The Spike Hypothesis: Why Normal Blood Sugar May Not Be Normal Enough

#1

Two Romanian cardiologists argue that continuous glucose monitors (CGM), the stick-on sensors built for diabetics, could work as precision instruments for cardiovascular prevention in people who are entirely healthy by conventional testing. Their case rests on a mechanistic chain: glucose spikes inside the so-called normal range injure blood vessels, those spikes vary enormously between individuals eating identical meals, and only a CGM can see them. They then propose stacking CGM-guided diet changes, intermittent and prolonged fasting, and off-label low-dose metformin or acarbose into a four-phase protocol. The paper is honest about what it is: a hypothesis, not a validated strategy.

For thirty years, cardiovascular prevention has run on averages. Population guidelines tell everyone to eat the same foods, and a fasting glucose or HbA1c declares you metabolically fine or not. This review argues both assumptions are leaking.

The first leak is individual variability. A landmark Weizmann Institute study of 800 adults showed that identical meals produce wildly different glucose responses in different people, driven by gut microbiome composition, meal timing, sleep and prior activity. Population-average glycaemic index tables cannot predict your response to a banana. A continuous glucose monitor can measure it.

The second leak is that the damage may sit in the swings rather than the average. Acute excursions, even within the non-diabetic band of 8 to 10 mmol/L (144 to 180 mg/dL), generate reactive oxygen species, suppress nitric oxide availability and switch on inflammatory adhesion molecules in the vessel lining within 60 to 90 minutes of a meal. Laboratory work suggests oscillating glucose kills endothelial cells more efficiently than steady mild elevation at the same mean. HbA1c, being an average, is blind to this. So is a single fasting draw, whether you read it as 5.4 mmol/L or 97 mg/dL.

From there the authors build outward. Glucose sits upstream of the four nutrient-sensing systems that dominate ageing biology: insulin and IGF-1, mTORC1, AMPK and the sirtuins. Anything that lowers glycaemic load, whether time-restricted eating, prolonged fasting, metformin or acarbose, pushes these pathways in the direction associated with longer life in model organisms. CGM becomes the shared instrument that lets a person titrate and verify each intervention on themselves.

The proposal is a four-phase model: two weeks of baseline CGM phenotyping, then dietary personalisation, then supervised fasting, then, only for higher-risk individuals under a physician, off-label metformin or acarbose.

Actionable Insights

Eat protein and vegetables before starch. Across CGM studies this cuts the post-meal glucose peak by 20 to 40 percent. On a typical 2.0 mmol/L rise that is a 0.4 to 0.8 mmol/L reduction, which in statistical terms is a moderate to large effect (roughly Cohen’s d 0.6 to 1.0). Cohen’s d simply expresses the change as a fraction of the normal spread between people: 0.2 is small, 0.5 moderate, 0.8 large.

Walk for 10 to 20 minutes within half an hour of eating. This reduces the same peak by 15 to 30 percent, a moderate effect of similar size to a drug, at zero cost and zero risk.

Time-restricted eating lowers systolic blood pressure by 4 to 8 mmHg in short trials, a small-to-moderate effect (d roughly 0.3 to 0.5). Epidemiologically, each 5 mmHg of systolic reduction associates with about 10 percent fewer cardiovascular events.

Acarbose is another option, and extended mouse lifespan by 22 percent in males, 5 percent in females. But, 74 percent of acarbose users get flatulence versus 29 percent on placebo (though avoiding a wheat-based diet seems to eliminate most of the flatulance risk). See: Acarbose - Details On Another Top Anti-Aging Drug

SGLT2 inhibitors are another possible option (in males at least), and force renal glucose excretion of roughly 60 to 100 g per day, about 240 to 400 kcal, producing a caloric restriction mimetic state independent of appetite. Fasting beta-hydroxybutyrate roughly doubles, typically from about 0.1 mmol/L to 0.3 to 0.6 mmol/L (1 to 6 mg/dL), the same ketone signal the review spends a full subsection praising in prolonged fasting. Canagliflozin, in the NIA Interventions Testing Program, started at 7 months in mice, increased lifespan by 14 percent in males. See: Canagliflozin - Another Top Longevity Drug

A CGM is a good teaching tool for two to four weeks. It is not yet a validated risk test in healthy people.

Context and Source

#2

Biomarker Data (Effect Size Extraction, Dual Units)

Framing note before the numbers: this paper generated no data. Every figure below is extracted from studies the authors cite. Where the source did not report a standardised effect size, I have derived an approximate one and labelled it as derived. Derived estimates carry more uncertainty than reported ones.

A short glossary. Hazard ratio (HR) and relative risk are multipliers on your baseline risk: HR 0.51 means roughly half the risk. Relative reductions sound larger than they are, so absolute risk reduction and number needed to treat (NNT, how many people must take a drug for one to benefit) matter more. Cohen’s d expresses a change as a fraction of the normal person-to-person spread.

The paper’s proposed CGM targets, in both standards

Table 1 of the review is the operational core of the framework. Here it is with US conversions added, which the paper does not supply.

Metric Definition (SI) Definition (US) Proposed target Status
Time in Range percent of readings 3.9 to 7.8 mmol/L 70 to 140 mg/dL above 97 percent Extrapolated
Time Above Range percent above 7.8 mmol/L above 140 mg/dL below 3 to 5 percent Extrapolated
Coefficient of Variation SD divided by mean, times 100 identical, unitless below 20 percent Author-proposed
MAGE mean of excursions exceeding 1 SD identical concept minimise No validated value
Delta-PPG peak minus pre-meal within 2 h same below 1.7 to 2.0 mmol/L Author-proposed
below 31 to 36 mg/dL

Two critical observations on this table. First, the international CGM consensus (Battelino et al.) defines time in range as 3.9 to 10.0 mmol/L, or 70 to 180 mg/dL. The paper’s 3.9 to 7.8 mmol/L (70 to 140 mg/dL) is a substantially tightened band that is not the consensus range, yet the evidence column says “extrapolated from international consensus TIR targets.” Both the range and the 97 percent threshold have been modified. [Confidence: High]

Second, the delta-PPG target of 1.7 to 2.0 mmol/L (31 to 36 mg/dL) must be read against device error. Current sensors (Abbott FreeStyle Libre 3, Dexcom G7, Medtronic Guardian 4) have a mean absolute relative difference of 7 to 9 percent. At a glucose of 7.8 mmol/L (140 mg/dL) that is roughly plus or minus 0.5 to 0.7 mmol/L, or 10 to 13 mg/dL, on a single reading. A delta-PPG target of 36 mg/dL is therefore only about three times the device’s own noise. Chasing it precisely is not defensible. [Confidence: High]

Human glycaemic effects, both standards

Intervention Effect on postprandial glucose SI US Derived Cohen’s d
Meal sequencing (protein and fibre first) 20 to 40 percent lower peak 0.4 to 0.8 mmol/L 7 to 14 mg/dL 0.6 to 1.0
Post-meal walk, 10 to 20 min 15 to 30 percent lower 0.3 to 0.6 mmol/L 5 to 11 mg/dL 0.4 to 0.8
Fibre-rich carbohydrate swap MAGE reduced 25 to 35 percent varies varies not estimable
Metformin 500 to 1000 mg/day delta-PPG reduced 0.5 to 1.0 mmol/L 9 to 18 mg/dL 0.5 to 1.0
Acarbose 50 to 100 mg three times daily excursion reduced 1.5 to 2.5 mmol/L 27 to 45 mg/dL 1.5 to 2.5
Moderate aerobic exercise, acute glucose decline during session 1 to 4 mmol/L 18 to 72 mg/dL not estimable
HIIT, transient sprint spike glucose rise 5 to 12 mmol/L 90 to 216 mg/dL not applicable
Prolonged fasting, ketoadaptation plateau stable low glucose 3.5 to 4.5 mmol/L 63 to 81 mg/dL not applicable

Bottom Line

The dietary and exercise components are free, safe, and supported by moderate-to-large measurable effects on postprandial glucose: roughly 7 to 14 mg/dL (0.4 to 0.8 mmol/L) from meal sequencing, 5 to 11 mg/dL (0.3 to 0.6 mmol/L) from a post-meal walk. Both are defensible without owning a CGM at all. The CGM itself is a reasonable two-to-four-week behavioural teaching tool with a real risk of promoting hypervigilance and no validated targets in healthy people under any national standard. The pharmacological layer has proven harms with tight numbers needed to harm and unproven benefits with confidence intervals wide enough to include nothing. The paper’s own conclusion, that this is a research agenda and not a protocol.