Where 18.0182 comes from
Glucose has a molar mass of about 180.16 grams per mole. A millimole per litre of glucose is therefore 180.16 milligrams per litre, which is 18.0182 milligrams per decilitre. Dividing mg/dL by 18.0182 gives mmol/L; multiplying goes the other way. Many sources round the factor to 18, which is fine for mental arithmetic and introduces an error of about one part in a hundred — enough to matter only when you are reconciling two printed values digit for digit.
The unit split is geographic rather than technical. The United States, and a handful of other countries, report glucose by mass. Most of the rest of the world reports it by amount of substance, which is the SI convention. Continuous monitors and phone apps generally have a unit setting, and the two scales are far enough apart that a mix-up is usually obvious: a value of 7 is a plausible mmol/L reading and an implausible mg/dL one. The one that catches people is A1C, which is a percentage and shares its numeric range with mmol/L readings.
A1C and the estimated average
A1C measures the fraction of haemoglobin that has glucose bound to it. Because red cells live roughly three months and accumulate that glucose over their lifetime, the figure reflects a weighted average of blood glucose over that period, with the most recent month contributing about half the total. It is a different kind of measurement from a fingerstick, not a substitute for one.
The conversion used here, estimated average glucose equals 28.7 times A1C minus 46.7, comes from a study that fitted a regression line to continuous glucose monitoring data from several hundred people. It is a population average with real scatter around it: two people with the same A1C can have average glucose values that differ by more than 15 percent, and the fit is looser at the extremes than in the middle. Treat the output as a translation between scales rather than as a measurement in its own right.
The relationship also breaks when red cell turnover is unusual. Anaemia, sickle cell trait and other haemoglobin variants, recent blood loss or transfusion, pregnancy, and advanced kidney disease all shift A1C away from what average glucose would predict. Laboratories know this and clinicians account for it; a converter cannot.
What the three measurements each see
Fasting glucose, a two-hour post-load value and A1C are not three ways of measuring the same thing. Fasting glucose largely reflects how much glucose the liver is releasing overnight. The two-hour value reflects how quickly the pancreas responds to a load and how readily tissue takes it up. A1C reflects the long-run average of everything. It is entirely ordinary for one of the three to sit in a different band from the others, and that discordance is itself information a clinician uses.
This is also why the printed bands cannot be collapsed into a single verdict. They exist as separate rows on the chart because they are separate tests with separate cut points, arrived at through separate lines of evidence about long-term risk. Where any individual's own targets sit — for fasting values, for post-meal values, for A1C — is set with a clinician and routinely differs from the chart, in both directions, depending on age, other conditions and the risk of lows.
Reading a home meter honestly
Consumer meters are permitted a wide tolerance against laboratory reference values, and the difference between fingertip capillary blood and venous blood is real and varies with how fast glucose is moving. Continuous monitors add a lag of several minutes between interstitial fluid and blood, which shows up most when values are changing quickly. Wash and dry hands first, because residue from fruit or a sweet drink on a fingertip can shift a reading by tens of mg/dL, and discard the first drop.
Given all of that, a converted number carries the uncertainty of the original plus nothing extra — the arithmetic here is exact, the input is not. If a value matters enough to act on, it is worth repeating, and worth taking to somebody who can put it next to everything else about you.
Questions people ask
What is 100 mg/dL in mmol/L?
5.55 mmol/L. Divide by 18.0182, or by 18 if you are doing it in your head, which gives 5.6 and is close enough for most purposes. The same arithmetic in reverse turns 5.5 mmol/L into 99 mg/dL and 7.0 mmol/L into 126 mg/dL, which is why those two numbers appear together on charts that print both scales. Values quoted to one decimal place in mmol/L carry about the same precision as whole numbers in mg/dL, so there is rarely a reason to convert to more decimals than the original had.
My A1C and my fasting readings disagree. Which one is right?
Both can be, because they measure different things. Fasting glucose is one moment in one state, mostly reflecting overnight liver output. A1C is a three-month weighted average of everything, including the hours after meals that a fasting test never sees. Someone whose fasting values look unremarkable but whose glucose rises steeply after eating can produce exactly that pattern. The reverse also happens, particularly where a haemoglobin variant or altered red cell turnover makes A1C an unreliable proxy. Discordance between the two is a normal reason for a clinician to look further rather than a sign that one of the numbers is wrong.
When exactly is the two-hour reading taken?
Two hours from the first bite or first sip, not from when you finish. In a formal glucose tolerance test the clock starts when the standardised drink is consumed, which is why it is measured that way. Glucose typically peaks somewhere around the first hour and is on the way down by the second, so a value taken at 60 minutes will usually be higher than the same reading taken at 120 minutes and cannot be compared against a two-hour chart. If you are timing meals at home, note the start time rather than trying to reconstruct it later.
Does this page tell me whether I have diabetes?
No. It converts between units and shows which printed band a number falls in. Diagnosis involves confirmed results, usually repeated, interpreted alongside symptoms and history by a clinician, and the cut points on the chart are population reference lines rather than personal thresholds. A value near a boundary is a value near a boundary. If a reading concerns you, the useful action is to take it to a clinician who can order the right confirmatory test, not to draw a conclusion from a converter.