Validate, Transfer or Establish: What Are You Doing with Your Reference Intervals?

Reference intervals are absolutely necessary for proper interpretation of laboratory tests, and yet obtaining appropriate reference intervals can be the bane of the laboratory. I mentioned establishing, validating or transferring reference intervals in an earlier blog post, but didn’t talk about exactly what these are and when to use which one.

Establishing a reference interval is exactly what it says. A reference interval must be established if a new assay has never been performed in the lab and there is no current reference interval to start with. Most often, laboratory developed tests (LDTs) that are developed from scratch will require the establishment of a reference interval. To do this, ideally 120 samples from healthy individuals for each sub-population (gender or age sub-group) is used, although there are methods available using smaller sample sizes. Samples used to establish reference intervals may be collected a priori, meaning they are collected from individuals for the express purpose of establishing a reference interval, with well-defined inclusion or exclusion criteria used, or a posteriori, meaning they are samples collected and analyzed first, with exclusion criteria applied after statistical analysis.

Validating a reference interval is the easiest way to obtain one, and is what is hoped for when a new method is introduced. Validation is usually used when a new instrument or method replaces an old one, and reference intervals are currently in place. A patient correlation study is done using at least 20 patients. The data is analyzed with regression, bias and correlation statistics. If the bias and regression are acceptable, the reference interval that is currently in place will also work with the new assay. The interval has been validated and can be used with the new method.

When a validation study is done for a new method and the results of the data analysis are NOT acceptable to validate the assay, then a transference study is necessary. A transference study is simply an extended correlation. More than 20 patients are used, enough to determine the amount of bias between the two methods. Then the old reference interval is adapted to fit the new method, using the amount of bias determined. For example if the new method measures 15 percent higher than the old method, the reference intervals will be increased 15 percent across the board. Transference is recommended to be performed once. If another new method is brought in for that analyte, rather than transfer the reference interval again, a new interval should be established.

All three of these methods for obtaining reference intervals are useful in the right situations. It is important to know when to use which method.



-Patti Jones PhD, DABCC, FACB, is the Clinical Director of the Chemistry and Metabolic Disease Laboratories at Children’s Medical Center in Dallas, TX and a Professor of Pathology at University of Texas Southwestern Medical Center in Dallas.

Reference Intervals vs. Reference Change Values

If we didn’t use reference intervals (RI), how would we know whether a person is “normal” or not? Or more accurately, how would we know whether a lab test result indicated health or disease? Reference intervals have been around as long as lab tests and they help clinicians diagnose and monitor a patient’s disease state. .

Most RI are developed using a specific patient population and should be used only with that population. However, some RIs are “health-based,” such as cholesterol and vitamin D. Both these analytes have RI that indicate what amount of the analyte should be present in a healthy individual, not how much is present in your specific population of patients. In general, health-based RI can be utilized in all populations, as long as the analyte assays are commutable. Thus these type of RI are often more useful than population-based intervals.

But should we be using reference intervals at all? One problem with population-based RI is that any given individual’s values may span a range that covers only part of the population RI due to biological variability. For example, an individual’s creatinine may be 0.6 – 0.9 mg/dL regularly. Since the RI for creatinine for his population is 0.4 – 1.4 mg/dL, a value of 1.2 mg/dL would not be flagged as be abnormal. However, 1.2 mg/dL may very well be an abnormal result for this individual We need to consider using reference change values (RCV) in addition to RI.

Reference change values are calculated values that are used to assess the significance of the difference between two measurements. Essentially, a RCV is the difference that must be exceeded between two sequential results for a change to be a significant change. The calculation requires knowledge of the imprecision of the analyte assay (CVA) and the biological variation (CVI) of the analyte. The formula for calculating RCV is: RCV=21/2 · Z · (CVA2 + CVI2)1/2 , where Z is the number of standard deviations for a given probability. Luckily, labs know the imprecision of their assays and there are tables available for biological variation.

It’s very likely that neither RI nor RCV by itself is adequate for interpreting analyte results. Using both may be a better alternative, especially using RCV for monitoring disease progression or therapeutic efficacy. Flagging sequential values that exceed the RCV—and reporting this change—should be considered.

-Patti Jones