Medical Tests:
What Does a "Normal Range" Mean?


Gary E Cordingley, MD, PhD
What is the difference between a laboratory test-result that is "normal" and one that
is "unremarkable?" The answer, it turns out, lies at the intersection of statistics and
clinical meaning.
We have a marvelous array of medical tests available to us. Many of them—typically blood-tests—even come
with results expressed in numbers signifying the exact quantity of something that was measured. On laboratory
reports these measured quantities are often accompanied by a "normal range" for what the laboratory
apparently thinks the value should have been, showing a lower number and a higher number. So if you had a
laboratory test, what does it mean when your measurement falls inside or outside this range?

It all depends on the particulars. First of all, it matters what is being measured and why it was measured in the
first place. In theory, a laboratory test is ordered when the doctor poses a question for which the laboratory test
is supposed to provide an answer. (If there was no question, how could the lab-test be an answer?) For
example, suppose a doctor is wondering if your hand tremors are due to too much thyroid-hormone in your
bloodstream. So the doctor's question is: Does this person have too much thyroid-hormone in the
bloodstream? A test measuring the thyroid-hormone would provide a clean answer if it was either too high (yes)
or within the normal range (no). Because thyroid-hormone levels that are too low do not produce tremors, a
measurement that was too low would be irrelevant to the question posed. It might still warrant consideration in
its own right as an "incidental finding," but is no different than a within-range outcome in answering the original
question.

How about the same blood-test, but with a different question? Suppose the doctor is trying to figure out why
you gained weight. The doctor knows that some people gain weight when their thyroid glands produce too little
thyroid-hormone. So the doctor's question is: Does this person have too little thyroid-hormone in the
bloodstream? This time, measuring the thyroid-hormone would provide a clean answer to the question if it was
lower than the normal range (yes) or within the normal range (no). Because elevated thyroid-hormone levels
do not usually cause weight-gain, a number higher than the normal range would produce an answer to the
original question no different than one that was within-range. (But even if this outcome was unexpected, it might
still be followed up.)

For some blood-tests the only meaningful result is in one direction. For example, a blood urea nitrogen (BUN)
measurement assesses kidney function. If the BUN measurement is too high, it could signify that the kidney is
impaired. But what does it mean if your BUN measurement is lower than the normal range? It means absolutely
nothing. It's a non-event. So then it's curious that a normal range for BUN even includes a lower number. How
did it get there?

These examples lead up to the question of how the normal ranges are created in the first place. They are
produced by statistics generated by measurements obtained in healthy volunteers. In the case of the BUN
measurement, for example, this substance might get measured in the blood of, say, 100 people without kidney
disease. An average number would be calculated by adding the numbers produced by all 100 people, and then
dividing by 100. This average would be the center of the normal range.

But the upper and lower numbers are produced by another method looking at how widely spread apart the BUN
measurements are in these 100 people. After all, it would be highly unlikely that all 100 people would produce
the exact same number-value. So how far from the average is still okay? The 100 measurements are plugged
into a mathematical formula to compute a "standard deviation," a widely-used statistic related to how widely the
numbers are spread apart. Numbers that are farther apart produce a larger standard deviation, while numbers
that are closer together produce a smaller standard deviation.

The next step is to decide how many standard deviations above and below average should be accepted as
normal. A typical choice for a blood-test is two standard deviations in either direction. It is known that
measurements falling within the range of two standard deviations above and below the average will include, on
average, 95% of the healthy people. It will also exclude or label as apparently abnormal the other 5% of healthy
volunteers. So if the "normal range" is generated in this fashion, one thing we already know is that it will be
wrong 5% of the time.

Another outcome of this statistical system is that if your doctor measures twenty different things in your
bloodstream, then, on average, one of them will be outside its normal range—
even if you are totally healthy!

Moreover, the statistical method automatically generates a lower and an upper value. This is done without
consideration of whether or not it means anything to be too high or too low. The normal range merely
expresses typical values obtained in healthy volunteers. It is up to your doctor to determine whether or not a
result outside the normal range is clinically meaningful. The "normal range" is not a judgment; it is merely a
statistical statement. You want the judgment to come from a trained clinician.

Your doctor has special words that are useful in navigating this intersection between statistics and clinical
meaning. One favorite word is "unremarkable." You might not like the idea of being described as unremarkable.
But this is a handy term that your doctor uses to lump together your laboratory results that fell within their
normal ranges with those that fell outside their normal ranges but were still considered benign. So in this case,
"unremarkable" is a good thing to be.


(C) 2005 by Gary Cordingley