Wrong numbers

First, Solem’s conclusion that only a limited number of studies has examined the therapies that are delivered in Norway is incomplete. He investigated one dataset that is publicly available on the website of our project(www.metapsy.org), but (as is clearly indicated on the website) this is not the most recent version of this dataset (we do not make the most recent versions publicly available). Solem worked on the dataset with the studies up to 1 January 2022 (415 trials). However, the current version of our database includes 552 trials (33% more). Furthermore, these are only the studies comparing therapies with control conditions. We also have several other datasets examining the effects of psychotherapies for depression (not (yet) available on our website), including 135 trials comparing psychotherapy with pharmacotherapy and combined treatments, a dataset of 152 trials comparing different types of therapy with each other, and a dataset of 44 trials in inpatients. So, the conclusion that there are only 11 trials on the representative therapies in Norway is, at best, based on outdated and incomplete information.

Wrong selection of studies

Second, the reasoning behind the selection of the 11 studies (individual therapy; > 10 sessions; only specific types of therapy) is highly disputable. The criteria for this selection are based not on evidence, but on the fact that this is what therapists in Norway are trained in. It is much better, however, to look at what the evidence says. There is no evidence that group therapies or guided self-help are less effective than individual therapies (Cuijpers et al., 2019). Nor is there any evidence to suggest that therapies with more than 10 sessions are more effective than those with fewer sessions (Ciharova et al., 2024; Cuijpers et al., 2013). We also conducted a large network meta-analysis of eight major types of therapy (examined in at least 10 trials) that found no significant differences between therapies (Cuijpers et al., 2021a). So why would one select studies based on criteria that have not been found to be associated with differences in effect size? The only criterion that makes any sense is to limit the selection to Western countries, because we have seen in several meta-analyses that therapies are somewhat more effective in non-Western countries (Cuijpers et al., 2018; Tong et al., 2024).

Although Solem did not conduct a meta-analysis of his selection of studies, we ran these analyses and compared the outcomes of his selection with those of the other studies in our database. We removed the studies conducted before 2022 from the most recent version of our database and made a subgroup of the studies selected by Solem (the databases are not exactly the same because we make minor changes in the continuous process of improving them). We found that the overall effect size of all 523 comparisons (from 427 trials) was SMD=0.69 (95% CI: 0.63; 0.74; I2=82; 95% CI: 81; 83). Solem’s subgroup of 15 comparisons (from 11 trials) between therapy and control conditions resulted in SMD=0.71. The p-value for the difference between this group of studies and the other studies was p=0.86. This means that there is no indication whatsoever that the studies selected by Solem differ significantly from the other studies in terms of outcomes.

Wrong focus on details

Third, Solem dives into the details of the studies included in his selection and those of some other studies from our database. Apart from the fact that he makes several methodological errors in looking at these details (too many to list here), he is right that the studies included in our database vary considerably. He recommends that readers of meta-analyses ‘should critically examine which studies are included’. But how does he propose to do that? Who can read 1,155 randomised trials (that is the number of trials currently included in our database of psychotherapies for depression) and make sense of that? There is a reason why Solem only explores the three studies with the largest and smallest effect sizes. Had he taken his own advice seriously, he would have looked into the 100 studies with the largest and smallest effect sizes. But, of course, that is not possible and would not have resulted in a useable result.

A better alternative would be to examine whether the differences between studies are related to the outcomes in meta-analyses. If Solem had read our meta-analyses, he would have seen, among other things, that all our meta-analyses include many sensitivity analyses (excluding extreme outliers, such as the Nigerian studies (e.g., Cuijpers et al., 2025)); that we have examined the differences between waitlist and care-as-usual extensively (and yes, waiting lists overestimate effect sizes (Cuijpers et al., 2021b; Cuijpers et al., 2024)); and that we always examine risk of bias (including the difference between completers and intention-to-treat data). He would also have seen that there are many trials that compare therapies with pharmacotherapies (‘good active control conditions’) and that there are more than 150 trials comparing different therapies directly with each other. And if one looks from a distance at this whole body of literature, the results all point in the same direction. All this is summarised in a systematic review of the more than 100 meta-analyses of psychotherapies for depression we have conducted since 2007 (Cuijpers et al., 2024). Yes, some differences between studies do have an impact on the outcomes, and we know about this impact on outcomes precisely because of the meta-analyses that have been conducted.

Wrong starting point

Fourth, had Solem not focused solely on our old meta-analysis (Cuijpers et al., 2014) and instead read more recent results and discussions of our meta-analyses, he would also have seen that our research does not support the conclusion that therapy has little effect. We have found convincing evidence that psychotherapy is as effective as pharmacotherapy in the short term but is more effective in the long term (after one year (Furukawa et al., 2021; Cuijpers et al., 2020)). Even one round of therapy in the acute phase is more effective after one year than taking antidepressants during that whole year. However, combined treatment is better than either of the two, in both the short and the long term.

Another important finding in this regard is that we examined the absolute response rates (defined as 50% symptom reduction from pre- to post-test) of therapies. We found that 42% of people receiving therapy responded, while 16% to 17% of people in the control groups responded (Cuijpers et al., 2021c). These numbers probably differ depending on baseline severity (higher response rates in more severe depression). This means that many people do not respond to therapy. Based on these numbers, one could say that while therapy is certainly effective for a considerable group of patients, it is not sufficient for many. However, in routine care, therapists can extend the therapy when there is no sufficient improvement, can recommend adding antidepressants or can switch to another therapy. In practice, this means that most patients do improve, though they may require several rounds of therapy to do so.

Conclusion

The suggestion that therapies have little effect cannot be drawn from our work, and certainly not from a small meta-analysis of a small subset of a large research field. Psychotherapies are certainly effective and have better outcomes than those for control groups who do not receive these therapies.

I certainly hope that, when talking to their patients, therapists will not discuss only the results of the studies in which they themselves were involved, as Solem suggests (incidentally, his own study also does not provide a good estimate of the true effect, due to several flaws). Meta-analyses are certainly superior to simply looking at single studies. Examining all single studies is impossible due to the huge number of trials currently available.

Solem also recommends that more research should be done on the therapies that are currently conducted in routine care in Norway. I see no need for such research, because there is no indication whatsoever that the usual therapies provided in Norway differ significantly from other therapies. Therapies do not need to be done individually, and are just as effective when the number of sessions is smaller.

It would be much better to take a critical look at the available (meta-analytic) research and innovate current practice in Norway using these results.