1. “Normal Values”: How are Normal Reference Ranges Established?
Module 5
“Normal Values”: How are Normal Reference Ranges Established?

2. Reference Ranges
Reference Ranges are required by professional accreditation and regulatory standards
Laboratory directors determine and evaluate reference or “normal” ranges reported with all test results.

3. Reference Ranges
A reference range is used to interpret a lab test result.
In most cases, a “normal range” is used as the test’s “reference range”.
But sometimes, what is normal isn’t healthy… So a reference (but not normal) range is used
e.g. total cholesterol desirable
<200mg/ml

4. Establishing Reference Ranges
When selecting the decision threshold or cutoff value (a limit above or below which a patient is considered affected by a particular disease, i.e., normal or abnormal), a variety of methods can be used.

5. Gaussian Distribution
If tests results from a normal healthy patient population fall into a bell-shaped, Gaussian, normal distribution, the central 95% is usually used as the test’s normal range.
For many (but not all) tests, this is how the range of tests results for normal healthy individuals is determined.

6. Some Normals are Abnormal …… and Vice Versa
The normal range encompasses the mean plus or minus two standard deviations or, again, about 95% of normal, healthy individuals’ test results.
However, 5% (roughly 1 out of 20) normal healthy patients may be outside the cutoff value.
Roughly 2.5% of normal people can be expected to have a result below and roughly 2.5% of normal people can be expected to have a result above the reported normal range.
This situation is encountered with almost all tests.
This is because the distribution of tests results from normal, healthy individuals overlaps (to varying degrees) with the distribution of test results from sick patients with the relevant disease.

7. Two Populations of Results
Theoretically, the better tests minimize this overlap between the distribution of normal and abnormal test results.
An ideal test would have no overlap at all and could perfectly discriminate between a normal and abnormal test result.
Lab experts continue to look for at least one test like this …….

8. Non-Gaussian Distributions
For non-Gaussian distributions, lab directors can use other nonparametric techniques to establish reference range limits
Example: set upper and lower limits of normal to include 95% of the population after all of the test results have been transformed into logarithms taking the central 95% of the transformed data.

9. Where’s the Threshold or Cut-off?
Ultimately, where the lab places the limits (threshold or cut-off) on a normal or reference range determines what level of result is considered “normal” or “abnormal”.

10. In the next slide, the values for the concentration of a hypothetical analyte were determined in a group of 100 healthy persons and in a group of 25 diseased persons. The raw data for the group were fitted to gaussian distributions.
A through D represent possible cutoff values that could be used to classify subjects based on the analyte values.

11. Distribution of a Test Result in
Healthy (n=200) and Diseased (n=50) Persons
Healthy Persons
Frequency
Cutoffs A B C D
Diseased Persons 10 20 30 40 50 60 70 80 90
100
110
120
130
Analyte Value (units)

12. What would be the advantages of selecting A as the cut off?
What about D?

13. Where’s the Threshold or Cut-off?
To reiterate, where the lab places the limits (threshold or cut-off) on a normal or reference range determines what level of result is considered “normal” or “abnormal”.
It also affects the distribution of values and how they are tallied in the predictive value grid shown on the next slide …..

14. Predictive Value Grid Test Result Disease or Sick No Disease
(Normal, Healthy)
Positive
Result*
True Positives
False Positives
Negative Result*
False Negatives
True Negatives
*“positive” usually refers to a test being abnormal, “negative” usually refers to normal

15. The Next Module - Entitled Module 6 “Common Laboratory Tests” - is a reference for you as youbegin interpreting laboratory data in MHD.