Editverse Free Calculator
Reference Range Calculator (CLSI C28-A3)
Establish a reference interval from healthy-subject data — the central 95% by the non-parametric percentile method, with confidence intervals on the limits — and copy the exact R command behind the result. A full guide follows below. Validated against R 4.3.3.
In Simple Words
A reference interval (often called a reference range or “normal range”) is the central span of values — conventionally the middle 95% — seen in a healthy reference population. Results outside it are flagged for attention, though they do not by themselves indicate disease.
The CLSI C28-A3 standard recommends the non-parametric approach: rank the reference values and take the 2.5th and 97.5th percentiles as the lower and upper limits, with a recommended minimum of 120 reference subjects so that confidence intervals on those limits are reasonably tight. This avoids assuming the data are normally distributed.
When to Use a Reference Interval
Establish or verify a reference interval when introducing a new analyte or method, or adopting one from another laboratory or the manufacturer. It is the basis for flagging patient results as low, normal, or high.
Define the healthy reference population and exclusion criteria carefully, since the interval is only as valid as the sample it comes from, and partition by sex or age where biology demands it. Distinguish a reference interval from a clinical decision limit, which is tied to disease risk rather than the healthy distribution. For the spread of a single result, see measurement uncertainty. Unsure how many subjects you need? Our Test Selector can help.
Expected format: Reference-population values
Drop a CSV here or click to choose. Parsed in your browser: nothing is uploaded.
Expected format: Reference-population values
x <- c({data})
quantile(x, c(0.025, 0.975), type = 6) # nonparametric (CLSI C28-A3)
mean(x) + c(-1, 1) * qnorm(0.975) * sd(x) # parametricUser guide
When to use. Establish a reference interval (central 95% of a healthy reference population) for a laboratory analyte.
Assumptions. Values come from at least 120 well-defined reference individuals (CLSI C28-A3); outliers have been reviewed, not blindly removed.
Common mistakes. With fewer than 120 values treat the limits as approximate. Partition by age or sex when the distributions clearly differ.
Worked example
Click Load example data above, then Run Analysis to see a fully worked result with figure, formula and interpretation for this test.
References
- CLSI. Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory; Approved Guideline C28-A3. 2008.
- Horn PS, Pesce AJ. Reference intervals: an update. Clin Chim Acta. 2003;334:5-23.
Worked Examples
Example 1 — Non-parametric limits
From 120 healthy reference subjects, ranking the values and taking the 2.5th and 97.5th percentiles gives the reference interval directly — no assumption of normality. Confidence intervals on each limit show how precisely they are estimated.
Example 2 — Verifying a borrowed interval
To adopt a manufacturer’s interval, a laboratory measures around 20 healthy subjects and checks how many fall outside the proposed limits. If no more than a small number fall outside, the interval can be transferred; if too many do, it must be established locally.
Assumptions to Check
A properly defined healthy reference population. Clear inclusion and exclusion criteria are essential; the interval reflects whoever was sampled.
Adequate sample size. The non-parametric method needs about 120 reference subjects for stable 2.5th and 97.5th percentiles with usable confidence intervals.
Appropriate partitioning. When values differ by sex, age, or other factors, separate intervals are needed; pooling masks real differences.
Common Mistakes
Confusing a reference interval with a decision limit
A reference interval describes the healthy distribution; a clinical decision limit is set by disease risk. They answer different questions and should not be interchanged.
Too few reference subjects
Estimating 2.5th and 97.5th percentiles from a small sample gives unstable limits. Follow the recommended minimum or use a validated transference approach.
Assuming normality unnecessarily
The non-parametric percentile method makes no distributional assumption; forcing a Gaussian model on skewed data distorts the limits.
Ignoring partitioning
Combining groups that genuinely differ (for example, by sex) produces an interval that fits neither group well.
Frequently Asked Questions
How many subjects do I need?
The CLSI non-parametric method recommends a minimum of about 120 reference subjects per partition for reliable 2.5th and 97.5th percentile estimates.
What is the difference from a clinical decision limit?
A reference interval describes the central 95% of a healthy population, whereas a clinical decision limit is set by the risk of a specific disease or outcome.
Can I just adopt the manufacturer’s interval?
You can transfer an interval after verifying it on a small local sample of healthy subjects; if too many fall outside, it must be established locally.
Does the result match R?
Yes. The calculator uses the non-parametric percentile method with confidence intervals on the limits and shows the R command so you can reproduce them.
Related Calculators
References
- Ozarda Y, Sikaris K, Streichert T, Macri J. 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. PMID 30047297.
- Koerbin G, Sikaris KA, Jones GRD, Ryan J, Reed M, et al. Evidence-based approach to harmonised reference intervals. Clin Chim Acta. 2014;432:99–107. PMID 24183842.
- CLSI. Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory. EP28-A3c (C28-A3). Wayne, PA: Clinical and Laboratory Standards Institute; 2010.
Need This Analysis Done for Your Manuscript?
Our team establishes and verifies reference intervals — non-parametric percentile limits with confidence intervals and appropriate partitioning per CLSI C28-A3 — with reproducible R output.
