LAUSR

I read with interest the article by Schultz et al.,1 which addressed a key issue faced by many clinicians treating patients with Huntington disease (HD)—the conundrum of whether or not to start tetrabenazine (TBZ) for treating chorea in those with a history of depression, given its black box warning of increased depression and suicidality risk. Nevertheless, I was puzzled by the study's finding that among the total group of patients with HD, especially the subgroup with a history of depression, TBZ use was associated with lower odds of both depression and suicidality.1 The most likely explanation for this finding is the phenomenon known as confounding by indication—a form of selection bias introduced by the fact that clinicians are simply less likely to prescribe TBZ to those with severe depressive symptoms due to the perceived detrimental effects of this drug on mood.2–4 Although the multiple logistic regression analysis was adjusted for a range of potential confounders,1 conditional regression estimates generally do not provide an unbiased estimate of true marginal effect in contexts where treatment selection is partially based on the outcome of interest (i.e., depression and suicidality).3 This issue can, at least partly, be addressed by applying inverse probability of treatment weighing using propensity scores.2,4 Therefore, it is important to know whether application of this method would result in similar findings or, conversely, support the notion of confounding by indication.4