Dr. Falk Leichsenring and his colleagues report the results of a methodologically rigorous randomized, controlled trial of supportive-expressive psychodynamic psychotherapy and cognitive-behavioral therapy (CBT) for generalized anxiety disorder in this issue of the Journal(1). The therapies were balanced for time, frequency, and therapist training. The investigators are to be praised for their scientific integrity in forthrightly presenting several findings that favor CBT over the psychodynamic therapy that they have championed—never an easy task. Dr. Leichsenring is well known for his meticulous reviews of psychodynamic psychotherapy outcome (2, 3), an effort undertaken in an attempt to delineate an evidence base for psychodynamic psychotherapy. The relative paucity of outcome research in this area results from a range of forces, including difficulties in operationalizing the therapies that describe themselves as “psychodynamic psychotherapy” and the relative lack of a tradition of randomized, controlled trials within psychoanalysis (4). The present study has many strengths, but it also highlights some of the design difficulties that have plagued investigators in the labyrinthine endeavor to develop a more credible clinical trials research base for psychodynamic psychotherapy (3).

The current study presents an opportunity to articulate challenging aspects of study design in randomized, controlled trials of psychotherapies. The field of psychotherapy outcome research is evolving, and the considerations that follow are not unique to this study. No study design is ever perfect; weaknesses always exist. As the authors note, budgetary limitations affected their design. This common consideration makes power calculations crucial during the planning phase of any study.

Most researchers test a favored psychotherapy: here, supportive-expressive psychotherapy (5). A more challenging decision facing investigators is the choice of which comparison condition to select in a trial for efficacy. Tests of new medication treatments face no such conundrum, as pill placebo is in many ways the perfect solution. Pill placebo is not an apt comparison for a psychotherapy, however.

In fact, no comparison condition for psychotherapy trials mirrors the parsimony of placebo for medication trials (6). A no-treatment or delayed-treatment condition is ethically questionable for anxious patients like those with generalized anxiety disorder. Furthermore, such a comparison predetermines disappointment and exaggerated symptoms by definition, as patients are told they have a disorder yet cannot qualify for immediate treatment. A waiting list may even be a nocebo(7). Unfortunately, at least in the United States, “treatment as usual” comparisons often amount to much the same thing. For patients with generalized anxiety disorder, who are sensitive to separation (8), it is essential to balance treatment time and therapist attention across conditions, as this study did. Therapist experience and training must be matched to avoid results that reflect an artifact of experience.

Ultimately, the first-line utility of any psychodynamic psychotherapy for any anxiety disorder will depend on its efficacy relative to CBT, which has a solid research evidence base in this area (9). There are serious disadvantages to conducting direct comparisons between these two treatments, however, before having initially established the efficacy of a psychodynamic psychotherapy for generalized anxiety disorder. Leichsenring and colleagues chose this design, comparing supportive-expressive therapy to a standard reference treatment (CBT), rather than to a minimal psychotherapy control. This strategy carries risks. If both treatments were to perform comparably, which was not quite the case in this report (CBT did somewhat better for some symptoms), the study could not distinguish whether this occurred because both treatments had been efficacious or because neither worked, even if patients improved, as they did here. This problem is called “assay sensitivity” (10, 11). Generally accepted margins for testing equivalence in generalized anxiety disorder have not yet been established, and this is a sine qua non for an equivalence or noninferiority study (12). The Food and Drug Administration (13) has made it clear: “In order to implement an equivalence or non-inferiority trial, the magnitude of [medication] effect must be stable and well-established in the literature, with consistent results seen from one trial to the next” (12, p. 32). Despite the experience accumulated with CBT, we have not yet reached this research benchmark for studies of generalized anxiety disorder, one of the most heterogeneous DSM disorders (14). While most researchers would likely agree that a 1-point difference on the Hamilton Anxiety Rating Scale (HAM-A) (15) does not constitute a clinically significant difference, not all of those who study generalized anxiety disorder would agree about the significance of 2- or 3-point differences in HAM-A outcome. Furthermore, the margin of equivalence (the mathematical point defining when two treatment conditions can be said to perform equally well—or equally poorly, for that matter) must be substantially smaller than the hypothesized treatment effect that is used to determine sample size in a superiority trial. So even if the field had agreed on a margin of equivalence, conducting a true equivalence study would require a very large study group, far exceeding the N of 57 in the trial by Leichsenring et al. Thus, the present study did not find an overall difference, nor did it establish that the two treatments are equivalent.

Pioneered by Lester Luborsky, Ph.D., supportive-expressive therapy represents a brilliant accomplishment. It is the first true psychodynamic psychotherapy successfully captured in manual form (16). This achievement made it possible to conduct much of the psychodynamic psychotherapy outcome research that has been performed thus far (17–19), including promising studies of generalized anxiety disorder (17, 20). Organized around the concept of the “core conflictual relationship theme,” which bears a relationship to the psychoanalytic concept of transference, supportive-expressive therapy is a specifically defined form of psychodynamic psychotherapy that encapsulates many, but not all, key aspects of “psychodynamic psychotherapy.” Other central aspects of psychodynamic psychotherapy more heavily emphasized in other psychodynamic interventions include attachment and reflection (21) and aspects of unconscious symbolic fantasy and development (8). The cost of this essential first manualization has been the sacrifice of some of the inherent flexibility of other forms of psychodynamic psychotherapy. In one study of personality disorders, this more constrained range may have resulted in a smaller effect for supportive-expressive therapy than for more broadly defined psychodynamic psychotherapy delivered by experienced psychodynamic psychotherapists (22). Patients with personality disorders, with wide-ranging psychopathology and underlying emotional conflicts possibly similar to those in generalized anxiety disorder, constitute a heterogeneous group that might potentially derive greater benefit from a more flexible therapeutic armamentarium within dynamic treatment.

What do these reflections imply about the efficacy of psychodynamic psychotherapy for generalized anxiety disorder? The two therapies were not significantly different overall, but CBT appeared superior to supportive-expressive therapy across a number of outcome measures in this trial. Despite the lack of difference on the primary outcome measure, differences were impressive, particularly as they emerged from a relatively small number of patients. Nonetheless, differences between patient groups and the vagaries of treatment deliveries mandate conducting more than one clinical trial to fully evaluate these outcomes. Furthermore, the study may teach us much about treatment specificity in evaluating their data for moderators of treatment effect. As Leichsenring et al. point out, CBT was well targeted to the symptoms of generalized anxiety disorder where it showed its better effects—trait anxiety, worry, and depression. The heterogeneity of generalized anxiety disorder makes me, as a clinical psychoanalyst and outcome researcher, very interested in investigating whether or not other variants within the domain of “psychodynamic psychotherapy” might be more readily targeted to patients with generalized anxiety disorder, for example, one that focuses more centrally on underlying fantasies and fears about relinquishing control (hence the need for constant alertness) and the ambivalence about personal autonomy/separation from close love objects that such a relinquishment would necessarily engender. For the moment, while CBT was superior to supportive-expressive psychotherapy in this trial, we do not yet have a definitive answer about the efficacy of “psychodynamic psychotherapy” for generalized anxiety disorder.

Address correspondence and reprint requests to Dr. Milrod, Weill Medical College, Cornell University, 525 East 68th St., New York, NY 10021; bmilrod@med.cornell.edu (e-mail). Editorial accepted for publication June 2009 (doi: 10.1176/appi.ajp.2009.09060810).

The author reports no competing interests.