You weren’t told you’re healthy. You were told you’re average.

“Normal” is a percentile, not a verdict

A reference range means your result landed inside the middle 95% of whatever population the laboratory used to build it. Not that you’re well. Not that you’re where you should be. Just that you’re near the middle of the pack.

The obvious question, which almost nobody is invited to ask: near the middle of which pack?

Reference intervals are built by taking results from a reference population and keeping the central 95% — the 2.5th percentile to the 97.5th. That’s the standard method, and it’s applied honestly. But notice what it doesn’t involve. Nobody followed those people for twenty years to see who stayed healthy. No outcomes were measured. The range describes where most people’s numbers land, and stops there.

Two consequences follow, and neither is usually explained. Five percent of completely healthy people fall outside the range by arithmetic — run twenty tests on a well person and one comes back flagged, on average. And the range is only ever as healthy as the population behind it.

The mirror moves

Reference populations are assembled from whoever is available. People who showed up at the lab. Blood donors. Participants in national health surveys. These groups are screened, at best, for the absence of a diagnosed disease — not for health.

The American adult population is not metabolically well. By CDC estimates, a large majority of adults carry excess weight, and a substantial share meet criteria for prediabetes — most of them undiagnosed, and therefore fully eligible for a reference population.

Testosterone: watch the floor move

For years, laboratories published lower limits for total testosterone that varied widely — many near 250 ng/dL or below, drawn from mixed-age, mixed-health male populations.

In 2017, Travison and colleagues built harmonized reference ranges from a deliberately selected group: non-obese, non-diabetic, healthy men aged 19 to 39, across four large cohorts. The result was 264 to 916 ng/dL.

Take out the men with obesity and diabetes, and the distribution shifts. That’s the whole argument in one paper. A man sitting at the bottom of an older range was being measured against a group that included a great many men who were not well. He was told he was normal. He was told the truth — it just wasn’t the information he came for.

TSH: the range contains the disease

The standard range for thyroid-stimulating hormone runs roughly 0.45 to 4.5 mIU/L. Analysis of NHANES III data examined what happens when you remove people with thyroid antibodies, known thyroid disease, goiter, or family history from the reference population. In that disease-free group, the distribution is substantially narrower, with the large majority well below the standard ceiling.

Read plainly: part of what defines the top of the “normal” thyroid range is people with early thyroid disease. They were in the population. They hadn’t been diagnosed, so they weren’t excluded. And their numbers helped set the ceiling you were measured against.

The test that was never run

Here’s the part that costs people the most years, and it has nothing to do with reference ranges at all. You cannot be out of range on a test nobody ordered.

A standard annual panel is a CBC, a metabolic panel, a lipid panel, and if you’re lucky, a TSH. It does not include fasting insulin, free testosterone, SHBG, ApoB, Lp(a), hs-CRP, ferritin, or a full thyroid panel. Every one of those can be badly wrong while your standard panel comes back spotless.

Fasting insulin is the one that should make you angry. Insulin resistance develops years — often a decade — before glucose becomes abnormal. The pancreas compensates by producing more insulin, and it succeeds for a long time. Fasting glucose and HbA1c stay normal through that entire window, because they measure the outcome the compensation is busy preventing. Measure insulin and you see the compensation itself, while it’s still reversible.

Fifteen minutes is a billing interval

The typical primary care appointment runs about fifteen minutes. That number wasn’t arrived at clinically. Nobody studied how long it takes to understand a person’s metabolic health and concluded the answer was a quarter of an hour. It’s an economic artifact — the interval at which a practice covers its overhead under current reimbursement.

In fifteen minutes a physician has to review your history, address what you came in for, document to a standard that satisfies a payer, and code the visit. What doesn’t fit is a conversation about why you feel worse at 44 than you did at 38 when every number on the page is inside its brackets.

This is not a failure of the person in front of you. Most primary care physicians are excellent, working inside constraints they didn’t design and can’t fix from the exam room. The system is doing exactly what it was built to do: find disease, efficiently, at scale. It’s very good at that. It was never built to answer the question you actually walked in with.