Methodology & review process

In the Field Lab I show you my own glucose curves: one person, one sensor, one meal. This page describes how I measure, read, assess and approve, so that every Field Lab article can link here instead of repeating the rules.

What an n = 1 series can show – and what it cannot

A personal measurement series can make patterns over time visible: when my value rises after a particular meal, how high, how long it stays up, and what happens if I walk afterwards instead of sitting. It cannot prove which mechanism in the body is behind it, and it says nothing about you. A curve is not a study: no control group, no blinding, no repetition across many people. That is why I call all my own values a “single-case observation (n = 1)” and never turn them into a general rule.

study contexts: established, plausible, open

Key factual statements in knowledge articles and Field Lab posts are assigned one of three levels, as text and not only as colour; older posts are being updated:

• Established: The statement is supported by guidelines, systematic reviews or several independent studies, numbered as sources at the end of the article.
• Plausible: The explanation fits known physiology and my timeline, but has not been isolated in my data, or the research base is thin – an experience value, not proof.
• Open: The data, repetitions or comparisons needed for a robust statement are missing. I raise the question but do not answer it.

A statement moves up only when new sources or a targeted comparison support it – I would rather rate too cautiously than too generously.

How I read CGM data

A continuous glucose monitor (CGM) measures not in the blood but in the interstitial fluid under the skin. Three rules follow for every curve:

• Time window: A 30-minute window answers a different question than two, four or twelve hours. Mixed meals rich in fat or protein, alcohol and intense exercise often act with a delay. I therefore always state the observation window and never judge a peak without the course before and after it.
• Sensor lag: During rapid rises or falls the reading trails blood glucose by several minutes. A fresh sensor, pressure on the site, dehydration and data gaps can distort the curve further; a value that does not fit the situation counts as uncertain.
• Influencing factors: Meal, exercise and sleep are the big three, joined by stress, infections, the starting value and the trend before eating. Not every factor can be held constant. What matters is to note deviations rather than explain them away afterwards.

The data sheet in every article

Newer Field Lab articles contain a data sheet with the same fields; older articles are being updated step by step:

• Sensor: FreeStyle Libre 3, values in mg/dl (mmol/l = value divided by 18).
• Period: Date and time of the start and end of the observation window.
• n: Number of observations in the article – usually one, sometimes two or three comparison days.
• Start value: Glucose and trend arrow immediately before the meal or event.
• Peak: Highest value in the window, with time and minutes since start.
• 2-h value: Glucose two hours after the start – the usual reference point for meal responses.

Each value notes whether it came directly from the sensor, a screenshot, a calculation from raw data or memory. I do not replace missing information with guesses.

Check questions before the conclusion

Before writing a conclusion I answer these questions; if one remains open, the conclusion becomes more cautious:

• Are the start value, trend and time documented?
• Are portion, ingredients and drinks fully described – including sauces and side dishes?
• Was the observation continued long enough to see a late second wave?
• Is there a comparison day under similar conditions?
• Is every statement labelled as established, plausible or open?
• Do the title, image and caption reflect the actual uncertainty?

Alternative explanations

Proximity in time is not causation. If my value rises after a beer, I first check what else was involved: the dressing, the portion, the stressful day, the missing walk, the starting value. Several factors can act at once and mask each other. I therefore never pin down a single factor as the “culprit” without a targeted comparison, and I also document the days when nothing unusual happened – showing only remarkable days would distort the picture.

Interpretation matrix

Every key statement belongs to one of four classes:

Approval control

No article appears without three steps. First, expert approval: a person checks sources, figures, units, study contexts and safety notes and sets the status “reviewed – approved”. Second, the content hash: at approval a checksum is calculated over text and images; any later change requires a new approval. Third, the revision log: every change is logged with a date, and articles carry “last reviewed” and “sources as of” dates. I correct errors visibly, not silently. Where AI helps – and where it decides nothing – is explained on How we use AI.

Safety framework

I live with type 2 diabetes without insulin or glucose-lowering medication – that is my starting point, not yours. So for everything in the Field Lab: do not stop medication, change a dose or postpone treatment because of an article. If you take insulin or other glucose-lowering drugs, target ranges, corrections and dietary changes belong with your care team. With marked symptoms, repeated hypoglycaemia, very high values or a curve that does not fit the situation, medical assessment matters more than any self-experiment. My curves should make you curious to explore your own patterns with medical guidance – not to copy my values.