Absolute and relative outcomes of psychotherapies for eight mental disorders: a systematic review and meta-analysis, 2024, Karyotaki et al

Absolute and relative outcomes of psychotherapies for eight mental disorders: a systematic review and meta-analysis
https://onlinelibrary.wiley.com/doi/10.1002/wps.21203

Psychotherapies are first-line treatments for most mental disorders, but their absolute outcomes (i.e., response and remission rates) are not well studied, despite the relevance of such information for health care users, providers and policy makers. We aimed to examine absolute and relative outcomes of psychotherapies across eight mental disorders: major depressive disorder (MDD), social anxiety disorder, panic disorder, generalized anxiety disorder (GAD), specific phobia, post-traumatic stress disorder (PTSD), obsessive-compulsive disorder (OCD), and borderline personality disorder (BPD). We used a series of living systematic reviews included in the Metapsy initiative (www.metapsy.org), with a common strategy for literature search, inclusion of studies and extraction of data, and a common format for the analyses. Literature search was conducted in major bibliographical databases (PubMed, PsycINFO, Embase, and the Cochrane Register of Controlled Trials) up to January 1, 2023. We included randomized controlled trials comparing psychotherapies for any of the eight mental disorders, established by a diagnostic interview, with a control group (waitlist, care-as-usual, or pill placebo). We conducted random-effects model pairwise meta-analyses. The main outcome was the absolute rate of response (at least 50% symptom reduction between baseline and post-test) in the treatment and control conditions. Secondary outcomes included the relative risk (RR) of response, and the number needed to treat (NNT).

Random-effects meta-analyses of the included 441 trials (33,881 patients) indicated modest response rates for psychotherapies: 0.42 (95% CI: 0.39-0.45) for MDD; 0.38 (95% CI: 0.33-0.43) for PTSD; 0.38 (95% CI: 0.30-0.47) for OCD; 0.38 (95% CI: 0.33-0.43) for panic disorder; 0.36 (95% CI: 0.30-0.42) for GAD; 0.32 (95% CI: 0.29-0.37) for social anxiety disorder; 0.32 (95% CI: 0.23-0.42) for specific phobia; and 0.24 (95% CI: 0.15-0.36) for BPD. Most sensitivity analyses broadly supported these findings. The RRs were significant for all disorders, except BPD. Our conclusion is that most psychotherapies for the eight mental disorders are effective compared with control conditions, but absolute response rates are modest. More effective treatments and interventions for those not responding to a first-line treatment are needed.

 

Aside from the general problem of GIGO in systematic reviews, those so-called response rates seem to mostly fall about where random noise would. They all fall short of IAPT's 50% recovery rate, and IAPT is applied to many health problems that clearly aren't mental disorders. And of course those studies always rely on far more expensive and lengthy courses than the McTherapy model that IAPT uses, which is closer to how things usually go in real life.

This is basically psychotherapy as applied for its intended purpose in unnatural settings, far above how it's used in real life, and it performs very poorly. Still, the framing is the usual "some people respond, we have to figure out how to succeed with the non-responders", and doesn't seem to consider the high probability that this is mostly noise and very little of this is actually effective at anything but employing people to do something ineffective. That there are no more responders than non-responders.

But, sure, it's 100% safe and effective at treating severe chronic issues like pain and post-infectious illness. Because why the hell not?

 

I'm pretty sure that IAPT chose 50% because it's the point at which the abysmal effectiveness of this model would become economical, if it worked. Which it doesn't.

But the people involve don't understand the economics involved here. They make the assumption that as they scale up, they get better returns. That if they treat more people, they will get the number-to-treat that makes it cost-effective. This is what you get with industrial scale. But this isn't a manufacturing industry, it's a direct contact service industry.

Therapy doesn't just scale linearly, it has a negative curve. The cost per unit never goes down, because the mode is always 1:1, and as you scale a linear model like this, you need more and more layers of management, coordination, administration, facilities and so on, all while the cost per unit never goes down. So the cost-effectiveness actually goes down as they scale up, and their only solution appears to be scaling up more, which only brings the RoI down further. But they promised that it would be cost effective, and after wasting billions on a failed boondoggle, no one has the courage to admit it.

 

It also allows them to claim more successes, leaving out that there are more failures too, but that's the fault of the patients not trying hard enough, so they don't count.

I agree that it sounds like random noise. I expect if they tried an absolutely nonsense therapy, they would get similar rates of success.