- Last Update On : 2016-07-24
Reference ranges for the same methods or instruments may differ between laboratories and geographic areas for the following reasons:
- Operating conditions are different.
- Criteria for selection of healthy subjects are different.
- Patient populations are different.
- Geographic areas have different temperatures, altitudes, barometric pressures, humidities, and time zones.
- Subject preparation and sample collection may differ.
Because of these differences, reference ranges should be established locally whenever a new test is introduced or an existing method is changed. NCCLS recommends testing at least 120 patient samples for establishment of a statistically significant reference interval. Other experts recommend a minimum of 200 samples to ensure stable lower and upper reference limits.
It may be necessary to subdivide reference ranges into partition groups. For example, hemoglobin reference ranges should be established by gender and age while cortisol ranges should be subdivided by the time of day. Each partition group should contain 120 samples.
Preanalytical factors may also need to be controlled in order to insure that accurate reference ranges are derived. Some factors to consider are: timing of collection; food, water, and alcohol consumption in the past 24 hours; venipuncture site; tourniquet time; and type of collection tube. Samples to be used for reference range determination should be collected and processed the same as routine laboratory work.
Two approaches, parametric and nonparametric, may be taken when analyzing reference range data. The parametric approach involves calculating the mean and standard deviation to determine the range of values that fall within the 95% confidence interval. Before this method can be used the population distribution must be examined to ensure that a Gaussian (bell shaped) distribution is present. If the population distribution is skewed, parametric statistics are not valid. The population distribution can be visualized by plotting a frequency histogram in a spreadsheet, such as Excel. Because this exercise is very time consuming, the nonparametric approach is often chosen.
The nonparametric approach involves establishing the values falling at the 2.5 and 97.5 percentiles of the population as the lower and upper reference limits. The first step is to enter the analyte values in a spreadsheet and sort them in ascending order of magnitude. The number of values in the data set is then multiplied by 0.025 and 0.975 to obtain the percentile ranks of the upper and lower reference limits, respectively. For example, if the data set consisted of 120 specimens, the 2.5 and 97.5 percentile ranks would be calculated by multiplying;
120 x 0.025 = 3
120 x 0.975 = 117
The lower reference limit would be the third number from the beginning (top) of the sorted list and the upper reference limit would be the third number from the (end) bottom. The reference range would be the central 95% of the data, which falls between the 3rd and 117th values.
Outliers can have a substantial effect on the calculation of reference ranges by this method and should be removed. Outliers are data points that lie outside the bulk of the population. Although computer programs and formulas exist to identify outliers, none is more effective than visual examination of the data. Outliers are results that differ from the mean by more than 3 SD or differ from other results by more than 30%.
Consensus Reference Intervals
Reference intervals for some analytes are determined by consensus of medical experts based on the results of clinical outcome studies. For example, the American Diabetes Association has developed consensus values for glucose and hemoglobin A1c. Other examples of consensus reference intervals are summarized in the following table.
|Analyte||Reference Interval||Consensus Group|
<100 mg/dL nondiabetic
100-125 mg/dL prediabetes
≥126 mg/dL diabetes
<200 mg/dL desirable
200-239 mg/dL moderate risk
>240 mg/dL high risk
|Triglycerides||<150 mg/dL||AHA, NCEP|
<7.0% target for diabetics
When consensus reference intervals are available, clinical laboratories report these values instead of determining their own reference range.
Verification of a Reference Interval
Verifying a reference range is different than establishing a reference range. For an FDA approved test method, the clinical laboratory can adopt the manufacturer’s stated reference range if its patient population gives similar results to those published in the manufacturer's package insert.
A total of 40 samples, 20 from healthy men and 20 from healthy women, should be tested and the results compared to the published reference range. The results should be evenly spread throughout the published reference range and not clustered at one end. If 95% of the results fall within the published reference range, it can be accepted for use. If the manufacturer's reference range cannot be validated, the laboratory needs to establish its own reference range.
Verification of the reference range is also useful when it is too difficult to collect a large number of samples to establish a new reference range. In this situation it is permissible to determine whether the manufacturer’s stated reference range, your existing reference range, or a reference range established by another neighboring laboratory using the same instrument and reagents is applicable.