The useful comparison asks four questions: for whom, compared with what, on which outcome, and for how long? Regulatory status and payment then add two more questions. They matter, but neither substitutes for the clinical result.

  • Digital does not mean untested—or automatically proven. The particular product and indication matter.
  • Improvement from baseline is not the same as benefit over a comparator. Both treatment and control groups may improve.
  • Adjunctive use is a meaningful outcome. A useful tool need not replace all other care.
  • Authorization, engagement and reimbursement are different tests. Passing one does not guarantee the others.

What separates a wellness app from a medical product?

A general wellness app and software intended to diagnose or treat a disease can have very different regulatory obligations. But there is no global rule that every medical function requires a large, double-blind randomized trial or that every prescription app has demonstrated superiority to medication. Jurisdiction, intended use, risk and the regulatory route affect the requirements.

Likewise, “over the counter” describes access, not the absence of medical-device regulation. The FDA cleared EndeavorOTC in June 2024 for a defined attention-function indication in adults with ADHD. Its labeling explicitly limits the claim and says it is not a replacement for other forms of treatment. FDA clearance and indications for EndeavorOTC.

A reader should therefore check the actual indication and study, not infer efficacy from a prescription requirement, a CE mark or a polished interface. Conformity or clearance applies to a specific product and use. It does not certify every claim subsequently made about it.

Insomnia: promising behavioral treatment, with limits to comparison

Digital cognitive behavioral therapy for insomnia is one way to deliver structured behavioral treatment. Its potential value includes making a program available without requiring every component to be delivered in a face-to-face appointment. It still requires engagement, and a software format does not make every program appropriate for every patient.

The DREAM study evaluated Somryst in a real-world, single-arm setting and reported improvements in self-reported insomnia severity. It did not randomize participants against medication or another treatment. Changes can be informative about use in practice while remaining vulnerable to selection, missing follow-up and other influences. The original DREAM report.

A subsequent sponsor-associated conference poster reported a mean Insomnia Severity Index of 12.0 among 103 respondents at the 24-month assessment; 266 participants had reached that assessment point. These are important denominators. They do not mean that every original participant had a documented durable response, and a symptom score cannot prove permanent changes in brain structure. The 18- and 24-month poster.

A network meta-analysis also compared digital therapy with face-to-face CBT-I and insomnia medicines indirectly across studies. Such an analysis can estimate relative effects, but its rankings depend on the comparability of populations, outcome timing and trial methods. It is not the same as assigning comparable patients directly to the app or a particular medicine. The insomnia network meta-analysis.

The defensible conclusion is that evidence supports investigating and using particular digital behavioral interventions within appropriate care. The claim that software permanently rewires the brain or is superior to every sleeping medicine goes beyond the measurements.

Chronic back pain: the comparison group changes the interpretation

A randomized trial of RelieVRx compared an eight-week home virtual-reality program with an active sham intervention. Its modified intention-to-treat analysis included 1,067 participants with chronic lower back pain. At the end of treatment, the advantage over sham was approximately 0.41 points for pain intensity and 0.52 points for pain interference on 0–10 scales. The randomized RelieVRx effectiveness trial.

The active program showed a larger improvement from its own baseline. That is not the same effect: the sham group improved too. Reporting only the treatment group's change makes the product appear to receive credit for changes that were also present in the comparison group.

The study did not compare VR with opioids and cannot establish that the program replaces them or produces medication-free relief. Absence of drug exposure avoids some medication-specific risks, but does not make a digital device risk-free or suitable for all users. Safety, contraindications, burden and clinical usefulness still require product-specific assessment.

This is a case in which modest additional benefit may still matter to some people. A careful account can recognize that without converting an effect estimate into a claim of universal superiority.

ADHD: an attention test is not every part of daily functioning

The STARS-ADHD randomized trial enrolled 348 children aged 8–12 and compared the digital intervention with a control. Its primary computerized attention measure favored the intervention. The study did not show significant between-group differences on several secondary behavioral and impairment measures. The original STARS-ADHD trial.

This difference between endpoints should be central to the explanation. A better score on a standardized attention task is a real measured result; it does not automatically mean improved classroom behavior, academic performance or every symptom of ADHD. Nor was this trial a comparison demonstrating superiority to stimulant treatment.

The pediatric prescription product and the adult OTC product must also be kept separate. An adult indication cannot be transferred to children, and changing a commercial strategy does not erase the conditions of a regulatory authorization.

Three examples, three different evidence questions

  • Digital CBT-I. Evidence described here: Real-world symptom follow-up; separate indirect comparative analysis What it cannot establish by itself: Permanent neural change or superiority to every medicine
  • Behavioral VR for back pain. Evidence described here: Randomized advantage over an active sham at eight weeks What it cannot establish by itself: Replacement of opioids or a guarantee of lasting relief
  • Digital ADHD intervention. Evidence described here: Randomized improvement in a computerized attention endpoint What it cannot establish by itself: Improvement in all behavioral outcomes or replacement of other care

The comparison does not rank the conditions against each other. It shows why the same word, “effective,” can conceal materially different claims. A clinical decision needs the actual population, effect, uncertainty and alternatives.

Why clinical evidence does not guarantee a viable business

A digital treatment can face several independent obstacles: eligible users may not start it, users may not complete it, clinicians may not have time to integrate it, and payers may not cover it. These are possible explanations to investigate, not a reason to assign every company failure to one cause.

Evidence of symptom improvement is also different from evidence of savings. To claim that an app reduces hospital use or medication spending, the analysis must measure those outcomes with an appropriate comparison. A before-and-after difference may reflect changes unrelated to the product.

The practical commercial question is who bears the cost of software, onboarding, clinical review and support—and who receives the benefit. Reimbursement is a contractual and policy question alongside the clinical one. It does not change a weak trial into a strong trial, and a strong trial does not automatically create coverage.

Germany's DiGA system and the US TEMPO pilot answer different questions

Germany's DiGA route connects eligible digital applications listed by BfArM with statutory health-insurance access. Permanent and provisional listing must be distinguished: an application can enter an evidence-generation period before the required positive healthcare effect has been fully demonstrated. Directory status should be read together with indication, eligibility and evidence status. BfArM's DiGA guide.

Section 134 of Germany's Social Code V requires outcome-related components of at least 20% in the relevant remuneration agreements from January 2026. That is not a rule that 20% of every individual prescription is paid only when that particular patient is “healed.” The legal mechanism and negotiated measures matter. SGB V section 134.

The FDA's TEMPO pilot has a different purpose. It allows manufacturers to request case-specific enforcement discretion for certain device requirements when their products are offered through the CMS ACCESS context. The agency began accepting statements of interest in January 2026 and expects participating manufacturers ultimately to seek the appropriate marketing authorization. Participation does not amount to ordinary clearance or unrestricted permission to market an unapproved use. FDA's TEMPO explanation.

Neither initiative proves that the industry has solved engagement, evidence quality or payment. In particular, utilization numbers from 2025 cannot demonstrate the effect of a pilot starting in 2026. Policy design should be assessed through results collected after implementation, with a suitable comparison.

The useful decision is narrower than pills versus code

The most informative questions are specific: what benefit did this product demonstrate, in which people, against which comparator, and with what follow-up? How many participants completed the assessment? What risks and burdens were measured? Is the proposed use the one actually authorized and studied?

A digital intervention can complement medication, deliver a behavioral component or address an access gap. It can also fail to add enough benefit, be difficult to use or remain unaffordable. Treating either software or medicines as a single category with one safety profile obscures those differences. Useful care depends on matching a particular intervention to a particular need, not declaring a technological winner.

Our digital psychotherapy evidence review separates symptom improvement from effects on suicidal behavior and mortality.