Biological variation, and when a change in your labs is real

Draw blood from a healthy person every week for ten weeks, at the same time, same lab, same conditions, and you will not get the same number ten times. Some of the spread is the analyser. Most of it, for many markers, is the person: the body holds each marker around a set point and drifts around it. Laboratory scientists call this within-subject biological variation, and it decides whether the difference between your last two reports means anything.

The two sources of noise

• Analytical variation (CV_A): how much the lab's measurement scatters when it measures the same sample repeatedly. Every lab knows this figure for every test it runs.
• Within-subject biological variation (CV_I): how much the true level scatters around your own set point over days and weeks. Callum Fraser, who did much of the foundational work, defines it as the average variation around the homeostatic setting point.

There is a third, between-subject variation (CV_G), which is how much people's set points differ from each other. It is why population reference ranges are wide, and why a change within the range can still matter for you.

The reference change value

The reference change value, sometimes called the critical difference, is the smallest change between two results that is unlikely to be noise. Fraser gave the formula in Clinical Chemistry and Laboratory Medicine in 2011:

RCV = √2 × Z × √(CV_A² + CV_I²)

The √2 is there because two results each carry noise. Z sets the confidence; 1.96 gives 95 percent for a change in either direction, and √2 × 1.96 is about 2.77. So at 95 percent, a change needs to exceed 2.77 times the combined coefficient of variation to count.

Worked examples

The within-subject figures below come from studies run to the European Federation of Clinical Chemistry and Laboratory Medicine's standards. Where a lab's analytical variation was not part of the study, the RCV shown assumes a perfect assay, so the real threshold is a little larger.

The creatinine rows come from the EuBIVAS study in Clinical Chemistry in 2017, which drew 91 healthy adults weekly for ten weeks at six European laboratories. They show something the formula makes plain: the same person needs a bigger change to stand out on a less precise method. Glucose and HbA1c come from a 2020 meta-analysis in Advances in Laboratory Medicine, and hs-CRP from a 2024 PLOS ONE meta-analysis of 60 studies.

Why HbA1c is quiet and hs-CRP is loud

HbA1c averages three months of glucose, so day-to-day swings are smoothed out and its within-person variation is about 1 percent. That makes it a good marker for detecting small real changes. hs-CRP is the opposite. It rises within hours of a cold, a dental procedure or a hard session, and falls within days. With variation around 44 percent, the simple RCV formula gives a threshold above 100 percent, which is a sign the formula's assumptions are strained: markers like this are skewed, so labs use a log-transformed version that gives a different threshold for a rise than for a fall. In plain terms, one hs-CRP that doubled is not evidence of anything. That is why the CDC/AHA statement recommends averaging two measurements about two weeks apart.

Making two draws comparable

The RCV assumes the only differences between draws are biological wobble and assay scatter. Anything else you change adds to the noise. To keep that to a minimum:

• Same lab, same method. The creatinine example shows how much the method matters. A different lab is a different series.
• Same time of day. Several markers follow daily rhythms. A morning draw is easiest to repeat.
• Same fasting state. Glucose, insulin and triglycerides depend on it; see the fasting guide.
• Same training pattern. No hard session for a few days, since AST, ALT and creatinine rise afterwards.
• Not while ill. Inflammatory markers and the white count move with any infection.
• Same point in the dosing cycle. For a weekly injection or a long-acting compound, note the days since the last dose.

When one value is enough, and when it is not

One value can answer "is this far outside the range?" It cannot answer "is this changing?" Two values can answer that only when the difference exceeds the RCV. Three or more values, under matched conditions, start to show a direction, and that is where most useful conclusions come from.

Where to look up a marker

The EFLM Biological Variation Database collects appraised within-subject and between-subject estimates for hundreds of tests and is the reference labs use. Your lab can tell you its own analytical CV for any test. The lab report guide covers the rest of the report, and how often to get blood work covers spacing draws so they can show a change at all.

On a timeline

In Peptide AI, each marker is drawn over time with the dose timeline underneath, and a note on each draw (fasted, time, days since dose, training) makes it easier to judge whether two points were taken under the same conditions. See bloodwork tracking and blood panels in the app.