Blood glucose units: mg/dL, mmol/L and g/L
mg/dL is the mass of glucose in a tenth of a litre of blood; mmol/L counts glucose molecules in a litre. One mole of glucose (C₆H₁₂O₆) weighs 180.156 g, so 1 mmol/L is 180.156 mg per litre — 18.0156 mg/dL. To convert mg/dL to mmol/L, divide by 18; to convert mmol/L to mg/dL, multiply by 18. Some laboratory reports use a third unit, g/L: multiply by 100 for mg/dL, so 1.26 g/L is 126 mg/dL, or 7.0 mmol/L.
Which unit you see depends on where you are. FDA guidance says meters sold in the US “should report results in mg/dL”. A 2010 manufacturer’s safety notice calls mmol/L “the standard unit of measure in the UK”, Diabetes Canada publishes its targets in mmol/L, and the IFCC recommends that laboratories report glucose in mmol/L. So converting American blood glucose figures to British or Canadian ones means dividing by 18, and going the other way means multiplying.
| mg/dL | mmol/L | g/L |
|---|---|---|
| 40 | 2.2 | 0.40 |
| 54 | 3.0 | 0.54 |
| 70 | 3.9 | 0.70 |
| 80 | 4.4 | 0.80 |
| 90 | 5.0 | 0.90 |
| 98 | 5.4 | 0.98 |
| 100 | 5.6 | 1.00 |
| 110 | 6.1 | 1.10 |
| 126 | 7.0 | 1.26 |
| 140 | 7.8 | 1.40 |
| 145 | 8.0 | 1.45 |
| 160 | 8.9 | 1.60 |
| 180 | 10.0 | 1.80 |
| 200 | 11.1 | 2.00 |
| 250 | 13.9 | 2.50 |
| 300 | 16.7 | 3.00 |
| 400 | 22.2 | 4.00 |
And the other way: 1.9 mmol/L is 34 mg/dL, 4.0 is 72, 4.7 is 85, 5.5 is 99, 7.0 is 126 and 10.0 is 180. A reading of 22 mg/dL is 1.2 mmol/L — far below the 54 mg/dL (3.0 mmol/L) the ADA treats as level 2 hypoglycaemia.
18 or 18.0156 — and why two apps can disagree by 0.1
18.0156 is the exact chemistry; 18 is the rounding almost everyone uses, and the one built into the ADAG study’s equations (its mg/dL formula divided by its mmol/L formula gives 18.0005). No guideline we found mandates either. The difference is under 0.1%, but at one decimal it can move a reading across a line: 181 mg/dL is 10.05 mmol/L divided by 18.0156, shown as 10.0, and 10.06 divided by 18, shown as 10.1. A change to the open-source Nightscout CGM software moved away from ÷18 for this reason — its readings disagreed with the Dexcom app at exactly these points, including 181.
This converter uses 18.0156, shows the ÷18 result beside it, and compares a reading with the threshold published in the unit you entered: 10.0 mmol/L is inside the 3.9–10.0 mmol/L time-in-range band even though it converts to 180.2 mg/dL, and 181 mg/dL is above 180 mg/dL even though it rounds to 10.0 mmol/L.
Is it mg/dL or mmol/L? The decimal-point check
In 2010 a meter maker wrote to UK users because some meters had been set to mg/dL by mistake. Its notice, which a Diabetes UK forum member posted in full, gives the rule: “Test results expressed in mg/dL do not contain a decimal point and are 18 times higher than the same results expressed in mmol/L. For example, a result of ‘6.0 mmol/L’ would display as ‘108 mg/dL,’ which might be misinterpreted as ‘10.8 mmol/L.’” A number with a decimal point is mmol/L (or g/L); a whole number above about 30 is mg/dL. The converter warns you when the number you type looks like the other unit.
Blood sugar chart: reference points in both units
Guidelines publish their thresholds in both units, already rounded. These are the figures behind the reference points the converter shows. They are for comparison only — what applies to you depends on your diagnosis, treatment and age, and on whether you are pregnant.
| Figure | mg/dL | mmol/L | Published by |
|---|---|---|---|
| Level 2 hypoglycaemia (“requires immediate action”) | below 54 | below 3.0 | ADA 2026 |
| Level 1 hypoglycaemia | 54 to below 70 | 3.0 to below 3.9 | ADA 2026 |
| Time in range (CGM) | 70–180 | 3.9–10.0 | International consensus 2019 |
| Very high (CGM) | above 250 | above 13.9 | International consensus 2019 |
| Before-meal target, many adults with diabetes | 80–130 | 4.4–7.2 | ADA 2026 |
| Peak after-meal target, 1–2 h after starting a meal | below 180 | below 10.0 | ADA 2026 |
| Fasting plasma glucose: prediabetes | 100–125 | 5.6–6.9 | ADA 2026 |
| Fasting plasma glucose: diabetes | 126 or more | 7.0 or more | ADA 2026, WHO 2006 |
| 2 hours into a 75 g glucose test: prediabetes | 140–199 | 7.8–11.0 | ADA 2026 |
| 2-hour test, or random glucose with symptoms: diabetes | 200 or more | 11.1 or more | ADA 2026 |
The diagnostic figures apply to laboratory tests, not to a meter or CGM. The ADA requires two abnormal results — two different tests, or the same test on two occasions — unless high blood glucose is unmistakable, and says there is “presently insufficient evidence” for using CGM to diagnose diabetes or prediabetes.
Same reading, different guideline: US, UK and Canada
Converting one country’s figure does not give you another country’s figure. The UK’s hypo line of 4.0 mmol/L is 72 mg/dL, while the US line is 70 mg/dL (3.9 mmol/L). The ADA’s 2-hour pregnancy target of 120 mg/dL converts to 6.7 mmol/L, but NICE’s is 6.4 mmol/L.
| Figure | ADA (US) | NICE and NHS (UK) | Diabetes Canada, WHO |
|---|---|---|---|
| Low blood sugar | below 70 mg/dL (3.9 mmol/L) | below 4 mmol/L (NHS) | — |
| Fasting glucose: higher risk / prediabetes | 5.6–6.9 mmol/L | 5.5–6.9 mmol/L (PH38) | WHO: 6.1–6.9 mmol/L |
| Before meals, with diabetes | 4.4–7.2 mmol/L | Type 1: 4–7 mmol/L; 5–7 on waking (NG17) | Canada: 4.0–7.0 mmol/L |
| After meals, with diabetes | below 10.0 mmol/L, 1–2 h | Type 1: 5–9 mmol/L, 90+ min (NG17) | Canada: 5.0–10.0 mmol/L at 2 h |
| Pregnancy, 2 h after a meal | below 6.7 mmol/L (120 mg/dL) | below 6.4 mmol/L (NG3) | — |
HbA1c: percent or mmol/mol
HbA1c measures how much haemoglobin has glucose attached, which reflects average glucose over roughly the previous two to three months. It is reported as a percentage (NGSP/DCCT) or in mmol/mol (IFCC), linked by the master equation: mmol/mol = (% − 2.15) × 10.929. The 2010 international consensus asks laboratories to report “mmol/mol, no decimals” and “%, one decimal”.
| HbA1c % | mmol/mol | Why the figure matters |
|---|---|---|
| 5.7 | 39 | ADA prediabetes range starts |
| 6.0 | 42 | NICE: high risk of type 2 diabetes starts |
| 6.5 | 48 | Diabetes threshold (ADA, WHO); a NICE target |
| 7.0 | 53 | ADA goal for many adults with diabetes: below this |
| 7.5 | 58 | NICE: intensify type 2 treatment from here |
| 8.0 | 64 | — |
mmol/mol is not mmol/L. An HbA1c of 48 mmol/mol and a glucose reading of 4.8 mmol/L have nothing to do with each other, and the scale is not proportional: on the Diabetes UK forum a member worked out that 40 is “two thirds of 60” — but 40 mmol/mol is 5.8% and 60 is 7.6%. UK laboratories reported both units from 2009 and switched to mmol/mol only in autumn 2011, after a last-minute postponement to 30 September 2011. Japan and Sweden have used national scales with their own equations, so a Japanese (JDS) or Swedish (Mono-S) percentage needs a different conversion.
How to convert A1c to average blood glucose (eAG)
The formula comes from the ADAG study (Nathan et al., 2008): eAG (mg/dL) = 28.7 × A1C − 46.7, or in mmol/L, 1.5944 × A1C − 2.594. It was fitted to 507 people — 268 with type 1 diabetes, 159 with type 2 and 80 without diabetes, from 10 centres — each with about 2,700 glucose values over three months (R² = 0.84). An A1c of 6.4% converts to 137 mg/dL (7.6 mmol/L); 7% to 154 mg/dL (8.6 mmol/L).
It is an average across people, and the spread is wide: for an A1c of 7% the paper gives 154 mg/dL with a 95% interval of 123 to 185 mg/dL. The ADA’s 2026 Standards note that “it is unclear how generalizable these estimates are” to modern CGM data. Calculators also round differently: one popular UK converter uses 1.59 × A1C − 2.59 and gives 8.5 mmol/L at 7%, where the paper’s table gives 8.6; the ADA’s own online table shows 10.1 mmol/L at 8%, the paper 10.2.
Can you convert a blood glucose reading to A1c?
Not a single reading. A1c reflects months of glucose, and one finger-prick or CGM value can be anything within the day’s swings. The ADAG line can be run backwards — A1C = (average mg/dL + 46.7) ÷ 28.7 — but only for an average of many readings. An average of 145 mg/dL gives 6.7%; 139 mg/dL gives 6.5%.
Even then, a meter average only covers the moments you tested. On the Diabetes UK forum, one member was told by a nurse that an A1c of 8.6% meant about 9.0 mmol/L, while the eAG formula gives 11.1. Long-standing members point out the opposite trap too: if you test mostly before meals, your meter average will run lower than your HbA1c suggests.
GMI: the A1c estimate from a CGM
CGM reports show a Glucose Management Indicator, defined by Bergenstal et al. (2018) from 528 people wearing Dexcom G4 sensors: GMI (%) = 3.31 + 0.02392 × mean glucose (mg/dL), or GMI (mmol/mol) = 12.71 + 4.70587 × mean glucose (mmol/L). It replaced the older name “estimated A1C” after the FDA decided the name needed to change, and the authors suggest at least 10, preferably 14 or more, days of data.
GMI and a lab HbA1c often differ. In the paper’s data they were identical 19% of the time and 0.5 percentage points or more apart 28% of the time. GMI and the ADAG line also part ways at low averages: for 154 mg/dL both give 7.0%, but for 100 mg/dL GMI gives 5.7% and ADAG 5.1%. Diabetes UK forum members comparing their Libre estimates with lab results counted nine underestimates and three overestimates in twelve pairs — an anecdote, not a study, but a fair warning. The international consensus links 70% time in range to an A1c of about 7% (53 mmol/mol), and 50% to about 8% (64 mmol/mol).
Meters, plasma and accuracy
Glucose is about 11% higher in plasma than in whole blood (WHO, 2006: 8% to 15% depending on haematocrit), and the IFCC recommends reporting the plasma value, converting whole-blood results with a factor of 1.11. FDA guidance expects US meters to report “plasma equivalents”. Meters are not laboratory instruments: ISO 15197:2013 asks for 95% of results within ±15 mg/dL below 100 mg/dL and within ±15% above it. So when the lab value is 180 mg/dL, a meter meeting the standard could read anywhere from 153 to 207; at 70 mg/dL, from 55 to 85. The WHO also notes that capillary (finger-prick) and venous samples agree when fasting, but capillary results run higher after eating.
Where blood glucose comes from: glycogen, lactate and protein
Excess blood glucose is converted into glycogen, a branched glucose polymer stored in the liver and muscles, and — with insulin promoting fat synthesis — into fat. Between meals the liver keeps blood glucose up by breaking glycogen down and by making new glucose (gluconeogenesis).
- Can muscle glycogen be converted to blood glucose? Not directly. Muscle holds the largest total glycogen store, but releasing free glucose needs glucose-6-phosphatase, which is generally restricted to the liver, kidney and intestine. Muscle uses its glycogen itself; the lactate it produces can travel to the liver and become glucose there.
- Can blood lactate be converted to glucose in the liver? Yes. Lactate is one of the precursors of gluconeogenesis, alongside glycerol, pyruvate, propionate and glucogenic amino acids — mainly in the liver and, to a lesser extent, the kidney.
- Can proteins be converted to blood glucose? Yes, partly: when protein is broken down, the glucogenic amino acids can be turned into glucose by gluconeogenesis.
Diabetes in numbers
The WHO counts 830 million adults living with diabetes in 2022, 14% of adults, up from 200 million in 1990. The IDF Diabetes Atlas puts adults aged 20–79 at 589 million, one in nine, rising to 853 million by 2050. In the US, the CDC estimates 40.1 million people (12.0%) had diabetes in 2023, 11.0 million of them undiagnosed, and 115.2 million adults have prediabetes. Diabetes UK says almost 6 million people in the UK live with diabetes, including about 1.3 million who may have type 2 diabetes without knowing it.
What this converter does not do
It converts units and places a figure next to published reference points. It does not diagnose, it does not know your targets, and it never suggests a dose. Children, pregnancy and older adults have different targets from the adult figures shown — follow your doctor, nurse or diabetes team. Next to blood glucose, weight is the other number people track most: the BMI calculator works in kg, pounds or stone.