Am I Underweight? What the Question Is Really Asking

What kind of thing a cut-off is

The 18.5 line was drawn by a World Health Organization expert committee whose report, published in 1995, was about anthropometry: how to measure bodies at scale and how to interpret the measurements for public health purposes. The classification it produced has eight tiers running from severe thinness below 16.0 up through class III obesity at 40 and above. Its function is to let a health service say how many people in a district sit below a line, and to tell a clinician when a further look is worth the time.

It follows that the line has no width and no meaning at its own edge. At 1.75 m, a BMI of 18.5 corresponds to 56.7 kg and a BMI of 18.3 to 56.0 kg — seven hundred grams apart, which is less than the difference between morning and evening on the same day, or between one pair of shoes and another. In feet and inches, at 5 feet 9 inches the line falls at 125.3 lb and a BMI of 18.3 at 124 lb: about a pound and a third. A single pound moves the reading by 0.148 units. Nobody's physiology changes at 18.5, and no calculator, including ours, can pretend otherwise.

What the mortality data below the line actually show

The largest analysis of the question pooled individual data from 239 prospective studies on four continents — 10.6 million adults in total. To reduce the two biases that wreck this literature, smoking and illness that causes weight loss, the primary analysis kept only never-smokers with no chronic disease at entry and at least five years of follow-up, leaving 3.95 million people. In that restricted group, and against a reference band of 22.5 to 25.0, the hazard ratio for a BMI of 18.5 to 20.0 was 1.13 and for 15.0 to 18.5 it was 1.51. The band of lowest risk was 20.0 to 25.0.

Two things follow, and they pull in opposite directions. The association is real and it survived a serious attempt to remove reverse causation, so it is not simply an artefact of sick people being thin. But an association measured across millions of person-years is a statement about a population, and a hazard ratio of 1.13 is a small effect on a small absolute risk. Neither of those sentences is a statement about you, and neither of them tells anyone what to do — which is the whole reason the clinical criteria in the next sections exist.

Blind spot one: what the body is made of

BMI is a mass divided by a squared height. It has no term for what the mass is. That is why clinical criteria for malnutrition do not stop at BMI: the GLIM consensus lists reduced muscle mass as a separate phenotypic criterion, and quotes the ESPEN fat-free mass index thresholds of below 17 kg/m² for men and below 15 for women, measured by DXA or bioimpedance rather than by a scale.

The arithmetic makes the gap visible. Take three people all 1.75 m tall and all 56 kg, so all with the same BMI of 18.3 and all classified the same way. At 8% body fat, fat-free mass is 51.5 kg and the fat-free mass index is 16.8. At 15%, it is 47.6 kg and 15.5. At 25%, it is 42.0 kg and 13.7. Against the ESPEN thresholds those three identical BMI readings sit on opposite sides of the line depending on which criterion is applied and to whom. BMI cannot distinguish them; a body-composition measurement can.

Blind spot two: the direction, which is the informative part

This is the part of the topic that a single reading structurally cannot reach, and it is where the citable clinical content lives. The GLIM criteria, agreed between the major clinical nutrition societies and published in 2019, treat weight loss as a phenotypic criterion in its own right: more than 5% within the past six months, or more than 10% beyond six months. The severity grading splits it further — 5 to 10% within six months or 10 to 20% beyond is stage 1, moderate; more than 10% within six months or more than 20% beyond is stage 2, severe. The MUST screening tool, quoted in the same consensus, uses 5 to 10% over three to six months for medium risk and more than 10% for high risk.

Work an example and the contrast is stark. A person 1.75 m tall who weighed 70 kg has a BMI of 22.9 — comfortably in the middle of the normal range. Six months later they weigh 66.5 kg. Their BMI is now 21.7, still normal, still nowhere near 18.5, and a calculator asked "am I underweight?" would say no. But they have lost 5% of their body weight, which is exactly the GLIM phenotypic criterion, and if there is also an inflammatory or intake-related cause the diagnosis of malnutrition can be made. In pounds: 154 lb falling to 146.3 lb, a BMI moving from 22.7 to 21.6. Two normal readings, one flag. The flag is the change.

Blind spot three: the same line does not fit everyone

The clinical criteria say so explicitly. GLIM's low-BMI threshold is below 20 for people under 70 and below 22 from 70 onwards — the line moves with age, because the relationship between body mass, muscle and outcome changes as people get older. GLIM also gives separate Asian values, below 18.5 under 70 and below 20 above it. And the WHO's own 2004 expert consultation on Asian populations concluded that the risk of type 2 diabetes and cardiovascular disease is substantial at BMIs below the existing international cut-off, identified additional public health action points at 23.0, 27.5, 32.5 and 37.5, and nevertheless kept the international classification in place for comparison purposes. That combination — keep the line for counting, move it for deciding — is the clearest statement anywhere of what these numbers are and are not.

What a low reading is a reason to do

One thing, and it is not a diet plan. It is a conversation with a doctor — particularly if the number has moved, if it moved without you intending it, or if anything else has changed alongside it. Unintentional weight loss is a symptom, and the list of things that produce it is long and includes thyroid disease, coeliac disease, inflammatory bowel disease, diabetes, infections, malabsorption, medication effects, depression and malignancy. None of those is diagnosed by a BMI, and none of them is treated by eating differently on the advice of a web page.

And there is a second reason people arrive at a page like this one, which deserves saying plainly rather than in a footnote. If food, weight or your body takes up a large part of your thoughts, if you are restricting, purging or exercising to control them, or if a number on a screen has the power to change how your day goes, that is worth telling a doctor or an eating disorder service about. Eating disorders are common, they occur at every body weight including a completely ordinary one, they are treatable, and outcomes are better the earlier treatment starts. Nothing on this page is a substitute for that conversation.

Frequently asked questions

My BMI is 18.3. Is that a problem?

A BMI cannot answer that, because the answer depends on things it does not measure. What a clinician would want to know is whether the number has changed, how fast, whether it was intentional, what your muscle mass looks like, whether you have symptoms, and what is normal for you historically. Those questions have answers; a single reading does not. If the number has fallen without you meaning it to, that is the part worth taking to a doctor.

Why do clinical criteria use 20 rather than 18.5?

Because they are answering a different question. The WHO classification is built to describe populations, where a conservative line keeps the count meaningful. GLIM is built to catch malnutrition in people who are already in a hospital or a clinic, where the cost of missing someone is high — so its low-BMI threshold sits at 20 under 70 years and 22 from 70, and it is only one of three phenotypic criteria, any one of which can be met.

Can someone be underweight by BMI and perfectly well?

Yes. A cut-off drawn across a population necessarily includes people for whom nothing is wrong — that is what a screening threshold is: a filter tuned so that few genuine cases are missed, at the price of flagging many people who turn out to be fine. Some adults sit naturally and stably below 18.5, have done so all their lives, eat normally and have no symptoms. The clinical criteria exist precisely to separate that situation from the ones that need attention, and they do it by looking at change, composition and cause rather than at the level alone.

Does frame size change the answer?

It changes the interpretation, not the classification. BMI has no frame term, so two people of the same height and weight get the same reading whatever their skeleton. The classical elbow-breadth and wrist-circumference methods are an attempt to add one, and we cover what they assume in a separate article; they are approximations built on old reference tables and they do not turn a screening number into a diagnosis. The measurements that clinicians reach for when the frame question matters are body composition and muscle mass, not a frame category.

How quickly does weight loss become clinically significant?

The published thresholds are percentages of body weight over a stated period, not absolute amounts. GLIM uses more than 5% within the past six months or more than 10% beyond six months as a phenotypic criterion, and grades more than 10% within six months or more than 20% beyond as severe. The MUST screening tool, quoted in the same consensus, uses 5 to 10% over three to six months for medium risk and more than 10% for high risk. All of those can be met at a completely normal BMI, which is the point.