Why repeat testing matters more than the first result

A single result is a point. Two make a direction.

This is the least glamorous article on the site and the most important one. Almost everything that goes wrong in metabolic medicine — overtreatment, undertreatment, expensive panels that change nothing — comes from reading one number as if it were a fact about you rather than one observation of a moving thing.

Every result contains two kinds of noise

Biological variation is you changing. Hormones follow daily rhythms, glucose responds to last night's dinner and this morning's stress, lipids drift with the season. Measure the same healthy person twice a week apart and the numbers differ, because the person differs.

Analytical variation is the laboratory. Every assay has an imprecision, reported as a coefficient of variation. A well‑run laboratory keeps it small; it never reaches zero.

Put together, a change between two results has to clear both before it means anything. For some tests that bar is low. For hormones measured by immunoassay, it is high enough that modest changes are frequently noise.

Regression to the mean, which fools everyone

Here is the trap that makes bad treatment look effective.

If you test a group of people and select those with the most extreme results — the highest glucose, the lowest testosterone — and then retest them, their numbers will on average move towards the middle even if nothing was done. Not because anything improved, but because extreme values contain more noise, and noise does not repeat.

So a clinic that tests you, finds an alarming number, prescribes something, retests and shows you an improvement has not demonstrated that the treatment worked. Some of that improvement would have happened anyway.

This is precisely why diagnostic criteria for diabetes require a repeat on a separate day, and why hypogonadism requires two low morning results. The repeat is not bureaucracy. It is the part that removes the illusion.

What makes two results comparable

1. The same laboratory. The single most important one. Different analysers, different assays, different reference ranges. A ten per cent shift in your testosterone after changing laboratory may be entirely the laboratory.
2. The same time of day. Testosterone, cortisol and TSH all follow daily rhythms. Morning both times, and ideally the same morning hour.
3. The same preparation. Fasted both times, or fed both times. Not one of each.
4. The same state of health. Not in the fortnight after an infection, an injury or a hard training block. Inflammation moves hs‑CRP and ferritin, illness moves thyroid tests, exercise moves liver enzymes.
5. Enough time. Repeating an HbA1c after three weeks tells you almost nothing — it describes three months, and most of it has not changed yet.

How long to wait

It depends on what the marker is made of rather than on a clinic's appointment schedule.

• HbA1c — about three months, because that is the lifespan of the red cells carrying it.
• Lipids — six to twelve weeks after a change in diet or medicine.
• Liver enzymes — weeks, and sooner if a new medicine is the question.
• Thyroid — six to eight weeks after any dose change, because the system is slow to settle.
• A suspicious single result — soon, and under better conditions than the first.

What this means for what a panel costs

If a second measurement is where the meaning is, then the panel you can afford to repeat matters more than the panel you can afford once.

That is the reasoning behind our seven markers at ₡101,500 rather than thirteen at ₡222,500, and behind a repeat panel of three markers at ₡53,000. A thirteen‑marker baseline that is never repeated produces a folder of numbers. Seven markers measured three times in a year produce three lines on a graph, and the lines are the clinical information.

The boundary

None of this says how often you personally should be tested. That depends on what was found, what is being treated and how fast it is expected to move — a conversation with a physician, who should be able to tell you not just when to come back but what they expect the second result to show, and what they will do if it does not.