Reference Range Headaches
I was recently asked to give a lecture on establishing a reference range. For me, this is an easy task on paper but a difficult task to carry out. By reference range I mean that range from a low value to a high value within which most “normal” people will fall. Usually the reference range includes only those people above the lower 2.5 percent and below the upper 97.5 percent of those people tested—the 95 percent range. Thus, by definition, 5 percent of “normal” people will be excluded from the normal range and be labeled abnormal. Sometimes, diagnostic tests will be run on these persons to verify that they are normal and not abnormal. Any definition, other than including everyone labeled “normal” will result in some people being labeled “abnormal.” The more diagnostics tests run on any person using a cutoff such as the 95 percent range will result in an increasing number of these “abnormal” patients. If 5 percent are abnormal by Test A and another 5 percent are abnormal by Test B, that yields 10 percent abnormal results.
The next problem is the definition of “normal.” A partial list of those factors that could be considered as part of the definition of normal might include:
-
Was the patient fasting?
-
What time of day was the sample drawn?
-
What time of month was the sample drawn?
-
Was the person on ANY medication (including any vitamin supplement, aspirin, etc.)?
-
Does the person drink, and how much?
-
Is the patient a smoker?
-
What's the patient's age and sex?
If we forget the first points and use only age and sex and ask for, say 100 data points for each sex and each decade beginning at age 20 through age 70, we have 2 times 6 time 100 or 1,200 persons that we much test for any analyte for which we want to establish a reference range. Adding any other variables increases the number of patients we need to test.
We might define as normal a person with a normal chem.-12 profile. If that is the only criterion for normal and we use as normal the 95 percent range mentioned earlier and eliminated anyone with an abnormal value, we would lose nearly 50 percent of the patients we tested as being abnormal. The 50 percent is 0.95 raised to the 12th power. Think of adding a normal CBC to this criterion and then add a normal urinalysis and imagine how few people for each sex and age and you can see the time, effort and cost involved in establishing a reference or normal range. It is a daunting task. Verifying a reference interval is easier but still involves consideration of the variables and thus the source of the data.