Which lipid measures actually move with 18% weight loss, and which barely budge?

Down 18% of body weight over ten months. My triglycerides have collapsed from 3.1 to 1.1 mmol/L, which is the change everyone predicted. My LDL-C has gone from 3.4 to 3.2, which is essentially nothing, and I was expecting it to fall proportionally. HDL-C is up from 1.02 to 1.18. My Lp(a) was 168 nmol/L before and 171 nmol/L now, which I assume is measurement noise on an unchanged value.

So the question: is this pattern normal, or is my LDL-C behaving oddly? I had assumed "lose weight, lipids improve" applied uniformly, and instead I have one dramatic change, two modest ones and one that is inert. If the different measures respond differently I would like to understand which mechanism drives each, so I stop expecting the wrong things.

Secondary: given the above, which of these numbers is worth repeating, and which should I stop paying for? I have been running a full panel every eight weeks and I now suspect most of that spend was buying me noise.

Expanding on Lp(a), because "essentially fixed" understates how unusual it is among the things on a lipid panel and there are two practical traps.

Why it does not move

Plasma Lp(a) concentration is determined predominantly by the LPA gene, with heritability estimates typically in the 70 to 90% range. The dominant determinant is the number of kringle IV type 2 repeats, which sets the size of the apolipoprotein(a) isoform and inversely correlates with plasma concentration, plus a set of common variants. Diet, weight, exercise and statins do not meaningfully change it. Statins may nudge it slightly upwards. Niacin lowers it and does not reduce events. The agents that lower it substantially are antisense oligonucleotides and small interfering RNAs currently in outcome trials, and until those report there is no established Lp(a)-lowering treatment with proven benefit.

So the correct mental model is not "a lipid to improve" but "a fixed risk multiplier to discover once". Its function is categorical: a high result moves you into a higher-risk stratum, which tightens the targets for the things you can change — chiefly ApoB and blood pressure — and flags first-degree relatives for testing, since it is inherited.

Trap one: the units

Lp(a) is reported either in mg/dL as a mass concentration or in nmol/L as a particle concentration, and there is no valid universal conversion between them. The reason is structural: mass depends on isoform size, which varies between people, so a fixed multiplier is wrong by a person-specific amount. The commonly used approximations of 2.0 to 2.5 nmol/L per mg/dL will place you on the wrong side of a threshold reasonably often.

Practically: your 168 nmol/L is a molar result and is comfortably above the widely quoted 125 nmol/L boundary for elevated risk. If a later result comes back as, say, 60 mg/dL, do not multiply it and compare — ask which assay was run, and if you need to compare, get the same assay repeated.

Trap two: reading the change as real

Your 168 to 171 nmol/L is a 1.8% difference. Analytical imprecision for Lp(a) immunoassays is in the 4 to 8% range, and there is modest within-person biological variation on top, so the reference change value is on the order of 20 to 30%. Your two results are the same result. This matters because people who see a small rise conclude something is wrong and repeat the test annually forever, and people who see a small fall conclude their intervention worked.

The one situation where a genuinely changed Lp(a) is worth knowing: it rises with substantial renal impairment, particularly nephrotic-range proteinuria, and in acute inflammatory states, and it falls with severe hepatic dysfunction and in hyperthyroidism. If a repeat comes back 60% different, look for one of those rather than assuming assay failure.

One addition on the HDL-C rise, since it is the measure most likely to be misread as a success.

HDL-C is largely reciprocal to triglyceride-rich lipoprotein burden. When VLDL and remnant particles are abundant, cholesteryl ester transfer protein shuttles cholesteryl ester out of HDL in exchange for triglyceride; the triglyceride-enriched HDL is then remodelled and cleared faster, so HDL-C falls. Reduce the triglyceride-rich pool and that drain closes, and HDL-C rises passively. This is why the HDL-C rise lags the triglyceride fall by months and why its magnitude is roughly predictable from the triglyceride change.

Which means your +16% HDL-C is not an independent finding. It is the same event as your −65% triglycerides, observed from the other side. Counting both as improvements double-counts one physiological change.

The reason this matters beyond bookkeeping: HDL-C has repeatedly failed as a treatment target. Pharmacological HDL-C raising, most notably with CETP inhibition, has produced large increases in HDL-C without corresponding event reduction, and Mendelian randomisation studies have generally found no causal relationship between genetically determined HDL-C and coronary disease, in sharp contrast to the findings for ApoB-containing lipoproteins [1].

The practical reading: HDL-C is a barometer, useful for what it tells you about the triglyceride-rich pool and about insulin sensitivity, and not something to optimise. A rise alongside a triglyceride fall is confirmatory. A rise in isolation, with unchanged triglycerides, is more likely to be alcohol than progress.