Dietary supplements are, for most people, largely unnecessary. Doctor friends I’ve spoken to say that, by and large, “you just piss them out”. Your body needs the various vitamins and minerals and so on, but in tiny quantities, so you get them naturally in your diet; all the rest just makes its way through1.
The dietary supplement industry knows this, but it wants you to buy its vitamins anyway. It can’t say “Buy our vitamins, they’ll make you healthier,” because that’s not true, and they can’t say “Buy our vitamins, you’ll piss them straight down the toilet,” because that’s not the most effective sales pitch. So they choose a third way: they make a true scientific statement which heavily implies, without ever quite saying, that the supplements will help.
Like what you’re reading? Get the free UnHerd daily email
Already registered? Sign in
For instance, they might say “iodine contributes to normal growth”, or “iron contributes to normal cognitive development”. Both of which are entirely true. But what they don’t say is “taking our iodine supplements will help your children grow”, because as long as your child is getting a healthy, varied diet, it won’t. Your child will essentially be acting as a middleman between Holland & Barrett and the sewage system. The supplement producers have not broken any advertising rules, but – I would argue – they have not been completely honest about their product’s effectiveness, either.
Recently, the Government agreed a £161 million deal with a British company called DnaNudge to provide 5.8 million Covid tests, as part of its “Moonshot” programme for mass testing of the population at the point of care. The CovidNudge test is “a rapid, accurate, portable and lab-free RT-PCR test that delivers results at the point of need and in just over an hour”, according to DnaNudge’s own promotional material. DnaNudge is a spinoff company of Imperial College London.
The CovidNudge test is what is known as a lab-on-a-chip, or chip microfluidics: it carries out a PCR (DNA) test at small scale, using vanishingly tiny quantities of sample. “Chip microfluidics are exactly what they sound like,” a microfluidics academic who has his own spin-off company told me. “It’s the classical image of a scientist with a pipette doing mixing different mixtures, heating, cooling, running it in a big instrument, except it’s doing it all on a single chip the size of a matchbox.”
CovidNudge, as I understand it, is a classical PCR test. Users take a swab sample from the mouth and test the mucus for the viral genome. There’s one difference – the SARS-Cov2 virus doesn’t have DNA, it has the related molecule RNA, so CovidNudge has to look for that instead. That means it has to include one more step that a normal PCR test doesn’t have, a reverse-transcription from DNA to RNA, so it’s called RT-PCR. None of this is new or revolutionary technology – it’s essentially next-generation sequencing, which has been around since at least the late 2000s – but it is solid.
I am a little concerned, though, for several reasons. One, while I’m happy to accept that the test does indeed detect Covid, there is a shortage of good evidence that it’s as effective as we would like or need. Two, there are a few questions over its procurement — the timeline is a bit strange. And three, DnaNudge’s other, non-Covid-related products, and those of its founder Christofer Toumazou, seem to be advertised in a way that is somewhat reminiscent of those vitamin supplements I was just talking about.
Let’s look at these in order. First, the evidence. Because DnaNudge is a private company, it – understandably – doesn’t release all its research. That’s fine, but when a company is in receipt of £161 million in taxpayers’ money for its scientific products, some sort of evidence that those products work is probably in order.
Luckily, a study was published last month in the journal Microbe, a subsidiary journal of the Lancet. It took two samples from each of 386 patients. One sample from each patient was tested for Covid-19 in a laboratory; the other was tested by the CovidNudge kit. The results were then compared, to see if CovidNudge spotted all of the positive cases that the lab tests did.
Superficially, the results were quite promising: 71 of the 386 patients tested positive in the lab; CovidNudge detected 67 of them, suggesting that it was 94% as sensitive as the full laboratory test. And it had no false positives.
There are some concerns, though. First, this is already a small sample; an effective sample size of 71. I’d be pretty wary of placing a £161 million investment on the back of it. Second, 94% sounds pretty good, but it’s far from perfect. A biotech investor and former academic I spoke to said “If you take today’s numbers, we had about 6,000 new cases [when I spoke to him in late September]. That means this test would tell about 360 [infected] people that they’re fine and don’t need to isolate.”
In their response, DnaNudge say that “symptomatic individuals should remain in isolation regardless of negative tests”. But NHS guidelines under the current testing regime say that if someone tests negative, then they don’t need to self-isolate unless they feel unwell (which is not the same as being symptomatic).
And there’s a more specific worry. The study validated its own work by not only checking for SARS-Cov2, the virus that causes Covid – it also looked for human RNA. If the test came back negative for that, then they decided that the sampling must have been ineffective. As well as the 386 valid samples, there were 32 invalid ones, 24 of which were being tested with CovidNudge. The other eight were samples tested in the lab.
That’s fine – sampling is imperfect. The concern is that those failed results were then excluded from the analysis. Of the 24 invalid CovidNudge tests, 22 had a corresponding lab test. Of those, apparently, “16/22 (73%) tested negative”. I take that to mean that six of them tested positive. If I’m right, that means that instead of CovidNudge detecting 67 out of 71, it detected 67 out of 77, reducing its sensitivity to about 87% rather than 94%. “I don’t know how it got through the Lancet, honestly,” the microfluidics expert told me, meaning Microbe, the subsidiary journal. DnaNudge say that the invalid tests can be redone, which, they say, would mean this is less of an issue.
And just as important, this testing is under ideal situations, run (presumably) by highly qualified testing staff in a laboratory situation. If 5.8 million of these tests are built and rolled out, then the overwhelming majority of them will not be used by experts but by ordinary people after a few minutes’ training. It is hard to see the errors in the sampling process becoming less of a problem. “PCR is a very sensitive technique,” said the biotech investor. “Once you let it loose in the hands of people who are not well trained, that 94% will get lower.” DnaNudge suggested the best case for sensitivity was actually 98% and that this figure would not get lower as the test was simple to operate.
None of this is devastating to CovidNudge, but – again – I’d want to ask a lot more questions before giving the company £161 million. DnaNudge say, in their response, that “CovidNudge is not being evaluated or deployed for community testing at this time,” which rather raises questions about what we’re spending all this money for.
Second, I mentioned concerns about the timeline of the procurement. According to the health minister Helen Whately, responding to a written parliamentary question from the shadow cabinet minister, Helen Hayes, the DnaNudge contract award was published in the Official Journal of the European Union on 31 July. Imperial College London announced it on 3 August.
Again, that’s fine, but I don’t know what the Government was basing its decision on. The Microbe paper – the only study so far carried out – wasn’t published until 17 September. It was published as a preprint on 15 August, but that’s still two weeks after the decision to award this contract was apparently finalised. It may well be that the Government had access to the data before it was made available publicly, but it’s hardly ideal, again with so much public money going into it.
I put these serious questions about the timeline to DnaNudge, who confirmed they had awarded the contract prior to and in the absence of any peer review because “the study results [had] already been made available to the key groups using the CovidNudge test in clinical practice”. But they didn’t make it clear whether the Government had seen the data. If the Government had not, it is hard to see how the necessary due diligence could have been performed on such a significant contract. And even if they had seen it, the data would not have been peer-reviewed: the Government’s decision to avoid peer-review is questionable, given how important it is to the due diligence process.
None of this would be so much of a problem if the company had a track record in providing diagnostic PCR tests for viral illnesses – which, for instance, Oxford Nanopore, the other company granted money in the Moonshot programme, does. Nanopore was providing real-time PCR tests for the Ebola virus back in 2016. “If you wanted to drop the Moonshot budget on something,” says the microfluidics expert, “[Nanopore] have a longer track record and the technology they use is more precise.”
The geneticist Adam Rutherford, who is on record expressing concerns about the contract award to DnaNudge, agrees: “[Nanopore] is tested in the wild, it’s a reputable and good company who do quick-turnaround DNA sequencing for scientists who use it.”
That is not true of DnaNudge, which Rutherford says on Twitter “tests the boundaries of where direct-to-consumer genetics meets snake oil”. Their main product is a DNA test (“NudgeBox”) that, they say, reveals “your genetic predispositions and risk levels – red, amber or green – to a number of nutrition-related health traits”, and uses that to tell you what foods you should avoid. It comes with a wristband (“colour-customisable DnaBands”) that you can then wear to the supermarket, and use to scan the barcodes of food products. It will then “[flash] red or green to tell you if it’s a good match for your biology”.
Which brings us back to the vitamin stuff. It’s not that our DNA tells us nothing about our diets. When scientists carry out enormous genome-wide association studies (GWAS), looking at tens or hundreds of thousands of subjects, they often find statistically significant correlations between certain “single-nucleotide polymorphisms”, SNPs – that is, single letters of DNA somewhere on our genome – and, for instance, our ability to digest certain foods. But they’re the sort of correlation that requires a sample size of tens or hundreds of thousands to detect.
“They work at a population level, not individual level,” says Rutherford. “So the identification of those variants might have literally no effect on the individual, but at population level might have some statistically significant role in processing sugars per carbohydrates or whatever.”
So while the DnaNudge website shows examples of its NudgeBox saying that Heinz Baked Beans are suitable for someone’s DNA profile but that Heinz Organic Tomato Ketchup isn’t, it’s just not plausible that a DNA test could tell us that, at an individual level.
DnaNudge say, in response, that their NudgeBox uses “well-established genetic markers that relate to the genetic risks for obesity, Type 2 diabetes, hypertension, and blood cholesterol”, and that by helping users make better food choices, they can postpone those diseases. But, says Rutherford, “Any geneticist would tell you that there’s nothing that this test can tell you that’s more personalised than ‘eat less doughnuts and more lettuce, and do more exercise’.”
DnaNudge also say that their product might recommend, for instance, that users replace “Paleo Foods Co. Honey & Pecan Grain-Free Granola” with “Essential Waitrose Porridge Oats with Wheatbran”, and that by doing so for a year, you would consume 135,200 fewer calories and expect to lose nearly 8lb in weight. But that would be true whatever your NudgeBox told you.
Toumazou, the founder, was previously the scientific adviser to another company called GENEU, now defunct. It had a similar idea – it would take a PCR test and use it to give personalised DNA information, except this time for cosmetics. They had a shop on New Bond Street in London and an “exclusive” deal with Selfridges; its creative director was the Duran Duran keyboardist Nick Rhodes.
The idea was that you would pay £200, do a DNA test, have a consultation, and then an hour later you’d get your genetic report and advice and your profile-matched skincare products. Again, this just isn’t something that DNA tests can reasonably do – or, at least, if your skin is greasy, you don’t need to know that you have some SNPs that make it more likely that your pores will overproduce sebum. You can already tell that your skin is greasy; the DNA test doesn’t add anything new.
Toumazou “is a great inventor”, the biotech investor I spoke to said. “The work he has done is groundbreaking and phenomenal on building models of human systems. He’s a world leader in that, no doubt.” But, he says, the problem is overhyping. “I’m a cheerleader for these technologies,” he says. “But I’m also the guy who feels I need to protect them from overhype. I’m in the business of investing for a long time, 10 years, and what we don’t want is that there’s a flash in the pan and everyone hates it and no one invests in the long run.” DnaNudge is “overstating what DNA can do for us”, he says.
The microfluidics expert agrees: “It’s an example of using a technology that doesn’t really work but they’re very precise on what they claim. ‘This [SNP] has been shown to be involved in diabetes,’ yes, but so are a million other things.”
Another microfluidics scientist, who works in the private sector, is a bit more cautious, but essentially agrees. She says it’s “not rubbish science, it’s more like lipstick science. ‘This is what you need to eat for your genetics’ comes across as bull, but there is genuine science behind it. But you need to look at hundreds of thousands of people to get a proper signal.”
The scientists I spoke to also raised concerns about its scalability, wondering whether they have the facilities to build the millions of testing boxes needed – DnaNudge made a loss of £9 million last year, which isn’t unusual for a relatively new startup, but does, according to the biotech investor, raise questions about how they’ll be able to “raise the capital to build the hardware”. DnaNudge did point us to the fact they have announced two manufacturing agreements which they say will facilitate production at the required speed and scale.
One scientist suggested that the use of point-of-care tests like CovidNudge are a strange way to go, when there are 150 or so laboratories around the country, in universities and testing facilities, that could perform the tests more rapidly if you can get the swabs to them quickly. “Ramp up the facilities, ramp up the testing,” he said. “I can get a McDonald’s in 10 minutes from Deliveroo, I can get those turnaround times down.” It’s not as sexy, it doesn’t have a beautifully designed box, but it would probably work. And then you’d have a testing infrastructure that would be in place for the next pandemic. The other scientists I spoke to all agreed.
Or if you do want to do point-of-care tests, then there are other companies that have been doing it for a long time. Oxford Nanopore is one; GeneDrive is another; the Cambridge researcher Helen Lee developed the “SAMBA” lab-on-a-chip years ago which tested for HIV in sub-Saharan Africa.
More important, though: the really key thing we need to sort out is track and trace. That’s “the cornerstone of fighting the pandemic,” according to Rutherford, and at the moment it doesn’t seem to be working.
So what I think we need to know is how the Government knew CovidNudge would work — and agreed to spend vast sums of money on it — before anyone had even seen the data. It’s one thing to piss a fiver away on vitamin supplements, but pissing £161 million on a Covid test from a company with no track record in disease diagnostics is quite another thing.