CGMs sit on chemist shelves from Auckland to Invercargill. A two-week sensor costs less than a decent pair of running shoes, and the app draws pretty curves on your phone. The curves are not the insight. The skill is knowing which spikes matter, which are noise, and what to change on Monday.

What a CGM actually measures in a non-diabetic adult

A continuous glucose monitor reads interstitial fluid glucose, not blood glucose. There is a 5 to 15 minute lag, and accuracy sits around plus or minus 15 percent against a venous draw. For a person without diabetes, this matters: a reading of 8.2 mmol/L after porridge might be 7.1 on a finger-prick, and neither number means you are pre-diabetic.

What the device does well is show you patterns. Shape, timing, recovery, and the relationship between meals, sleep, training, and stress. Single numbers lie. Two weeks of curves do not.

For diagnosis, HbA1c and a fasting insulin via your GP remain the reference. We work alongside your GP on that side. The CGM is a behavioural tool.

The four numbers worth watching

Most NZ apps show an average glucose, time in range, and a standard deviation. For a non-diabetic adult, these are the ones we coach to:

  • Fasting glucose on waking: ideally 4.4 to 5.5 mmol/L. Persistently above 5.6 across a fortnight is a conversation with your GP.
  • Post-meal peak: a rise of less than 2.0 mmol/L above pre-meal is a well-tolerated meal. A rise above 3.0 mmol/L repeatedly on the same food is a signal.
  • Time to baseline: returning to within 0.5 mmol/L of pre-meal by the 2-hour mark suggests good disposal. Still elevated at 3 hours means the meal, the context, or both need work.
  • Standard deviation across 14 days: under 1.0 mmol/L is smooth. Over 1.4 means your day is a rollercoaster, and you will feel it.

Average glucose alone is a poor headline. Two people can both sit at 5.4 mmol/L average. One is steady. The other is spiking to 9 and crashing to 3.5. The second person feels terrible by 3pm and does not know why.

Meal patterns we see in client data

Across 1,380+ clients and a food dataset of 2,846 items, the same handful of patterns repeat in CGM trials. They are worth naming.

The white-carb solo: toast alone, rice alone, a banana alone. Sharp peak, fast crash, hunger inside 90 minutes. Add protein and fat to the same carb and the curve flattens by roughly a third in most people.

The second-meal effect: a high-carb breakfast produces a bigger spike at lunch than the same lunch eaten after a protein-led breakfast. Your morning meal sets the tone for the day.

The stress spike: glucose climbs 1.5 to 2.5 mmol/L during a hard meeting, no food involved. Cortisol and adrenaline mobilise liver glycogen. This is normal physiology, not a metabolic fault.

The bad-sleep tax: one night under 6 hours typically lifts the next day's average glucose by 0.3 to 0.6 mmol/L and worsens every post-meal response. The CGM makes this visible in a way nothing else does.

The evening fade: the same kūmara that sits flat at lunch spikes harder at 8pm. Insulin sensitivity drops through the day. Carb timing is real.

The CGM does not tell you what to eat. It tells you how your body responded to what you ate, in the context you ate it. Those are different questions.

Where CGMs beat a finger-prick

A finger-prick gives you one data point. A CGM gives you 96 per day, every day, for two weeks. For a curious adult, that delivers three things a glucometer cannot:

  • Overnight data: dawn phenomenon, nocturnal dips, and the glucose cost of a late dinner or a glass of pinot.
  • Exercise response: how your morning run, your afternoon weights session, and your Sunday hike each move glucose differently.
  • Real-world meals: the actual flat white and muffin at the Christchurch café, not a lab-controlled test meal.

The trade-off is accuracy. If you need a precise number, prick a finger. If you need a pattern, wear a sensor.

Where CGMs mislead

Three traps catch most first-time wearers.

A reading of 7.8 mmol/L after a meal is not "a spike" in any clinical sense. Healthy adults routinely touch 7 to 8 mmol/L after carbohydrate. The disease threshold is 11.1 at two hours on an oral glucose tolerance test. Context matters.

Compression lows are common. Sleep on the sensor and you will see a "reading" of 2.8 mmol/L that never happened. If you feel fine, you were fine.

And cherry-picking single meals is a waste of a sensor. One pasta dinner tells you nothing. The same pasta dinner across three Tuesdays, with and without a walk afterwards, with and without a hard day at work, tells you a lot.

What to do this week

  • Buy one sensor, not four. Two weeks of attention beats two months of passive wear.
  • Log every meal in the app with a photo. Untagged data is useless data.
  • Run two A/B tests: the same breakfast with and without 30g of protein, and the same dinner with and without a 15-minute walk after.
  • Note your sleep hours and stress each morning. Overlay them on the curves at the end of the fortnight.
  • Take your fasting average and standard deviation to your GP if anything sits outside the ranges above. The device starts the conversation. It does not finish it.