by Michael D. Anestis, M.S.
The article in Newsweek published last week on why so many psychologists ignore science prompted a powerful response from many perspectives in a variety of forums. The article itself prompted a bit of an uproar in the comments section on Newsweek's website. Various outlets where I and others posted the article itself or commentary on its contents (e.g., Reddit) experienced similarly contentious debates. Another forum in which I encountered quite a few differing perspectives on the idea that many practicing psychologists are either unaware of scientific data on the relative utility of various forms of psychotherapy or simply ignore such information was the listserv for the Society for a Science of Clinical Psychology, which includes many prominent psychologists (and, in fact, one of the authors of the article upon which the Newsweek article was based has been a frequent contributor to the conversation). I highly encourage readers interested in the debate to consult and engage with these resources to further the discussion, regardless of your stance on the place of science in clinical psychology.
One point that came up many times across the different forums was the idea that empirically supported treatments (EST's) are a useless phenomenon because, in actuality, the alliance between therapist and client is the true determinant of outcome. Today, I would like to address this claim in some detail in an effort to dispel any misunderstanding people might have regarding the topic. I suspect some readers will vehemently disagree with the content of this article, but hopefully if that is the case, the end result will be a fruitful and respectful conversation in the comment section. In previous PBB articles, Joye has outlined data on numerous common factors - variables that are not specific to one particular form of therapy, but rather inherent aspects of any therapeutic environment - including therapeutic alliance and I encourage everyone to consult those articles for a more in depth look at this and other related topics. Today, I plan to discuss an influential article published in 2005 by Robert DeRubeis, Melissa Brotman, and Carly Gibbons of the University of Pennsylvania. In this article, the authors addressed why the widely held idea that therapeutic alliance is the primary mechanisms of change in therapy might be as ill advised as the Dodo Bird Hypothesis.
The main crux of the DeRubeis et al (2005) paper was that many researchers and clinicians have concluded that common factors are the driving factor behind improvement in therapy based upon two pieces of evidence:
- Studies that have found equivalence between different forms of therapy
- Studies that have shown a correlation between therapeutic alliance and outcome
Unfortunately, the data used to support the common factors assumption is highly flawed and DeReubeis and colleagues (2005) sought to explain to us exactly what those flaws are and how they might lead to erroneous conclusions.
Equivalence between forms of therapy
I have touched on this topic before, but DeReubeis et al (2005) gave a nice overview, so I thought I would summarize their thoughts here as well. Essentially, the majority of individuals who claim ESTs are a meaningless concept and that all treatments work the same base that claim on the Dodo Bird Hypothesis (Smith & Glass, 1977). This is problematic for a number of reasons. First of all, rather than comparing individual treatments to one another, the authors of the study upon which the Dodo Bird phenomenon is based actually grouped them into two arbitrary and, frankly, bizarre groups. Different forms of cognitive therapy - which made up the bulk of the well-researched forms of therapy, particularly in the late 70's - were included in each group, resulting in a comparison of two random groups that, in reality, were not divided into distinct categories.
In addition, as DeRubeis et al (2005) pointed out and I have mentioned on PBB before, the research question in the Dodo Bird paper and similar follow-ups has been, "which is better, Therapy A or Therapy B?" This question does not mention a specific problem and the authors noted that it is akin to asking which is better, insulin or an antibiotic, without specifying the condition for which the patient seeks to be treated. We would never do that in a medical situation and it is unclear why the rules of logic would somehow be altered specifically for clinical psychology. The better question is whether a particular treatment is better for treating a particular disorder than is another form of treatment (and that question can be further subdivided in order to see if the answer changes depending upon a particular characteristic trait or environmental factor).
There have been some cases in which well-designed research studies have compared two different forms of therapy for a particular diagnosis and found equivalent results. For instance, Elkin and colleagues (1989) found that cognitive therapy and interpersonal psychotherapy produced equivalent results in the treatment of depression. Some individuals look at that finding and conclude that common factors must explain the outcome. In other words, because the specific treatments did not lead to different outcomes, the things that the two therapies have in common must have been the driving factor in the outcome. DeRuebeis and colleagues (2005) pointed out, however, that this logic would never fly in other areas. For instance, if two different medications that work on different biological systems result in equivalent improvements, would we assume that they worked in the same manner? Alternatively, if a medication and psychotherapy produced equivalent results in the treatment of a mental illness would we assume that they somehow worked in the same manner? Certainly placebo effects could play a role there, but it would be silly to assume that a pharmacological treatment and psychosocial treatment work through the same mechanisms. As such, assuming that the failure to find a statistical difference between two forms of therapy means that common factors are all that matters is simply not a valid way of thinking about things and interpreting data.
The bottom line is, there are many forms of therapy that have demonstrated superior results for particular diagnoses relative to other approaches. DeRubeis et al (2005) mentioned four of them, but a more complete list and explanation can be found on the APA Division 12 website (see our earlier post detailing how to navigate that site), which is an easy to read resource for professionals and non-professionals seeking information on mental illness and psychotherapy.
A point that DeRubeis and colleagues (2005) surprisingly did not mention was the fact that treatment studies are almost universally underpowered. Statistical power refers to the degree to which you have enough people in your sample to actually determine if a difference is significant (to find an effect, small, medium, or large). Think of it this way: if you flip a coin four times and you get heads three times, you won't assume anything is odd about the coin. If you flip that coin four hundred times and you get heads three hundred times, however, you will likely think about the situation differently even though the percentage of times that the coin landed on heads was no different. In other words, we needed enough coin flips - power - to detect the difference. It was still there when we only flipped the coin four times, but we could not detect it statistically. In treatment studies, we don't have enough coin flips. Because these studies are so expensive and difficult to implement, the samples are generally very small so, unless one particular treatment absolutely destroys the other in terms of results, we are unlikely to be able to actually see any effects that are there. That does not mean there is no difference between the treatments, it simply means that the degree to which they are different requires a larger sample to fully understand. As such, we are likely to get a LOT of false negatives, erroneously indicating equivalence between therapies.
Therapeutic Alliance
Okay...finally onto the topic of alliance, which was what prompted me to write this article in the first place. Many people believe that the quality of the alliance between the therapist and the client is the most important factor in determining the outcome of treatment. A good alliance enhances the chances of a positive outcome while a negative alliance decreases those chances. This makes intuitive sense and, in fact, several studies have found a correlation between alliance and outcome. Horvath and Symonds (1991) found a correlation of .26 between alliance and outcome. Martin and colleagues (2000) found a correlation of .22 between alliance and outcome. What does that mean? It means that there is a small (Cohen, 1988) relationship between the two that indicates that when alliance is strong, outcomes are better. Keep in mind, however, that correlation does not mean causation. This is particularly important in this case, as DeRubeis and colleagues (2005) explained in detail. I'll come back to this point in a moment.
Proponents of the therapeutic alliance argument are, in large part though not universally, psychdynamically oriented or "eclectic." That being said, it makes sense to see whether the correlation between alliance and outcome is consistent across forms of therapy. In fact, it is not. With respect to cognitive therapy, some studies have found a positive correlation (e.g., Krupnick et al., 1996) and some have not (DeRubeis & Feeley, 1990). In other words, for some forms of therapy, the quality of the therapeutic alliance might be very important whereas, for others, simply following the protocol and properly implementing the therapy is what truly matters. Certainly a therapist with no social skills who can not make any sort of connection with a client is unlikely to see much improvement in his or her clients, but this does not mean that therapeutic alliance is the driving factor in outcome. Punching clients in the face is likely to result in bad outcomes too, but I don't think anyone would use that point to conclude that the key to therapy is not punching your client in the face.
Okay...back to the correlation does not equal causation point. A question that somehow has been overlooked in most of the literature on therapeutic alliance is whether a good alliance is a cause or a symptom of therapeutic improvement. In other words, do we get better because we have a good alliance with our therapists or do we have a good alliance with our therapist because we get better? This is an empirical question - one that we can measure with the right data. In studies by DeRubeis and Feeley (1990) and Feeley and colleagues (1999), early therapeutic alliance did not predict symptom change; however, symptom change predicted alliance late in therapy. In other words, whether or not the therapist and client had a strong alliance had no impact on whether or not the client improved, but whether or not the client improved had a strong relationship with whether or not the alliance was strong at the end of therapy. Similarly, Tang and DeRubeis (1999), in a study of cognitive therapy for depression, found that sudden improvements in symptoms were not predicted by alliance just prior to the symptom change, however, alliance quality was significantly higher in the session following sudden improvements. These studies indicate that we might form better alliances with therapists who help reduce our symptoms and poor alliances with therapists with whom we are less successful in treatment.
Very few studies have collected data capable of answering the question of whether alliance is a cause or symptom of outcome. To be fair, some studies have found results supportive of the alliance argument. Specifically, Barber et al. (2000) found that alliance predicted outcome even controlling for prior symptom improvement in brief dynamic therapy and Klein et al (2003) found a small alliance outcome relation, with early alliance predicting improvement and early improvement not predicting subsequent alliance, in CBASP. So what do we know? Well....not enough.
Conclusions
Quite frankly, this debate is a clear sign that more research is needed. We can't talk our way to a conclusion. More studies with large sample sizes and methodological rigor need to compare specific forms of treatment to other specific forms of treatment for particular diagnoses. Data need to be collected frequently throughout treatment and independent observers need to rate the degree to which the therapists adequately administer the particular forms of therapy. Outcome measures need to be agreed upon ahead of time so that we can legitimately determine relative success and samples need to include a diverse sample of clients in order to ensure that the findings can generalize to a broad population and not simply be dismissed by clinicians as irrelevant to their day-to-day work.
As it stands, however, the evidence does not support the Dodo Bird hypothesis, nor does it support the idea that therapeutic alliance or any other common factor is the mechanism of change in therapy. Such conclusions are based upon a misunderstanding of the data and of the proper methodological concerns of a rigorous scientific study.
I know today's article was a bit dense, but I am rather fired up about this topic these days and feel that it is important for people to understand these issues clearly in order to make informed decisions regarding their mental health care. If you have questions, please feel free to post them as comments or to email them to me. If you disagree with any portion of this article, please feel free to post that comment and know that all perspectives are welcome as long as they are respectful and base their claims on evidence rather than philosophy. If you do agree but think I overlooked something or could explain something in an alternative manner, I would love to hear that as well. In the coming days, I fully intend to touch on related topics, so hopefully an interesting dialogue amongst PBB and readers will emerge.
If you are interested in learning more about empirically supported treatments, we recommend the following items, all of which are available in our online store of science-based psychological resources:
- A Guide to Treatments that Work
by Peter Nathan and Jack Gorman
- Clinical Handbook of Psychological Disorders, Fourth Edition: A Step-by-Step Treatment Manual
by David Barlow
- Science and Pseudoscience in Clinical Psychology
by Scott Lilienfeld, Steven Lynn, and Jeffrey Lohr
- Cognitive Therapy Techniques: A Practitioner's Guide
by Robert Leahy
Mike Anestis is a doctoral candidate in the clinical psychology department at Florida State University
Subscribe in a reader
