Making Sense of Blood Work

What a reference range actually is

The interval printed next to your result is a reference interval, and it is built the same way almost everywhere. A laboratory measures the marker across a group it has defined as its reference population and publishes the middle band — conventionally the central ninety-five percent of that group. The range answers one question: what do results from these people usually look like.

That is descriptive statistics, not a verdict on health. It was not derived from outcomes and it makes no claim that people inside it are well or that people outside it are sick. Two things follow, and both matter more than they sound.

The first is arithmetic. If a range is the middle ninety-five percent of a healthy population, then by construction one result in twenty from a perfectly healthy person falls outside it. On one marker that is a small chance. On a panel of a few dozen, the chance that at least one comes back flagged is close to expected. A flag on a long panel is not, on its own, evidence of anything — it is the panel behaving as designed, which is why chasing every marked line is how a well person acquires a file of follow-up tests.

The second is about who was in the room. A reference range inherits its population, and that population is usually whoever the laboratory could recruit — a cross-section of a country, not a cross-section of the well. Where a condition is common the range moves with it: insulin resistance, a lipid pattern associated with cardiovascular risk, insufficient vitamin D through a winter, all of it sits inside the distribution the range was drawn from. “Within range” there means typical for the people you were compared to, which is a weaker claim than it looks.

One practical consequence: a result only means anything against the range printed on that report. Laboratories differ in assay, reference population and range, and some stratify by age or sex while others do not. Comparing a number from one laboratory against a range found somewhere else is a category error, not interpretation.

Optimal and normal, from both sides

Once you know a reference interval only describes a population, the case for reading results more tightly than the printed range almost makes itself — which is the argument the “optimal range” idea rests on. It deserves both a serious hearing and a hard look, because it is at once the thing that catches real problems early and the most common way a wellness clinic turns a healthy person into a patient. Both are true, and neither cancels the other.

The honest case for it. Some ranges are wide enough to hide genuine, early, treatable problems. Iron stores can be depleted a long way before a blood count shows anemia, so someone can be exhausted, losing hair, and technically normal. Thyroid ranges are broad, and the top of one is a different clinical situation from the bottom. Reading position within a range, not only membership of it, finds things a strict in-or-out reading will not. That is ordinary clinical reasoning, not fringe practice.

The honest case against it. The phrase carries an authority it has usually not earned. Ask where a particular optimal band came from and the answer is often an author, a conference slide, or a clinic's own house style rather than a study tying that band to an outcome. Narrow a range far enough and everyone is deficient in something, and the recommended response is reliably something the clinic can supply. The incentive runs one way: no business model depends on telling you that you are fine.

Two questions cut through most of it. Is that band tied to an outcome — symptoms that improve, events that become less likely — or only to an assertion of rigor? And is the proposal to treat a number, or to treat a person whose symptoms the number helps explain? Treating a number in someone who feels well is where this goes wrong, and it goes wrong quietly, because nothing about it feels reckless.

You will notice this page never says what a result should be. That is deliberate. Useful targets depend on age, sex, history, symptoms, what else is on the panel and what is being treated, and they are set by a clinician who has all of that in front of them. A universal target published for everyone tells you about the publisher, not about your physiology. Why two clinicians can read the same result differently goes further into this.

Why a number moves when nothing has changed

Two results from the same person a fortnight apart will not be identical, and the difference is usually not news. Every result carries two kinds of variation: analytical variation, which is the measurement itself — assay, instrument, laboratory — and biological variation, which is the marker genuinely moving within the person as a matter of normal function. For some markers the second is substantial.

The things that move a result are mostly mundane. Time of day matters a great deal for cortisol and for testosterone, both of which follow a daily rhythm; a morning draw and an afternoon draw are not comparable. Fasting state changes glucose, insulin and triglycerides, and “I only had coffee with milk” is enough to matter. Hydration shifts anything measured as a concentration, which is why a dehydrated draw can nudge a blood count upward. Recent illness, infection or vaccination moves inflammatory markers, and ferritin is an acute-phase protein that rises with inflammation regardless of iron stores — a high reading taken during a bad week can mean almost nothing. Hard training in the days before a draw affects muscle enzymes and can influence iron markers and a blood count. For anyone still cycling, where you are in the cycle determines what a reproductive hormone result even means. Even posture and how long the tourniquet was on have measurable effects.

Laboratories differ too. For some markers the assays are not standardized across them at all, so switching provider between two draws does not produce a trend. It produces two numbers that were never on the same scale.

So the habit worth building is to ask, before reacting: is this shift larger than the amount this marker moves on its own? Some are stable enough that a modest change is real; others swing so much between draws that only a large move means anything. A change inside the noise is not a trend, and treating it as one is how people end up adjusting a plan every few weeks in response to nothing.

A trend beats a snapshot

A single result is a point estimate of something that is always moving. Three results taken the same way over a couple of years show direction and speed, and direction is often the more interesting fact. A marker sitting comfortably in range but climbing steadily year on year can matter more than one mildly out of range that has sat there, unchanged, for a decade.

Getting a usable trend means controlling the things above. Same laboratory where possible, same time of morning, same fasting state, and not in the week after an illness or a hard training block. It costs nothing and it is the difference between a series you can read and one you cannot.

The corollary is that testing more often is not the same as knowing more. Most markers do not change meaningfully over weeks, and repeating them at that interval mostly samples the noise — producing a scatter of results, at least one alarming flag, and a decision made on a number that would have read differently on a different morning. Sensible intervals depend on the marker and on what is being treated. How Often Blood Work Is Actually Worth Repeating After 45 works through the ones that come up most.

The markers worth knowing about that standard panels often omit

A routine panel is built around a small set of common questions and does that job. Several markers with a genuine argument behind them are simply not on it. What follows is what each one is for, not what yours should say.

ApoB and Lp(a), for cardiovascular risk. A standard lipid panel reports how much cholesterol is being carried; ApoB counts the particles carrying it, which is closer to what interacts with an artery wall. The two can disagree in ways that change how a person is classified. Lp(a) is a distinct particle, largely inherited and barely moved by diet or exercise — which is why it is worth measuring once. It either is or is not part of your risk picture. What each one tells you sets them side by side.

Fasting insulin and HOMA-IR, for metabolic health. Glucose is held steady for a long time by rising insulin, so it can look unremarkable for years while the effort of keeping it there climbs. Measuring insulin alongside glucose — and combining the two, which is what HOMA-IR does — shows the effort, not only the result.

hs-CRP, for inflammation. A high-sensitivity C-reactive protein reading detects low-grade inflammation an ordinary CRP would not resolve. Its weakness is that it is entirely non-specific — infection, injury and a great many other things raise it — so a single elevated reading during a bad month is not a finding. Repeated away from acute illness, it carries more.

Ferritin, for iron status. Ferritin reflects iron stores, and stores fall before a blood count does, which is why low ferritin can explain fatigue and hair shedding in someone whose count is normal. It cuts both ways: ferritin also rises with inflammation, so a high reading is not automatically iron overload.

The full thyroid picture rather than TSH alone. TSH is a pituitary signal — a measure of what the brain is asking the thyroid for, not of what the thyroid delivers or the tissues receive. It is a good first-line test and a poor complete one. Free T4 and free T3 describe the hormones themselves, and TSH can read unremarkably while the wider picture does not.

One caution, because this is easy to misread as a shopping list. More markers is not automatically better information. Every added test is another chance of an incidental flag, and some flags start a cascade of scans and repeat draws with its own cost and anxiety. These markers have an argument behind them. That is not an argument for testing everything measurable, which is a different proposition sold in the same language.

When every lab is normal and you still feel unwell

This is one of the most common situations in midlife medicine and it deserves a straightforward answer rather than a mystical one. Being told everything came back normal while still sleeping badly and running out of energy by mid-afternoon is not a contradiction needing an exotic explanation. There are several ordinary ones.

The relevant marker may not have been measured. A routine panel is a small selection. If nobody looked at iron stores, at the thyroid beyond a single signal, or at insulin, those results are not reassuring — they are absent.

The result may be technically normal and personally unusual. A wide reference interval accommodates a lot of movement. Someone who sat near one end of a range for years and has drifted to the other end never left it, and has changed considerably. Only a previous result makes that visible.

The cause may not be a blood marker at all. Much of what makes people feel badly in midlife does not appear in serum. Obstructive sleep apnea, common and very often undiagnosed. Depression and anxiety. Chronic pain, and the sleep it costs. Deconditioning. Alcohol. Side effects of a current prescription. Perimenopause, which is diagnosed from symptoms and pattern rather than a hormone level, because those hormones swing so widely during the transition that one measurement can mislead.

The timing may have been wrong. A morning-rhythm hormone drawn in the afternoon, or a reproductive hormone drawn at an arbitrary point in a cycle, is accurate and uninterpretable at once.

What it does not mean is that it is in your head — and it does not mean the answer is an unusual test or a stack of supplements. Those are the two failure modes either side of this, and the second is sold far harder. The productive next step is usually a careful history, the pattern over time, and a few specifically chosen additions — not another broad panel, and not a protocol. When Every Lab Is Normal and You Still Feel Wrong takes the same question article-length.

What blood work cannot do

Blood work is a good tool regularly asked to do a job it cannot do, so the boundaries are worth stating plainly.

It cannot diagnose most of what makes people feel unwell. What most commonly accounts for fatigue, low mood and poor concentration in midlife is mostly not visible in serum. A clean panel narrows the list; it does not close it.

It cannot tell you how you feel. Symptoms are data, and not a lesser class of it than a number. A result that does not match the symptom is a reason to look further, not a reason to discount the person.

It cannot predict an individual future. Risk markers are statements about probability across populations. A favorable lipid profile is not a guarantee and an unfavorable one is not a sentence; both shift the odds, which is useful and is not a forecast.

It cannot interpret itself. The same numbers mean different things in someone who has just finished a hard training block, a woman mid-transition, and someone recovering from a virus. Context is not a nicety layered over the result. It is most of the interpretation.

And it cannot, by itself, justify a treatment. A number is not a symptom. The decision to treat comes from the whole picture — history, symptoms, risk, preference, and what the result adds to it. That is the standard worth holding anyone who reads your labs to, including us.

Frequently asked questions