How We Measure What Actually Works

Plenty of vendors in this category claim behavior change reduces GLP-1 dependence. Few back that up with more than a testimonial. Levels ran an IRB-approved study for three years, the largest independent dataset connecting real-world glucose, food, and lab data to metabolic outcomes that we're aware of: 62,236 participants and 1.15 billion glucose readings. It gave us a detailed look at weight and metabolic trajectories in people without diabetes, one that applies directly to how GLP-1 prescribing and coverage decisions for weight-loss-related prescriptions get made.

Here's what we found, and where the real caveats are.

What we found

At one year:

• 22% of participants who started in the obesity category, and 33% who started overweight with an additional qualifying risk factor, no longer met FDA indication criteria for GLP-1 therapy. Published natural-history data on untreated populations put the comparable rate at roughly 2% and 5%.¹
• 43% of participants who started with prediabetes reached a normal A1C.
• The cohort showed a 67% lower rate of progression from prediabetes to diabetes than published natural-history data would predict for an untreated population.²

In plain terms: a meaningful share of people moved out of GLP-1-eligible territory, or off the path toward it, at a rate several times higher than we'd expect to see with no intervention at all.

Why we think this happened

Nobody in the study was on a fixed diet or a prescribed protocol. They had continuous visibility into how their glucose responded to specific meals, sleep, stress, and activity, plus guidance on what to do with that information: which patterns were worth changing, and how. That combination, seeing the effect and getting help acting on it, is different from willpower-based dieting, where the feedback loop between a choice and its effect can take weeks to show up, if it shows up at all. Continuous data plus guidance collapses that loop to hours. Our working explanation for the results above is that this tighter loop, information paired with a clear next step, is what let people find and stick with changes that actually worked for their own biology, rather than a generic plan applied to everyone.

We can't fully isolate that mechanism from everything else happening in participants' lives over 3.5 years, which is exactly why the caveats below matter.

What this study doesn't prove

This is an observational study, not a randomized controlled trial. No matched control arm ran alongside our participants. The natural-history figures above come from separately published research on different populations, not from a comparison group inside our own study, so the gap between our numbers and theirs is a signal worth taking seriously, not a proven causal effect.

Our population is also self-selected: people who chose to use a CGM and coaching product. That's a real limitation of any dataset built this way, ours included.

Why we're publishing this level of detail

Most vendors in this category report outcomes from internal case studies or self-reported measures, with no stated methodology attached. In the same spirit of scientific rigor we expect from any vendor's numbers, we'd rather you can check ours than take our word for them. If you want the full study design and statistical detail behind any figure here, we'll walk through it directly.

These results matter for GLP-1 strategy even before anyone starts the medication, but they matter after, too. The same continuous feedback and guidance that helped people avoid needing a GLP-1 in this study is, by the same mechanism, what we'd expect to protect the metabolic benefit for someone already taking one throughout therapy and after they taper off.