Metabolic Markers and the CGM Question

Metabolic dysfunction is the quiet engine underneath most of what shortens healthspan: cardiovascular disease, kidney disease, several cancers, and dementia risk all track it.

The good news is that it announces itself years in advance, through cheap and standardized tests. The complicated news is that a newer, more expensive, far more engaging tool has arrived, and the evidence for it in people without diabetes is thinner than the marketing suggests.

Start with the boring test

HbA1c reflects average blood glucose over roughly the preceding three months, by measuring how much haemoglobin has been glycated. It requires no fasting and one blood draw.

The thresholds are simple (CDC):

• Normal: below 5.7%
• Prediabetes: 5.7% to 6.4%
• Diabetes: 6.5% or above

The prediabetes band is the interesting one, because it is where intervention has the best evidence and where most people are told “your sugar is a little high” and sent home.

Prediabetes is reversible, and the trial evidence is unusually strong

The Diabetes Prevention Program randomized 3,234 adults with elevated fasting and post-load glucose to three arms: placebo (n=1,082), metformin 850 mg twice daily (n=1,073), or intensive lifestyle intervention targeting at least 7% weight loss and at least 150 minutes of physical activity per week (n=1,079). Mean follow-up was 2.8 years (DPP Research Group, 2002).

Diabetes incidence per 100 person-years:

Arm	Incidence	Risk reduction
Placebo	11.0	—
Metformin	7.8	31% (95% CI 17–43%)
Lifestyle	4.8	58% (95% CI 48–66%)

Number needed to treat over three years to prevent one case: 6.9 for lifestyle, 13.9 for metformin.

Treat seven people for three years, prevent one case of type 2 diabetes. That is among the most favourable numbers in preventive medicine, and it comes from a behavioural intervention that outperformed the drug by a wide margin.

A number in the 5.7–6.4% band is not a warning to file away. It is the entry criterion for the most effective preventive trial we have.

The two targets are specific and worth memorizing: 7% weight loss, 150 minutes of activity per week.

What about fasting insulin?

A reasonable objection to HbA1c is that it is a late signal. Insulin resistance develops years before glucose rises, because the pancreas compensates by secreting more insulin. By the time HbA1c moves, that compensation is failing.

This is why fasting insulin and derived indices such as HOMA-IR attract interest. There is supporting evidence that they move meaningfully with intervention: in the CALERIE trial of two years of caloric restriction, fasting insulin fell by about 1.71 μIU/mL and HOMA-IR improved by about 0.36, alongside a significant increase in the insulin sensitivity index.

The caution is that fasting insulin lacks the standardization HbA1c has. Assays vary between laboratories, reference ranges are inconsistent, and there is no equivalent of the 5.7% threshold with an intervention trial attached to it. It is a reasonable additional measurement and a poor primary one.

The CGM question

Continuous glucose monitors are now sold over the counter in several countries, including the US, to people without diabetes. They produce a compelling stream of data: a graph of your glucose responding to every meal, night’s sleep and stressful meeting.

The question is whether that data changes outcomes.

Problem one: experts cannot agree on what the reports mean. Eighteen clinicians with extensive CGM experience were asked to evaluate 20 Dexcom G6 Pro reports from individuals without diabetes, alongside HbA1c and fasting glucose, and recommend whether follow-up was needed (Spartano et al., 2025).

Agreement was poor: Fleiss kappa = 0.36. Only 3 of 20 reports produced universal agreement. The one discernible trend was that a majority recommended follow-up when readings exceeded 2% time above 180 mg/dL. Beyond that, experts disagreed substantially about which patterns warranted intervention.

The authors conclude that CGM has limited immediate utility for screening non-diabetic populations without established guidelines, and call for research to identify which CGM features actually indicate progression risk.

If eighteen experts cannot agree on what a report means, an app’s interpretation of your report is not carrying the authority it implies.

Problem two: as a behaviour-change tool, the effect is modest. A systematic review and meta-analysis of 25 randomized trials with 2,996 participants — predominantly type 2 diabetes (68%), plus type 1 (12%), gestational diabetes (12%) and obesity (12%) — found (Richardson et al., 2024):

• HbA1c: −0.28% (95% CI −0.42 to −0.15, P < 0.001), 23 trials, high heterogeneity (I² = 88%)
• Weight: −0.7 kg (95% CI −1.4 to 0.0, P = 0.066) — not significant
• Time in range: +7.4% (95% CI 2.0 to 12.8, P = 0.008)

The authors describe modest improvements in glycaemic control and non-significant effects on weight and BMI.

Note that this population was mostly people with diabetes — where CGM has a clear clinical role. The evidence base for metabolically healthy people wearing one is thinner still.

There is a genuine counterargument. For some people, seeing a glucose spike after a specific meal produces a behaviour change that no lab value ever did. That is real, and a −0.28% HbA1c reduction is not nothing. But it is a fraction of the 58% risk reduction produced by a structured lifestyle programme, and it costs considerably more.

What to do with this

• Get an HbA1c. It is cheap, standardized, requires no fasting, and has clear thresholds. If you have had no metabolic testing, this is the first thing to fix.
• Treat 5.7–6.4% as actionable, not informational. That band is the DPP entry criterion, and the intervention has a number needed to treat of 6.9.
• Use the trial’s actual targets: 7% weight loss, 150 minutes weekly. These are not generic wellness goals; they are the prescription that produced a 58% risk reduction.
• Consider fasting insulin as a supplementary early signal, understanding that it is less standardized and has no threshold with trial evidence behind it.
• Be honest about why you want a CGM. If it is motivation, it may be worth the money. If it is diagnosis, the expert-agreement data (kappa 0.36) argue against relying on it.
• Do not over-interpret individual spikes. No established normative framework exists for CGM patterns in people without diabetes, and specialists disagree on which patterns matter.
• If you wear one, use it as a time-limited experiment. Two to four weeks to learn how your body responds to specific meals, then act on what you learned. Continuous monitoring of a metric nobody can interpret is an expensive habit.

The uncomfortable summary is that the cheapest test in this article has the strongest evidence and the most expensive has the weakest. That ordering is common in longevity, and it is worth expecting.