Clinical evidence

Built on science, not promises

Every claim we make is backed by peer-reviewed research conducted at independent academic institutions. Here's what the evidence shows — and what it doesn't.

Key to Health was founded by physicians and scientists. From the start, we committed to proving our program works through the same standards used to evaluate pharmaceuticals: randomized controlled trials with independent oversight, pre-registered analyses, and peer-reviewed publication.

Our evidence base now spans four published or submitted studies, from a controlled clinical trial to real-world data on more than 11,000 adults over two years.

Published & submitted research

Published

Randomized controlled trial

Falkenhain K, Locke SR, Lowe DA, et al. Obesity. 2021;29(10):1606–1614.

A pragmatic, parallel-arm RCT comparing Key to Health (Mediterranean-style low-carbohydrate diet with breath acetone biofeedback) against WW (calorie-restricted low-fat diet app) — an evidence-based active comparator, not placebo. Conducted at the University of British Columbia with independent biostatisticians. Funded by the Canadian Institutes of Health Research.

155
participants randomized
−5.6 kg
weight loss at 12 weeks (KTH group)
2.2×
greater weight loss vs. comparator at 12 wks
p < 0.001
between-group difference

The published study also reported improvements in HbA1c and liver enzymes in the Key to Health group. No concerning changes in cholesterol or other lipids.

Read on PubMed
Published

1-year follow-up of the RCT

Falkenhain K, Locke SR, Lowe DA, et al. Obesity Science & Practice. 2025;11(2):e70106.

Pre-specified secondary analysis of the same RCT, evaluating weight loss durability at 48 weeks — well beyond the 12-week primary intervention period. Uses a linear mixed-effects model incorporating all available daily weight measurements.

48 wks
follow-up duration
~10%
total body weight loss (KTH group)
Sustained
superiority vs. comparator at 48 wks

Weight loss was maintained through 48 weeks without additional intervention, demonstrating durability beyond the active trial period.

Read on Wiley
Published

Mediation analysis: why biofeedback works

Falkenhain K, Locke SR, Lowe DA, et al. JMIR mHealth and uHealth. 2022;10(3):e33940.

A secondary analysis of the RCT intervention arm (n=75) examining which user behaviors predicted weight loss. Used correlation, regression, and causal mediation analyses to trace the pathway from biofeedback to outcomes.

Biofeedback engagement
User breathes into device, sees real-time metabolic response to food choices
Dietary adherence
Feedback reinforces healthy choices, creating a self-correcting behavior loop
Weight loss
Sustained outcomes driven by changed behavior, not novelty or placebo

The biofeedback didn't magically cause weight loss — it changed behavior, and changed behavior caused weight loss. This matters because most digital health interventions can show that they work but not why. Mechanism evidence gives confidence that results are reproducible.

Read on JMIR

Real-world evidence: 11,365 adults over 2 years

Submitted to Obesity and ObesityWeek Journal Symposium, 2026.

A pre-registered, observational cohort study of 11,365 adults with BMI ≥25 who used Key to Health in real-world conditions. Analyzed using mixed-model repeated measures (MMRM) with a pre-specified statistical analysis plan. Intention-to-treat analysis — no cherry-picking completers.

11,365
adults enrolled
2 years
of follow-up data
Pre-registered
analysis plan & endpoints

This study tests whether the RCT findings hold at scale in real-world conditions — 73× the enrollment of the original trial, with 2-year follow-up. Specific outcomes will be published here upon peer-reviewed publication.

What the evidence shows — and its limits

We believe presenting evidence honestly is more credible than overclaiming. Here's a fair assessment of our research.

Strengths

  • Randomized, controlled design with an evidence-based active comparator (not waitlist or placebo)
  • Independent academic investigators at UBC with third-party biostatisticians
  • Publicly funded (Canadian Institutes of Health Research)
  • Pre-registered trial protocol and statistical analysis plans
  • Real-world evidence at 73× the scale of the RCT with 2-year follow-up
  • ITT analysis in both RCT and RWE — results include all dropouts
  • Published mechanism evidence explaining why biofeedback drives outcomes

Limitations

  • RCT sample size (N=155) is moderate — not a large multi-site trial
  • RCT population was predominantly female (71%) and based in California
  • RWE study is observational, not randomized — no control group
  • RWE study is currently under peer review (not yet published)
  • No head-to-head comparison with GLP-1 medications or bariatric surgery
  • Long-term outcomes beyond 2 years are not yet available
FDA update — March 2026

510(k) premarket notification submitted

We have submitted a 510(k) premarket notification to the FDA for the Key to Health breath acetone sensor, classified under product code JIN (Class I). Our predicate device is the Abbott breath acetone analyzer.

510(k) clearance means the FDA finds a device to be substantially equivalent to a legally marketed predicate device. It does not constitute FDA approval or endorsement.

Current status: Submitted March 2026, under review. This page will be updated when a determination is received.

Interested in the evidence?

We're happy to walk through our data with employers, health plans, and clinical partners.

Get in touch