Written and maintained by CASRAI Editorial Board
Last updated
General laboratory information for lab and clinic staff, not medical advice. Individual results are interpreted by a clinician or the reporting laboratory.
Short answer
A reference interval is the interval between, and including, two reference limits, derived from the results of a sample of well-defined reference individuals. In the nonparametric method, the limits are the 2.5th and 97.5th percentiles of those results. A laboratory either establishes its own interval, with at least 120 reference individuals per partition, or adopts one and verifies it with a small study of 20.
Definition and scope
The IFCC definition, as given by Ceriotti and Henny (EJIFCC 2008), is that a reference interval is the interval between, and including, two reference limits, which are values derived from the distribution of results obtained from a sample of the reference population. A reference individual is a person selected for testing on the basis of well-defined criteria.
The same paper explains why the everyday phrase “normal range” is avoided: “normal” is not a scientific term. It also separates three different yardsticks that are easy to confuse:
- Reference interval, built from the results of reference individuals.
- Decision limit, which depends on the type of decision to be made.
- Reference change value, used where a person’s results are compared with their own earlier results.
This page covers the first: how a laboratory arrives at the interval printed on a report. CLSI guideline EP28-A3c, Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory (3rd ed., published October 2010; the CLSI catalogue page lists it as reaffirmed in April 2020), is the guideline for quantitative laboratory tests. It is written for diagnostic laboratories and manufacturers and covers establishing, transference and verification.
How labs establish a reference interval
Horowitz (EJIFCC 2008) recommends 120 reference individuals. With 120 observations there is enough data to determine the central 95% of the distribution. Each partition the laboratory wants separate limits for, such as sex, age bracket or race, needs its own 120 observations.
Worked example: the nonparametric method with 120 values
- Rank the 120 results from lowest to highest.
- Rank 3 is the 2.5th percentile, the lower reference limit.
- Rank 118 is the 97.5th percentile, the upper reference limit.
- Each limit has its own uncertainty. A 90% confidence interval for the 2.5th percentile runs from rank 1 to rank 7; for the 97.5th percentile, from rank 114 to rank 120.
That last step shows that a limit is an estimate. Ceriotti and Henny note that a different sample of reference individuals would give slightly different limits; the range of that swing narrows as more individuals are selected and widens for skewed distributions or when unhealthy people are wrongly included. For the idea behind the 90% band, see confidence interval.
Transference and verification
Many laboratories do not run a 120-person study for every analyte. They adopt an interval from another source and check that it fits their population. Two points from the published sources:
- Evaluate the original study first. A CAP TODAY Q and A (March 2024, answered by Brian Naidrin Chang, MD, citing CLSI EP28-A3) says a laboratory should evaluate the original study, including demographics, sample size and methods, before adopting an external interval.
- Verify with 20 samples. Both Horowitz and CAP TODAY describe a verification study using samples from 20 people. The interval is acceptable if no more than 2 of the 20 results fall outside it.
When verification does not pass
The sources describe the failure branch differently. Horowitz says that when 3 or more of 20 fall outside, more data should be collected. CAP TODAY gives a longer ladder: if 3 or 4 of 20 fall outside, collect an additional 20; if 3 or more of the second set fall outside, reconsider the interval; and if more than five fall outside, consider an alternative interval.
Establish, transfer or verify: side by side
| Route | What the sources say it involves | Source |
|---|---|---|
| Establish | New study of at least 120 reference individuals per partition; nonparametric limits at the 2.5th and 97.5th percentiles | Horowitz 2008 |
| Transfer | Adopt an interval from another study after evaluating its demographics, sample size and methods | CAP TODAY 2024; CLSI EP28 scope |
| Verify | Samples from 20 people; pass if no more than 2 fall outside the interval | Horowitz 2008; CAP TODAY 2024 |
Where it matters for labs and suppliers
- Partitions multiply the work. Because each partition needs its own 120 observations (Horowitz), separate limits for sex or age groups multiply the number of reference individuals required.
- Intervals depend on the reference population. Ceriotti and Henny note that a different sample of reference individuals would give slightly different limits, so the interval a laboratory applies should come from a population and method suited to its own.
- Limits are not sharp lines. Ceriotti and Henny describe a “grey zone” around the reference limits.
- Related lab documents. Intervals appear on reports and requisitions, see laboratory requisition, and are distinct from alert thresholds, see critical value reporting policy.
Common misunderstandings
- Outside the interval does not mean disease. Reference intervals include about 95% of unaffected individuals, so a person has roughly a 5% statistical probability of falling outside the limits even when unaffected (Ceriotti and Henny).
- A reference interval is not a decision limit. The first describes a reference population; the second relates to disease or risk.
- One interval does not fit every laboratory or person. The interval depends on the reference population, the method and the partition.
- Test performance is a separate question. How well a test separates disease from non-disease is described by measures such as positive predictive value, not by the reference interval.
FAQ
What do reference intervals mean?
They describe the spread of results in a defined group of reference individuals, usually the central 95%. They show where results from that group usually fall, not whether a given person is healthy or ill.
What is a reference value?
It is a result obtained from a reference individual. The reference limits are values derived from the distribution of results obtained from a sample of the reference population, and the reference interval lies between them.
What is the normal reference range?
“Normal range” is the everyday term for the reference interval printed on a report. Ceriotti and Henny call “normal” not a scientific term and use reference interval. Which interval applies depends on the laboratory, the method and the partition.
Sources
- Horowitz GL. Reference Intervals: Practical Aspects. EJIFCC 2008;19(2):95-105. https://pmc.ncbi.nlm.nih.gov/articles/PMC4975204/ (checked 2026-11-01).
- Ceriotti F, Henny J. “Are my Laboratory Results Normal?” Considerations to be Made Concerning Reference Intervals and Decision Limits. EJIFCC 2008;19(2):106-114. https://pmc.ncbi.nlm.nih.gov/articles/PMC4975205/ (checked 2026-11-01).
- CLSI. EP28-A3c, Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory, 3rd ed. Catalogue page: https://clsi.org/shop/standards/ep28/ (checked 2026-11-01).
- CAP TODAY Q and A, March 2024: What source should a laboratory use for reference intervals for analytes? https://www.captodayonline.com/qa-column-0324/2/ (checked 2026-11-01).
