Adverse Responders and Antidepressant Effectiveness

If researchers used better patient selection methods when testing antidepressants, the effectiveness of those medications would be more accurately measured and confirmed.


 Richard J. Metzner, M.D.

Clinical Professor 
Semel Institute for Neuroscience and Human Behavior at UCLA

 A serious flaw in the way antidepressants have been tested in clinical trials is the assumption that patients fall into only two groups: responders and non-responders. The possibility of a third group: adverse responders, for whom a particular medication may not only be ineffective but also contraindicated, is now being considered. [1]

In most clinical studies adverse responses are merely footnoted as a random occurrence associated with dropouts and side-effects, but identifying possible adverse responders should be of paramount importance to prescribers.[2]

Here are some facts:
- Adverse responders do better on placebos than antidepressants to which they are intolerant.
- Antidepressants and placebos are similarly ineffective for non-responders.
- Adverse responders and non-responders may respond better to other antidepressants.

Moderators, i.e., patient characteristics that might better guide individualized treatments, are finally receiving attention in psychiatry.[3] Without incorporating knowledge of moderators into patient screening, studies of antidepressants are likely to include good, neutral and bad responses to medications being tested. Several recent studies using statistical methods called trajectory-based, latent class or growth modeling have examined this mixture of responses more deeply. Muthén et al[4] found previously unrecognized differences in patterns of response to antidepressants when dropout data were treated as non-random occurences. Gueorguieva et al[5] studied patients receiving either antidepressants or placebos and found that, while placebo patients all had a single response curve over time, those taking antidepressants displayed two divergent patterns. Patients who responded to antidepressants did better than those responsive to placebos as time progressed, but patients who didn’t respond to antidepressants had worse treatment trajectories than those non-responsive to placebos. The inclusion of patients in clinical trials for whom the adverse effects of an antidepressant make a placebo preferable may cancel out the positive data derived from others and explain the failure of antidepressant medication to demonstrate superiority to placebo in those studies.

Using evidence-based clinical biomarkers and profiles to screen out adverse responders ahead of time may not only be more ethical than allowing research subjects to suffer unnecessarily, but may also be the key to determining the true effectiveness of antidepressants when matched to the right patients.The following illustrative example based on data collected using the Targeted Treatment Depression Inventory (TTDI)® screening tool[6] demonstrates how this might happen:

- Half of each group randomly receives antidepressant (AD) or placebo
- SCREENED GROUP excludes potential adverse responders using TTDI algorithm
- UNSCREENED GROUP does not exclude potential adverse responders
- SCREENED GROUP RESULTS: 60% AD responders; 30% placebo responders
- UNSCREENED GROUP RESULTS: 30% AD responders; 30% placebo responders.

Note that when adverse responders are screened out, the ratio of responders to non-responders increases for antidepressants, but not placebos. Adverse responders who remain in clinical trials decrease the percentage of antidepressant responders relative to those who don't respond. Matching antidepressants to patients using personalized treatment methods increases the number of those who receive the right treatment and therefore respond at higher rates.

Experienced clinicians know that antidepressants and psychotherapy can make some people better, keep some the same and cause some to get worse.[7] For example, Patient A who can't stop worrying and wants to die to escape his thoughts usually needs a different treatment than Patient B, who is too apathetic to care about anything. Yet the DSM-IV and all tests for major depressive disorder identify A and B as the same disorder. Most published guidelines for treating depression make no distinction between treatments for A and B. Placebos may help both A and B, but the treatment that calms A may sedate B. The treatment that energizes B may agitate A. Active treatments are worse than placebos when they are the wrong treatment. Mix patients A and B together as all antidepressants trials do, and it’s no surprise that antidepressants appear no more effective than placebos in some studies. Would it "bias" research results in favor of antidepressants to screen out the adverse responders ahead of time? No more than it biases research on antibiotics to obtain cultures and sensitivities of organisms being treated. The hallmark of good medicine is to tailor the treatment to the disorder. When specific classes of antidepressants are matched to specific subtypes of depression it is simply better medicine.

The problem, many say, is that depression is too complicated and multi-faceted for anyone to figure out who needs what. Our own experience suggests otherwise. Empirically-based algorithms enable researchers to evaluate antidepressants more usefully and clinicians to utilize them more effectively.




[1]Sato Y Laird NM Yoshida T.Biostatistic tools in pharmacogenomics--advances, challenges, potential.”Curr Pharm Des.2010;16(20):2232-40

[2]Hazell L Shakir SA. “Under-reporting of adverse drug reactions : a systematic review” Drug Saf. 2006;29(5):385-96.

[3]Shelton RC Trivedi MH. “Using Moderator-Based Algorithms and Electronic Medical Records to Achieve Optimal Outcomes in DepressionJ Clin Psychiatry. 2011; 72(7):e24.

[4] Muthén B Asparouhov T Hunter AM Leuchter AF.  “Growth Modeling With Nonignorable Dropout: Alternative Analyses of the STAR*D Antidepressant Trial” Psychological Methods 2011; 16(1):17–33

[5] Gueorguieva R Mallinckrodt C Krystal JH “Trajectories of Depression Severity in Clinical Trials of Duloxetine: Insights Into Antidepressant and Placebo Responses” Arch Gen Psychiatry 2011;68(12):1227-1237

[6]Metzner RJ and Ho AP “Can clinicians improve antidepressant remission rates with better treatment algorithms?” American Psychiatric Association Scientific Meeting presentation No. 54, San Francisco, CA 2009.

[7] Lilienfeld, S.O. “Psychological treatments that cause harm” Perspectives on Psychological Science, 2007;

2, 53–70.



Copyright 2012, Scaled Psychiatric Systems, Inc. All rights reserved.