Why Digital Therapeutics Are Failing Independent HTA Reviews — and How to Avoid It

Independent health technology assessment has become the hardest gate in digital health. Not FDA clearance, which most digital therapeutics either obtain or sidestep. Not app store distribution. The gate is the moment an independent reviewer reads your evidence and decides whether the effect is real, clinically meaningful, and durable enough to pay for.

Over the past two years those reviews have gone badly for a lot of products. The pattern is consistent enough to be predictable, which means it is also avoidable.

What the reviews keep finding

The Peterson Health Technology Institute publishes independent assessments of digital health solutions using a framework built with ICER. Its evaluations of digital diabetes management, hypertension and musculoskeletal tools reached broadly similar conclusions: measurable clinical effects in some categories, smaller than vendor claims, and in several cases not large enough to offset the cost of the technology to the health system.

Its synthesis work on remote monitoring went further, finding that benefits are frequently time limited. Hypertension tools can get patients to control quickly, within three to six months, after which continued monitoring adds cost without adding much clinical value.

NICE reaches similar conclusions by a different route. Its early-use HealthTech route exists precisely because digital technologies arrive with thin evidence; the recommendation is usually conditional use while evidence is generated, not routine adoption. Germany's DiGA fast track is built on the same logic — a provisional listing, then a positive healthcare effect demonstrated inside twelve months or removal.

None of this is anti-digital. It is what happens when a product designed around engagement metrics meets reviewers who are looking for clinical outcomes.

The five failure modes

Weak comparator. The trial compares the app against a waitlist or usual care defined as nothing. Reviewers replace that with the real alternative — an existing programme, a pharmacological option, a nurse-led clinic — and the incremental benefit collapses.

Surrogate endpoints. Engagement, app opens, self-reported symptom scores at eight weeks. These do not persuade an assessor looking for HbA1c, blood pressure control, avoided admissions or validated functional outcomes at a clinically relevant horizon.

Short follow-up. A twelve-week randomised study cannot answer a question about a chronic condition managed over years. When reviewers extrapolate, the uncertainty they add is usually fatal to the economic case.

Selected populations. High-motivation enrollees who completed onboarding are not the population a payer will cover. Reviewers apply real-world dropout, and effect sizes shrink.

Economic claims not grounded in the clinical result. A model that assumes admissions fall by a figure the trial never measured is the fastest way to lose credibility on every other claim in the dossier.

What a dossier that survives looks like

Name the comparator your reviewers will name, not the one that flatters you. If standard care in your target system is a structured nurse programme, that is your comparator. Losing this argument early is cheap; losing it in assessment is not.

Choose endpoints that a clinician would accept as clinically meaningful, and state the minimal important difference in advance. Reviewers notice when a threshold appears after the data.

Run long enough to speak to durability, and plan the evidence-generation phase explicitly. Both NICE's early-use route and the DiGA provisional listing are designed as evidence-generating periods, not as approvals. Treat them that way, with a protocol, a data source and a named analysis, and you convert a conditional recommendation into a permanent one. We covered how the German model is tightening in Germany's DiGA reform.

Report your dropouts honestly and analyse by intention to treat. A modest effect in a realistic population beats a large effect in a filtered one, because the first survives review.

Build the economic model on the clinical result you actually have, with sensitivity analysis around the variables that matter. Say what would change your conclusion. Reviewers trust analyses that acknowledge their own weak points.

Design for de-adoption. If the benefit of your intervention concentrates in the first six months, say so and price accordingly. Products that propose a stopping rule are treated very differently from products that assume indefinite use — a point that runs straight into the forever codes debate in remote monitoring.

Regulatory status is not evidence of value

A recurring confusion in pitch decks is that FDA clearance or a CE mark answers the value question. It does not. A 510(k) demonstrates substantial equivalence to a predicate; it says nothing about cost-effectiveness. Our guides to FDA regulatory pathways and EU MDR and UKCA compliance set out what each authorisation does and does not assert.

The same applies to the reverse argument. Some digital therapeutics avoid device regulation entirely as wellness products, then discover that HTA bodies and payers will not assess an unregulated product at all. Clearance is often the entry ticket to the assessment, not a substitute for it.

A practical sequence

Before your pivotal study: agree the comparator and endpoints with a clinician and a health economist who have sat on the reviewing side. Pre-register.

During: collect resource use and cost data alongside clinical endpoints. It costs almost nothing at this stage.

After: publish, including the neutral or negative subgroup results. Assessors read the literature you did not send them.

At assessment: submit a dossier that answers the standard questions in the standard order — population, comparator, incremental benefit, uncertainty, budget impact, and who bears the cost.

MedTech Compass maps which markets run formal assessments, what evidence they expect and how long each takes, so you can sequence launches around the assessments you are ready for. For evidence and documentation control, DevicePath keeps the underlying records audit-ready.

The companies clearing these reviews are not the ones with the best apps. They are the ones who designed their evidence for the reviewer from the beginning.

Sources

1. Peterson Health Technology Institute: https://phti.org/ 2. PHTI — Digital Diabetes Management Solutions assessment report: https://phti.com/wp-content/uploads/sites/3/2024/03/PHTI-Digital-Diabetes-Mgmt-Assessment-Report-v1.0.pdf 3. Peterson Center on Healthcare — Evolving Remote Monitoring: https://phti.org/evolving-remote-monitoring/ 4. NICE — HealthTech programme manual, early-use assessments: https://www.nice.org.uk/process/pmg48/chapter/early-use-healthtech-guidance-assessments 5. NICE — Evidence standards framework for digital health technologies: https://www.nice.org.uk/about/what-we-do/our-programmes/evidence-standards-framework-for-digital-health-technologies 6. BfArM — DiGA fast-track guide: https://www.bfarm.de/SharedDocs/Downloads/EN/MedicalDevices/DiGA_Guide.html 7. BfArM — DiGA directory: https://diga.bfarm.de/ 8. HAS — Digital medical devices assessment: https://www.has-sante.fr/jcms/p_3118283/en/digital-medical-devices

This article is general information about evidence and assessment practice, not legal, regulatory or financial advice.