Biomarkers

Triglyceride to HDL Ratio: Good, Bad, and Why Your Lab Units Change the Answer

On this page
  1. Why the units decide the number
  2. What the ratio was validated to do
  3. Most of what the ratio does is the triglyceride moving
  4. The fasting question, which your lab may have quietly changed
  5. Where it fails, and that part is not a footnote
  6. What the guidelines actually target
  7. What to do with a number you already have

The triglyceride to HDL ratio is one of the very few numbers on a lipid panel that changes value when your lab changes units. The widely quoted cutpoint of 3.0 comes from a 2003 study that worked in mg/dL. The same blood, reported in mmol/L, gives roughly 1.3, because triglycerides and HDL cholesterol convert between the two unit systems by different factors. A report in mmol/L read against 3.0 looks about two and a half times better than it is, and that is the single most common way this ratio is misread.

ValueIn mg/dL unitsSame thing in mmol/LWhere it comes from
Ratio cutpoint for insulin resistance3.0~1.3The 2003 ROC analysis, in overweight non-diabetic adults. The only ratio figure here with a primary source
Triglycerides alone, same study130 mg/dL1.47 mmol/LPerformed about as well as the ratio, which is worth knowing before computing anything
"1.8 in SI units", as printed in that paper beside 3.0would be ~4.11.8Does not convert into 3.0. See the next section: one of the two is the one to use, and we cannot tell you which the authors meant
3.5, 4.0 and similar "high risk" lines3.5 / 4.0~1.5 / ~1.7Calculator and blog conventions. No primary source found for either as an insulin-resistance threshold

Two definitions before anything else. The ratio is fasting triglycerides divided by HDL cholesterol, both taken from the same blood draw, and it is a unitless number only in the sense that the units cancel. They cancel into a different number depending on what they were. And it is not a cardiovascular risk score: what it was actually validated against is a research measurement of insulin resistance, in a specific population, which is a narrower and more interesting claim than most pages about it make.

You could be asking this about yourself

how have my own markers moved since my last test?

Connect BioTrakk to Claude, ChatGPT or Gemini. Log a reading in one sentence, and your own history becomes something you can ask about — in the chat you already have open.

Free. No BioTrakk account to make first. Claude takes one click.

Why the units decide the number

Total cholesterol divided by HDL is unit-proof: both are cholesterol, both convert by the same factor, and the ratio survives. Triglycerides and HDL cholesterol are different molecules with different molecular weights, so they do not.

The standard lipid conversion factors are 88.57 for triglycerides and 38.67 for HDL cholesterol, both from mmol/L to mg/dL. Divide one by the other and you get the whole problem in a single figure: a ratio computed in mg/dL is about 2.29 times the same ratio computed in mmol/L.

So one blood draw produces two legitimate ratios. Triglycerides of 150 mg/dL with HDL of 50 mg/dL is a ratio of 3.0. The same sample in mmol/L is 1.69 and 1.29, a ratio of 1.31. Nothing changed except the column heading.

Worth checking what your own report uses. Spain, the United States and much of Latin America print mg/dL, so 3.0 applies as written. The United Kingdom, Ireland, Canada, Australia and most of northern Europe print mmol/L, where the comparable line is about 1.3. France reports lipids in g/L as well as mmol/L, and a ratio taken from two g/L figures equals the mg/dL ratio, because both are masses per volume: 1.50 g/L over 0.50 g/L is the same 3.0.

There is an untidy detail in the source paper itself, and it is better said than hidden. The 2003 study lists its optimal cutpoints as 1.8 in SI units and 3.0 in traditional units, presented as the same threshold. Under the conversion factors above they are not: 3.0 traditional is 1.31 SI, and 1.8 SI is 4.12 traditional. We cannot tell from outside which of the two the analysis produced, and it is a real fork, because at a true ratio of 2.5 in mg/dL one version flags you and the other does not. Our reading is to use 3.0 on a mg/dL report, since that is the version the paper recommends in its own conclusion and the version the later literature has carried, and to treat anything near the line as near the line rather than as a verdict. If you have the full text and can settle it, we would like to know.

What the ratio was validated to do

The 2003 Annals of Internal Medicine study asked a narrow and practical question: in overweight adults without diabetes, which cheap routine measurement best identifies the ones who are insulin resistant?

The comparison was against the insulin suppression test, with insulin resistance defined as being in the top tertile of steady-state plasma glucose, and overweight as a BMI of 25 or above. Receiver-operating characteristic analysis then picked the best cutpoints from the routine candidates. Three came out: triglycerides at 130 mg/dL (1.47 mmol/L), the triglyceride/HDL ratio, and fasting insulin at 109 pmol/L. The authors reported that the predictive power of these markers was similar to that of the ATP III metabolic syndrome criteria, and concluded that in the absence of a standardised insulin assay, the most practical approach is to use the triglyceride cutpoint or the ratio cutpoint.

Three consequences follow, and all three are routinely dropped:

  • Triglycerides alone did about as well. The ratio's appeal is that it feels like it extracts more from the panel. In the study that gave it its number, it did not do much more than the numerator did by itself.
  • The population was overweight adults without diabetes. A cutpoint chosen by ROC analysis is tuned to the prevalence and spread of the group it was chosen in. It is not a law of physiology that travels unchanged into lean people, athletes, or people already on lipid-lowering treatment.
  • It identifies, it does not measure. The output is a flag for "worth looking at insulin resistance", not a quantity of insulin resistance. If you want the quantity, fasting insulin and HOMA-IR are the ones with a scale, and HbA1c and fasting glucose answer a different question again.

Most of what the ratio does is the triglyceride moving

Fasting triglycerides are among the least stable routine blood measurements, and HDL cholesterol is among the most stable. Put the volatile one on top of the steady one and nearly all the movement in the ratio is the numerator.

A one-year study of repeated measurements in healthy adults found within-person variation across subjects running from 12.9% to 40.8% for triglycerides, against 3.6% to 12.4% for HDL cholesterol. Later work points the same way: the EFLM biological variation database carries a within-subject estimate near 5% for HDL cholesterol, and a 2022 study of monthly draws reported ~25.7% within-subject variation for triglyceride-rich lipoprotein cholesterol.

What that does to a ratio is arithmetic rather than biology, so treat the next number as ours and not the papers'. Combining a numerator at ~25% and a denominator at ~7% in the usual way gives a ratio with a within-person coefficient of variation around 26%. At a true long-run ratio of 3.0, readings of about 2.3 and about 3.9 are both ordinary. One of those is under the cutpoint and one is well over it, and nothing has changed about you between the two draws.

The practical form of that: a single ratio is a measurement of a morning. Two or three fasting panels, same lab, over months, are a measurement of you. This is the same trap as a single hs-CRP reading, with the difference that nobody has written a guideline telling you to average this one.

The fasting question, which your lab may have quietly changed

Every cutpoint above was derived on fasting blood. Meanwhile the default has moved. The 2016 EAS and EFLM joint consensus statement concluded that fasting is not routinely required for a lipid profile and that non-fasting sampling is preferred for routine testing, since most lipids barely move after a meal.

Triglycerides are the exception, which is exactly the problem here. The consensus flags non-fasting triglycerides as elevated at 2.0 mmol/L (175 mg/dL) against the fasting line of 1.7 mmol/L (150 mg/dL), a deliberate allowance for the postprandial rise. The statement is a sensible piece of medicine and it is not in dispute. It does mean that a panel drawn at 4pm after lunch inflates the numerator of a ratio whose threshold was fitted on fasting samples, and nothing on the report will tell you the ratio you just computed is not comparable.

So the first thing to check is not the number, it is whether you had eaten. If the draw was non-fasting, the ratio is not the one the 3.0 line was built for. Write down which it was, because in two years you will not remember.

Where it fails, and that part is not a footnote

A 2005 study in Archives of Internal Medicine is titled as flatly as a paper can be: fasting triglyceride and the triglyceride-HDL cholesterol ratio are not markers of insulin resistance in African Americans. In 125 black participants, neither fasting triglycerides nor the ratio identified insulin resistance, while fasting insulin, waist circumference and BMI did. The accompanying commentary and later work from the same group attribute it to lipoprotein lipase activity: triglyceride-rich lipoproteins are cleared efficiently even in an insulin-resistant state, so the triglyceride level stays normal and the marker has nothing to see.

The evidence is not unanimous, and the honest version says so. A large NHANES-based analysis across three racial and ethnic groups found the ratio the best of the lipid predictors of hyperinsulinaemia and suggested the null result in black participants may partly reflect the small sample. A 2024 review of the whole literature lands in between: predictive power varies by ethnicity and by sex, with the published thresholds performing better in white, Asian and Hispanic populations than in African Americans, and better in men than in women.

Whichever way that resolves, the general lesson is already usable. A marker that works in one population and not in another is not a universal number, and a ratio of 2.1 does not mean the same thing for everyone who can compute it. If this ratio is the only thing being used to rule insulin resistance out, it is being asked for more than it has ever delivered.

What the guidelines actually target

Not this. The 2019 ESC/EAS dyslipidaemia guidelines keep LDL cholesterol as the primary target and add non-HDL cholesterol and ApoB as secondary targets, specifically recommending both for risk evaluation in people with raised triglycerides. For high-risk patients the stated goals are non-HDL-C under 2.6 mmol/L (100 mg/dL) and ApoB under 80 mg/dL; for very-high-risk patients, 2.2 mmol/L (85 mg/dL) and 65 mg/dL.

Lipoprotein ratios are not discussed in that document at all. Silence is not disapproval, and we are not going to dress it up as one. What it does mean is concrete: there is no guideline target for the triglyceride/HDL ratio, so there is no established benefit to chasing it as a number, and if a panel is being used to inform a treatment decision, the figures that decision rests on are LDL-C, non-HDL-C and ApoB.

What to do with a number you already have

If a lipid panel has been run, the ratio is free to compute and genuinely informative within its limits. A few things make it worth more:

  • Check the units before the value. mg/dL or g/L, compare against ~3.0. mmol/L, compare against ~1.3. This one step removes the most common misreading of this marker.
  • Check whether you had eaten. The cutpoints are fasting. A non-fasting panel is fine medicine and a poor input to this particular arithmetic.
  • Treat a value near the line as near the line. With a within-person swing of roughly a quarter, a 2.8 and a 3.4 are the same result wearing different labels.
  • Keep the two raw numbers, not just the ratio. A ratio of 3.0 from triglycerides 180 and HDL 60 is not the same situation as 3.0 from 120 and 40, and the ratio throws that away.
  • Do not use it alone to rule anything out, and know that its performance differs by population and sex.
  • Take any decision to a clinician. There is no dose and no treatment on this page. What a raised ratio means for one person depends on their history, their other markers and their absolute risk, which is not a question a cutpoint answers.

The failure here is usually clerical rather than medical. Panels arrive as PDFs months apart, in whichever units that lab uses, sometimes fasting and sometimes not, and two years later there is a folder of files and no trend. Logging each panel in BioTrakk (photograph the lab report or type the values in) keeps one line per marker with the units and the context recorded beside it, so a real move is distinguishable from a lunch. The blood test tracker page describes how that storage works, and if you want to test something against your own baseline rather than against a population cutpoint, that is what an n-of-1 experiment is for. On a marker this noisy, honest software will often report "not distinguishable from chance", which is frequently the correct answer.

BioTrakk is not a medical device and this is not medical advice. A lipid panel is a clinical laboratory test: interpreting yours, and deciding anything on the basis of it, is a conversation with a clinician. Nothing here diagnoses, treats, monitors or prevents any condition, and this page recommends no treatment and no dose.

Frequently asked questions

What is a good triglyceride to HDL ratio?

On a report in mg/dL, the only threshold with a primary source behind it is 3.0, from a 2003 Annals of Internal Medicine study that used ROC analysis in overweight non-diabetic adults to find the best routine marker of insulin resistance. On a report in mmol/L the comparable figure is about 1.3, because triglycerides and HDL convert by different factors (88.57 and 38.67), so a ratio in mg/dL is about 2.29 times the same ratio in mmol/L. Values like 3.5 or 4.0 circulate widely as "high risk" lines without a primary source we could find.

Why do different sites give different cutoffs?

Partly units and partly provenance. A 3.0 derived in mg/dL becomes 1.3 in mmol/L, and quoting one against the other produces apparent disagreement that is really arithmetic. The 2003 source paper itself prints its cutpoint as 1.8 in SI units and 3.0 in traditional units, and those two do not convert into each other: 3.0 traditional is 1.31 SI and 1.8 SI is 4.12 traditional. Beyond that, several widely repeated thresholds trace back to calculators and blogs rather than to a study.

Does the ratio measure insulin resistance?

It flags it, in the population it was fitted in. The 2003 study compared routine markers against the insulin suppression test, with insulin resistance defined as the top tertile of steady-state plasma glucose, and reported that triglycerides at 130 mg/dL, the triglyceride/HDL ratio and fasting insulin at 109 pmol/L were the best of them, with predictive power similar to the ATP III metabolic syndrome criteria. Triglycerides alone did about as well as the ratio. If you want a measured quantity rather than a flag, fasting insulin and HOMA-IR have a scale.

Does it need to be a fasting blood draw?

The cutpoints were derived on fasting blood, so yes if you intend to compare against them. Routine practice has moved the other way: the 2016 EAS and EFLM joint consensus statement concluded that fasting is not routinely required for a lipid profile, and flags non-fasting triglycerides as elevated at 2.0 mmol/L (175 mg/dL) against 1.7 mmol/L (150 mg/dL) fasting, to allow for the rise after a meal. That is sound medicine and it makes a non-fasting panel a poor input to this particular ratio, because the numerator is inflated and the report will not say so.

Does the ratio work for everyone?

No, and the exception is well documented. A 2005 Archives of Internal Medicine study of 125 black participants found that neither fasting triglycerides nor the triglyceride/HDL ratio identified insulin resistance, while fasting insulin, waist circumference and BMI did; the proposed mechanism is efficient lipoprotein lipase activity keeping triglycerides normal in an insulin-resistant state. The evidence is not unanimous, and a larger NHANES analysis found the ratio the best lipid predictor of hyperinsulinaemia across three groups, arguing the earlier null may reflect sample size. A 2024 review reports that accuracy varies by ethnicity and sex.

Should I try to lower my triglyceride/HDL ratio?

There is no guideline target for it. The 2019 ESC/EAS dyslipidaemia guidelines keep LDL cholesterol as the primary target and add non-HDL cholesterol and ApoB as secondary targets, recommending both in particular for people with raised triglycerides. Lipoprotein ratios are not discussed in that document at all, which is silence rather than disapproval, but it means no treatment decision rests on this ratio and no trial has shown a benefit from moving the ratio as such. Any decision belongs with a clinician, and this page recommends no treatment and no dose.

Track it. Test it. Know for sure.

Log food, supplements, sleep and biomarkers by chatting with BioTrakk on Telegram, WhatsApp or the web — then let it find what actually moves your numbers.

Free on Telegram & Web. No app to install.