Robert Elliott & Elizabeth Freire
University of Strathclyde
Among his various accomplishment, Carl Rogers was a pioneer in the field of psychotherapy research. Following in his footsteps, we have just completed a major project to integrate 60 years of research on the effectiveness of Person-Centred and related therapies. Our results have confirmed, strengthened and extended previous results, using much larger sample of more 180 scientific outcome studies. With a few important exceptions, these results are good news for therapists and counsellors working within the Person-Centred Approach, because they provide multiple lines of evidence demonstrating that these therapies are highly effective.
We presented our results this past July at the Person-Centred/Experiential (PCE) therapies conference in Norwich, England. Although we used more rigorous methods than in previous analyses, we found some of the results so different from widely-held beliefs in the mental health field that we felt compelled to go back and re-do them, just to make sure we had not made a mistake somewhere. The pace of research on PCE therapies continues to accelerate, making it difficult to keep up with, and the data set we have assembled so far is rich enough to keep us busy for several years. At this point, however, the main findings are clear and can be summarized as follows:
Conclusion 1: PCE therapies are associated with large pre-post client change.
To establish this, we looked at 203 samples of clients, from 191 studies, amounting to more than 14,000 people. In these studies, the researchers had provided enough quantitative information about how the clients were doing before and after therapy that we could calculate effect sizes, in this case the difference between how the clients were doing on average before they started therapy and how they were doing on average after therapy. Because different researchers use different outcome measures, we converted all these pre-post differences to a common metric by dividing them by an estimate of the variability of the scores (a standard deviation, sd for short).
What did we find? Similar to previous analyses of smaller samples of clients, we obtained an average effect size of 1.01 standard deviation (sd) units. Social scientists consider this to be a very large effect, many times larger than effects typically found for common medical procedures or medications. In other words, on average, PCE therapies make a big difference for clients. Furthermore, contrary to what many PCE therapists believe, this is particularly true for symptom measures like the CORE-OM, as indicated by the two large UK-based studies by Stiles et al. (2006, 2008).
Conclusion 2: Clients' large posttherapy gains are maintained over early & late follow-ups.
Next, we looked to see if clients retained the benefits of PCE therapy over time. The answer to this question is yes also. In fact, our analyses indicate that if anything clients in PCE therapies show slight further gains within the first year after therapy (effect size: .99 immediately after therapy vs. 1.12 sd for follow-ups less than a year after therapy). Furthermore, these gains are maintained at and beyond the one-year mark (effect size 1.13 sd). This stability of post-therapy benefit is consistent with the PCE philosophy of enhancing client self-determination and empowerment, indicating that clients continue to develop on their own after they have left therapy.
Conclusion 3: Clients in PCE therapies show large gains relative to clients who receive no therapy.
In order to show that there is causal relationship between PCE therapy and client change, it is necessary to compare clients who get therapy to those who don't. For example, some clients might be asked to wait a period time before starting therapy, so researchers can see whether they would have changed on their own without therapy; this is called a "waitlist control design". These studies are most convincing when the assignment to therapy or no-therapy (or waitlist) is random (making it a "Randomized Clinical Trial" or RCT). This is because randomization tends to make the two groups of client roughly equivalent to start with.
We analyzed data from 60 studies, involving more than 2100 PCE clients compared to more than 1900 controls. For each of these studies we first measured the amount of pre-post change in the PCE therapy clients, and then calculated how much people in the no-therapy or waitlist group changed in the absence of therapy. Finally, we measured how much more or less clients in PCE therapies changed in comparison to no-therapy clients. This difference in the amount of change is the controlled effect size. We found a controlled effect size of .81 sd, which is considered to be a large effect size. (Clients who received therapy showed very little change: .19 sd.)
About half of these controlled studies did not randomize clients to receive PCE therapy or not; these studies are generally dismissed by scientific review panels like those charged with developing and revising the NICE guidelines. For this reason, we ran the same analyses for the 31 Randomized Clinical Trials within our sample (some 550 PCE clients), and found that randomization made almost no difference (controlled effect size: .78 sd). Scientists tend to believe that these kinds of careful results allow us to conclude that therapy causes client change. This provides the second main line of evidence for the effectiveness of PCE therapies.
Conclusion 4: PCE therapies in general are clinically and statistically equivalent to other therapies.
How do PCE therapies stack up against other therapies? To answer this question, we first assembled a large collection of 109 studies, including 134 comparisons between PCE and other therapies, among them CBT; these studies contained data from more than 10,300 clients. As with the controlled studies just described, we first calculated how much clients changed in PCE therapy, then how clients seen in other therapies changed, and finally how much more or less PCE clients changed than the other clients. Overall, there was virtually no difference between PCE and other therapies (comparative effect size: .01 sd), indicating equally large amounts of change. That is, in general PCE therapies were neither more nor less effective than other therapies. Once again, we weeded out the nonrandomized studies, leaving 91 so-called "gold standard" RCTs, with virtually identical results.
Conclusion 5: PCE therapies in general might be trivially worse that CBT.
It is commonly assumed by CBT therapists, government officials, and the general public that CBT has clearly better outcomes than other therapies such as PCE therapies. The pre-post and controlled results we have just described do not address this issue. Therefore, we looked at the 78 studies in which PCE therapies were compared to CBT, including 63 RCTs. When all these PCE therapies were pooled together they at first appeared to be slightly but trivially less effective than CBT (78 studies; effect size: -.18 sd = a small effect; for the 63 RCTs the effect size was -.16). However, this effect disappeared when we statistically controlled for the theoretical orientation of the researcher (referred to as researcher allegiance), a frequent source of bias in treatment research.
Conclusion 6: So-called "Supportive" therapies have worse outcomes than CBT but other kinds of PCE therapy are as effective or more effective than CBT.
Why should controlling for the researcher's theoretical allegiance make the trivial superiority of CBT go away? In order to understand what was going on, we divided the PCE therapies into four types:
(1) Pure Person-centred therapy, following Carl Rogers, including both classical (non-directive) or relational (as practiced in the UK) and broader new forms of Client-centred therapy (as practiced in Europe).
(2) Therapies typically labelled by researchers as "supportive" or "nondirective-supportive"; further investigation of these revealed them to be watered-down, typically non bona fide versions of PCE therapies, commonly used by CBT researchers, especially in the USA
(3) Process-Experiential (also known as Emotion-Focused Therapy), developed by Greenberg, Rice and Elliott, and recently recognized as an empirically supported therapy in the USA
(4) Other experiential therapies, including Gestalt, Focusing-oriented, expressive and so on.
What we found when we did this was that the small effect in favour of CBT could be accounted for the presence of the "supportive" therapies. That is, studies in which these therapies were used have substantially worse outcomes when compared (by CBT researchers) to CBT (38 studies; effect size: -.35 sd; for the 33 RCTs the effect size was -.29).
In contrast, once the supportive therapies were removed, the effects of the bona fide PCE therapies could be seen more clearly: Pure PCT appeared to be statistically equivalent in effectiveness to CBT (22 studies, including 18 RCTs; effect size: -.09 sd for each), even without controlling for researcher allegiance. In a small number of studies, the newer Process-Experiential Therapies for individuals or couples actually appeared to be more effective when compared to CBT (7 studies; effect size: .35 sd; for the 4 RCTs the effect size was .55sd). (However, this effect may also be due to researcher allegiance.) (Other experiential therapies were also equivalent to CBT: 10 studies: -.14sd; including 7 RCTs: -.07sd.)
What are the Implications of these analyses?
In fact, these results are uniformly good news for Person-Centred/Experiential practitioners: Clients use our therapies to make large changes in themselves; these changes are maintained over time and are much larger than our clients would have experienced without therapy. Furthermore, our clients show as much change as clients seen in other therapies, including CBT, but only if bonafide Person-Centred, Process-Experiential and Other Experiential therapies were involved.
From a policy point of view these data support the proposition that Person-Centred/Experiential therapies are empirically supported by multiple lines of scientific evidence, including "gold standard" RCTs and recent very large RCT-equivalent studies in the UK (e.g., Stiles et al., 2006, 2007). This body of research suggests that the NICE Guidelines need to be updated, and that PCE therapies should be offered to clients in primary care, NHS, and other mental health settings. Relying on multiple lines of evidence, such as provided in the present study, provides a sound basis for establishing public mental health policy. The shortfall in the availability of psychological therapy in the NHS could be instantly resolved if health authorities were to draw upon the large body of trained Person-Centred counsellors and psychotherapists throughout the UK.
For those of us in the PCE tradition, the moral of this story is that we do not need to be afraid of quantitative either outcome research or RCTs. However, if we let others define our reality by studying watered-down versions of what we do, we are going to be in trouble. For this reason, it is imperative that PCE therapists do our own outcome research - including RCTs -- on legitimate versions of our therapies. As Carl Rogers said, "The facts are friendly."
Stiles, W.B., Barkham, M., Mellor-Clark, J., Connell, J. (2007). Effectiveness of cognitive-Behavioural, person-centred, and psychodynamic therapies as practiced in UK primary care routine practice: replication in a larger sample. Psychological Medicine. Published online 10 September 2007. doi:10.1017/S0033291707001511.
Stiles, W.B., Barkham, M., Twigg, E.., Mellor-Clark, J., Cooper, M. (2006). Effectiveness of cognitive-behavioural, person-centred and psychodynamic therapies as practised in UK National Health Service settings. Psychological Medicine, 36, 555-566.
Author note: Robert Elliott can be contacted at [fac0029 @ gmail.com]; Beth Freire can be contacted at [elizabeth.freire @ strath.ac.uk].