In this issue of the Journal, Harmer and colleagues (1) report interesting and potentially paradigm-changing findings that may shed new light on a mechanism for antidepressant action. That these findings might suggest a common mechanism of action for both pharmacologic and psychotherapeutic interventions only adds to the possibilities of this line of research. Harmer et al. conducted a tightly controlled experiment in which 33 outpatients with major depressive disorder and 31 healthy comparison subjects received either a single dose of 4 mg of reboxetine (a highly selective norepinephrine reuptake inhibitor) or placebo. They found that in the depressed participants, administration of reboxetine essentially normalized three laboratory correlates of the negative information processing biases that characterize depression: diminished perception of facial expressions of happiness, decreased response to positive self-referent items, and poorer recall of positive personality descriptors. Perhaps the most intriguing aspect of these findings is the rapidity of the beneficial effects (i.e., 3 hours after administration of the drug), along with the observation that such dramatic early changes in information processing were not associated with appreciable changes in mood or other symptoms. The authors suggest that such changes in information processing may represent a common mechanism for both antidepressant pharmacotherapy and cognitive behavior therapy and speculate that the effect is too rapid to involve "downstream" adaptations in receptors or gene activity. Their hypotheses also challenge some conventional wisdom about the process of change during antidepressant therapy, such as the assumption that improvements in neurovegetative symptoms typically precede improvements in cognitive processes.

This team of investigators has previously conducted research evaluating the impact of antidepressants on information processing in healthy individuals, including studies using the selective serotonin reuptake inhibitor citalopram and the serotonin-norepinephrine reuptake inhibitor duloxetine (2). Across a series of studies, these antidepressants were found to have positive effects on various indices of emotionally relevant information processing. The report in this issue thus represents an extension of this line of research to include a clinical population. Although the magnitude of the reboxetine effect tended to be greater for the depressed participants than for the healthy volunteers, only the effect on the emotional memory task showed a significant treatment-by-group interaction. Thus, the drug effect on emotional information processing does not appear to be specific to depression, and the differences in magnitude of effect observed in this study may simply reflect the greater ease of improving a low score than a high one.

Whether or not the finding is specific to depression, the rapidity of the effect suggests that these changes in information processing are the direct and immediate result of increasing synaptic monoamine concentrations by inhibiting uptake transporters. When coupled with the apparent lack of specificity (i.e., the effects are observed in both patients and healthy volunteers), one must ask whether effects that are detectable within only 3 hours are directly and sequentially linked to eventual antidepressant effect 4, 6, or 8 weeks later. On the "pro" side, there is no doubt about the relevance of negatively biased information processing to the psychopathology of depression, including increased vulnerability to illness onset and relapse as well as the characteristic cognitive symptoms of depressive episodes (3). It is also true that all three medications that have been shown to exert these effects in highly controlled experiments have antidepressant effects in clinical populations. On the "con" side, the dose of reboxetine that was used in this experiment (4 mg) is only half of the minimum daily therapeutic dose, and this report does not include follow-up data. However, elsewhere the investigators have reported results of a treatment study in which patients with major depressive disorder showed significant improvements in accuracy of recognition of facial emotions after only 2 weeks of treatment with either reboxetine or citalopram, and these changes were predictive of clinical outcome after 6 weeks of randomized, open-label therapy (4).

Obviously, much additional research is needed to properly test the hypothesis that improved information processing is an important pathway for antidepressant activity, although the findings to date are certainly promising. Of course, the highest research priority is to confirm that the rapid effects observed in this study are predictive of eventual clinical benefit. With respect to pharmacologic specificity, it also will be valuable to study the effects of other drugs that rapidly enhance monoaminergic neurotransmission but do not have reliable antidepressant effects, such as atomoxetine (another highly selective norepinephrine reuptake inhibitor, approved for the treatment of attention deficit hyperactivity disorder); psychostimulant medications, such as methylphenidate and amphetamine; and drugs of abuse, such as cocaine. One also might wonder about the relevance of these effects in people with bipolar depression and whether rapid changes in the positivity of affective processing might even be linked to the propensity of antidepressants to provoke treatment-emergent affective switches in individuals with bipolar disorder (5).

Another interesting aspect of this research is that the observed effects of reboxetine on negative perceptual bias were not only rapid but relatively large, meaning that most of the group that received reboxetine showed a significant improvement. This observation must be contrasted with the relatively small specific effects that reboxetine and other, more widely used, antidepressants show in contemporary placebo-controlled trials (6, 7). Since the depressed individuals in this study had mild to moderately severe depressive syndromes, it is highly unlikely that this group was in some way uniquely responsive to antidepressant medication. Could this mean that a rapid improvement in affective information processing is a necessary but not sufficient element of a therapeutic response? Is it possible that drug effects on affective information processing make it easier for depressed people to perceive happy faces, attend to positive information, or recall good things about themselves but that the individual must have the opportunity to use these assets in rewarding social interactions to obtain more sustained improvements in mood and other depressive symptoms? Or could it be that changes in affective processing are simply a behavioral assay of the immediate effect of the drug at the synaptic level and that other, downstream intracellular accommodations are still necessary to produce a durable antidepressant response?

Assuming that the next wave of studies demonstrates relatively specific and sustained effects that are predictive of eventual antidepressant outcome, it will be important to examine individual differences in response, including the factors, both clinical and genetic, that might account for these differences. What are the characteristics that distinguish individuals who do not show significant changes in information processing from those who show large effects? Do some individuals experience no effect with one type of antidepressant and a large effect with a different type? What might characterize those individuals who experience a large immediate improvement in information processing but whose depression does not subsequently respond to the antidepressant?

Having spent a large majority of my career working with David Kupfer and colleagues at the University of Pittsburgh Medical Center, I am reminded of an equally interesting line of research that began over four decades ago involving changes in REM sleep in depression (8). Like negative perceptual biases, reduced REM latency was also rapidly "normalized" by several types of antidepressants, and this effect also was evident within hours of the first administration of low doses of amitriptyline (e.g., 50 mg). Moreover, prospective studies showed that the magnitude of this rapid REM suppression was predictive of subsequent treatment outcome. Although the promise for use of polysomnographic assessments of perturbations of REM sleep to improve the care of patients in everyday clinical practice was not realized, it is nevertheless true that alterations in REM sleep were among the first well-replicated neurophysiologic correlates of depression and have been linked to other relevant changes in brain function by studies using newer neuroimaging techniques (9).

With respect to future research testing the common mechanism hypothesis—that is, that changes in affective information processing account for the benefits of both pharmacotherapy and cognitive-behavioral therapy—one must be guided by the understanding that both interventions also share another common mechanism, namely, the placebo-expectancy response. It is therefore important that future studies be large enough to take into account this more nonspecific contributor to treatment outcome. With respect to mode-specific mechanisms of action, it is now widely believed that medications initiate therapeutic response from "below" the cortex (that is, via effects mediated by serotonergic and noradrenergic tracts ascending from brainstem nuclei), whereas psychotherapies work from "above" by helping the patient engage compensatory executive cognitive control mechanisms (3). If the findings of Harmer et al. are relevant to these processes, then it is plausible that the pathways of change will converge with dampening the negative information processing in circuits involving limbic and medial and dorsolateral prefrontal cortical structures. The speed of effect advantage that sometimes favors pharmacotherapy over psychotherapy in comparative studies could well be mediated by the rapidity of the impact of antidepressants on these mechanisms. Finally, Harmer et al. suggest that the clinical benefits of the improvements in information processing following administration of antidepressant are not immediate because it takes patients time to learn to interact with the world with less biased perceptions. Yet, it is also true that depressed patients who benefit from antidepressants remain at great risk for relapse after treatment is discontinued, even after a year of continuous remission (10). Does this mean that the learning is state-dependent, or could learning strategies to suppress negative information processing biases remain a fundamental difference between pharmacotherapy and cognitive-behavioral therapy?

Address correspondence and reprint requests to Dr. Thase, Department of Psychiatry, University of Pennsylvania, 3535 Market Street, Rm. 689, Philadelphia, PA 19102; thase@mail.med.upenn.edu (e-mail). Editorial accepted for publication August 2009 (doi: 10.1176/appi.ajp.2009.09081199).

Dr. Thase has received research funding from Eli Lilly, GlaxoSmithKline, NIMH, and Sepracor; he has served in an advisory or consulting capacity to, or received speakers honoraria from, AstraZeneca, Bristol-Myers Squibb, Eli Lilly, Forest Laboratories, GlaxoSmithKline, Janssen Pharmaceutica, MedAvante, Neuronetics, Novartis, Schering-Plough, Sepracor, Shire, Supernus, Transcept Pharmaceuticals, and Wyeth; he has equity holdings in MedAvante; he has received royalties from American Psychiatric Publishing, Inc., Guilford Publications, Herald House, and W.W. Norton; and his spouse is employed by Advogent (formerly Cardinal Health). Dr. Freedman has reviewed this editorial and found no evidence of influence from these relationships.