Why Laboratories Report Slightly Different Ranges

Why Laboratories Report Slightly Different Ranges

A fasting glucose of 5.4 mmol/L can be flagged high on one report and left unflagged on another, printed weeks apart, from the same vein. The number did not change. The reference interval did. This page walks through an annotated specimen of a typical lab report, line by line, so that a reader comparing two documents can see exactly where the divergence enters — the assay, the population, the interval method, the units — and can tell the difference between a meaningful shift and a clerical one.

The specimen below is a composite, assembled from the structure of routine panels rather than copied from any single laboratory. It is deliberately ordinary: a basic metabolic panel plus lipids, with the flag column doing most of the work.

The specimen: one report, four ranges, four philosophies

Every line contains at least four separate decisions: which assay platform produced the figure, which population was sampled to set the interval, which statistical method trimmed that population, and which units the laboratory chose to print. Change any one of the four and the flag column can flip while the underlying biology stays put. That is the entire phenomenon. The rest of this page takes the specimen apart.

Line by line: what each element is actually telling you

Take the ALT line first, since it carries a flag. The interval "< 40 U/L" is what the literature calls an upper reference limit, not a threshold of disease. It is usually set so that the central 95% of a chosen healthy population falls inside it — which means, by construction, one person in twenty from that same population sits outside it while being entirely well. A result of 41 is one unit past a statistical fence, and the fence was built from other people's numbers, not from the reader's own history.

The glucose line shows a different pattern. "3.9 – 5.5" is a two-sided interval, and the upper end sits close to where professional bodies begin discussing impaired fasting glucose. Move to a laboratory using a different calibration standard and the same plasma sample might read 5.3, which prints without a flag. Move again to one that reports in mg/dL and the number becomes 97, a figure that looks unfamiliar but describes the same concentration. Unit convention alone accounts for a large share of the confusion people describe when they compare an older report with a newer one.

• Assay platform and calibration standard, which shift absolute values by small but real margins.
• Reference population: age band, sex, ethnicity, and how "healthy" was defined at recruitment.
• Statistical method — parametric, non-parametric, or a fixed decision limit set by guideline.
• Units and decimal convention, which change the appearance of the figure without changing the sample.
• Rounding rules applied at print time, which can hide a value that sits within a hair of the limit.
• Whether the interval is two-sided or a one-sided "less than" cut-off.

Those six factors are not equally weighted. In practice, the reference population and the statistical method dominate; units are a cosmetic problem that disappears once a reader converts them. A professional body publishing a guideline interval and a laboratory publishing its own locally derived interval may disagree by a full decimal place on the same analyte, and both can be defensible. Neither is "the correct one" in any absolute sense.

A reference interval is a statement about a population, not a verdict about a person; it describes where most well people sit, not where any individual ought to be.

That distinction has practical weight when a result sits just outside the fence. A single flagged value in isolation carries little information; what carries information is the trajectory. A liver enzyme drifting upward across three reports matters more than one report straddling a cut-off, and the drifting pattern is visible only if the same laboratory and the same assay are used over time. This is the main argument for keeping serial testing in one place, and it is also why clinicians often ask for the previous report rather than reading a new one in isolation.

The interval is not the same thing as a decision limit

Many reports mix the two without saying so. A reference interval is descriptive — it summarises a sampled population. A decision limit is prescriptive — it is a point chosen by a guideline panel because action at that point was judged to do more good than harm, on the strength of outcome studies rather than distribution alone. Cholesterol cut-offs are largely of the second kind. Haemoglobin intervals are largely of the first. The specimen above does not label which is which, and most reports do not either.

Vitamin D is the clearest illustration of the gap. Laboratories differ on whether to report in nmol/L or ng/mL, and professional bodies differ on where sufficiency begins, because the evidence linking a specific level to a specific outcome is weaker than the evidence for, say, blood pressure thresholds. A result of 58 nmol/L may be printed without a flag by one laboratory and with a low flag by another, and the disagreement reflects a genuine unsettled question rather than an error at either site.

What an interval cannot tell you

An interval cannot say whether a change is clinically meaningful for a particular person, because it was never built to do that. It cannot account for biological variation within a single individual across a day, which for some analytes is large enough to move a result across a boundary on its own. It cannot tell you whether the assay changed between two reports, which is one of the most common reasons a serial trend looks like a jump. And it cannot substitute for a trend line assembled from consistent conditions.

1. Interval boundaries describe a sampled population, not a target for any individual.
2. Decision limits come from outcome evidence and may sit inside or outside the interval.
3. Biological and analytical variation can move a single result across a boundary unaided.
4. Assay or platform changes between reports can mimic a real shift in the underlying value.
5. Units and rounding change presentation, and presentation is what most comparisons actually see.

Practical reading: comparing two reports without guesswork

The first thing to check when two documents disagree is whether the units match. The second is whether the reference interval printed alongside the result is identical; if it is not, the flag columns are not comparable and should be set aside. The third is the assay or method note, which is often printed in small type at the foot of the page or available from the laboratory on request. Only after those three checks does it make sense to ask whether the underlying value actually moved.

For anyone tracking a long-term picture — body composition, waist-to-height ratio, morning energy score, or a vitamin D level followed across seasons — the useful practice is to record the interval alongside the value rather than the value alone. A lab results glossary kept in the same place as the reports, updated as new analytes appear, removes most of the guesswork later. A symptom log or habit journal serves the same purpose for context that the report itself does not capture, such as sleep latency on the night before a draw.

When to take the interval at face value, and when to ask

Copy the interval verbatim into any personal record. It is the only way to reconstruct later whether a flag was produced by the value or by the fence. Adapt it — that is, treat it as an approximation rather than a target — whenever the result sits within a few percent of a boundary, whenever the analyte is one where professional bodies disagree, and whenever the report crosses laboratories or platforms. In those cases the interval has done its job by telling you where the population sits; what it cannot do is tell you what any single reading means.

What is known is that intervals differ for explainable reasons and that those reasons are usually printed somewhere on the report. What is less certain is how much of any individual's risk sits in the decimal place between two neighbouring cut-offs, and the literature is honest about that gap. Reading two reports side by side with the units, the interval, and the method in view turns a confusing discrepancy into an ordinary piece of documentation — which is all a range was ever meant to be.