This manuscript utilizes the STARD trial patients to examine the attrition rates potentially secondary to the frequency, intensity, and burden of side effects related to citalopram, which was employed as the first arm of the trial. Overall this is a very interesting article and is well done. However, there are a couple of elements that must be clarified or remarked upon prior to publication.
1. The authors remark that the STARD trial used the PRISE scale at baseline, but do not list the rates (%) of the patients reporting these symptoms. I bring this up as many of the reported symptoms from various organ systems (gastrointestinal, sleep, sexual function, etc.) could have been from the depression itself, rather than the medication. Having the baseline data would be helpful to know if or how much of a worsening occurred of these symptoms/side effects.
a. Further, if this information is known, then could it be used to compare the rates of attrition in various weeks versus baseline?
2. It is noted that the STAR*D trial protocol increased the dose of citalopram to 40 mg/day on week 4, and 60 mg/day on week 6, a time in which many patients seemingly had an increase in the frequency, intensity, or burden of their reported side effects. Thus, it should be noted that clinicians be very cautious when increasing the dose for a patient who is already having side effects. This was not mentioned within the discussion section.

Agree with Reviewer #2's sentiments. Overall the authors have done a good job with revisions. While I appreciate the authors clarifying that a baseline PRISE was not completed, this should be mentioned within the Limitations section of the manuscript. The fact we do not know what kinds of physical symptoms the patients may have been incurring due to depression is troublesome. The QIDS baseline and subsequent sections on sleep and weight could potentially be used to glean attrition information on the rates of side effects from antidepressants. However, if the STAR*D Trial did not separate out individual components from the reported total score then I realize this comparison might be impossible.