First, what this article is not about. If you are here because you feel foggy and flat at three in the afternoon, that is a different subject and we have written it up separately: the post-lunch dip in healthy people, which is largely circadian rather than dietary, and where the measured outcome in most trials is glucose rather than cognition.
This article is about the other end of the same axis — sustained insulin resistance and type 2 diabetes, and what they do to the brain over decades. Type 2 diabetes is one of the Lancet Commission's modifiable dementia risk factors. We have cited that list repeatedly and never written this. Different timescale, different evidence base, different conclusions.
The association, and its actual size
This is one of the better-quantified relationships in the field.
A pooled analysis of 14 studies covering 2,310,330 people and 102,174 dementia cases found type 2 diabetes associated with roughly a 60% higher risk of dementia of any kind — relative risk 1.62 (95% CI 1.45 to 1.80) in women and 1.58 (1.38 to 1.81) in men. Split by subtype, the pattern is informative: vascular dementia 2.34 (1.86 to 2.94) in women and 1.73 (1.61 to 1.85) in men; non-vascular dementia 1.53 and 1.49. Women carried a 19% greater excess risk of vascular dementia than men (ratio of relative risks 1.19, 1.08 to 1.30, p<0.001).
Two honest notes. The risk is roughly doubled for the vascular subtype and clearly smaller for the rest, which tells you where most of the mechanism probably sits. And estimates vary with the comparison group and the adjustments — a UK Biobank analysis of 447,931 people reported a hazard ratio of 2.85 for all-cause dementia against people with no diabetes at all. Anyone quoting a single precise multiple for this is overstating the precision of the literature.
It is not only diabetes. In the Adult Changes in Thought cohort, 2,067 people without dementia were followed for a median 6.8 years with 35,264 glucose measurements. Among the 1,835 participants without diabetes, an average glucose of 6.4 mmol/L compared with 5.5 mmol/L carried an adjusted hazard ratio for dementia of 1.18 (95% CI 1.04 to 1.33, p=0.01). Among those with diabetes, 10.5 versus 8.9 mmol/L gave 1.40 (1.12 to 1.76). A graded relationship with no obvious threshold — which is the strongest argument that this is about glucose and insulin handling rather than a diagnostic label.
The mechanism, in the order the evidence supports it
Vascular damage first, because the subtype data points there. Diabetes accelerates atherosclerosis in large vessels and damages small ones, producing the same arteriolar disease, white matter lesions and reduced perfusion that hypertension produces. These two rarely travel alone, and the clustering is why the Commission's factors overlap rather than adding up.
Brain insulin signalling second. Insulin receptors are densely expressed in the hippocampus and cortex, and insulin has signalling roles there beyond glucose uptake, including in synaptic plasticity. The hypothesis is that chronic peripheral hyperinsulinaemia produces central insulin resistance, degrading those functions. The mechanism is plausible and reasonably well characterised in animals. It is much less well established that it is a major driver in humans.
Then the supporting cast: advanced glycation end products, chronic low-grade inflammation, and competition for insulin-degrading enzyme, which clears both insulin and amyloid-beta. Each is real biochemistry. None has been shown to be the rate-limiting step in a person. We have written about how far the inflammation story can honestly be pushed.
One more finding worth knowing: in that UK Biobank analysis, diabetes onset before age 55 trended toward higher vascular dementia risk than later onset, and people with diabetic complications had roughly double the risk of those without. Duration and severity appear to matter, not just presence.
"Type 3 diabetes": useful frame or overreach?
The phrase was proposed in 2008 to capture the idea that Alzheimer's disease is fundamentally a brain-specific insulin resistance disorder. It is contested, and we should say so plainly rather than borrowing its authority.
What it gets right: it draws attention to a real metabolic dimension of Alzheimer's that was underweighted, and it generated testable predictions.
What is wrong with it as a claim: it is a hypothesis, not a classification. No diagnostic system recognises type 3 diabetes — not the WHO's ICD, not the American Diabetes Association's classification of diabetes. Alzheimer's disease is defined by amyloid and tau pathology, and most people with it do not have diabetes while most people with type 2 diabetes do not develop Alzheimer's. An association of the size described above is not the same thing as identity.
And the frame made a prediction that has been tested. If Alzheimer's is brain insulin resistance, delivering insulin to the brain should help. Intranasal insulin does that, bypassing systemic circulation. A 2026 meta-analysis of five randomised trials and 540 participants with mild cognitive impairment or mild-to-moderate Alzheimer's found no significant benefit on any cognitive or functional outcome — ADAS-Cog13 mean difference −1.09 (95% CI −4.89 to 2.71), ADCS-ADL 0.06 (−0.33 to 0.45), and nulls on CDR-SB, delayed recall, and CSF amyloid-beta 42, total tau and phosphorylated tau. Gastrointestinal side effects were more common with insulin (risk ratio 1.57, 1.15 to 2.14). The authors concluded current evidence does not support clinical use.
Verdict: useful as a pointer to a real mechanism, an overreach as a description of what Alzheimer's disease is. Treat anyone using it as a settled fact with caution, particularly if they are selling something.
The glucose-lowering trials, and the result people do not expect
Here is the part that should change how you read this topic. If high glucose damages the brain, lowering it should protect the brain. That has been tested directly, and it did not work.
ACCORD-MIND enrolled 2,977 people aged 55 to 80 with type 2 diabetes, HbA1c above 7.5% and high cardiovascular risk, randomised to intensive glycaemic control targeting HbA1c below 6.0% versus a standard target of 7.0 to 7.9%. The primary cognitive outcome was the Digit Symbol Substitution Test at 40 months, in 1,378 intensive versus 1,416 standard participants.
There was no significant difference: mean difference 0.32 (95% CI −0.28 to 0.91, p=0.2997).
The imaging outcome did favour intensive treatment. In the MRI subset, total brain volume was greater in the intensive arm by 4.62 cm³ (95% CI 2.0 to 7.3, p=0.0007) — a real, statistically robust difference. And it made no difference to how people actually performed. That divergence between a structural marker and a functional outcome is the single most useful thing in this literature, and it is why surrogate endpoints deserve suspicion.
The context is worse than a null. The intensive arm of the parent ACCORD trial was stopped early in February 2008 because of increased mortality. The MIND authors' own interpretation was explicit: combined with the non-significant effects on other outcomes and the excess deaths, their findings do not support intensive glucose lowering to reduce diabetes' effects on the brain.
Lifestyle intervention fared no better on cognition. Look AHEAD randomised 3,751 overweight and obese adults with type 2 diabetes to ten years of intensive lifestyle intervention for weight loss versus diabetes support and education, then tested cognition 10 to 13 years after enrolment. No significant difference overall (p=0.10) or in any domain (all p>0.10), with cognitive scores unrelated to changes in weight or fitness — and some signal of modest harm in participants with higher BMI or prior cardiovascular disease.
None of this means glucose control is pointless. It prevents retinopathy, nephropathy and neuropathy, and those are reasons enough. It means the specific claim "controlling your blood sugar protects your brain" is not currently supported by randomised evidence, and the observational association has not translated. Reasonable readings include that the damage accrues earlier than these trials intervened, or that the association is partly confounded. We do not know which.
The honest position on supplements
No Proco product treats, prevents or manages insulin resistance, prediabetes or type 2 diabetes, and we do not sell anything that does. We are not going to imply otherwise, and the section above is the reason it would be indefensible: the most powerful glucose-lowering interventions ever tested — intensive drug therapy and a decade of supervised lifestyle change — did not improve cognitive outcomes. A capsule is not going to outperform them.
The ingredients marketed for this are chromium, cinnamon, berberine and alpha-lipoic acid. None is in our range. The pattern to watch for is the same one we have documented with curcumin absorption claims: a real mechanism, small trials with surrogate endpoints, and a conclusion that outruns both. Our own comparison of exercise against supplements is the relevant honest benchmark.
If you are here because your thinking feels slower, the useful move is not a supplement. An HbA1c is one blood test, and it belongs alongside the short panel worth asking for. Knowing when fog needs a doctor rather than a purchase matters more than any ingredient, and the wider list of causes is broader than metabolism. For where diabetes sits among the other modifiable factors, the 45% figure and what it actually means is here.
This article is not medical advice and has not been evaluated by the FDA. Nothing here is intended to diagnose, treat, cure or prevent any disease, or to start, stop or change a prescribed medication. Glycaemic targets are individual and belong with a clinician who knows your history.
Sources: Chatterjee S, et al. Diabetes Care. 2016;39:300-307 · Crane PK, et al. N Engl J Med. 2013;369:540-548 · Launer LJ, et al. Lancet Neurol. 2011;10:969-977 (ACCORD-MIND) · Espeland MA, et al. J Am Geriatr Soc. 2017 (Look AHEAD) · Intranasal insulin for MCI and Alzheimer's disease: systematic review and meta-analysis. Rev Neurol. 2026 · Sex-specific associations between diabetes and dementia. Biol Sex Differ. 2023 (UK Biobank) · Livingston G, et al. Lancet. 2024;404:572-628