Lipoprotein(a): nmol/L, mg/dL, and why the two do not convert

Lp(a) is the only marker in TherapyLog that the app flatly refuses to convert. That refusal is not fussiness: mg/dL and nmol/L measure two different physical things, and the factor that translates one into the other is different in you than it is in the person standing next to you.

The one result this app will not convert

Lipoprotein(a) is an LDL particle with a second protein bolted on. Apolipoprotein B-100 sits inside it exactly as it does in ordinary LDL, and a molecule of apolipoprotein(a) — apo(a) — is joined to that by a disulphide bond. Labs report the result in one of two units. Milligrams per decilitre is a mass: how much lipoprotein(a) material is in the sample. Nanomoles per litre is a count: how many apo(a) molecules, and so how many Lp(a) particles, are in it. Those are different quantities, not two scales for the same quantity.

Established clinical use Turning one into the other requires knowing what a single particle weighs, and that weight is not the same in everyone. The apo(a) protein contains a segment called kringle IV type 2 which is present as a variable number of identical copies — one to more than forty per allele — so apo(a) exists in over forty sizes across the population, and a large-isoform particle is simply heavier than a small-isoform one. The European Atherosclerosis Society's 2022 consensus statement therefore declines to endorse any fixed conversion factor and asks that results be reported in molar units wherever the assay allows it.¹ When molar and mass results were measured head to head in 1,635 samples across five commercial assays, the ratio between them was not a constant: it ran from roughly 1.8 in samples below 75 nmol/L to roughly 3.6 in the highest samples, and varied by method as well as by isoform size.²

That is the entire argument for the refusal. Every "multiply by 2.5" table online applies a population average to an individual, and the error is not random noise — it is systematically largest at the top of the range, where decisions actually get made, because high Lp(a) travels with small apo(a) isoforms. Run it in the direction people actually do. A mass result of 45 mg/dL, multiplied by the usual 2.5, reads 113 nmol/L — inside the grey zone and comfortably below the 125 nmol/L line. But at the ratios genuinely observed near and above that line, the same sample can be anywhere from about 85 to 126 nmol/L depending on the assay pair and the isoform, and the top of that spread is over the threshold rather than under it. The conversion did not tell you where you stand; it told you where an average person with your mass result would stand. So TherapyLog’s Lp(a) field is nmol/L only. Hand its converter an mg/dL value and it answers no valid conversion and declines to store a number, rather than approximating one. That is deliberate: two results in different units are two facts, and they are not a trend.

The assay is part of the result

Ask which assay produced the number. Mass assays measure the whole particle and report mg/dL. Molar assays are calibrated to count apo(a) and report nmol/L. Nearly all routine assays are immunoassays, and nearly all of them use polyclonal antibodies raised against apo(a) — which means part of the antibody population binds the repeated kringle IV epitopes, the very feature that differs in number between people.

Established clinical use This is a measurement problem, not a theoretical one. Work published in Clinical Chemistry in 1995 built three ELISAs differing only in the detecting antibody, calibrated them against a single serum containing apo(a) with 21 kringle 4 domains, and ran them across 723 people typed for apo(a) size. Assays whose antibody bound the repeated domain read high in carriers of large isoforms and low in carriers of small ones.³ Since small isoforms are the ones found in people with high Lp(a), a size-sensitive assay is least reliable in precisely the group whose result matters most.

The fix is to measure something that does not repeat. An IFCC-endorsed reference measurement procedure quantifies apo(a) by mass spectrometry using proteotypic peptides drawn from the kringle 5, kringle 9 and protease domains — regions present once per molecule — with calibrators traceable to the former WHO/IFCC reference material SRM 2B, which makes it independent of the size polymorphism.⁴ Almost no routine lab runs it, but it is what molar assays are increasingly standardised against. The practical version for you is short. Record the assay and the lab beside the number, and treat a result from a different lab as a different measurement rather than a change in you. A printout that says nmol/L has not, by itself, told you the assay was calibrated in molar units; the consensus asks labs to report in the units their assay was actually calibrated in.¹

Why one number is worth having at all

Established clinical use Lp(a) concentration is among the most strongly inherited quantitative traits in humans. Something close to 90% of its variance is controlled at the LPA locus, and the kringle IV type 2 copy number alone accounts for roughly 40 to 70% of it, with larger isoforms tracking lower concentrations.⁵ Adult levels are reached early and are broadly stable for the rest of adult life. Nothing you eat, lift or inject was ever going to be the main determinant of this one.

The reason to know it is that the association with cardiovascular disease looks causal rather than incidental. Because apo(a) size variants are allocated at conception and independently of how anyone lives, they function as a natural randomisation. In three Copenhagen cohorts, genetically raised Lp(a) carried a hazard ratio of about 1.22 for myocardial infarction per doubling of Lp(a) on instrumental-variable analysis, closely matching the observational estimate — which is the signature of a causal exposure rather than a bystander marker.⁶ That is also why Lp(a) sits outside the panels the app tracks for any compound: it is not something a protocol moves, so nothing in the app monitors it.

Guidelines have followed the genetics. The National Lipid Association's 2024 focused update asks for Lp(a) to be measured at least once in all adults and treats the result as a continuum, with below 75 nmol/L low risk, 75 to 125 nmol/L an intermediate grey zone, and 125 nmol/L or above high.⁷ TherapyLog's generic 0–75 nmol/L range is a fallback for reports that carry no range of their own. Your own lab's stated range wins, because it belongs to the assay that produced your number. There is no optimal band for Lp(a) and this page will not invent one — a result of 80 nmol/L is not a finding, it is a reason to look harder at everything else.

Does TRT or a GLP-1 move it?

Off-label or community practice Testosterone does move Lp(a), and this is the part the general Lp(a) pages leave out. In a study in normal men published in the American Journal of Cardiology in 1996, average Lp(a) fell by 37% on testosterone alone and by 28% when an aromatase inhibitor was given alongside it, which the authors read as an androgenic effect rather than one mediated by conversion to oestradiol.⁸ The studies are small, short and decades old, they used a surrogate endpoint, and no trial has asked whether lowering Lp(a) with an androgen lowers anybody's risk of anything. Lowering Lp(a) is not an indication for testosterone and should not be treated as one. What it does mean is concrete: if you were already on TRT when your first Lp(a) was drawn, that number may not be your untreated baseline, and a fall after starting therapy is not evidence your arteries are safer.

Animal-only or theoretical GLP-1 receptor agonists are a different case, mostly because there is not yet a case. I could not find published Lp(a) data from the large semaglutide or tirzepatide outcome trials. The nearest human evidence points the wrong way: across four cohorts of people with and without type 2 diabetes, three to four months of energy-restricted dieting improved almost every conventional risk factor while Lp(a) rose, by about 15 nmol/L on average, with the rise correlating with the amount of weight lost.⁹ Extrapolating from diet-induced weight loss to drug-induced weight loss is exactly that, an extrapolation, but it is the only direction the data currently point. Expecting a GLP-1 to fix an Lp(a) result is not supported by anything published.

Statins deserve a paragraph because the community is confident about something the literature is not. A 2020 analysis in the European Heart Journal pooled six randomised trials, 5,256 patients, and found statin therapy raised Lp(a) by around 11% relative to placebo, with supporting cell work showing increased apo(a) production.¹⁰ A network meta-analysis of 39 randomised trials and 24,448 patients, published two years later, found no clinically important difference between any statin and placebo.¹¹ Both are real analyses and they disagree. What they agree on is the point that matters here: statins do not lower Lp(a), so an Lp(a) problem is not one that LDL therapy has quietly solved for you already.

What to do with a number you cannot change

Log it once, properly. The unit exactly as printed, the assay if the report names it, the lab, the date. Do not convert it, do not average it against a result in the other unit, and do not read a trend across units — the app will not draw one for you, and that is the point.

The measure-once advice is guideline consensus, and it is being questioned. A 2025 study of 1,263 patients with two Lp(a) measurements taken at least a year apart found a median intra-individual change of 16.7%, with 44% shifting by 20% or more and 14% by 50% or more, and its authors argue the measure-once recommendation should be revised.¹² Read that as a caution about precision rather than a licence to trend. A repeat that differs by a fifth is telling you about assay and biological variation, not about something you did last quarter.

Animal-only or theoretical There is at present no approved drug whose purpose is to lower Lp(a), and no completed outcome trial showing that lowering it prevents events. Pelacarsen, an antisense oligonucleotide directed at apo(a), is being tested for exactly that question in a phase 3 trial of 8,323 people with established cardiovascular disease and elevated Lp(a), running monthly injections against placebo over roughly six years, with a primary endpoint of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke or urgent coronary revascularisation.¹³ At the time of writing, follow-up has finished and no outcome result has been published. Until one is, the case for lowering Lp(a) rests on genetics — a strong argument, and not a proven one.

None of this is a reason to do anything on your own. A high Lp(a) is a reason to look harder at every risk factor you can move, and to do that with the clinician who manages your care rather than against a threshold read off a web page.

So the honest use of a high Lp(a) is as a multiplier on the things you can actually change: apoB and LDL cholesterol, blood pressure, smoking, and — if you are on testosterone — the haematocrit and blood pressure effects that come with it. Every one of those alters what someone might prescribe, and none of them is a decision to make from a number on a screen, least of all one whose meaning depends on which machine measured it. Take the printed report, with its unit, its range and its assay named, to the clinician who prescribes for you, and let the number change that conversation rather than your protocol.

What the app records for this marker

Questions people actually ask

Can I convert my Lp(a) from mg/dL to nmol/L?

Not reliably, and TherapyLog will not do it for you. mg/dL measures the mass of Lp(a) material; nmol/L counts apo(a) molecules. The weight of a single particle depends on how many kringle IV type 2 repeats your apo(a) carries, which is genetic and differs between people, so the factor that is correct for one person is wrong for another.

When both were measured on the same 1,635 samples, the observed ratio ran from about 1.8 at low concentrations to about 3.6 at high ones, and also varied by assay. A fixed factor of 2.5 is a population average being applied to an individual, and it goes most wrong at the high end where the answer actually matters.

My old result was in mg/dL and my new one is in nmol/L. Can I compare them?

No. Store both, keep each with the unit and lab that produced it, and treat them as two separate facts rather than two points on a line. The app deliberately refuses to plot them together.

If you want a comparable pair, ask for the repeat to be run by the same lab on the same assay in the same unit. Anything else confounds a change in you with a change in method.

Does TRT lower Lp(a)?

Small older studies say yes — average falls of roughly 25 to 40% have been reported in men given testosterone, apparently through an androgenic rather than an oestrogenic route. No trial has tested whether that reduction changes any clinical outcome, and lowering Lp(a) is not an approved reason to prescribe testosterone.

The practical consequence is about interpretation. If your first Lp(a) was drawn while you were already on testosterone, it may sit below your untreated baseline, and a drop after starting therapy is not evidence that your risk has fallen. Discuss what your particular result means with the clinician who prescribes for you.

Will a GLP-1 lower my Lp(a)?

There is no published Lp(a) result from the large semaglutide or tirzepatide cardiovascular outcome trials that I could find. The closest human evidence comes from diet-induced weight loss, where Lp(a) rose modestly even as almost every other risk factor improved.

So the honest answer is that it probably will not, and it might nudge it the other way. That is precisely why knowing your Lp(a) matters independently of the rest of your panel — it is the one line that your protocol is not quietly managing.

Do I need to repeat the test?

Guidelines say once in adulthood is enough, because Lp(a) is largely set by the LPA gene and is broadly stable across life. That advice is now being questioned: a 2025 study found nearly half of patients shifted by 20% or more between two measurements taken at least a year apart.

Read that as a statement about measurement precision, not as a reason to trend Lp(a) quarterly. A repeat that differs by a fifth is more likely assay and biological variation than a real change, and any decision to retest belongs with your prescribing clinician.

My Lp(a) is 80 nmol/L. Is that bad?

It is in the intermediate zone that the EAS and NLA describe — above the 75 nmol/L rule-out threshold, below the 125 nmol/L rule-in one. That is not a diagnosis and there is no optimal band for Lp(a) that this page will offer you.

The generic 0–75 nmol/L range in the app is a fallback for reports that carry no range of their own; if your lab printed a range, that one wins, because it belongs to the assay that produced your number. A result in the grey zone is a reason to take your other, modifiable risk factors more seriously — a conversation for the clinician who prescribes for you.

Sources

Every claim above that is not a definition is either labelled by evidence tier or carries a numbered reference to one of these.

1. Kronenberg F, Mora S, Stroes ESG, et al. Lipoprotein(a) in atherosclerotic cardiovascular disease and aortic stenosis: a European Atherosclerosis Society consensus statement. European Heart Journal 2022;43(39):3925–3946. The EAS panel declines to endorse a fixed mg/dL-to-nmol/L conversion factor, asks for molar reporting where available, and sets the 75 and 125 nmol/L rule-out and rule-in zones.
2. Marcovina SM, et al. Relationship of lipoprotein(a) molar concentrations and mass according to lipoprotein(a) thresholds and apolipoprotein(a) isoform size. Journal of Clinical Lipidology 2018;12(5):1313–1323. Head-to-head molar and mass measurement in 1,635 samples across five assays: molar/mass ratios are threshold-, method- and isoform-dependent, so a single conversion factor is not appropriate.
3. Marcovina SM, Albers JJ, Gabel B, Koschinsky ML, Gaur VP. Effect of the number of apolipoprotein(a) kringle 4 domains on immunochemical measurements of lipoprotein(a). Clinical Chemistry 1995;41(2):246–255. Establishes apo(a) isoform-size bias in immunoassays: antibodies binding the repeated kringle IV domain over-read large isoforms and under-read small ones against a single calibrator.
4. An LC–MS-based designated comparison method with similar performance to the Lp(a) reference measurement procedure to guide molar Lp(a) standardization. Clinical Proteomics, 2024. Describes the IFCC-endorsed mass-spectrometry reference procedure using proteotypic peptides outside the repeat region, traceable to WHO/IFCC SRM 2B and independent of apo(a) size polymorphism.
5. Lipoprotein(a) beyond the kringle IV repeat polymorphism: the complexity of genetic variation in the LPA gene. Atherosclerosis, 2022. Source for heritability approaching 90%, the KIV-2 copy number variation generating over forty isoforms, and the inverse relationship between isoform size and concentration.
6. Kamstrup PR, Tybjærg-Hansen A, Steffensen R, Nordestgaard BG. Genetically elevated lipoprotein(a) and increased risk of myocardial infarction. JAMA 2009;301(22):2331–2339. Mendelian randomisation across three Copenhagen cohorts giving a hazard ratio of about 1.22 for myocardial infarction per doubling of Lp(a) on instrumental-variable analysis.
7. A focused update to the 2019 NLA scientific statement on use of lipoprotein(a) in clinical practice. Journal of Clinical Lipidology, 2024. National Lipid Association recommendation to measure Lp(a) at least once in all adults, with the risk continuum expressed in nmol/L (<75, 75–125, ≥125).
8. Zmuda JM, Thompson PD, Dickenson R, Bausserman LL. Testosterone decreases lipoprotein(a) in men. American Journal of Cardiology 1996;77(14):1244–1247. Testosterone administration in normal men lowered Lp(a) by an average of 37%, attributed to an androgenic rather than an oestrogenic effect.
9. Berk KA, Yahya R, Verhoeven AJM, et al. Effect of diet-induced weight loss on lipoprotein(a) levels in obese individuals with and without type 2 diabetes. Diabetologia 2017;60(6):989–997. Across four cohorts, energy-restricted dieting improved conventional risk factors while Lp(a) rose, with the increase correlating with weight lost.
10. Tsimikas S, Gordts PLSM, Nora C, Yeang C, Witztum JL. Statin therapy increases lipoprotein(a) levels. European Heart Journal 2020;41(24):2275–2284. Pooled analysis of six randomised trials (5,256 patients) reporting an increase in Lp(a) on statin therapy relative to placebo, with supporting hepatocyte data.
11. de Boer LM, et al. Statin therapy and lipoprotein(a) levels: a systematic review and meta-analysis. European Journal of Preventive Cardiology 2022;29(5):779–792. Network meta-analysis of 39 randomised trials and 24,448 patients finding no clinically important difference in Lp(a) between statins and placebo — the counterweight to the 2020 analysis.
12. Burzyńska M, Jankowski P, Banach M, Chudzik M. Is a single lipoprotein(a) measurement enough? Results from the STAR-Lp(a) study. Medical Sciences 2025;13(4):320. 1,263 patients with repeat Lp(a) at least a year apart: median intra-individual change 16.7%, with 44% shifting by 20% or more — challenges the measure-once recommendation.
13. Design and rationale of Lp(a)HORIZON trial: assessing the effect of lipoprotein(a) lowering with pelacarsen on major cardiovascular events in patients with CVD and elevated Lp(a). American Heart Journal, 2025. Design of the 8,323-patient phase 3 outcome trial of pelacarsen — the trial that will decide whether lowering Lp(a) prevents events.