicate that suicidal ideation prior to antidepressant therapy was not associated with the dose prescribed after adjusting for our propensity score, indicating that prior suicidality may be reasonably approximated by accounting for the patient characteristics at our disposal. Furthermore, our bias analyses demonstrate that an unmeasured confounder that could explain away the heightened risk of deliberate self-harm among youth initiating high-dose therapy would need to be more than 10-times as imbalanced across dose levels than were our measures of psychiatric comorbidity and more strongly associated with deliberate self-harm than our strongest predictor of deliberate self-harm (ie, a history of deliberate self-harm), even after accounting for the other risk factors included in our propensity score. Our bias analysis allows readers to speculate about the extent to which any purported unmeasured confounder satisfies these requirements.
Miller M, Swanson S, Stürmer T. Antidepressant Dose and Risk of Deliberate Self-harm—Reply. JAMA Intern Med. 2015;175(3):464. doi:10.1001/jamainternmed.2014.7180