In 2022, for the first time in more than a decade, the number of published psychotherapy trials fell. Two hundred forty-one randomized controlled trials appeared that year, down 13 percent from 2021. The dip did not last, and 2023 brought a partial recovery, but it interrupted a growth curve that had run uninterrupted through the 2010s.
That number comes from a trend analysis by Ian Sivel, Winfried Rief and Cornelia Weise, psychologists at Philipps-University of Marburg and Friedrich-Alexander-Universität Erlangen-Nürnberg. They searched PubMed, Web of Science and PsycINFO for psychotherapy trials published between January 2019 and December 2023, screened more than 21,000 records down to 1,223 eligible studies, and sorted each one by the kind of therapy tested, the country it came from and the age of the people in it. Their question was narrow and specific: when the pandemic made in-person research nearly impossible, what did the field study instead?
The headline answer is that it studied screens. Across the 1,324 treatment arms in the dataset (a single trial can test more than one therapy, so arms outnumber trials), eHealth interventions accounted for 43 percent of activity, edging past cognitive behavioural therapy at 37 percent. The authors use eHealth to mean asynchronous, mostly self-guided internet programs, and telehealth to mean live sessions over video or phone. Together, those two categories rose from 40.1 percent of all treatment arms in 2019 to 2021 to 54.6 percent in 2022 and 2023. They were the only approaches whose share grew significantly. Everything else, including psychodynamic therapy, systemic therapy, mindfulness-based approaches and acceptance and commitment therapy, held under 7 percent apiece and moved essentially not at all.
But the reversal is smaller than it first appears, and the authors say so plainly. Of the 554 eHealth treatment arms they counted, 70.9 percent tested internet-delivered CBT. No other therapeutic model exceeded 8 percent of the digital corpus. What changed was the delivery channel, not the theory being delivered. CBT-derived treatment, whether a therapist offered it in a room or an app offered it on a phone, remained the empirical center of the field throughout. The team's reading is that manualized, well-documented protocols with a track record before 2020 were simply the ones that could be moved online fast, and iCBT fit that description better than anything else.
The 2022 decline deserves a caveat the authors are careful to attach. They projected forward from the 2019 to 2021 growth trend using Monte Carlo simulation, and both 2022 and 2023 fell below the resulting 95 percent prediction intervals. That suggests the downturn was more than random noise. Still, the baseline was only three years long, so the team calls the projection exploratory and descriptive rather than confirmatory. Publication year is also not the year a trial ran, which means these are trends in the published record, not a direct readout of what researchers were doing in real time.
Some of the sharpest findings are about who is missing. Trials came from 52 countries, but 73.4 percent originated in high-income settings, with Europe contributing 41.5 percent and North America 31.9 percent. Of the 1,183 trials with a clear age focus, 978 targeted adults and 205 targeted children or adolescents. Adults were 4.77 times more likely to be the population under study. A few countries stood out against the average: China was heavily overrepresented in mindfulness trials, the United States in telehealth, Sweden and Switzerland in eHealth, Norway in CBT. The authors note that searching only English and German publications may itself have deepened the geographic tilt.
Why it matters
Which treatments get trialled is not an academic bookkeeping question. Randomized trials feed clinical guidelines, and guidelines shape what insurers cover, what clinicians train in, and what a person is offered when they finally ask for help. A therapy that is widely practiced but thinly trialled sits in an awkward position, and the authors point out that the limited RCT presence of psychodynamic, systemic and integrative approaches makes their evidence-based standing harder to argue, even where clinicians use them daily.
The same logic applies to the demographic gaps. Children, adolescents and people in low-resource settings carry substantial unmet mental health need and appear least often in the trials that define what counts as evidence. The team links the child gap to concrete obstacles: stricter ethics review, complicated consent, thin institutional support for involving young people in research. None of that makes the existing evidence wrong. It does mean conclusions drawn from a narrow slice of the world should be generalized carefully.
Whether the digital shift sticks is genuinely unresolved. Mindfulness-based therapies rose sharply in the 2010s and then plateaued, and the authors raise that precedent as a reason for caution. Their own framing is modest: this is an early diagnostic signal from a five-year window, useful for spotting a turn, not for predicting where it ends.