Medical laboratory tests are an important feature in the treatment of many conditions. In psychiatry we often need to measure the metabolic effects of some medications, the effect and blood concentration values of lithium and others, the relationship of thyroid hormone levels to depression and many others. We look to see if a measured value in a patient is “normal” or not. “Normal” means falls with a “reference range,” meaning above a lower value and below an upper value. “Abnormal” are values that fall above the upper value—too high—or below the lower value—too low.” This is usually translated, communicated and understood as meaning, for example with respect to the TSH test (Thyroid Stimulating Hormone) either, “Your thyroid levels are too low,” “Your thyroid levels are normal” or “Your thyroid levels are too high.”
However, with respect to any given test, this is inaccurate for as many as 5% of people—one person out twenty, often with significant clinical implications. This is because reference ranges are not absolute boundaries but statistical ones based not a black and white definition of “abnormal” but on a “distribution” of clinically normal and abnormal values.
What is a Distribution?
In statistics, a distribution portrays the frequency of various outcomes within a data set. It is basically just a list of different values, in order, along with the number of times each value occurs, like this:
| Value | N |
| 0.0 | 0 |
| 1.0 | 1 |
| 2.0 | 1 |
| 3.0 | 4 |
| 4.0 | 10 |
| 5.0 | 12 |
| 6.0 | 7 |
| 7.0 | 6 |
| 8.0 | 3 |
| 9.0 | 1 |
| 10.0 | 1 |
| 46 Total |
This might represent an experiment where 46 people without X-related disease had their blood drawn and the level for a protein called Y is measured. People must have a certain amount of Y to be healthy. If they have too little they develop “hypoX” and if too much, “hyperX.” A real world example of this would be thyroid hormone levels in people.
In the above graph, we see that :
- the most common value of the marker in healthy people is 50.0 with 14 of 55 people having that value (25%).
- 40.0 and 60.0 are common as well—amounting to 36% of the test population.
- We observe, then, that 25% + 36% of the test population = 61% have X values between 40% and 60% inclusive.
- 30.0 and 70.0 amount to another 15% of the test population, so that 25% + 36% +15%= 76% of the test population have values between 30% and 70% inclusive.
- 7 people have values 20.0 or 80 (13%) so that 89% of the values lie between 20% and 80% inclusive.
- Finally 3 people have values 10.0 or 90.0 (5%) so that 95% of the population have values that lie between 10.0 and 90.0 inclusive and this defines “normal” because “normal” is certain kind of bell-shaped distribution of values.
We now perform a second test on people all of whom have hypoX and a third test on people with hyperX. We find similarly roughly bell-shaped distributions in these tests as well.
In the end, we discover from these the statistical range of values typically found in people who have no X-related illness, people with hypo-X and people with hyper-X.
The way statistics define matters (the definition is a matter of choice, it isn’t found “in nature” as it were) “in a group of normal people without X, 95% of them have values between 10.0 and 90.0,” which we call the reference range.
However, in above chart look at the value 100.0. There are 2 people (4%) who have no X-related illness but they have values that are very high and that fall within the typical range for people with hyper-x. In fact for every laboratory value, because of how the reference range is defined, for any condition X:
- There will be a small number of healthy people—5% or one out of twenty—with lab values that are defined as indicating disease
- There will be a small number of ill people—5% or one out of twenty—with lab values that are defined as indicating no disease
- 5% or one out every twenty laboratory values are expected to be abnormal just by chance.
For laboratory values in a healthy population, these measurements typically form what is known as a “normal distribution” or “bell curve.” A smooth curve arises when the number of healthy individuals tested run into the hundreds or thousands. This curve shows that most values cluster around the average or median, while fewer values fall towards the outer “tails”.
Key Points
Individual Variability: Some healthy people fall outside a reference range due to age, sex, race, diet, exercise habits, and genetic dispositions.
Context is Crucial: Simply because a lab value falls outside the reference range doesn’t necessarily mean sickness. Similarly, values within the range don’t guarantee health. Laboratory values must be interpreted in a context that includes symptoms, signs, history, other diagnostic information and an understanding that the laboratory value boundaries are clinically fuzzy.
Customized Care: Treatment and monitoring must be tailored to each individual’s unique baseline, rather than to the generalized reference range.
In summary, medical lab test reference ranges are guides and not outright indicators of either health or disease. They are statistical tools that, while invaluable, require interpretation within the broader context of an individual’s overall condition and circumstances.