Digital therapies: inside a 254-trial research map

Mental-health conditions account for about half the studies in a newly published map of digital-therapy trials. That finding describes a selected ClinicalTrials.gov dataset collected through January 2025, rather than the current worldwide market or a list of treatments proven to work.

The PLOS Digital Health paper, published on 14 October 2026, is useful for understanding where this research was concentrated. Its definitions and exclusions are just as important as the headline count of 254 trials.

Which software made it into the map

The authors looked for patient-facing software whose main digital function delivered a therapeutic intervention. Their criteria excluded general wellness products, software used only for diagnosis or monitoring, clinician-facing tools and systems concerned solely with communication.

These boundaries explain why “digital therapy” is narrower here than “health app”. A symptom-reporting service or a clinician’s decision-support tool cannot automatically be added to the same count.

Mobile apps were the most common technology in the selected studies. The map also included web applications, virtual reality, video games and apps combined with wearables. These describe delivery technologies, not a ranking of clinical effectiveness.

The largest research areas

The paper’s therapeutic-area classification puts psychiatric conditions first, including studies addressing insomnia, depression and anxiety. Nervous-system disorders form the next largest group. The table below reproduces four categories from the paper; it is a selected view, not the complete classification.

There were 254 entries overall. The first row is approximately 51.6% of that total. These are trial entries, not necessarily unique commercial products, and the counts cannot establish which disease group benefits most from digital treatment.

Not every entry is a randomized trial

The supplied supporting workbook contains 254 unique trial identifiers. Checking its allocation field gives 214 marked randomized and 40 marked not applicable.

That matters when describing the evidence base. A single-arm study and a randomized comparison can both add information, but they do not answer the same causal question. The overall count should not be presented as 254 randomized demonstrations of benefit.

The workbook also includes completed, recruiting and not-yet-recruiting studies. A registration records a research plan and its reported status; it is not equivalent to a published positive outcome. A blank registry results-date field, in turn, cannot prove that no journal paper exists.

What the map leaves outside its frame

The study searched one registry and excluded terminated and withdrawn studies. Trials recorded only in other registries may be absent, while excluding discontinued work limits what the dataset can say about unsuccessful development.

Its data cutoff also predates publication by many months. The map cannot establish current availability, reimbursement or the number of active studies today. Differences between countries may reflect registration practices and the search definition as well as research activity; the counts do not establish what caused those differences.

A clinical judgement about an individual therapy still depends on its results, comparator, outcomes and limitations. The map’s descriptions of registration and recruitment status cannot supply those missing details.

For a service considering a particular digital treatment, the next useful unit of evidence is therefore the specific intervention and its evaluation. A research map identifies where to look. It does not replace the product evidence or the separate assessment an NHS buyer undertakes.