Time to Feel Good

Practicing in an academic setting makes one aware of certain common clinical practices that are out of step with research findings. In this column, I hope to periodically to address some of these “disconnects” between clinical practice and research evidence.

In my last column, I discussed the apparent disconnect between the scientific evidence for the practice of increasing doses of SSRI in depressed patients and the common clinical wisdom and practice in this regard. In this month's column, I shall discuss a related issue with profound clinical and research implications, the evidence regarding the temporal aspects of a clinical response to antidepressant medications. Investigators studying the time course of antidepressant response have often employed a methodology called “pattern analysis.”

In this method, each individual who exhibits a therapeutic response to treatment is classified on two dimensions: Early versus late onset (with two weeks as the usual dividing line) and sustained versus nonsustained response.¹ Using this approach, some classic analyses of antidepressant clinical trials conducted about 20 years ago^1,2 concluded that placebo is far more likely than an active antidepressant to produce an early nonsustained response; whereas, the most likely pattern to be produced by an active antidepressant was a late-sustained response. For many decades, this concept has held sway among prescribers of antidepressants (for example see Dr. Hanau's letter to the editor in this issue of Psychiatry 2007) causing them to inform their patients that there is typically a delay of several weeks before therapeutic benefits can be perceived and to suspect that patients who reported early benefits from antidepressants were likely to be experiencing a placebo effect that is unlikely to persist. Likewise, research into the mechanism underlying antidepressants' therapeutic effects searched for biological processes that began after several weeks of antidepressant exposure.

In the last two years, some important new analyses of antidepressant time course have turned this “dogma” regarding the time-course of the antidepressant response upon its head.^3–5 A good example of these new analyses is the recent paper by Taylor, et al.⁵ This meta-analysis of 50 randomized, controlled trials comparing SSRIs and placebo concluded that, on average, treatment with SSRIs resulted in significant symptomatic improvement of depression by the end of the first week of the treatment. In fact, the largest weekly improvement in symptoms occurred during the first week of treatment, and on the whole, it represented about one third of the ultimate therapeutic improvement achieved over the course of these multiweek trials. The magnitude of average weekly improvement diminished steadily after the first week. It is important to realize that this temporal response was not a nonspecific or “placebo-like” phenomenon that would be expected to contribute to any active treatment response, especially one in the context of participating in an organized clinical trial. This is because the time course of improvements that I described have already factored out this nonspecific, placebo effect because the meta-analysis by Taylor, et al., used the average response seen with active-depressant minus the average response seen with placebo for each week. Thus, the time course of the improvement described can be directly attributable to the bioactive effects of the antidepressants.

This exciting meta-analysis and other recent ones that corroborate it obviously have profound implications for clinical practice and at the same time raise many provocative questions. For example, why did previous analyses and many individual studies fail to find strong evidence for an early, sustained response pattern for active-antidepressant treatment? Taylor, et al., suggest this may have been due to the way individual studies were designed and powered. Furthermore, if the new analyses are correct and the typical patient experiences substantial improvement in the first week, why have clinicians not railed against the previous dogma that response is typically delayed instead of generally abiding by it and incorporating it into their practice? Reading this, many clinicians may protest that they have often witnessed patients exhibiting an early sustained response to antidepressant treatment. However, I think this might be somewhat disingenuous since few, if any, clinicians I am aware of would have been willing to say, at least up to now, that an early sustained response, rather than a delayed sustained response, was the most common pattern of response among patients they treat.

Perhaps part of the explanation for this apparent disconnect between the new evidence and standard clinical wisdom is related to expectations that are created by research findings that then become “self-fulfilled” in the clinician's office. Alternatively, part of the disconnect may have to do with the important distinction between a statistically significant effect and a clinically meaningful effect. It may take a greater average absolute level of improvement than is typically seen in the first week to reach a threshold at which the typical patient will spontaneously report feeling better. Therefore, even though a patient's absolute symptomatic improvement is relatively smaller in the second and third week, compared to the first week, cumulative improvement may only reach a critical threshold of perception at the end of the third week or later, allowing the patient to perceive an overall improvement only from that point onward. Yet another contributing factor may be related to the differences in the way symptoms are measured in a clinical research trial versus clinical practice. In a clinical study, symptom burden is primarily determined using rating scales, which elicit information about specific symptoms of depression and do so in a highly consistent and systematic manner. Changes in a depressed patient's clinical status are rarely gauged that way in the clinician's office. In the clinical setting, the assessment is typically less systematic and more “gestalt” oriented (“How are you feeling?” “Are you feeling any better?”). While this may be supplemented by a few specific questions, the process is rarely as systematic and comprehensive as in clinical trials. Patients' gestalt impressions about their clinical status may not be simply the sum of their awareness of the status of individual depressive symptoms. I have, on more than one occasion, encountered patients whose initial impressions were that their depression had not changed, but upon systematic questioning of their individual symptoms they recognized, often to their surprise, that they have indeed experienced clinically meaningful symptom changes. This speaks to the importance of utilizing standardized ratings or scales to aid in diagnosis and evaluation of treatment response, and many thought-leaders in our field have been advocating this over the past several years. Interactive voice recording technology, the subject of two research articles in this issue of Psychiatry 2007, is of interest and of relevance to this topic as it may be the kind of tool that will facilitate eliciting evidence of early sustained response.

Conclusions from recent analyses of the temporal course of antidepressant response, such as from Taylor, et al.,⁵ hold out the highly appealing possibility of reducing the average duration of an antidepressant treatment trial in clinical practice. An argument can be made, based upon these new analyses of the temporal data, for decreasing patient suffering by accelerating the critical “go/no-go” decision point of an antidepressant trial. While this approach will result in some delayed onset responders needlessly receiving augmentation or undergoing medication switches, if the new analyses are correct, it would nevertheless decrease the time required to achieve a meaningful therapeutic benefit for majority of depressed patients who are prescribed antidepressants.

I will conclude with the interesting observation that often in science a zeitgeist of dogmatic re-examination occurs resulting in (or resulting from) analogous lines of iconoclastic evidence emerging simultaneously from different disciplines or databases. The revolutionary recent evidence regarding the time course of therapeutic effects from antidepressants that I describe in this column has been remarkably paralleled by similar recent evidence regarding the therapeutic time course of the antipsychotic drugs. That, however, is material for another column.