[Skip to Content]
[Skip to Content Landing]
Invited Commentary
June 2014

Initial Dose of Antidepressant and Suicidal Behavior in Youth: Start Low, Go Slow

Author Affiliations
  • 1Departments of Medicine and Health Sciences, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
  • 2Western Psychiatric Institute and Clinic, University of Pittsburgh, Pittsburgh, Pennsylvania
  • 3University of Chicago, Chicago, Illinois
JAMA Intern Med. 2014;174(6):909-911. doi:10.1001/jamainternmed.2013.14016

In this methodologically exemplar pharmacoepidemiological study by Miller et al,1 data on patients with depression were assessed for prospective risk for deliberate self-harm (DSH; ie, suicide attempt) according to whether the patients had been initiated on therapy with the modal dose vs higher than modal dose of antidepressant. Using propensity-matched analysis, initiation at a higher than modal dose of antidepressant resulted in a 2-fold increased risk of deliberate self-harm in patients aged 10 to 24 years, especially in the first 3 months of treatment, while there was no such effect found in those aged 25 to 64 years.

One of the many strengths of this study was that the outcome was actual suicidal behavior rather than the more commonly used outcome “suicidal events,” which includes worsening in suicidal ideation, such as “thoughts of death” as well as actual suicidal behavior. These analyses examined unselected patients at much higher risk for suicidal behavior than the carefully screened participants in clinical trials. The prospective risk for suicidal behavior was assessed in patients treated for the same indication, namely, depression. These patients were new initiators of selective serotonin reuptake inhibitor (SSRI) therapy who had not been treated with an antidepressant in the past year. The use of propensity score matching allowed for contrasts between groups that were well balanced in terms of observed potential confounders. Finally, sensitivity analyses showed that a possible unmeasured confounder would have to be larger than the effect of prior DSH on future DSH, providing additional confidence for the validity of the results.

One obvious question is: why were 18% of patients treated with high initial doses of antidepressants, counter to clinical guidelines?2 The recommended practice is to initiate treatment at the equivalent of 10 mg/d of fluoxetine hydrochloride for 1 week, and then increase the dosage to 20 mg/d for the next 3 weeks, and only then to consider a dose escalation.2 The decision to initiate treatment at a higher dose (eg, fluoxetine hydrochloride at a dose > 20 mg/d) suggests there was something different about these patients that may have also put them at greater risk for DSH, such as a history of treatment nonresponse or a history of a positive response only at a higher than modal dose. It is also possible that patients started on higher doses experienced more adverse effects, discontinued the medication, and that it was the discontinuation, rather than the higher dose per se that put the patient at increased risk for DSH. In one study,3 discontinuation, titration up, and titration down of antidepressants were all associated with an increased risk of suicide attempt.

There are several other analytic approaches to this data set that could be informative. The dose response analysis examining the relationship of antidepressant exposure to DSH could have been broadened by comparing propensity-matched patients with no antidepressant treatment, those treated with below-modal dosage, along with those treated with modal- and high-dose antidepressants.

Although the authors chose not to emphasize preexposure vs postexposure designs, their requirement of a depression diagnosis and propensity matching allows for a comparison of the rate of DSH before and after treatment initiation in modal- and high-dose groups. In their matched samples, the rates of suicide attempts in the year prior to SSRI initiation were 1.31% for the modal dose and 1.46% for the high dose (assuming each patient was followed for 1 year). Following treatment initiation, the rates were 1.47% for the modal dose and 3.15% for the high dose, yielding relative risks of 1.12 and 2.16, respectively, for DSH. This analysis suggests that youth who were treated with an antidepressant at a modal dose were not at increased risk of DSH after medication, whereas those who were started on high-dose antidepressants were at more than twice the risk than they were before medication was initiated.

This data set could also shed light on the risk of DSH after dose escalation consistent with clinical guidelines by comparing propensity-matched sets of patients who were (1) started and maintained at a modal dose, (2) had a dose increase after initiation at a modal dosage, and (3) were initiated and maintained at a higher dose. In the article by Miller et al,1 patients who had a dose adjustment were censored, but inclusion of these patients in future analyses may have the best chance of addressing the risks, if any, of dose escalation in the current, clinically accepted manner.

In the context of considering the risk-benefit ratio for the use of antidepressants in youth, the authors raise concerns about the “modest” effects of antidepressants among youth and that “antidepressant dose is generally unrelated to therapeutic efficacy.” A comparison of the efficacy of fluoxetine between youth and adults actually shows a stronger initial effect on reduction on depressive symptoms in youth compared with adults.4 While clinical trials in youth show similar response rates in active medication to adults, the placebo response rate is higher in studies of depressed youth, particularly in children.5 If one compares effect sizes of treatment in patients of comparable severity, the efficacy of antidepressants is similar in youth and adults.

With regard to antidepressant dose and response, it may be more accurate to say, as is noted in one of the reviews6 cited by the authors, that there are significant methodological flaws with nearly all of the previous studies examining the relationship of dose and treatment response. In addition, there are studies that show that blood concentration, rather than dose of SSRI antidepressants, such as fluoxetine and citalopram hydrobromide, may be related to eventual antidepressant response in depressed adolescent and adult patients.7,8 In fact, in one study7 of treatment-resistant adolescents, participants treated with citalopram or fluoxetine who received a dose increase at 6 weeks because of nonresponse and whose drug concentration changed from below the geometric mean (GM) for the sample to above the GM were much more likely to show a response at 12 weeks than were those whose postincrease drug concentration continued to be below the GM. While it seems that most individuals treated at modal dosages will show serotonin reuptake inhibition at a level considered to be necessary for response, individual differences in drug metabolism could necessitate a higher dose in some patients to attain adequate blood level and receptor binding to result in response.

In summary, Miller et al1 are to be commended on a thoughtful and careful analysis of the effects of initiating antidepressants at higher than modal dosages. Their findings suggest that higher than modal initial dosing leads to an increased risk for DSH and adds further support to current clinical recommendations to begin treatment with lower antidepressant doses. While initiation at higher than modal doses of antidepressants may be deleterious, this study does not address the effect of dose escalation. Moreover, while definitive studies on the impact of dose escalation in the face of nonresponse remain to be done, there are promising studies that suggest in certain subgroups, dose escalation can be of benefit. Finally, it should be noted that in this study, there was no preexposure to postexposure increase in suicidal behavior after the initiation of antidepressants in youth treated at the modal dosage.

Back to top
Article Information

Corresponding Author: David A. Brent, MD, Western Psychiatric Institute and Clinic, University of Pittsburgh, 3811 O’Hara St, BT 311, Pittsburgh, PA 15213 (BrentDA@upmc.edu).

Published Online: April 28, 2014. doi:10.1001/jamainternmed.2013.14016.

Conflict of Interest Disclosures: Dr Brent receives royalties from Guilford Press; has or will receive royalties from the C-SSRS royalties from the electronic self-rated version of the C-SSRS from ERT Inc; and serves as an UpToDate Psychiatry Editor. Dr Gibbons is president and founder of Psychiatric Assessments Inc, a corporation that uses the trade name Adaptive Testing Technologies, through which the CAT-MH will be made commercially available. In the past 3 years, Dr Gibbons has been an expert witness for Pfizer Pharmaceuticals in cases related to suicide and adverse neuropsychiatric events for the drugs Neurontin and Varenicline. No other disclosures are reported.

Funding/Support: Drs Brent and Gibbons receive grant funding from the National Institutes of Mental Health. This Invited Commentary is not associated with grant-funded projects.

Role of the Sponsor: The National Institutes of Mental Health had no role in the preparation, review, or approval of the manuscript, and decision to submit the manuscript for publication.

Miller  M, Swanson  SA, Azrael  D, Pate  V, Stürmer  T.  Antidepressant dose, age, and the risk of deliberate self-harm [published online April 28, 2014].  JAMA Intern Med. doi:10.1001/jamainternmed.2014.1053.Google Scholar
Birmaher  B, Brent  D, Bernet  W,  et al; AACAP Work Group on Quality Issues.  Practice parameter for the assessment and treatment of children and adolescents with depressive disorders.  J Am Acad Child Adolesc Psychiatry. 2007;46(11):1503-1526.PubMedGoogle ScholarCrossref
Valuck  RJ, Orton  HD, Libby  AM.  Antidepressant discontinuation and risk of suicide attempt: a retrospective, nested case-control study.  J Clin Psychiatry. 2009;70(8):1069-1077.PubMedGoogle ScholarCrossref
Gibbons  RD, Hur  K, Brown  CH, Davis  JM, Mann  JJ.  Benefits from antidepressants: synthesis of 6-week patient-level outcomes from double-blind placebo-controlled randomized trials of fluoxetine and venlafaxine.  Arch Gen Psychiatry. 2012;69(6):572-579.PubMedGoogle Scholar
Bridge  JA, Birmaher  B, Iyengar  S, Barbe  RP, Brent  DA.  Placebo response in randomized controlled trials of antidepressants for pediatric major depressive disorder.  Am J Psychiatry. 2009;166(1):42-49.PubMedGoogle ScholarCrossref
Ruhé  HG, Huyser  J, Swinkels  JA, Schene  AH.  Dose escalation for insufficient response to standard-dose selective serotonin reuptake inhibitors in major depressive disorder: systematic review.  Br J Psychiatry. 2006;189:309-316.PubMedGoogle ScholarCrossref
Sakolsky  DJ, Perel  JM, Emslie  GJ,  et al.  Antidepressant exposure as a predictor of clinical outcomes in the Treatment of Resistant Depression in Adolescents (TORDIA) study.  J Clin Psychopharmacol. 2011;31(1):92-97.PubMedGoogle ScholarCrossref
Ostad Haji  E, Tadic  A, Wagner  S,  et al.  Early improvement and serum concentrations of citalopram to predict antidepressant drug response of patients with major depression.  Pharmacopsychiatry. 2013;46(7):261-266.PubMedGoogle ScholarCrossref