The T2D Reversal Evidence

The strongest human evidence in this whole topic is not in longevity or athletics — it is in type 2 diabetes, and it comes from supervised clinical programs rather than do-it-yourself keto. This page walks through what those programs actually achieved at one and two years, what the word "reversal" means inside the data, and why the results are not a promise to someone adjusting their own medication at home.

What the Program Actually Did

The headline study was an open-label, non-randomized, controlled evaluation of a continuous remote care model for adults with type 2 diabetes (Hallberg et al., 2018). Participants got something closer to a clinic than a diet: nutrition coaching, connected monitoring (scale, blood pressure, glucose), ketone tracking, and — the part headlines skip — physician-managed medication adjustment. The comparison group received usual care, and the design was not randomized, which matters for how much weight the results can carry.

At one year, the numbers were remarkable by the standards of diet studies: HbA1c from about 7.6% to about 6.3%; among the 204 participants with HbA1c reported at 1 year, 60.3% met the program's reversal criterion, and average weight loss was around 12%. Participants were asked to eat few enough carbohydrates to maintain nutritional ketosis while keeping protein adequate, and to stay in touch with the care team between visits — the diet and the monitoring were packaged together by design. These are averages from a program, not outcomes anyone can self-administer, which is the theme this page keeps returning to. The underlying physiology of the diet itself is covered by the ketosis physiology page.

One design limit deserves its own sentence: this was not a randomized trial. Participants chose the program, and the comparison group received usual care without being randomly assigned — so motivation, engagement, and baseline differences can all inflate the contrast. The results are real; the causal credit is shared with everything else that distinguished these participants from the people they were compared against.

What "Reversal" Means Inside the Data

"Reversal" in this literature is a defined outcome, not a vibe: an HbA1c below 6.5% while off glucose-lowering medication except metformin. Metformin staying on the table is a detail worth noticing — the result is real, and it is not the same thing as medication-free. Formal remission definitions used elsewhere (for example, an HbA1c under 6.5% maintained for at least three months with no diabetes medication at all) are stricter than the study's criteria, so comparisons across trials need care.

The Two-Year Picture

One-year diet results are notoriously generous. The two-year follow-up (Athinarayanan et al., 2019) is therefore the more honest document: improvements in HbA1c, weight, and medication use were largely sustained across two years, but the paper itself shows some regression and continuing attrition. Read plainly, the arc is "substantial early gains, partially held with support, eroding without it" — a pattern that matches almost every long-running weight and metabolic intervention ever studied.

That attrition is not a scandal; it is the data. Anyone selling a version of this story where everyone sails through two years with no dropout is selling a different study than the one that exists. The glucose numbers themselves — fasting glucose, HbA1c — are interpreted through the wider diabetes literature; the Glucose 101 topic owns that context.

Medication Adjustment Is the Protocol

Here is the sentence that separates the evidence from the marketing: for participants on insulin or sulfonylureas, medication doses were reduced as part of the intervention, under physician supervision, precisely because carbohydrate restriction lowers blood sugar and the old doses could otherwise cause hypoglycemia. The de-prescribing was not a side effect of the diet working — it was an active, planned clinical task.

That task cannot be improvised. A person on insulin who starts a very-low-carb diet without an adjustment plan is running an uncontrolled experiment on their own blood sugar. For anyone in that position, the risk math lives on the safety page in this series, and the adjustment plan belongs to their prescriber.

The Evidence at a Glance

Read the table as one evolving dataset rather than four independent confirmations: two of the rows are follow-ups of the same program, and the randomized row is a different dietary route entirely. The pattern that survives all four is the modest one — glycemia improves under several forms of serious dietary change, and the maintenance problem outlasts the enthusiasm of the first months.

The comparison honesty matters: DiRECT achieved a similar headline number through total diet replacement, a randomized design, and primary-care delivery — no ketosis required. The meta-analytic view (Goldenberg et al., BMJ, 2021) is blunter still: low-carb approaches improve glycemia over months, the trials are mostly small and short, and the evidence certainty is low. Type 2 diabetes responds to several dietary strategies; this one is among the best documented, not the exclusive property of ketones.

Questions, Answered Briefly

• ⏱️ How fast do numbers move? Meaningful HbA1c changes are usually described over months, not days; the medication adjustment, however, starts immediately — which is why supervision starts immediately too.
• 🔁 Is this a permanent reversal? "Reversal" here is a state maintained while the diet and the support hold. The two-year data show it eroding for some participants over time; it is a managed state, not a finished one.
• 🏠 Does it work without the program? That version has not been studied at this scale. The published results describe continuous care, and self-directed low-carb attempts elsewhere have not matched them in the trials that exist.
• 💊 Can I cut my own medication if my readings improve? No — this is prescriber territory. The adjustment sequence is part of the treatment, and getting it wrong has a named failure mode: hypoglycemia.
• ⚖️ Is it better than other diets for T2D? The honest summary is "competitive, not crowned". Head-to-head evidence is heterogeneous, and several dietary patterns produce meaningful improvements.
• 📉 What does the average hide? Range. Some participants met the full reversal criteria, others saw partial improvement, and some dropped out before the one-year mark — the headline percentages describe a distribution, not a personal forecast.

Where This Evidence Stops

Everything here is about type 2 diabetes over one to two years, under care. It says nothing yet about hearts, strokes, or lifespan — and nothing at all about what the same diet does to a person without diabetes, including the subgroup whose LDL cholesterol rises sharply on carbohydrate restriction. That last piece is the next page's territory, and the outcome claims are the one after that.

⚠️ Medical caution: Educational content only — not medical advice. Ketogenic diets require clinician supervision in type 2 diabetes (medication adjustment), and are unsuitable for some people, including those on SGLT2 inhibitors, during pregnancy, or with kidney disease or an eating-disorder history. Discuss any major diet change with a qualified professional.

Related Topics