Reference Interval Verification Calculator
Reference Interval Verification Calculator
Apply the CLSI EP28 verification procedure. Test twenty reference individuals against a published or manufacturer’s interval, count how many fall outside, and find out whether you may adopt it.
Reference Interval Verification
n tested, n outside → verdictA laboratory adopting the manufacturer’s reference interval tests 20 reference individuals, and 1 result falls outside the interval. This is the first set
The EP28 verification procedure
≤ 2 outside the interval → verified
≥ 3 outside → repeat with a further 20
≥ 3 outside again → the interval must be established independently
establishing an interval: ≥ 120 reference individuals per partition
- 20
- the number of reference individuals CLSI EP28 specifies for a verification. It is small because verification is a much weaker claim than establishment: you are asking whether an interval someone else derived is consistent with your population, not deriving one
- 2
- the acceptance limit. A correct interval should exclude 5% of reference individuals by construction, so 1 of 20 is expected; 2 is unremarkable; 3 or more is more than chance comfortably explains. With a true 5% outside rate, about 7.5% of sets of 20 will nonetheless show 3 or more — which is why one repeat is allowed rather than immediate rejection
- the repeat
- a fresh set of 20, not the original 20 topped up. If the second set also fails, the interval is rejected for your laboratory
- 120
- the minimum for ESTABLISHING a reference interval non-parametrically, per partition — so a separate 120 for each sex, age band or other partition you intend to report against. This is the number that makes verification worth doing
- reference individuals
- people who meet your own stated definition of health, selected to resemble the population your laboratory actually serves. This is the part that is genuinely difficult, and getting it wrong invalidates the procedure however carefully the counting is done
- what counts as outside
- below the lower limit or above the upper limit, counted together. Someone who turns out not to meet your definition of a reference individual should be excluded and replaced, not counted as a failure
- one interval at a time
- verification applies to one interval, on one method, for one population. A new analyser, a new method for the same analyte, or a materially different patient population all mean verifying again
Worked example
A laboratory adopting the manufacturer's reference interval tests 20 reference individuals, and 1 result falls outside the interval. This is the first set
20 reference individuals tested, 1 result outside the candidate interval
The acceptance limit is 2 of 20 — no more than 10% of the sample
1 ≤ 2, so the interval is verified and may be adopted
Note that 1 outside is exactly what a correct interval predicts: a reference interval is defined to exclude 5% of reference individuals, which is 1 in 20
Had 3 fallen outside, the verdict would have been to repeat with a fresh set of 20 — not to reject the interval, because with a genuinely correct interval about 7.5% of first sets of 20 will show 3 or more outside by chance alone
Had that second set of 20 also shown 3 or more outside, the interval could not be verified and would have to be established independently, which takes at least 120 reference individuals per partition
Verifying an interval against establishing one
| Verification | Establishment | |
|---|---|---|
| Reference individuals needed | 20, with one repeat of 20 allowed | At least 120 per partition |
| The question asked | Is this published interval consistent with my population and method? | What are the 2.5th and 97.5th centiles in my population? |
| What you get | Permission to adopt somebody else’s limits | Your own limits, with confidence intervals around them |
| Partitions | Each partition you report against needs its own verification | Each partition needs its own 120 |
| Effort | A week of work for most analytes | A project, often a multi-centre one |
| When it is not enough | When two sets of 20 fail, or no suitable published interval exists | — |
Reading the pattern of the failures
| Pattern in 20 subjects | Verdict | What it suggests |
|---|---|---|
| 0–2 outside, either side | Verified | Nothing to investigate |
| 3 or more, all above the upper limit | Repeat | A one-sided difference — your method reads high against the source’s, or your population genuinely differs. Check calibration and traceability before repeating |
| 3 or more, all below the lower limit | Repeat | The mirror image, and the same investigation |
| 3 or more, split both sides | Repeat | The interval is too narrow for your population, or your subjects are more heterogeneous than the source’s. Re-examine the selection criteria |
| 3 or more, and the subjects were not truly healthy | Repeat after replacing them | Not a failure of the interval at all. Exclude and replace anyone who does not meet your stated definition of a reference individual |
Twenty subjects, and what they can honestly establish
Almost no laboratory establishes its own reference intervals. Doing it properly requires at least 120 reference individuals for every partition you intend to report against — separately for each sex, each paediatric age band, and any other subdivision — which for a medium-sized department across a full chemistry repertoire is an undertaking nobody has the samples or the money for. So laboratories adopt intervals from manufacturers, from published studies or from national harmonisation projects, and CLSI EP28 provides the procedure for checking that an adopted interval is not obviously wrong for the population in front of you.
The procedure is small on purpose. Test 20 reference individuals. If no more than 2 of their results fall outside the candidate interval, it is verified and may be adopted. If 3 or more fall outside, repeat with a fresh set of 20; if that set fails too, the interval cannot be verified and has to be established independently. The arithmetic behind the numbers is worth knowing, because it explains why a repeat is allowed at all. A reference interval is defined to exclude 5% of reference individuals, so 1 result outside in 20 is exactly what a correct interval predicts. With a true 5% outside rate, the probability of seeing 3 or more outside in a set of 20 is about 7.5% — so roughly one verification in thirteen fails on a perfectly good interval, and rejecting on a single failure would be unjust to the interval.
Verification is a much weaker claim than establishment, and that has to be understood rather than glossed over. Twenty subjects can detect an interval that is badly wrong for your population; they cannot refine one that is nearly right, cannot tell you where the 2.5th centile really sits, and cannot support any confidence interval around the limits. What you get from a successful verification is permission to use somebody else’s numbers, not evidence that those numbers are correct for you to any precision.
Which makes the choice of the 20 people the part that actually matters. They must meet your own stated definition of a reference individual and they must resemble the population your laboratory serves — in age, sex, ethnicity, diet and the ordinary comorbidity of the people whose samples you will be applying the interval to. Twenty members of laboratory staff are convenient, and they are usually younger, healthier and less diverse than the patients who will be measured against the interval, which biases the verification towards passing. Sampling conditions matter for the same reason: fasting state, posture, time of day and tourniquet time all shift some analytes more than the difference you are trying to detect. And when a verification does fail twice, that is a finding and not an obstacle — it says the published interval does not describe your population, or your method disagrees with the one it came from, and the direction of the failures usually says which.
Frequently asked questions
How many reference individuals do I need to verify a reference interval?
CLSI EP28 specifies 20. The interval is verified if no more than 2 of their results fall outside it. If 3 or more fall outside, a second set of 20 is tested, and if that also fails the interval must be established independently rather than verified.
Why is 3 out of 20 enough to fail a verification?
Because a correct reference interval excludes only 5% of reference individuals by construction, so 1 result outside in 20 is expected and 2 is unremarkable. With a true 5% outside rate, the chance of seeing 3 or more in 20 is about 7.5%, which is small enough to be a signal but large enough that one repeat should be allowed before the interval is rejected.
What is the difference between verifying and establishing a reference interval?
Verification asks whether an interval someone else derived is consistent with your population and method, and needs 20 reference individuals. Establishment derives the interval from your own data and needs at least 120 reference individuals per partition. Verification gives you permission to adopt limits; it does not give you evidence about where those limits should sit.
Who counts as a reference individual?
Someone who meets your own written definition of health for this analyte and who resembles the population your laboratory serves in age, sex, ethnicity and general health. That last condition is the one most often broken: twenty members of laboratory staff are convenient and are usually healthier and less diverse than the patients the interval will be applied to, which biases the verification towards passing.
What should I do if the verification fails twice?
Treat it as a finding. Two failed sets of 20 say that the published interval does not describe your population, or that your method disagrees with the one the interval was derived on. Look at whether the failures were one-sided, which points at a calibration or traceability difference worth investigating, then either establish an interval yourself with at least 120 subjects per partition, or find a source whose population and method resemble yours more closely and verify that.
Does a verified interval stay verified?
Only for the method, platform and population it was verified on. A change of analyser, a change of method for the same analyte, a reformulated reagent or a materially different patient population all mean the verification has to be repeated. It is a statement about one combination of circumstances, not a permanent property of the interval.
Related calculators
References
- CLSI EP28-A3c. Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory; Approved Guideline. 3rd ed. Clinical and Laboratory Standards Institute; 2010.
- Horowitz GL. Reference intervals: practical aspects. EJIFCC. 2008;19(2):95–105.
- Ozarda Y, Sikaris K, Streichert T, Macri J; IFCC Committee on Reference Intervals and Decision Limits. Distinguishing reference intervals and clinical decision limits — a review by the IFCC Committee on Reference Intervals and Decision Limits. Crit Rev Clin Lab Sci. 2018;55(6):420–431.
- ISO 15189:2022. Medical laboratories — Requirements for quality and competence. International Organization for Standardization; 2022 — biological reference intervals and their periodic review.
Medical Disclaimer: The tools and content provided here are for educational and reference purposes only. They are not intended to substitute for professional medical advice, diagnosis, or treatment. Clinical decisions should always be based on the comprehensive assessment of a qualified healthcare professional.