Deirdre Parkinson, Clinical Naturopath, Herbalist, Nutritionist.

20/09/2026

Two people on a statin can show the same normal LDL cholesterol and carry different cardiovascular risk, and the variable that separates them is not the amount of cholesterol in the blood but the number of particles carrying it.

Researchers drew on the Copenhagen General Population Study, took 13,015 adults who were already on statin therapy, measured both their LDL cholesterol and their apolipoprotein B, and followed them for a median of eight years of heart attacks and deaths. When the two markers disagreed, the people whose apoB ran high while their LDL cholesterol sat in the normal range had a 49% higher rate of myocardial infarction and a 21% higher rate of death from any cause than the people whose two markers were both low.

The finding that gives this study its weight is the mirror image. The intuitive assumption is that any elevated marker means elevated risk, so a high LDL cholesterol should be as dangerous as a high apoB. It was not. When LDL cholesterol ran high but apoB stayed normal, the risk did not rise at all, for either heart attack or death. Non-high-density-lipoprotein cholesterol, which sums the cholesterol across every atherogenic particle and so tracks the particle burden more closely than LDL cholesterol does, behaved the same way apoB did. The cholesterol concentration on its own carried no signal once the particle count was normal. It was the number of particles, not the amount of cholesterol they carried, that separated the people who had events from the people who did not.

The reason the two markers can diverge is that they measure different things. Every atherogenic particle, whether LDL, VLDL, IDL, or lipoprotein(a), carries exactly one apoB molecule, so apoB is a direct count of how many such particles are in circulation, while LDL cholesterol measures only the mass of cholesterol those particles happen to be carrying. The two readings split apart most in people whose particles are cholesterol poor, the small dense LDL common with high triglycerides and insulin resistance: each particle carries less cholesterol, so LDL cholesterol can fall into the normal range on a statin while the particle count stays high, because the cholesterol measure understates how many particles are actually there. It is the particles themselves, not their cholesterol cargo, that cross the arterial wall and initiate plaque, which is why a marker that counts them tracks disease more faithfully than one that weighs their contents.

This is an observational cohort, so discordant apoB cannot be declared a cause of the events it tracked; it may partly mark people whose metabolic health is worse in ways that independently raise risk, and no observational design fully removes that possibility. The population was specifically statin-treated, so the finding describes residual risk in people already on therapy and does not extend to untreated adults, in whom LDL cholesterol and apoB agree more often. Discordance was defined by a median split rather than a clinical threshold, which sorts people into high and low relative to this cohort rather than to any absolute cutoff. And the study measured association with events, not the effect of lowering apoB, which is a different question a cohort cannot answer.

What the evidence supports, is that in people on statin therapy, apoB and non-high-density-lipoprotein cholesterol reflect a residual cardiovascular risk that LDL cholesterol alone can miss, and that a normal LDL cholesterol does not rule that risk out when the particle count is still high. What it does not support is a promise that lowering apoB to a target will lower any individual's event rate, because demonstrating that requires a treatment trial rather than a cohort. The practical reframe is still worth stating: lipid management overwhelmingly targets the cholesterol number, the value printed on every standard panel, while the particle count that this study flags as the marker carrying the residual risk is one most people on a statin have never had measured. The open question is whether treating patients to an apoB target, rather than an LDL cholesterol target, changes hard cardiovascular outcomes, or whether apoB is simply the more honest readout of a risk that is already there.

Johannesen et al., J Am Coll Cardiol 2021
Sniderman et al., Circ Cardiovasc Qual Outcomes 2011
Marston et al., JAMA Cardiol 2022

07/09/2026

06/09/2026

A prescription is treated as something that ends when the course does. In 2,509 Estonians, drugs stopped years earlier were still visible in the gut.

That antibiotics leave a mark is old news. Four people given a week of one antibiotic were tracked against four controls for two years, and their bacterial communities never returned to what they had been. The newer and stranger idea is that ordinary medicines do something similar. When more than 1,000 marketed drugs were screened against 40 gut bacterial species, 24 percent of those aimed at human targets rather than bacteria stopped at least one species growing, drawn from every therapeutic class.

The most direct demonstration of that overlap is in this study. The team trained a pattern-recognition model on what one antibiotic class does to the gut community, then asked it to look at people who had never taken that antibiotic. It picked out, better than chance, who was taking metformin for diabetes and who was taking antidepressants. To the gut, those drugs leave a signature resembling an antibiotic's. The reason such damage may not repair is that gut species feed on each other's waste products, so removing one member can collapse a chain that does not simply reassemble once the drug stops.

Linking stool samples to five years of dispensing records rather than asking people to remember, the study tested 186 medicines with at least 20 users. Of those, 167 were associated with the gut community. Seventy-eight were still detectable in people who had stopped more than a year earlier, and for two antibiotic families, sleeping and anxiety tablets, and antidepressants, the mark was still there beyond three years. Benzodiazepines, the anxiety and sleep class, affected overall composition more broadly than several antibiotic classes did.

Several things constrain this. It is observational, and a prescription collected from a pharmacy is assumed to have been swallowed. The condition being treated shadows every drug: someone on antidepressants for years differs from someone never prescribed them in diet, sleep, stress and movement, and no statistical adjustment removes an exposure that tracks with an entire way of living. The cohort is one Estonian volunteer sample, 70 percent women, and biobank volunteers are healthier than the population they stand for. Only prescriptions were captured, so everything bought over the counter is invisible. And nothing here measures anyone's health.

One result is harder to dismiss than the rest. Among 328 people who gave a second stool sample about four years later, starting a drug moved the community one way and stopping it moved the community back. Effects that reverse on withdrawal are much harder to blame on confounding than effects that merely travel alongside each other.

The finding with the longest reach is also the bleakest. Microbial richness in people with any antibiotic history never caught up with people who had none, regardless of how much they had taken or how long ago they had stopped. There was no dose low enough and no interval long enough for the gap to close.

The idea is not that anyone should stop taking their medication, and not that a course of antibiotics five years ago is harming them today. Nobody has shown that. It is that the gut community is a record of everything a person has ever taken rather than a snapshot of what they are on now, and that this record is written by far more than antibiotics. Two-thirds of this cohort were on at least one prescription. Whatever the gut is doing for them, it is doing it with a history.

Aasmets et al., mSystems 2025;10(10):e00541-25 · PMID 40910778
Maier et al., Nature 2018;555(7698):623-628 · PMID 29555994
Jernberg et al., ISME J 2007;1(1):56-66 · PMID 18043614
Krigul et al., Gut Microbes 2024;16(1):2377570 · PMID 39034613

02/09/2026

Earlier today, we shared research that followed the effects of glyphosate exposure through 𝘁𝗵𝗿𝗲𝗲 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗼𝗻𝘀 𝗼𝗳 𝗿𝗮𝘁𝘀..

Now another newly published study leaves us with a very different — but equally interesting — question.

𝗛𝗼𝘄 𝗹𝗼𝘄 𝘄𝗼𝘂𝗹𝗱 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿𝘀 𝗵𝗮𝘃𝗲 𝘁𝗼 𝗴𝗼 𝗯𝗲𝗳𝗼𝗿𝗲 𝗴𝗹𝘆𝗽𝗵𝗼𝘀𝗮𝘁𝗲 𝘀𝘁𝗼𝗽𝗽𝗲𝗱 𝗽𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝘀𝗶𝗴𝗻𝘀 𝗼𝗳 𝗴𝗲𝗻𝗲𝘁𝗶𝗰 𝗱𝗮𝗺𝗮𝗴𝗲 𝗶𝗻 𝗵𝘂𝗺𝗮𝗻 𝗰𝗲𝗹𝗹𝘀?

They tested five concentrations.

100 mg/L.
10 mg/L.
1 mg/L.
0.1 mg/L.
And finally, just 𝟬.𝟬𝟭 𝗺𝗴/𝗟.

At every concentration tested, glyphosate produced a significant increase in micronuclei — a recognised marker of genotoxicity, including chromosome damage or disruption during cell division.

Even at the lowest concentration, the effect was still there.

So where was the point at which the effect disappeared?

𝗪𝗲 𝗱𝗼𝗻'𝘁 𝗸𝗻𝗼𝘄.

The researchers didn't go any lower.

This was a laboratory study using human intestinal cells — not a study of people — so it cannot tell us that everyday glyphosate exposure will damage someone's DNA.

But perhaps that isn't the only question worth asking.

When evidence of genotoxicity is still being detected at the lowest concentration researchers choose to test, 𝘀𝗵𝗼𝘂𝗹𝗱𝗻'𝘁 𝘀𝗼𝗺𝗲𝗯𝗼𝗱𝘆 𝘄𝗮𝗻𝘁 𝘁𝗼 𝗳𝗶𝗻𝗱 𝗼𝘂𝘁 𝗵𝗼𝘄 𝗹𝗼𝘄 𝘄𝗲 𝗻𝗲𝗲𝗱 𝘁𝗼 𝗴𝗼 𝗯𝗲𝗳𝗼𝗿𝗲 𝗶𝘁 𝗱𝗶𝘀𝗮𝗽𝗽𝗲𝗮𝗿𝘀?

New article now published.

🔗 Link in the comments.