It’s 10 at night and your phone buzzes. New lab results. You open the portal and scan down the column, and one line is in red. Your ALT, a liver enzyme, came back at 48. The range printed beside it stops at 44.
Four points over. You are now awake until 1 a.m.
Nobody ever explains where that 44 came from. So here it is. At some point, a laboratory measured that enzyme in a group of people it had decided were healthy, lined up the results, and drew two lines: one that cut off the lowest 2.5% of that group and one that cut off the highest 2.5%. What’s left in the middle, the central 95%, became the range printed on your report.
Which means the range was built to leave healthy people on the outside. One in 20 of them, by design. If you are the one in 20, your report says the same word it would say if something were wrong.
Where the two numbers in the margin come from
MedlinePlus, the consumer service run by the National Library of Medicine, puts the construction plainly: reference ranges are based on test results from large groups of healthy people. Some tests carry separate ranges for separate groups, children and adults being the obvious split.
The deeper history is more interesting, and it explains why the word “normal” causes so much trouble. In a 2018 paper in Postgraduate Medical Journal, Martin Whyte and Philip Kelly trace the term “reference interval” to Grasbeck and Saris, who introduced it in 1969 specifically because the older phrase “normal range” was considered flawed. The point of the newer term was to give a clinician something to compare a result against, not to define what a normal human being looks like.
Whyte and Kelly also deal with the 95% convention head on. There’s no biological theory saying the middle 95% of a healthy group is the physiologically correct slice. The figure is pragmatic. Two standard deviations from the mean was judged to be far enough from the center, an idea borrowed from statistical hypothesis testing rather than from physiology. You could draw the lines at the central 90% instead, which would flag more healthy people. You could draw them at 99%, which would miss more disease. The 95% version is a compromise, and like every compromise it gets some things wrong on purpose.
They raise a second problem. The whole approach assumes results pile up in a bell curve, and plenty don’t. Fasting triglyceride is lopsided, with its most common value sitting off to one side rather than in the middle. And the phrase “normal distribution” itself is a historical accident. The statistician Karl Pearson adopted “normal” because the bell-shaped pattern turns up everywhere in nature, not because it identifies normal people.
MedlinePlus doesn’t dodge this. It states that “the term ‘normal’ can be misleading,” and says outright that “it’s common for healthy people to sometimes have results outside the reference range.”
Run enough tests on a healthy person and something gets flagged
This is the part almost nobody is told, and it’s just arithmetic.
If a range is drawn so that 19 out of 20 healthy people land inside it, then each individual test carries roughly a 1 in 20 chance of flagging a healthy person. That’s 5%. Fine for one test.
Now count what’s on your panel. A comprehensive metabolic panel, one of the most commonly ordered blood tests in the country, measures 14 separate substances, according to MedlinePlus. Glucose, creatinine, several electrolytes, two or three liver enzymes, a couple of protein measures and so on. Fourteen chances to land outside a line.
The chance that all 14 come back inside their ranges is 0.95 multiplied by itself 14 times, which works out to about 0.49. So the chance that at least one comes back flagged is roughly 51%.
A coin flip. On a 20-item panel it climbs to about 64%. Add a lipid panel, a thyroid test and a vitamin D level to your annual physical and you are, statistically speaking, shopping for a red line.
One honest correction, and Whyte and Kelly make it themselves: this math assumes each test is independent of the others, and they aren’t. Liver enzymes move together. Electrolytes move together. Because the results are correlated, the real-world chance of at least one flag is somewhat lower than the raw calculation suggests. The direction of the point survives the correction. The more things you measure in a healthy person, the more likely it becomes that at least one of them falls outside a boundary that was drawn to exclude healthy people in the first place.
So the flag is the lab doing its job correctly. It means “this value sits outside the middle of our healthy sample.” It does not mean “something is wrong with you.”
Your lab’s range is not the lab across town’s range
Two labs can measure the same vial of your blood and report different numbers, with different ranges, in different units. Neither of them is broken.
MedlinePlus is unambiguous about the consequence. Because labs use different testing methods, they publish different ranges, and you cannot compare results across labs. If you’re tracking something over time, try to use the same lab every time.
There’s hard evidence behind that advice. A 2019 study in Clinical Chemistry by Monsurul Hoq and colleagues took serum from healthy newborns and children, split each sample, and ran the pieces across five different types of analyzer. They reported reference values for 30 common biochemistry measurements and quantified exactly where results from one child could and could not be compared across machine types. For most of what they looked at, the variation between analyzers stayed inside acceptable error limits, but the study exists because the question is real and the answer is not “it doesn’t matter.”
Labs are regulated, which is worth knowing even though regulation doesn’t make two labs interchangeable. The Clinical Laboratory Improvement Amendments, usually called CLIA, give the federal government authority over certification and oversight of clinical laboratories, and a lab needs the right certificate before it can accept a human sample at all. Three agencies split the work. The Centers for Medicare and Medicaid Services issues the certificates, runs inspections and enforces compliance. The FDA categorizes tests by how complex they are. That leaves the CDC, which writes laboratory practice guidelines, sets technical standards, and runs the federal advisory committee on lab improvement. The FDA’s page on this was current as of July 2023.
What that means for you is narrow and practical. If your primary care office sends blood to one lab and a specialist sends it to another, a change between the two visits may be the machine and not your body. Check the lab name at the top of each report before you conclude anything moved.
What moves a result besides disease
MedlinePlus lists several ordinary causes of an unexpected number. Here they are grouped by when they happen, which is more useful than an alphabetical list.
The day before. A hard workout can shift results. So can what and when you ate and drank.
What you take. Prescription drugs, over-the-counter medicines, vitamins and supplements can all change a value. Tell whoever orders the test about everything you take. Then do not stop any of it on your own. That call belongs to your provider, and MedlinePlus is firm about it.
Where you are in a cycle. Having a menstrual period at the time of the test affects certain results.
Whether you followed the prep. A fast means water only, and the length of it is set by the test, which is why asking how many hours beats guessing. Fluids cut both ways: some blood draws go better after extra water, and certain urine collections come with their own instruction to drink a short while before you arrive. MedlinePlus warns that even a small departure from the instructions can have a large effect on results. If you broke the prep, say so before the draw.
There’s a subtler source of movement, and it runs the other way. Whyte and Kelly point out that variation within one person is usually much smaller than variation across a whole population. Your personal range is narrower than the printed one. So a result can sit comfortably inside the reference interval and still represent a real change for you. Their example: a man whose hemoglobin reads 130 is inside a range that runs from 130 to 180, but if his usual value has always been 170, something happened.
That idea is badly underused in practice. A 2020 survey in Biochemia Medica by Humeyra Ozturk Emre and colleagues questioned 399 clinicians about variation in lab results. About 60% understood the variation that comes from sample handling and from the analyzer. Only 3.5% had working knowledge of biological variation, the natural fluctuation within an individual. Not one of them reported using it to interpret results. Yet 79.4% agreed that a difference between two results, both sitting inside the reference interval, could still be significant. It’s a single-country survey, so treat the exact percentages as local rather than universal. The gap it describes is familiar to anyone who works in a lab.
A range is not a threshold, and the difference is everything
This guide covers the distinction that resolves most confusion about lab reports, and almost no one explains it.
Some numbers printed beside your result are reference intervals, built the way described above from a sample of healthy people. Others are decision limits, chosen by expert consensus because of what happens to people on either side of a particular number. They look identical on the page. They mean completely different things.
Hemoglobin A1C, the three-month blood sugar average, is the cleanest example. The National Institute of Diabetes and Digestive and Kidney Diseases gives three bands: below 5.7% is normal, 5.7 to 6.4% is prediabetes, and 6.5% or above is diabetes. Those boundaries are not the edges of a healthy sample. They’re lines drawn where risk and treatment benefit change.
NIDDK adds several conditions that readers should know about. A diagnosis requires confirmation with a second measurement, taken on a different day, unless there are clear symptoms. The test should be run with a method certified by the National Glycohemoglobin Standardization Program, and point-of-care finger-stick versions should not be used to diagnose. And A1C can give false readings in specific situations: very low iron, as in iron-deficiency anemia, can push the number falsely high, hemoglobin variants more common in people of African, Mediterranean or Southeast Asian descent can interfere with some versions of the test, and kidney failure or liver disease can throw it off. That NIDDK page was last reviewed in April 2018, so confirm current practice with your clinician.
The practical consequence: a value can sit inside a reference interval and still be on the wrong side of a decision limit, or sit outside a reference interval while being clinically unremarkable. Asking “is this number a population range or a treatment threshold?” is one of the most clarifying things you can say to a clinician.
Be careful with “optimal range”
A familiar pattern in consumer health: a company sells a test, prints your result against a narrower “optimal” band of its own invention, and marks you suboptimal on a value your actual lab reported as normal. The recommended correction is frequently a product the same company sells.
Narrowing a band isn’t automatically wrong. The A1C thresholds above are narrower than a healthy population’s spread, and they’re narrower for good, published reasons about who develops complications. The question is whether the narrower number comes attached to evidence about health outcomes, or just to a cart.
Two facts help you judge. First, as Whyte and Kelly explain, there is no underlying theory that makes the standard 95% interval physiologically correct, so a different band isn’t heretical on its face. It just needs a reason. Second, the FDA states that tests sold for non-medical purposes, general wellness or low-risk medical uses generally are not reviewed by the agency before they go on sale, and that direct-to-consumer tests are not a substitute for a regular health care evaluation. The FDA’s consumer page on this is dated December 2019.
When the same party defines the range, sells the test and sells the fix, that’s three roles that normally sit with three different parties. Name it for what it is and ask for the evidence. Consumer lab services deserve their own full treatment, and they get one elsewhere in this series.
How to read your own lab report, line by line
Seven steps. Do them in this order.
-
Start at the top, not at the red line. Write down the collection date and the name of the laboratory. Those two details are what make today’s number comparable to next year’s.
-
Read across, never down. For each result, the range in the adjacent column is the one that applies to that measurement, from that lab, on that day. A range you saw on a different report, or on a website, is not yours.
-
Check the units. The same substance reported in different units produces numbers that look dramatically different. A number that seems alarming next to a figure you remember is often the same value in different clothing.
-
For anything flagged, measure the distance. One point past the line and double the upper limit are not the same event and shouldn’t produce the same reaction.
-
Find your own previous value for that same test from that same lab. Direction and size of change usually carry more information than the single number does.
-
Write down what you did before the draw. Fasting or not, workout, illness, a new supplement, time of day. Put it next to the result while you still remember.
-
Bring the whole page. Not a photo of the one red line. Clinicians read the pattern across values, and a single flagged enzyme means something different depending on what the other 13 did.
When a flagged value shouldn’t wait
Slowing down is the right default. These are the exceptions.
Call the same day if someone from the office phones you rather than posting the result. Someone made a judgment that it shouldn’t sit in a queue.
Call if a flagged value comes with symptoms. The symptom is the reason to call, not the number. Anything new, worsening or interfering with daily function counts.
Call if the value is far outside the range rather than just over the edge, or if several values in the same system are flagged together. A group of abnormal liver or kidney markers is a pattern, and patterns carry more weight than any single flag.
Call if a value keeps moving in the same direction across repeat draws at the same lab, even when each individual result stays inside the range. That’s the situation reference intervals are worst at catching.
And one in the opposite direction. MedlinePlus notes that a result inside the range isn’t a guarantee of good health, and that you may still need more testing, particularly if you have symptoms. A clean panel is not a reason to stop describing what’s happening to you.
The one thing to take away
The reference range is a description of a sample of other people. It is not a statement about you.
When a value gets flagged, two questions tell you nearly everything you need: how far outside the line is it, and which way is it moving compared with your own earlier results from the same lab. A number that’s barely out and hasn’t budged in three years is a very different object from one that’s doubled since last spring, even if both print in red.
And don’t let the red line drive a purchase. A flagged value is a prompt for a conversation, not a diagnosis, and certainly not a shopping list.
Sources
- MedlinePlus, National Library of Medicine. “How to Understand Your Lab Results.” medlineplus.gov
- MedlinePlus, National Library of Medicine. “Comprehensive Metabolic Panel (CMP).” medlineplus.gov
- MedlinePlus, National Library of Medicine. “How to Prepare for a Lab Test.” medlineplus.gov
- MedlinePlus, National Library of Medicine. “What You Need to Know About Blood Testing.” medlineplus.gov
- National Institute of Diabetes and Digestive and Kidney Diseases. “The A1C Test & Diabetes.” niddk.nih.gov
- U.S. Food and Drug Administration. “Clinical Laboratory Improvement Amendments (CLIA).” fda.gov
- U.S. Food and Drug Administration. “Direct-to-Consumer Tests.” fda.gov
- Postgraduate Medical Journal (PubMed). “Whyte MB, Kelly P. "The normal range: it is not normal and it is not a range." Postgraduate Medical Journal, 2018. PubMed. DOI: 10.1136/postgradmedj-2018-135983.” doi.org
- Clinical Chemistry (PubMed). “Hoq M et al. "Reference Values for 30 Common Biochemistry Analytes Across 5 Different Analyzers in Neonates and Children 30 Days to 18 Years of Age." Clinical Chemistry, 2019. PubMed. DOI: 10.1373/clinchem.2019.306431.” doi.org
- Biochemia Medica (PubMed). “Emre HO et al. "Utilization of biological variation data in the interpretation of laboratory test results – survey about clinicians' opinion and knowledge." Biochemia Medica, 2020. PubMed. DOI: 10.11613/BM.2021.010705.” doi.org