The Evidence Behind the Method

Every idea in these programs traces back to peer-reviewed research or major clinical bodies. Where an idea is still a hypothesis, this page says so. This page gathers the load-bearing studies in one place. It is not exhaustive, and it is honest about where evidence is strong and where it is still emerging.

Ultra-processed food drives overeating

In a tightly controlled NIH ward study, the same people ate about 508 more calories a day on an ultra-processed diet than on a matched whole-food diet, and gained weight, without trying. Participants were told to eat as much or as little as they liked. A randomised inpatient crossover trial: twenty adults, two weeks on each diet, every meal supplied and measured. (Hall 2019; randomised crossover trial, 20 adults, two weeks per diet; moderate reliability.) It was a strong design but small and short, so it shows what can happen, not what happens to everyone.

Hall KD, et al. Cell Metabolism, 2019.

Fast carbohydrate increases later hunger

In a small crossover study, twelve teenage boys were hungrier and ate about 81% more at the next meal after a high-glycemic meal than after a low-glycemic one of identical calories. A small study, so read it as a signal rather than a law. But it is the crash and craving loop, measured.

Ludwig DS, et al. Pediatrics, 1999.

A fast-digesting meal was followed by eating more in one small study

In a small crossover study (Ludwig 1999; small crossover study, 12 participants; low reliability), twelve teenage boys ate about 81% more over the next five hours after a high-glycemic lunch than after a low-glycemic one of identical calories. A small study, so read it as a signal rather than a law. It measured hunger and how much the boys ate afterward, which is one piece of the crash-and-craving idea.

Well-formulated keto can improve type 2 diabetes

In the Virta Health trials, a supervised ketogenic approach reduced blood sugar and medication use in a large share of participants at one year, with the group followed out to two and then five years. (Hallberg 2018; Athinarayanan 2019; McKenzie 2024; non-randomised trial funded by the program's company; low to moderate reliability.) At five years the average effect was smaller and about a third of participants remained at target on no medication or metformin alone. Of the 122 people still in the study (262 started), about a third (39) had an A1C below 6.5% while taking no glucose-lowering medication, or only one common first-line medication.

Read these with their limits in view. The trials were funded by Virta Health, the company providing the treatment, and participants chose their own group rather than being randomised. That does not make the findings worthless. They are among the largest datasets in the field. It is why they belong in the promising column rather than the settled one.

In these trials, doctors adjusted each person's medication as their blood sugar changed. Coaching is not that, and it isn't medical care. If you take medication to lower your blood sugar or blood pressure, talk to your prescriber before cutting back on carbs or starches, because your dose may need to change.

Hallberg SJ, et al. Diabetes Therapy, 2018 (1 year) • Athinarayanan SJ, et al. Frontiers in Endocrinology, 2019 (2 years) • McKenzie AL, Athinarayanan SJ, et al. Diabetes Research and Clinical Practice, 2024 (5 years).

How you eat a meal changes its effect

Eating protein and vegetables before starch significantly lowered the blood-sugar and insulin response to the identical meal. Order is a free tool.

In a pilot study of 11 adults with type 2 diabetes (Shukla 2015; pilot study, 11 participants, order not randomised; low reliability), eating protein and vegetables 15 minutes before bread and juice lowered the blood-sugar and insulin rise after the identical meal. The order wasn't randomised and the study only followed two hours, so read it as a signal. In my experience, meal order is one easy thing to try, if you choose to.

Shukla AP, et al. Diabetes Care, 2015.

Habits form through repetition, not willpower

New habits take on average about 66 days to become automatic, and missing a day does not undo progress. In a study of 96 volunteers, a new daily habit took a median of 66 days to feel automatic, with a range of 18 to 254 days, and missing one day did not materially set people back (Lally 2010; observational study; moderate reliability). This is why the programs build one small change at a time, and why a slip is never a fall.

Lally P, et al. European Journal of Social Psychology, 2010.

Plain-language notes on four lab numbers

These notes explain what four common blood tests measure. They are general education, not a reading of your results. Your doctor is the right person to talk to about what your own numbers mean.

Fasting insulin. This measures how much insulin is in your blood after an overnight fast. It is not part of routine diabetes screening in Canada, and Canadian guidelines do not set a normal or target range for it. A lab that runs it prints its own reference range on the report. Some researchers think insulin can rise for years before blood sugar tests change. This is a hypothesis, not settled practice (Crofts 2015, a review proposing a hypothesis; low certainty). Whether the test would be useful in your case is a good question for your doctor.

A1C and its trend. A1C reflects your average blood sugar over about the past three months. Diabetes Canada uses these bands for adults: 6.0% to 6.4% is prediabetes, and 6.5% or higher is in the diabetes range, which a doctor confirms with a repeat test (Diabetes Canada 2018 Clinical Practice Guidelines, chapter 3; clinical practice guideline; high reliability). In my experience, the trend over several years can tell you more than a single result. Your doctor can tell you what yours means.

Triglycerides. Triglycerides are a type of fat carried in your blood. Canadian cholesterol guidelines treat a level above 1.5 mmol/L as the point where your doctor looks at non-HDL cholesterol or apoB instead of LDL cholesterol, because LDL becomes a less reliable guide above that level (Canadian Cardiovascular Society 2021 dyslipidemia guidelines; clinical practice guideline; high reliability).

HDL. HDL is often called "good" cholesterol. Canadian guidelines use it to help estimate heart risk, for example in the Framingham Risk Score, where a lower HDL raises the estimated risk (Canadian Cardiovascular Society 2021). You may see the triglyceride-to-HDL ratio discussed online. It has no agreed cutoff in Canadian guidelines.

Where the evidence is still emerging

Some mechanisms, such as the finer details of mitochondrial overload, are active models rather than settled facts, and carnivore-style elimination is largely anecdotal. We say so plainly, and we do not oversell.