Health guidance can begin with a simple question: what does my result mean, and what next step makes sense for me? For a service developer, this introduces a second task. Initial guidance needs to become a coherent journey, with a starting point, an agreed goal and opportunities to revisit progress or difficulties. This dossier examines research as a foundation for that development challenge.
What matters
- A development strategy should define the population, intended change and support process together.
- Automation needs a specific scope and an appropriate comparison of quality.
- A testable service connects progress, feedback and clearly assigned responsibility for next steps.
01
Prevention needs a process
The Diabetes Prevention Program randomised 3,234 adults with raised blood glucose. Over an average 2.8 years, intensive lifestyle support reduced diabetes incidence by a relative 58% against placebo. It combined dietary and activity goals with individual support. This demonstrated an effect on new diagnoses, applying to the complete programme and the population at risk studied. [1]
Our editorial interpretation for service development is to start with a clearly defined change that the service aims to support. This means specifying the population, the intended journey and the service to be delivered. Accessible information can provide an entry point. A prevention service should also explain how it intends to support implementation and how meaningful progress would be recognised.
02
Digital delivery is a design challenge
A meta-analysis of 28 randomised prediabetes trials found an additional mean weight loss of 1.74 kilograms against non-digital comparators; the 95% confidence interval was 1.11–2.37 kilograms. Evidence certainty was moderate. Formats varied, including fully digital programmes and combinations with personal support. The pooled result therefore does not isolate an automation effect. [2]
This does not point a new service towards one particular interface. The design decision is which tasks digital delivery should reliably support: understanding the starting situation, agreeing an action, providing a reminder or reviewing progress. Making these distinctions turns a development idea into something that can be tested. It also provides a basis for assessing each feature against its intended contribution.
03
Automation can be tested directly
A JAMA trial compared full automation with human online coaching in 368 adults with prediabetes and overweight/obesity. At twelve months, 31.7% versus 31.9% met the composite weight, activity and glycaemic endpoint. The one-sided 95% lower bound, −8.2 percentage points, met the predefined −15-point non-inferiority margin. This establishes non-inferiority under that criterion, neither equivalence nor reduced diabetes incidence. The AI used reinforcement learning, not a language model. [3]
This sharpens the development question for a business: what specific task should an automated process perform, and what comparison would meaningfully test it? Such an evaluation requires a firm description of the service. It also requires deciding in advance what difference in outcomes would remain acceptable. Automation can then be judged against an explicit standard of quality.
04
Support must work in everyday life
A randomised trial in 1,062 people at increased cardiovascular risk combined game elements with financial incentives. Activity rose by an average 868 daily steps against a control group receiving step goals and feedback across twelve intervention months. Six months afterwards, the difference remained 576 steps. The outcome was movement, not a reduction in cardiovascular events. [4]
For an editorial consideration of continuous support, the period after someone joins deserves particular attention. A service should be able to describe how it intends to handle interruptions, changing goals and missing feedback. The end of a programme also needs its own design: what guidance will remain available, and how might someone return at a later stage?
05
Feedback gives measurement a practical purpose
A UK randomised trial in 622 people with poorly controlled hypertension combined home measurements, digital feedback and clinician-directed medication adjustments. At twelve months, systolic blood pressure was 3.4 mmHg lower than with usual care; the 95% confidence interval was −6.1 to −0.8. This result concerns a hybrid care model. Participation required internet access. [5]
Our interpretation is that every measurement should have a clear purpose in the service design. Who receives the feedback, who interprets it, and who has authority to act on it? These responsibilities belong in the initial plan. For investors, this would reveal the operation behind the interface: the service itself, its handovers and the conditions required to deliver it consistently.
PERSPECTIVE FOR HUMAN NEXUS
A development perspective for Human Nexus — editorial interpretation
For Human Nexus, we see a development direction in connecting digital health guidance, personalised prevention and operationally defined services. The vision is a support process that understands individual starting points, establishes clear goals and revisits progress. Each service could share common processes while adapting content, intensity and responsibilities to its particular purpose.
For investors, this perspective should be assessed through concrete development milestones: a well-founded target population, a coherent service model, practical testing and subsequent outcome evaluation. Use, health improvement and economic viability should be treated as separate questions. This would allow a stepwise assessment of which form of support is sustainable and under what conditions it could expand. The studies discussed offer ideas for this work; they do not demonstrate the effectiveness of a Human Nexus service.
Original sources
Every source links to the original scientific publication. Source dates refer to the research; the editorial update is shown at the start of this article.
- New England Journal of Medicine ·
Reduction in the Incidence of Type 2 Diabetes with Lifestyle Intervention or Metformin
Diabetes Prevention Program Research Group · RCT · 3,234
DOI: 10.1056/NEJMoa012512Funding & context
Public and industry support; one author disclosed manufacturer shareholdings.
Clinical endpoint: diabetes incidence.
- Journal of Diabetes Science and Technology ·
The Effectiveness of Digital Health Lifestyle Interventions on Weight Loss in People With Prediabetes: A Systematic Review, Meta-Analysis, and Meta-Regression
Tanja Fredensborg Holm et al. · Systematic meta-analysis · 28 RCTs
DOI: 10.1177/19322968241292646Funding & context
Unfunded; three authors are pharmaceutical company employees and shareholders.
Online 2024; issue 2026.
- JAMA ·
An AI-Powered Lifestyle Intervention vs Human Coaching in the Diabetes Prevention Program: A Randomized Clinical Trial
Mathioudakis et al. · Non-inferiority RCT · 368
DOI: 10.1001/jama.2025.19563Funding & context
NIH-funded; providers paid, without outcome analysis; external author fees disclosed.
Unmasked.
- Circulation ·
Effect of Gamification, Financial Incentives, or Both to Increase Physical Activity Among Patients at High Risk of Cardiovascular Events: The BE ACTIVE Randomized Controlled Trial
Fanaroff et al. · RCT · 1,062
DOI: 10.1161/CIRCULATIONAHA.124.069531Funding & context
NIH-funded; one author co-owns a behavioural consulting company.
Wearable measurement.
- BMJ ·
Home and Online Management and Evaluation of Blood Pressure (HOME BP) using a digital intervention in poorly controlled hypertension: randomised controlled trial
Richard J. McManus et al. · RCT · 622
DOI: 10.1136/bmj.m4858Funding & context
NIHR-funded; discounted devices and manufacturer collaboration disclosed.
Baseline characteristics corrected in 2022: DOI 10.1136/bmj.m2216.