Report 058 · Supplements
Two federally funded case series ran 60 supplements through a mass spectrometer and compared what came out against what the label promised. Most of the labels were wrong. But the test asked a narrower question than the headline number suggests, and understanding which question it asked changes what you should do about it.
Most of the reports I write on this beat are about whether an ingredient does anything. Does lion's mane improve cognition, does cordyceps improve endurance, does magnesium help you sleep. Those are questions about biology, and they get argued out in trials.
This one sits underneath all of those, and it is a question about chemistry rather than biology. Before you can ask whether the compound works, you have to know the compound is in the capsule. That turns out to be an open question far more often than it should be, and unlike the efficacy debates, it has a clean answer available from an instrument.
The same research group ran both of these: the Consortium for Health and Military Performance at the Uniformed Services University, working with the National Center for Natural Products Research at the University of Mississippi. Both were published in JAMA Network Open. Both used liquid chromatography-mass spectrometry to determine what was actually present, then compared that against the Supplement Facts panel.
The first, published in August 2022, bought 30 immune-support supplements from Amazon in May 2021, selected from "featured" results. Thirteen of the 30 had accurate labels. Of the 17 that did not, 13 listed ingredients that were not detected in the analysis, which makes them misbranded. Nine contained substances that were not claimed on the label at all, some of which the authors note may be considered adulterated. Five products managed both failures at once. Not one of the 30 carried a third-party certification seal on its packaging.
The second, published in May 2024, is the one that should get a service member's attention. The team bought 30 weight-loss supplements in June 2023 from 12 online companies that advertise military discounts, found by searching for exactly that. The results:
25 of 30 products, 83%, had inaccurate labels. 24 listed ingredients on the label that were not detected. Seven contained hidden components not present on the label. Ten contained substances on the Department of Defense Prohibited Dietary Supplement Ingredients List, either declared or hidden. Nine contained phenethylamine or its derivatives, stimulants prohibited in sport under the World Anti-Doping Agency. Applying the Operation Supplement Safety Risk Assessment Scorecard, all 30 products rated as risky.
I want to sit with the framing of that second study for a moment, because it is not incidental. These products were found by searching for a military discount. The discount is the marketing hook, and the population being hooked is one where a positive drug test ends a career. Ten of 30 carried something on the DoD prohibited list, and in some cases it was not on the label, which means a service member reading the label carefully and doing everything right could still fail a test. That is not a nutrition problem. It is a targeting problem.
Here is the part that most coverage of studies like these leaves out, and it is the part I care about most, because reading an analytical result correctly is the actual skill.
Both studies say plainly that they were not quantitative. The 2022 paper puts it this way: "this analysis was not quantitative in nature; therefore, we cannot comment on whether the amount of each ingredient listed on product labels matched what was detected from analysis." The 2024 paper states the same limitation, noting that the amount of each ingredient detected was outside the scope of the project.
So the question the mass spectrometer was asked was: is this molecule present at all? Not: is it present in the amount claimed.
That cuts in two directions, and both matter.
It means "83% inaccurate" is not a claim about underdosing. The dominant failure in both datasets is an ingredient on the label that the instrument could not find anywhere in the product, which is a starker problem than a weak dose. Something the label sold you was, as far as the analysis could tell, simply not there.
But it also means the pass rate is softer than it looks. A product could contain a trace of every labeled ingredient, nowhere near the declared quantity, and still land in the "accurate label" column. Thirteen of 30 immune products had accurate labels by this standard. That standard is presence, not potency. If you are a consumer, potency is what you were actually buying.
This is the same distinction I keep returning to in what a claim technically licenses versus what it implies. A negative result from an instrument means "below the detection limit of this method," not "zero." A positive means "present," not "present in a useful amount." Neither study overstates this. Both flag it explicitly. The overstatement happens downstream, when the number gets repeated without the method attached.
There is a second scope limit worth naming, and it is a straightforward matter of instrument physics rather than any failing of the researchers.
The 2024 paper states that fat-soluble vitamins including A, D, E and K, lipids and oils, and elements cannot be detected or have low sensitivity by liquid chromatography-mass spectrometry. Minerals in particular are not what this technique is for. If you want to know how much magnesium or zinc is in a capsule, LC-MS is the wrong instrument and you would reach for something like inductively coupled plasma mass spectrometry instead.
My own research background is microwave spectroscopy, which is a different technique answering different questions entirely, and nothing in my published work bears on supplements. But the general discipline transfers exactly: every analytical method has a window, and everything outside that window is unmeasured rather than absent. A "not detected" is a statement about the method as much as about the sample. Both of these papers are careful about this. Read a marketing page that cites them and you will not find the caveat anywhere.
Suppose a manufacturer clears all of the above. Every labeled ingredient present, nothing hidden, nothing prohibited. There is still a legal gap you should know about, and it is not a loophole so much as a written-down rule.
Under the federal Supplement Facts regulation, ingredients inside a proprietary blend must be "declared in descending order of predominance by weight" and indented under the blend name, and the quantity given for the blend "shall be the total weight of all other dietary ingredients contained in the proprietary blend." What the regulation does not require is the amount of each individual ingredient. Operation Supplement Safety, the Defense Department's own supplement-education program run out of the Uniformed Services University, states it directly: the specific amount of each individual ingredient in a proprietary blend does not have to be listed, only the total combined amount in the blend.
The practical consequence is what the industry calls fairy dusting. A blend can lead with a cheap filler and include a trace of the expensive active that the whole product is named after, in full compliance, and the label will look complete. Descending order by weight is genuinely useful information and it is the one lever you have here: if the marquee ingredient is listed last in a blend, you have learned something. But you cannot compare a proprietary blend against a clinical dose, because the number you would need is not on the panel and does not have to be.
This is exactly why I have argued in previous reports that a dose you cannot verify is not a dose. It is now also why I think label accuracy and label usefulness are two separate problems. The studies above measure the first. The second is unregulated by design.
Three things, in order of how much they buy you.
Look for a third-party certification mark. NSF Certified for Sport, Informed Sport, or USP Verified. These programs test finished product against the label, which is the exact failure mode both studies documented. Zero of the 30 immune-support products carried any third-party seal, and that is not a coincidence, it is the signal. OPSS makes the same point for service members: laboratory testing is the only reliable way to verify actual contents. If you are subject to a drug test, treat this as mandatory rather than as a nice-to-have.
Prefer a panel with individual amounts over a proprietary blend. Not because blends are inherently dishonest, but because a blend removes your ability to check the one thing that determines whether the product could plausibly work. A company confident in its dose has no reason to hide it.
Be most suspicious where the marketing is most targeted. The military-discount finding is the sharpest result in either paper. A discount aimed at a specific group is a signal about the seller's marketing strategy, and in this sample it correlated with an 83% label failure rate. Affinity marketing is not quality assurance. If anything, it identifies sellers who compete on trust signals rather than on product.
Both studies are case series of 30 products each, not random samples, and both say so. The 2024 authors state directly that their 30 products are not meant to be representative of all weight-loss supplements online, nor of all advertising targeted at service members. The 83% figure describes those 30 bottles. It is not a national prevalence rate and I am not presenting it as one.
Both are also snapshots, from May 2021 and June 2023 respectively. A specific product tested then may have changed. And as covered above, neither measured quantity, and neither could see minerals or fat-soluble vitamins by the method used.
What I think survives all of that: two independent product categories, purchased two years apart from different channels, tested by the same competent lab, both showed most labels failing a presence-only check. That is a consistent pattern in a sector where the manufacturer is responsible for its own quality control, and it is enough to justify changing how you shop even though it is not enough to put a number on the whole market.
Not medical advice. This is educational analysis, not a recommendation — a study is not a prescription. Talk to a qualified clinician before acting on anything you read here. Full disclaimer →
The supplement conversation almost always runs at the level of ingredients. Does this compound do the thing. That is the interesting question and it is the wrong place to start, because it silently assumes the manufacturing question was already settled.
It often is settled, by companies that test and certify and will show you the certificate of analysis. But it is settled by the company, not by a regulator checking before the bottle ships, and the two datasets here are what it looks like when nobody checks: an ingredient on the label that the instrument cannot find, a stimulant in the capsule that the label never mentioned, and a discount code aimed at people who cannot afford either surprise.
Ask what is in the bottle first. The efficacy debate is downstream of that, and it is a debate about a product you might not be holding.
Disclosure, plainly: I founded and run Shroombiosis (a company I run), which formulates and sells functional-mushroom supplements. That is a direct commercial stake in the industry this report is about, and the testing standard I am describing here is the standard my own products have to meet. Shroombiosis does not sell a weight-loss or immune-blend product of the kind either study tested, and nothing here is sponsored or earns a commission; here's the full policy. A recommendation with no stake at all: for performance nutrition, Die Tryin Co. is a fellow combat-veteran-owned brand I'm glad to point people to. I don't own it and earn nothing from the link.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.