The NORMAL RANGE for most blood tests is defined as the range where 95% of healthy people are. This means that, statistically, one of every twenty tests done on a healthy person can be expected to fall outside this range. A number outside the normal range can be the sign of illness or simply be due to “statistics”. We have to look for patterns, such as which other tests are outside the range, and how the test values change over time in order to decide if an abnormal blood test needs to be pursued or not. It is important to keep in mind that in most cases an abnormal result does not mean there is something wrong.
Ideally we would like to see healthy people in a range that never overlaps the range of people with disease, but things don’t usually work that way in medicine.
We can set arbitrary boundaries, such as with blood sugar. One year the cutoff between diabetes and no diabetes was 140 mg/dl (7.8 mmol/L) and the next it was 126 mg/dl (7 mmol/L). Soon the U.S. will follow those countries that have dropped that level to 115 mg/dl (6.4 mmol/L).
Does such a definition really mean that a person just below the limit is a true non-diabetic and a person just over it is a full-blown diabetic? I certainly wouldn’t treat them completely differently based on such a definition, even though one technically is a diabetic and the other is not.
It gets more complicated when you look at a test like PSA, Prostate Specific Antigen. We use this test when looking for prostate cancer. A “normal” PSA level is 0 – 2.5 under age 45 and 0 – 4 in older men, although many urologists allow for levels of 6 or greater with increasing age. The problem is that cancer can be found well down into the normal range, while levels of 10 or more can be seen in older men with harmless, age related enlargement of the prostate. Change over time, here too, can be an indicator of when to worry and when not to.
PSA screening is a very good example of the concepts of sensitivity and specificity.
Sensitivity is the likelihood that a disease will show up as an abnormal test result. A test with 100% sensitivity will identify every sick person, and therefore will have no false negative results.
Specificity is the likelihood that a negative test really means that the person is actually healthy.
Imagine we have a test with 90% sensitivity and 90% specificity.
90% of all sick people would have an abnormal test, while 10% have a false negative test. At the same time, 90% of healthy people would have a normal test, but 10% would have a false positive test.
This may sound fairly good, but it isn’t. Imagine that 1% of people have a bad disease, and our test for the disease has 90% specificity and 90% sensitivity. Then out of 1000 people, there would be 10 sick ones and 990 healthy ones. Of the 10 sick ones, 9 would have an abnormal test and one would have a normal, false negative test. Out of the 990 healthy people, 99 would have an abnormal, false positive test. 891 healthy people would have a true normal test.
In this scenario, 9 sick people and 99 healthy people would have an abnormal test. The odds that a positive test would mean the person was sick would be 9/108, or 8.3%. This is called Positive Predictive Value. In other words, a bad test means no disease more than 90% of the time. That simply isn’t good enough for a screening test.
Let’s make the sensitivity and specificity in our example both 99%. Then all the sick people would have an abnormal test, so we would find them all, but of the 990 healthy ones, 10 would have an abnormal test. The odds that an abnormal test would mean we had found a sick person would be 10 in 20, or 50%.
We would still need good backup tests and a good dose of medical judgement to make a definitive diagnosis. Tests alone rarely do the job.