This is the third in my series teaching you how to read and interpret scientific studies. This time, we’re looking at meta-analyses. These are one of the most important types of statistical analyses that you’ll come across, and there are a few really simple things you need to understand to not get tricked by them.

Part One

Part Two

These pieces are for paid subscribers only. If you like my work, and want to learn more about interpreting biomedical research, consider subscribing!

Subscribe now

Read more

Smoking was a remarkable public health success. A single, simple behaviour that on its own increased your risk of dying by a truly staggering amount. Despite the addictive nature of nicotine it’s still relatively straightforward to get people to change a single facet of their lives.

These days, everyone is trying to find the new smoking. We all want to identify another simple trait that we could fix to save millions of lives. One thing that pops up regularly is the idea that sitting is the new smoking—that spending long periods of time sitting down is a simple, easy-to-change behaviour that makes us sick.

The newest study seems to support this idea. Sitting down is, it seems, causing us to get cancer and die more often. If you just replace an hour of sitting with some movement every day, you’d live a lot longer, according to the headlines.

There are many issues with this idea. Yes, sitting down a lot probably isn’t ideal for your health, but there’s no reason to buy a standing desk just yet. Let’s look at the science.

The Data

The new paper that’s making headlines is a large observational study based on the large UK Biobank cohort. The authors had data on about 100,000 people who wore accelerometers—things that measure your movement, similar to a Fitbit—for 7 days between 2013 and 2015. In a model correcting for a range of common confounding factors, they found that sedentary behaviour—sitting down—was associated with an increased risk of dying from various cancers. They also found that people who had more “interrupted sedentary time”—defined as getting up frequently during periods of sitting—were less likely to die from cancer.

The numbers were, however, a bit unimpressive. People who spent the most time sitting down—on average 13+ hours per day—had a 2.8% risk of dying from cancer in the decade of the study. People who spent the least time sitting down—less than 10.7 hours per day—had a 1.3% risk of dying. That’s not nothing, but it’s not quite the sort of health impact that you’d see from smoking.

The findings were also quite complicated. The authors ran what’s called a replacement analysis where they looked at what might happen if people were to replace one activity with another. They found that replacing prolonged sedentary time with interrupted sedentary time actually had no impact unless the interrupted sedentary time included exercise. That is, standing up was not associated with better health unless you went for a walk or did some star jumps.

This sort of study also has all of the usual issues that you see in observational research. For one thing, the authors only had movement data at one time point. They got 7 days of physical activity for each person sometime between 2013-2015, and had to use that to define their movement for the entire study.

This creates some obvious issues. Reverse causality, for example, is when your outcome variable—in this case, cancer—is causing the intervention variable. In other words, people who were experiencing the early stages of cancer or other health problems that would eventually lead to cancer might have already been spending more time sitting down at the start of this study. That would make the results here largely meaningless, and it’s a possibility the authors do not appear to have looked into.

There’s also the ever-present issue of unmeasured confounding. People who spend less time sitting were are different in lots of ways to people who don’t sit down much at all. In this study, the people who spent the least time sitting down were younger, smoked less, drank less alcohol, exercise far more, were thinner, and were generally much healthier than people who sat down a lot. It’s likely that there were some things that the authors simply could not measure which also influenced the cancer risk, which makes it hard to know if the increased risk of cancer related to sedentary time was caused by the sitting or by something else that the study couldn’t control for.

Exercise Is…Good?

The bottom line here is that the study doesn’t really tell us anything new. People who get more exercise are healthier than people who get less. There’s probably some causal relationship there, because we know that exercise improves your health in a multitude of ways.

Whether this is really related to sitting down or not is up for debate. I think you could argue that the results from the replacement analysis prove that sitting has very little to do with the issue—instead, it’s how much exercise people are getting and how much of that exercise is vigorous vs light.

At best, the study shows that if you sit down a lot it’s probably better to go for a walk once every 90 minutes or so than to stay sitting down the entire time. That is, I think, not really news to anyone. Yes, sitting down for your entire day probably isn’t ideal for your health, but there’s still no strong evidence that standing up for the same period of time is good for you. The real trick is to get more exercise, but unfortunately that sort of boring health advice rarely makes headlines.

In the world of science, the main process we have to deal with seriously bad research is called retraction. This hearkens back to the old days of scientific publishing when articles were in paper journals read by a small handful of academics. They would issue a retraction notice in the newest issue of the journal to let everyone know that the editors no longer trusted an article for whatever reason—obviously, in paper journals, you couldn’t simply remove everyone’s access to the article from a previous issue.

These days, the process is much simpler, because most scientific articles are published online. While some journals do still exist in print, and will issue old-style retractions, most will simply update the online page to note that the journal has retracted the paper and slap a big red RETRACTED watermark on the PDF. This process is used for everything from studies with serious mathematical errors to plagiarism to numeric fabrication—it’s mostly reserved for the biggest mistakes or crimes of academia.

I’ve written before about how this whole process is a bit broken. It often takes years for journals to do anything about even the most obvious misconduct, and sometimes they simply never do. I was once told by an editor that he wouldn’t retract a paper even after the author sent through a clearly fake dataset because he thought readers would still be interested in the work.

It’s a system that relies on trust to the point that you might call it painfully naive.

But what about the worst of the worst? People who’ve had dozens of retractions for all sorts of problematic issues, and whose papers are undeniably unreliable? Not all of their work gets retracted, largely due to the endless swathe of red tape involved, but surely no one would cite their work?

We set out to look at precisely that question. The answer is a bit depressing.

Read more

Positivity is everywhere these days. You can’t go more than a mile without seeing a bunch of posters extolling the benefits of always looking up, or having an activewear-clad smiler tell you that all you need to do is put a smile on your face and the world will improve. This is not helped by the now-debunked—but still common—belief that simply smiling would make you feel more happy.

People love the idea of positivity. If all you need is a sunny outlook to make your life improve, that makes everything so much easier. Which means that journalists leap on any study showing even a slight benefit for positivity like starving wolves sighting a lame elk.

The most recent example of this phenomenon is a series of headlines saying that being positive is the one secret trick to aging well. Apparently all you need is a “healthy mindset” and you can improve your brain and body as you get older.

Unfortunately for everyone who is reading this with a forced smile on their face, the evidence is incredibly weak. There’s no reason to believe that simply changing your thoughts can prevent age-related decline.

Let’s look at the data.

Positive Age Beliefs

The new paper that has everyone all agog about positive thinking is a large epidemiological study looking at “positive age beliefs” in a cohort of people in the United States. The authors took a group of people enrolled in the Health and Retirement Study, and analyzed those participants who had scores at baseline and follow-up for cognitive function—brain health—and physical activity. While the original study included more than 20,000 people aged 65 and above and was nationally representative, this new paper only had brain measurements for 11,000 and physical measurements for 4,000 of the initial sample.

The first thing that the authors noted was that a reasonable portion of the sample improved either physically or cognitively. Indeed, 45% of those that they had data for showed a higher score at the end of the study than at the start on either physical or cognitive scales.

But that’s not the headline finding. At the start of this study, people were asked a series of five questions about how they feel about aging. Some examples of these: “The older I get, the more useless I feel and I am as happy now as I was when I was younger”. The authors dichotomized the responses—splitting people into either high or low positivity—and looked at the correlation between these five questions and their long-term physical and cognitive ability. In a model that controlled for various confounding factors, they found that people who had more positive age-related beliefs had 4% (p=0.049) higher odds of having physical and/or cognitive improvement over the 12 years of the study than people who had fewer positive beliefs.

And from this, headlines.

There are so many huge gaps here that it’s quite hard to know where to start. The entire thing is a bizarre example of what happens when no experts are involved in communicating scientific studies at any point in the process.

For one thing, there’s the issue of selection bias. The authors only had information on half of the study for any outcomes—what happened to the other 9,000+ people? It’s quite likely that people who experienced serious health problems dropped out of the cohort, which would cause a huge problem for this sort of analysis.

Also, while 45% of people saw some improvement in either of their scores over time, the average results dropped. A lot of people had some improvement in one, but a drop in another. So even in this highly-selected group, the overall result was a decline in both physical and cognitive ability over a decade of aging.

The result is also hilariously weak. A 4% increase in odds that you have any improvement in physical or cognitive ability is basically meaningless. It’s the sort of effect size that’s only noticeable in huge cohorts, because it’s so tiny. Also, the authors didn’t include key confounding factors like cancer and respiratory disease in their statistical model, so there’s every chance that minor changes to the modelling here would render the results null.

There’s the issue of reverse causality. People who are 65 and already have a negative view of aging are probably not doing that great. It’s perhaps entirely understandable that such people would end up doing worse on measures of physical and brain health over a decade.

The positivity scale is very inadequate. It mostly measures how people feel about aging, but that’s closely tied to all sorts of things. Whether this actually measures a positive outlook or not is entirely up for debate.

And on, and on. It’s not a very good study, and at most it tells us that there is a very weak correlation between answers on this particular subscale of the Philadelphia Geriatric Center Morale Scale and long-term physical/cognitive scores in this highly-selected cohort of people. What that means for aging is anyone’s guess.

Healthy Aging

We’ve known the “secret” to aging well for decades. It’s the same simple, boring health advice that you hear all the time—exercise more, don’t drink alcohol, don’t smoke, eat less food. Most other things turn out to be more noise than signal, and this looks to be no different.

If you want to try and have a positive outlook on life, go ahead, but the evidence that thinking positively will actually improve your health is not particularly strong. This study in particular mostly just shows that mediocre research can still make headlines as long as no one looks too closely at the data.

Subscribe now

It feels like there’s a new cure for aging every week. Where once we talked about health in general terms, or wanted stuff that enhanced our wellbeing, the new quest is to break the bonds of time and stop people from getting older. Or, at least, to stop us all from experiencing the health problems that happen when you reach a certain age.

The newest story that has everyone very excited about no longer getting older is about a nasal spray. A new study has, it seems, shown that we can prevent age-related disease with nothing but a couple of puffs up our noses. If true, this would be revolutionary for anyone who is concerned about their own mortality.

Sadly, the data does not hold up to even the most basic of scrutiny. This is yet another example of extremely early research that is unlikely to ever be used in real people to improve their health.

Let’s look at the science.

The Study

The recently published piece of research from the journal of extracellular vesicles has the rather fascinating and difficult to pronounce title “Intranasal Human NSC-Derived EVs Therapy Can Restrain Inflammatory Microglial Transcriptome, and NLRP3 and cGAS-STING Signalling, in Aged Hippocampus” .

The specifics of what the researchers did are quite convoluted, but in essence they looked at what happens when you give nasal sprays of human stem cells to areas of the brain. They gave these sprays in old age, and examined how that impacted the brains of the treated patients. When dissecting the brains at the end of the study, they found that those who had received the nasal sprays of these stem cells had less inflammation and fewer negative brain changes than those who had received another type of treatment that acted as the control.

That sounds impressive, or it would if the patients were people. Of course, you can’t really do dissections of brain tissue in most human studies, so the clever among you have probably already realized the issue here: the study was in mice. Specifically:

“The study comprised two cohorts of C57BL/6 mice: young adult (3 months old) and late middle-aged (18 months old).”

So what we can say is that there are some interesting findings—and to be clear, the findings are quite interesting—that show that these intranasal sprays may have an impact on mouse brains. This has, unfortunately, very little meaning for the average person who is interested in their brain health.

Most treatments that are tested in mice fail to make it to humans. Sometimes the mice just react differently, sometimes the treatments are unexpectedly deadly to people for some reason, and sometimes we don’t really know why they aren’t working—whatever the explanation, fewer than 1% of treatments that show some benefit in rodents end up treating real humans in the real world.

Bottom Line

I’m not trying to attack basic science here. We need studies of new chemicals and techniques in rodent brains otherwise we’d never develop anything that works in people. It’s vital that we do studies like these, because they are what eventually leads us to things that work.

That being said, this is very early days. Giving some mice nasal sprays is the very first step in a decade-long process that might, if the researchers are very lucky, lead to something that works. Chances are it won’t end up curing aging either, because real life is rarely that neat. The most likely outcome by far, however, is that this does not pan out and we never hear about it again.

I’ve covered dozens of these stories in my decade working in the science communication space. The main problem is that we hear about them at all. It’s interesting research, but of the thousands of similar stories only one or two will end up being anything other than a footnote in a report about failed treatments that we’ve already tried.

In 2023, I started a series for paid subscribers looking at some basics of epidemiology and evidence appraisal. Stuff that you probably wouldn’t know unless you’ve done a masters of public health or similar, but which is really important for understanding how the world of scientific evidence works. For various reasons—mostly other stuff coming up—I stopped writing the series and haven’t picked it up since. Fortunately for you all, a devoted reader looked at those early pieces and asked if I’d resume the series. I think it’s a great idea, so thanks reader!

If you haven’t read part 1 of systematic reviews you can find it here:

I’d recommend reading that first. I go into the history of the subject, and why you should care about this sort of scientific study. Now, on to the meat of the issue. If you come across a systematic review, what should you do?

Read more

It’s been two years since the Cass review into gender identity services in the UK came out. Since then, the UK has progressively restricted access to gender clinics of all kinds for trans teens, now offering only a handful of appointments a year. The new clinics exclusively focus on psychological and psychosocial interventions, with the UK government having banned the use of puberty blockers entirely based largely on Cass’ recommendations.

The latest foray into removing access to medical interventions for transgender youth is the new NHS policy on hormone therapy. The draft policy proposition from NHS England states:

“NHS England has carefully considered the evidence reviews conducted by Solutions for Public Health (2026) and has identified and reviewed any further published evidence available to date. We have concluded that there is not enough evidence to support the safety or clinical effectiveness of feminising and masculinising medicines to make the treatment routinely available to children and young people at this time. This includes both individuals who identify with a binary, or a non-binary gender. “

This is based largely on a series of reviews conducted by Solutions for Public Health (SPH, an NHS evidence reviewing body), which are also linked in this policy proposal. The organisation has conducted a remarkable 10 systematic reviews and meta-analyses looking into a relatively simple question, and come to the conclusion that there is no good evidence that hormones are beneficial for transgender young people.

This is interesting, because the previous NHS-commissioned review, conducted by the University of York as part of the Cass review, found that there was “moderate quality” evidence suggesting a benefit for hormones. These new systematic reviews are, in theory, an update on that data, but somehow found substantially less evidence that hormones help transgender teens.

I was commissioned by Gender Plus, a private clinic for gender identity treatments in the UK, to look into these studies and provide a report on them. They are, in my opinion, largely useless as a guide for policy in the UK or anywhere else. They have serious weaknesses, and do not allow us to make useful decisions about, say, whether to ban hormones for transgender teens in the UK’s gender clinics.

Methodological Weaknesses

The first thing that I found in my report is that the reviews are quite weak from a methodological perspective. Specifically, they don’t report quite a lot of the important things that you’d expect systematic reviews to report. A systematic review is a way of aggregating all of the evidence together on a specific topic, and the methodology is therefore basically just how the authors identified studies and grouped them together. Usually, you report things like how you downloaded the studies, who did the extraction, and how exclusions were made.

These 10 reviews do not report any of those things. Instead, they refer to an internal NHS document called “Guidance on conducting evidence reviews for Specialised Services Commissioning Products’ (2020)”. There’s no information on what that document contains, and I could not find it online, which means that we simply do not know important things about these reviews.

For example, one common issue in systematic reviews is that it is not simple to decide whether a study meets or does not meet the inclusion criteria. It can be easy to miss specifics that might cause a study to be dropped from the analysis. That’s why we usually get two people to independently review the studies we’ve identified in our searches, and if they disagree on what gets included we have a third person on standby to be the deciding vote.

But in these SPH reviews, they don’t say that they did this. There’s just that reference to an internal document that may or may not include this important step. It’s possible the authors did do their reviews using standard methodology—it’s also possible that they did everything completely wrong. We just don’t know.

There are many basic measures of good systematic review practice that these SPH reviews don’t describe. That’s a serious deficiency, and makes it much harder to trust the results at the outset.

Excluding Most Evidence

The methodological weaknesses are bad, but much worse was the way the reviews put together their questions. The studies that they actually included. In all 10 reviews that they conducted looking at hormones for gender diverse teenagers, they included a total of just 11 studies. 6 of the reviews included no studies at all. In contrast, the York University review included 53 studies on the same broad question. How is this possible?

The issue comes down to the specific questions the SPH authors asked. A systematic review is fundamentally about finding the literature looking at a medical question. The method used by SPH to define these questions is called PICO—Population, Intervention, Comparator, Outcome. For example, we might want to know what impacts metformin, a diabetes medication, has on mental health compared to a placebo in people with pre-diabetes. That would be:

P: people with pre-diabetes

I: metformin therapy

C: placebo pills

O: mental health

There’s an important point here though. It is quite easy to make your PICO framework too broad—there are probably quite a few studies on the above question, making it a VERY tedious process to try and comb through them—but it’s also possible to make the PICO too narrow. If we changed the above PICO framework to something rather more specific:

P: women aged 40-55 with pre-diabetes and no history of mental health disorders who do not drink alcohol, take a statin, and exercise regularly

I: metformin taken twice a day

C: matched placebo pills taken in the same schedule as the metformin

O: a composite outcome including depression and anxiety scores

We would probably end up finding no studies at all looking at our new question. But the issue here is not that there is no research which might inform us, it’s that we’ve chosen a ridiculously narrow question to ask of the data. There are probably no randomized studies of metformin on this extremely specific population, but we have lots of randomized trials including many different groups that would almost certainly give us a strong answer as to what metformin would do in this situation.

In the SPH reviews, they made a simple mistake. Their questions were FAR too narrow. The example I picked in my report was review 2417k. This review had the PICO:

P: Children and young people (up to their 18th birthday) who have gender incongruence as defined by the study and identify as non-binary and wish partial physical masculinisation.

I: Masculinising medicines comprising testosterone monotherapy.

C: One or a combination of: 1. Psychological and psychosocial support 2. Social transitioning to the gender with which the individual identifies OR 3. No intervention

O: Impact on gender incongruence, Impact on mental health, Impact on Quality of Life, Masculinising physical changes, Psychosocial impact, Fertility, Feasibility of masculinising genital surgery, Cognitive outcomes, Detransition after receipt of masculinising medicines, Regret after receipt of masculinising medicines, Safety, Cost-effectiveness

This is patently absurd. For one thing, I have not been able to find anyone else who uses the phrase “partial physical masculinisation”. The concept of giving hormones to non-binary youth to cause partial changes in their gender expression is mostly theoretical at this point, and there are no studies looking at what this might mean in practice.

In addition, the ages are defined quite tightly. There’s little practical difference between hormone use for 17 year olds and 18 year olds, but the authors decided that they would exclude any study that looked at people over the age of 18 (this is also bad in another way—we’ll come back to that).

On top of all of this, they are only looking at “testosterone monotherapy”. That means that kids who ever got puberty blockers are excluded from this and all of the other SPH reviews that discussed monotherapy. The SPH team did run 4 reviews on combination therapy—hormones and puberty blockers—but they only included studies where the hormones and puberty blockers were provided concurrently for all patients in the paper.

This is ridiculous because any study where some portion of children got puberty blockers before hormones was automatically excluded from all the SPH reviews. That’s a large portion of the evidence-base on hormone use in trans teens. Worse still, it was standard NHS policy to require puberty blockers be given before hormones for children in the UK for nearly a decade, meaning that SPH automatically excluded all data from their own country as part of their review process. The reviews noted that “The generalisability of the findings to the UK is limited.”, but this was entirely due to the authors disregarding any UK evidence as a key part of their protocol.

It’s also problematic because most studies combine both gender transitions in their analyses. That is, they look at the impact of gender-affirming hormones for both young trans men and women at the same time. The biggest single cohort of trans youth taking gender-affirming hormones, for example, does not break out its analysis by gender, which means that it was excluded from all of the SPH reviews as well. One of the largest analyses of trans youth—which showed a reduction in suicidality for teenagers prescribed gender-affirming hormones compared to those who did not get hormones—was excluded because results were not published separately for those receiving testosterone vs oestrogen.

As I pointed out in my report, there are at least a dozen papers that show benefits for hormone use that SPH excluded in this manner. They defined their questions so narrowly that basically no studies could be included in the reviews, and then concluded that there was insufficient evidence to support the treatment. But the lack of evidence was just due to the mistakes in their own review process.

Inconsistencies

The final issue with the reviews that I discussed in my report was the inconsistency in how they applied their own inclusion criteria.

Remember the thing about age from above. Technically, the authors stated in their methodology that they would exclude any study that had people over the age of 18 in it. However, 7 of the 11 included studies had people >18 as part of the sample. What the SPH team appears to have done is decided that they could include studies with an average age <18. So a paper where the average age was 17.92, and which included a few dozen people over 18, would be included, but a paper where the average was 18.01 would not.

This is, obviously inconsistent. SPH was extremely rigorous with their application of the hormone prescription criteria. There were dozens of papers excluded because those papers had some teenagers in them who were given puberty blockers before hormones with no question about the average child.

But when it came to age, the reviews were much less stringent. There’s no information on how and why this choice was made.

Similarly, the reviews were divided into binary and non-binary groups. Every question was asked both for a binary physical transition—say, male to female—and for a non-binary physical transition (which was not defined by the SPH team). But there are virtually no studies that separate out their patients into those who wanted a binary and those who wanted a non-binary physical transition—they do occasionally just exclude non-binary people entirely—largely because the phrase “non-binary physical transition” appears to have been introduced by the SPH authors as part of these reviews.

The authors noted for all 11 of their included studies that they did not report whether these teenagers were given hormones for a binary or non-binary transition. But they still included them.

Had the SPH authors treated all of their PICOs as similarly important, they would’ve included no studies at all in any of their reviews. That’s rather problematic.

Should Teenagers Be Able To Access Hormones?

In my opinion, these SPH reviews are a total waste of time. You can’t possibly produce 10 reviews including just 11 studies and expect people to take the work seriously. The authors excluded most of the work looking at hormones for transgender teens, and ended up publishing what I would call completely useless systematic reviews.

They also excluded all evidence on simple, boring questions like whether hormones cause physical changes. This lead to the remarkable statement that the SPH team was unsure of whether giving children cross-gender hormones would cause changes to their bodies. Does giving testosterone to someone assigned female at birth change their body? How could we possibly know???

Which brings us to the real question: what does the data say?

We could go back to the York review, which found that there was some positive evidence showing that hormones improve the mental health of trans teens. We could also look at some of the data that’s come out since then, which also appears to be quite supportive of providing hormones to 16 and 17 year olds who want them. We could even read the only randomized trial of hormone therapy for transgender people, which was in young adults and found a fairly large benefit for mental health, and extrapolate to these older teens.

Unfortunately, the evidence is still just not that strong either way. As I’ve said many times before, I have no firm opinions on medication use for teenagers with gender dysphoria. The current best evidence is certainly supportive of hormone use for trans teens, but we still need better data before we can be really sure.

Regardless, what we can say with a great deal of certainty is that NHS England is currently relying on terrible evidence to make decisions about trans healthcare. It’s not just the SPH reviews—the NHS policy guidance document says:

“NHS England commissions the specialist NHS Children and Young People’s Gender Service. The service provides a focus on psychosocial, psychological and psychoeducational support following individual and family assessment.”

But their own evidence review—conducted as part of the Cass review—found that there was literally no evidence looking at these psychological, psychoeducational, or psychosocial interventions. Based on their own data, the NHS is implementing the treatment paradigm with the lowest quality evidence, all while publicly arguing that evidence is key.

I don’t know whether giving 16 year olds with gender dysphoria hormones is necessarily the best thing to do, but what I can say is that it has a lot more evidence behind it than the current NHS proposal.

The question of how many people get Long COVID has always been contentious. The reason for that is pretty obvious—as COVID-19 has decreased in acute severity, causing far fewer hospitalizations and deaths, the argument has become almost entirely about Long COVID and the possible impacts it may be having on our health. If COVID-19 causes few deaths, but leaves lots of people sick with a new chronic disease, then it may still be something that we need to spend time on even now in 2026.

The newest estimate of Long COVID rates has just come out, and it’s remarkably high. A new paper has estimated that 1 in 6 people who catch COVID-19 end up with Long COVID, and that the number is actually increasing over time. If true, this would mean at least a billion people with long COVID worldwide, likely more.

The number is, however, absolutely ludicrous. There is absolutely no plausible way that this many people get Long COVID after an infection—even in 2020 the high-quality data showed much lower percentages.

The Study

The new paper that has caused a furore on Bluesky is called “Long COVID Persistence and Surveillance Gaps Across 58 US Hospitals”. It’s a fairly complicated calculation of Long COVID rates based on a dataset of nearly 500,000 people across 58 hospitals in four US states.

The authors developed what they describe as a “custom artificial intelligence algorithm” which identifies whether people had Long COVID or not. Based on this algorithm, they classified 16.28% of the patients in their cohort as having had Long COVID after their infection.

This estimate is one of the worst that I’ve seen in years.

The first issue is the algorithm. While the authors describe it as artificial intelligence, in practice it’s quite a simple mathematical calculation. The approach initially measures temporal correlations between various conditions. Basically, it identifies whether two things are likely related in the cohort in question, so for example whether chronic heart failure is related to chest pain.

The algorithm itself then uses these correlations to determine whether disease codes—the way that hospitals record the health problems that you have—are related to something an individual had before their COVID-19 infection. For conditions on the list of things that the authors have decided constitute Long COVID, if there is no other defined cause—so for example if someone has chest pain but no chronic heart failure beforehand—and they last for 2 months or more, they are considered to be Long COVID.

This sounds complicated, but in practice it’s really quite simple. There’s a list of conditions that the authors think are all part of Long COVID. If someone experiences one of these conditions for 2+ months after their COVID-19 infection, and it can’t be explained by another health condition that they had before their infection, it is considered Long COVID.

This means there is absolutely no assessment of causality in the study. This is just an estimate of the proportion of people who experience one of the conditions that the authors put on their list after having a COVID-19 infection. There’s a little bit more to it than that, but in essence that’s what the study shows.

In addition, there’s the list itself. While some of the things on the list make sense—for example, chronic fatigue—some of the conditions are truly bizarre. Here are some examples of International Classification of Diseases (ICD) codes that are included in the authors definition of “Long COVID”:

F14.180 Cocaine abuse with cocaine-induced anxiety disorder

O90.4 Postpartum acute kidney failure

D84.821 Immunodeficiency due to drugs

H53.50 Unspecified color vision deficiencies

E86.0 Dehydration

R19.6 Halitosis

E55.0 Rickets, active

J15.5 Pneumonia due to Escherichia coli

The list goes on. While you could argue in some cases that COVID-19 could be a contributor to some of these things, most of them are very straightforwardly not related to a coronavirus infection in any plausible way. Dehydration has a vast array of potential causes. Pneumonia due to E. Coli is, well, caused by E. Coli. Postpartum disease is by definition related to the pregnancy.

This is one of the inherent issues with Long COVID that I’ve been writing about for years. If you define a disorder as pretty much any symptom that could occur, you’re going to find a lot of that disorder in any population. But most of these things are almost certainly not related to COVID-19, and it makes no sense to include them in a disease definition for Long COVID.

On top of this, the authors have proven in their paper that the algorithm they used was terrible at estimating prevalence. The methods includes this paragraph:

Algorithm Validation Through Distributional Robustness

In the absence of site-specific EHR review, we assessed algorithm validity through distributional robustness testing.^28,29 Consistent performance across demographically divergent populations provides evidence of robustness,^30,31 whereas systematic miscalibration would be expected to manifest as divergent prevalence estimates across sites with different demographic compositions.³² The 58 hospitals across 4 regions represent substantially different populations in terms of racial and ethnic composition, comorbidity burden, and health care system scale, providing a rigorous test of algorithm generalizability beyond the development site.

In essence, they are validating their “artificial intelligence” algorithm by checking that different populations end up with similar numbers of Long COVID patients. If you have one group of people with COVID-19 where the algorithm finds a rate of 50%, and another where the rate is 5% then there are probably significant issues in how the algorithm is working.

The authors never again refer to this paragraph in the paper. Which is shocking, because the algorithm clearly fails the test of distributional robustness. Despite similarities in age and disease characteristics, nearly double as many people were classified as having Long COVID in California than Pennsylvania using the algorithm. In New England, only 5% of Long COVID ‘cases’ had endocrine issues, while in Texas 15% of ‘cases’ had these problems. The p-values for these differences—the likelihood that they would happen if the regions actually had the same distributions—are essentially 0.

In other words, the data fails the authors own stated test for robustness, which they never bothered to check. This makes the estimates even more dubious even as an interesting description of the diagnostic codes that people who once had COVID-19 eventually received.

COVID-19 Is On The Decline

I wrote about Long COVID rates using the existing best evidence way back in 2023. At that time, the data indicated that around 1% of people who were being infected with COVID-19 were going on to have long-term symptoms of the disease. The number appeared to be declining, but it was hard to say where exactly it would end up.

That analysis hasn’t really changed. The best evidence on Long COVID rates remains the high-quality studies conducted in the UK in 2021, 22, and 23, which suggested that around 10% of people in the initial 2020 outbreaks had long-term symptoms. This declined to around 1% of people by the Omicron waves, likely due to a combination of infection-derived immunity and vaccination.

I could try and estimate the current incidence of Long COVID—the percentage of infections that result in long-term symptoms. I would be surprised if it was higher than it was in 2023. But there’s really not that much point, because COVID-19 is on the decline anyway.

COVID-19 infections have plummeted in recent years. In 2022, there were around 1.5 infections per person, meaning almost everyone was infected once and many people had more than one infection. In 2023, the number was around 1. In 2024, we stopped recording really good statistics on the number of infections, but every indicator we had showed another huge decline. By 2026, we are looking at record low rates of COVID-19 across hospitalizations, deaths, wastewater data, and cases.

It’s hard to know precisely how many people have had a COVID-19 infection in the last 12 months, but we can look at the data to get a clue. The number of COVID-19 deaths has been halving every year since 2022, and is still falling. It’s likely that only 10,000 people will die of COVID-19 in the US in 2026, compared to half a million in 2021. There are currently no people hospitalized in Germany due to COVID-19, something that has not been true since before the virus initially broke out. Wastewater detections of COVID-19—the number of viral particles showing up in sewerage—show rates that are less than 0.1% of the 2022 waves in Denmark, the US, and elsewhere.

While we can’t put exact figures on the number of COVID-19 cases over the last year, we can say with a great deal of certainty that it is very low. And falling.

That makes all of this discussion about Long COVID incidence a bit pointless. It’s highly unlikely that even as many as 10% of people who catch COVID-19 will experience long-term symptoms, but even if the rate was that high it would still not be especially concerning with COVID-19 rates this low. More realistically, your risk of getting Long COVID today is quite close to 0%.

None of this is reassuring for people who have Long COVID from previous waves. I really feel for the community of people who have been permanently disabled by a disease that is no longer of great interest to most of the world. It sucks to have a condition that very few people want to spend money researching.

That being said, the ludicrously high estimates of Long COVID help no one. The 16.28% figure doesn’t pass the basic sniff test, and it’s clearly wrong. The true rate of Long COVID in 2026 is likely to be at least an order of magnitude lower, if not more.

One of the most common beliefs you’ll find about health is that your life is basically under your control. That your actions largely define how sick you are, and your daily routine can be the difference between dying young from diabetes and running marathons at age 85. It’s an idea that you can see promoted by everyone from Gwyneth Paltrow to RFK Jr., and it’s so seductive for the simple reason that it’s empowering.

If the things you do can make you healthy, then anyone has health within their grasp.

This has reportedly been demonstrated by science. Headlines are proclaiming that people are responsible for 80% of their own ill health in old age, implying that you can take actions to remove most illness from your own life. Simple actions such as exercise and taking supplements can apparently keep you healthy well into your golden years.

The problem here is that this is untrue. It’s also not based on any evidence. While there are of course some actions you can take that impact your long-term health, the evidence shows that many factors which impact your wellbeing are entirely outside of your ability to control.

Oxford Longevity Project

The news about how we can all ensure that we are healthy well into old age comes from a rather surprising source. While the media is describing this all as based on a “study”, in fact the newest estimate of how much of our health is under our direct control comes from a report written by the Oxford Longevity Project. This is a charitable organisation founded by two retired academics—a famous linguist and a biologist—, a medical doctor, and the owner of a supplement company. The project is funded by the supplement company, called Oxford Healthspan, which appears to mostly sell several versions of a compound called spermidine.

The report itself is basically a series of essays. Each of the founders of the project have a few pages where they voice their opinions on healthy aging, with an additional chapter written by a former NHS executive and medical doctor. As far as I can tell, the document does not cite a single scientific study—there are only a handful of references in the document, and none appear to be to any specific evidence.

Where does the 80% number come from, you ask? Well, Sir Christopher Ball, one of the founders of the project and renowned linguist and university administrator, appears to have pulled it out of thin air. He sets out seven lessons to achieve healthy longevity. Number 1 reads (bolding added for emphasis):

“The first explains the nature and causation of the major diseases of longevity, which may prevent us from ‘living longer, better’. We must, above all, try to avoid heart disease, cancer, dementia, ‘diabesity’ (obesity, typically accompanied by type 2 diabetes), falls & accidents, and auto-immune breakdown. These are not primarily, or normally, caused by infection or our genetic inheritance (or just bad luck!) as many believe: the main cause of the diseases of longevity is our own behaviour, our mindset and lifestyle. At least 80% of the responsibility for ill health in old age lies in how we choose to live our lives. Choose wisely!”

Sir Ball appears to be a highly intelligent, lovely man with strong opinions about healthy aging. Among other things, he took up marathon running at age 67, and appears to be generally a fairly remarkable man. The sort of person that we all generally listen to when they say something about our long-term wellbeing.

He’s also straightforwardly wrong. Take diabetes, for example. Type 1 diabetes is obviously highly heritable, as it is an autoimmune disease caused by specific genetic factors. Type 2 diabetes is generally thought to be lifestyle induced, because it is caused by obesity, but that’s not entirely true. Genetic factors explain a large proportion of your risk of type 2 diabetes as well. Depending on study, anywhere between 25-72% of type 2 diabetes is related to heritable factors, which are generally largely outside of individual control.

In addition to genetics and heritability, diabetes is also strongly influenced by other factors that you can’t personally change. For example, your risk of developing diabetes in adulthood is associated with where you grew up and the jobs your parents had when you were a child. These are not things that people can generally control themselves, and are therefore not our responsibility per se. They are, if anything, a societal failing that we could remedy as a group if we really wanted to.

The same is true for most chronic disease. Dementia, heart disease, and cancer are all heavily impacted by both genetics and societal factors beyond individual control. I’m actually a perfect example of this—I am currently the fittest I’ve ever been, and am aiming to run a marathon in the next twelve months at the age of 36. I’m running at least 30km a week, although my 5-month-old daughter is making it somewhat challenging. I’ve also lost around 20kg since my heaviest weight, and kept it off for more than a decade. Despite this, my cholesterol is slowly ticking up as the years pass. I have a very strong genetic predisposition to high cholesterol—every man on my father’s side of the family has elevated cholesterol by their 40s. No matter how much I exercise or how much weight I lose, chances are I’ll need to take a statin before I turn 50.

It is very hard to accurately estimate the proportion of ill health in old age that is attributable to forces outside of our control. It varies from disease to disease—for some conditions, 100% of the disease is uncontrollable, but for others it’s a much smaller percentage—but there are very few diseases where the combination of genetics and environmental factors does not add up to at least 50%. That is, at least half of your risk of chronic disease at any age—or dying of a chronic disease—is likely due to things that you cannot directly control.

There are certainly things that individuals can do to improve their wellbeing. The easiest example is to stop smoking and drink less alcohol—just those two things will increase your healthy life expectancy by a fair bit. If you can do at least two hours of exercise a week and keep yourself within a BMI of around 20-25, you’ll probably live longer as well. But those are not the biggest influences on your life, and they are also impacted by where you grew up, the opportunities you had access to, and a whole wealth of other factors that are not things you can personally control.

It’s nice to think that we are masters of our own destiny. I can understand why Sir Ball, after putting an enormous amount of time and effort into his own health, believes that his work has directly led to his current wellness. But the data does not really support this. While there are some things that people can do to improve their health, a great deal of your overall wellbeing is simply out of your hands.

Subscribe now

I’ve written many times about how supplements are probably a waste of money. That’s because in most cases, common supplements you can buy in the supermarket or pharmacy aisle are either disproven—the evidence shows that they don’t work—or unproven. In most cases, buying them only harms your wallet, because supplements themselves are pretty innocuous health-wise unless you start taking truly absurd amounts of vitamin A.

But recent headlines about omega-3 supplements are challenging that argument. According to recent news, omega-3s—most of which are fish oil tablets—may be destroying your brain tissue and causing dementia.

Fortunately for the large portion of the world that takes a daily pill, this is probably not true. Omega-3 supplements are very unlikely to improve your wellbeing, something we’ve seen proven in study after study, but they are also pretty safe. While the new research is very well-done, it does not prove that fish oil is going to damage your mind.

Let’s look at the data.

Read more

It’s been a remarkable 6 years since the start of the COVID-19 pandemic. Throughout that time, we’ve seen viral misinformation spread across the globe, moving at a pace so rapid that it’s hard to even know where the myths started in the first place. There’s been the consistent promotion of ineffective medications such as ivermectin, the idea that the COVID-19 mortality rate was lower than that of influenza, and a whole swathe of arguments about “lockdowns” that rarely even attempt to define what a lockdown is.

There’s one myth in particular that has been astonishingly robust over the last half decade—the idea that COVID-19 only kills people who are already unwell. This persistent argument suggests that no one really died due to COVID-19, because they were all already very sick to begin with, usually from chronic diseases. It’s not that we had a pandemic, it was the existing issues with diabetes, heart disease, and similar that made the death toll so high.

You can hear this exact argument from RFK Jr. as recently as this April. Here’s the full quote from 1:41:56:

“during COVID, we had the highest death rate of any country in the world. And when you ask CDC, why is that true? They say, well, it's because we're the sickest population. The average American who died from COVID had 3.8 chronic diseases. What was killing them? COVID or the chronic disease? You have to understand infectious disease has a very hard time killing healthy person.” - RFK Jr. April 17th 2026

It’s a remarkable statement. Imagine if we could’ve gotten rid of the pandemic by eliminating chronic diseases! That would be a truly wonderful world.

It’s also total nonsense. The idea has been going around since mid-2020, but it’s never been even a tiny bit accurate. While chronic diseases do certainly increase your risk of death from COVID-19—or anything, really—and many people who died in the pandemic were older and may have had chronic diseases, both the numbers that RFK Jr are citing and the underlying concept are entirely wrong.

Comorbidities

The first thing that we can easily fact-check is that 3.8 number. Where does it come from?

This is something I last wrote about in 2020. The underlying facts have not changed. If you look at the CDC website relating to COVID-19 deaths, it discusses “Comorbidites and other conditions”. The CDC notes that:

“For over 5% of these deaths, COVID-19 was the only cause mentioned on the death certificate. For deaths with conditions or causes in addition to COVID-19, on average, there were 4.0 additional conditions or causes per death.”

This isn’t quite the 3.8 number that RFK Jr. cites, but looking at archived versions of the site I could see quite a bit of variance around that number. The exact figure of 3.8 comorbidities appears to have been sourced from the website in early 2021, as you can see using the Wayback machine. It also appears in RFK Jr.’s book about the pandemic, which he was writing at around that time.

The problem with this 3.8/4 number is that it is not chronic diseases. You can tell this immediately if you look at the table on the CDC website:

The issue is pretty obvious. For example, the second most common condition that occurred in someone who died of COVID-19 was “respiratory failure”. This is not a comorbidity or a chronic disease in the traditional sense, it’s something that COVID-19 directly causes. Calling it a comorbidity is a bit like calling bleeding a comorbidity for someone who was just stabbed.

This all has to do with how we record deaths. The US uses standardized death certificates like this one. Most countries are very similar. These have entries for the immediate cause of death—the thing that finally killed a person—as well as underlying causes of death. So, for example, if someone has a car accident, needs major surgery, and finally dies a week later from a secondary infection in ICU, the infection would be the immediate cause of death, surgical complications would be the secondary underlying cause, and the accident would be the underlying cause.

In addition, there’s a space for “other significant conditions contributing to death”. This is usually where you’d put comorbidities such as diabetes or chronic respiratory issues. But in their reporting, the CDC did not distinguish between a condition contributing to death and a secondary underlying cause—they just reported all of the conditions that were listed as either contributing or causing death.

If you take another look at the table, you can see how ridiculous RFK Jr’s argument is. There were a total of 1,146,242 deaths from COVID-19 when this website was archived. Of these deaths, just under 15% had diabetes listed as a contributing issue. That’s only slightly higher than the US national diabetes prevalence, which is estimated at 12% (or around 15% of adults).

We also know that most of the deaths attributed to COVID-19 were caused by the virus. The CDC has run careful audits to confirm that COVID-19 was the underlying cause of death on individual death certificates rather than some other issue. You can see this in the death statistics, which are still available on the CDC website: in 2021, for example, at least 90% of COVID-19 deaths were directly caused by COVID-19*.

There are also other sources of information that confirm this fact. Excess mortality, for example, shows an increase during COVID-19 waves that is not attributable to other causes of death. The authors of this recent paper demonstrated this by showing that neither non-pharmaceutical interventions (“lockdowns”), war or natural disasters, or other causes of death explained the increased mortality during the pandemic. The deaths were solely caused by infections of the virus itself.

The Neverending Story

I don’t expect this myth to die. It’s not as if RFK Jr. is going to check his facts on any major issue any time soon, and the many people who’ve lied about the pandemic for years aren’t changing their tune either. No one has any great interest in reconsidering their deeply held beliefs about COVID-19, even when those beliefs are built on total nonsense.

But it’s still important to push back on these falsehoods. Yes, some people who died of COVID-19 had comorbidities. But we know, due to the hard work of many scientists, that the comorbidites are not what killed them. The virus is.

Being “healthy” is great, but it doesn’t protect you from all disease. The biggest factor for chronic diseases is age, and it’s hard to escape the clutches of time. It’s true that the US might’ve had a slightly lower death toll from COVID-19 if the obesity rate was a bit lower, but the impact would’ve been extremely small. Probably in the range of thousands of deaths, perhaps even fewer than that.

On the other hand, if there was no COVID-19, there would be at least 1,146,242 fewer deaths between 2020 and 2023.

*Note: interestingly, you can also see that this number goes down dramatically over time. It’s been 66/67% since 2024. That’s actually what we’d expect, because the virus has become dramatically less dangerous. So not only are there fewer COVID-19 deaths than in 2021, there are fewer deaths where COVID-19 was the chief underlying cause. These days, at least 1/3 of COVID-19 deaths are actually caused by something else, which was not the case at all in 2020 and 2021.

As with any disease outbreak, this is a developing situation. It’s possible that the data will change, and we will have to start stressing. I’ll update this piece if there are any major changes to what’s happening with the disease, and please do leave a comment if there’s anything that looks wrong!

Since the start of the COVID-19 pandemic people have been understandably on edge about outbreaks of infectious disease. Everyone is on tenterhooks worrying about the next pandemic that will destabilize the world and kill millions of people. The latest disease to hit the news in a big way is hantavirus, which has infected at least 8 on a cruise ship in the Atlantic and killed 3 people so far.

This is, without a doubt, scary. There’s an eerie similarity to the early days of COVID-19, when we were hearing daily updates about people dying on a very different cruise ship. People are wondering if this is the next pandemic that we will have to live through.

Fortunately, while nothing is ever certain, the chances that hantavirus is going to infect the world is very small. The hantavirus outbreak will probably have no impact at all on the lives of most of the people who are reading this article.

What’s A Hantavirus?

The post-COVID focus on infectious disease has proven to be a fascinating experience for me personally. Every six months, I have to dig in my notes from my masters of public health to read up on another uncommon disease because suddenly its in the news.

Hantaviruses are a group of viruses that mostly infect animals. The majority of these pathogens rarely infect humans, and of the 30 or so that have caused human outbreaks, most die out very quickly. That’s because cases of hantavirus are almost always caused by exposure to animal feces—there are very few documented cases of human-to-human transmission. We will get back to that, because it’s important.

There are two very different syndromes caused by hantaviruses—one mostly impacts the kidneys, the other is a respiratory disease. The kidney syndrome has never been documented to transmit between humans, while the respiratory syndrome has in some cases. The respiratory disease is caused by hantaviruses from the Americas, while the kidney disease is mostly from viruses in other parts of the world.

Hantaviruses have a very wide range of disease severity. While the case fatality rate in documented outbreaks is quite high—somewhere in the region of 1/3 of people who are known to have an infection die—there are almost certainly many more cases out there that we don’t know about. One study in China found that the case fatality rate between 1950 and 1997 was only 3.5%, and even that is an estimate based on people who tested positive to the virus. There are likely people who had less severe disease who did not get tested, and therefore the true case fatality rate—or the infection fatality rate—is probably lower, at least for the two strains of hantavirus that cause infections in China.

There is some direct evidence that the true fatality rate from hantaviruses is much lower than the case fatality rates would suggest. In a 1999 outbreak in Panama, there were 11 cases of respiratory hantavirus and 3 deaths. This is a case fatality rate of 27%, which is of course very high. But the team looking into this outbreak did a big survey of all of the contacts of the hantavirus patients, looking for antibodies to the disease. An additional 40 people in these households had antibodies to the infection. That means that the true infection fatality rate—the percentage of infections that resulted in death—would be more like 5.9%. That’s still high, but it’s substantially lower than the initial figure, and there were probably even more cases out there that the team never identified.

Infectiousness

The other important thing to know about hantaviruses is that they don’t spread easily between people. In that 1999 outbreak in Panama, which was of respiratory disease, there were only two possible case of human-to-human spread. One doctor who cared for patients, and another operating room assistant.

In fact, the only hantavirus out of the 30 where we’ve directly identified human-to-human transmission is the Andes virus. This causes respiratory disease, and is responsible for a large outbreak in 2018/19. In this outbreak, 34 people were confirmed to have been infected, and 11 died. The initial outbreak happened when an infected person attended a party with more than 100 guests, and managed to pass on the disease to 6 other people.

In this outbreak, the disease spread was very slow. There are two main reasons for that:

1. The infectious period started when people became symptomatic. In 52% of cases, the investigation into this outbreak confirmed that transmission happened when the infected person became feverish. Many people didn’t pass on the disease at all, likely because they stayed home when they started to feel sick.

2. The incubation period for the infection was very long. The serial interval—the time from the first symptoms appearing in an infectious person to the time that those symptoms appeared in the next patient—was 23 days. That means that on average the disease took nearly a month to incubate and start spreading each time it infected a new person.

This is why, despite there being two months of uncontrolled disease spread before the Panamanian health authorities and the CDC started intervening, there were only 34 cases in this outbreak total. For context, in the initial days of COVID-19, two months of uncontrolled disease spread was enough to cause somewhere in the region of 1 million infections. This is primarily because COVID-19 had a serial interval of 2-7 days in early 2020, and can be passed on before people started experiencing symptoms.

Epidemiologically, it’s hard to know exactly how much contact is required for someone to get hantavirus. In that 2018/19 outbreak, 6 people were infected from the index case—the first person identified to have the virus—during a 90-minute attendance at a party. That’s not a great deal of close contact, meaning that we know at least in theory that the virus may be spread quite readily in high-risk settings.

Share

The Current Outbreak

This all brings us to the current outbreak. As the World Health Organization is reporting, the cases are centered around a cruise ship which left Argentina in early April. Of the 147 passengers aboard, 8 have become unwell and sadly 3 people have already died. The cruise ship is currently quarantined off the Canary Islands on the west coast of Africa.

In addition, there were a number of passengers who disembarked the cruise ship in St Helena, and island in the Atlantic. These people have all already returned home, which means that there are potential cases spread across the globe.

The virus has also been confirmed to be the Andes strain of hantavirus. This is not ideal, because as I mentioned it is the only hantavirus that is known to cause human-to-human transmission. It is therefore quite likely that we will see at least a few more cases in the coming weeks.

For the people on the cruise ship, the next few weeks are probably going to be awful. My heart goes out to all of them, because they are stuck on a small ship with few facilities, waiting to see if they get a nasty virus that has already killed several people that they were travelling with. It’s not a great situation to be in.

For everyone else, however, the current situation is very low risk. Yes, the Andes virus can pass between humans, but it takes a very long time for that to happen. Unless there has been some major mutations that reduce the incubation time and make people infectious before they become symptomatic, it’s very unlikely that there will be an explosion in cases.

And to be clear, while such mutations are possible, I don’t think they’re very likely. Just look at the cruise ship data—the ship left Argentina, where the initial infections most likely happened, on April 1st. It is now May 7th as of writing, and there are only 8 confirmed cases so far. This fits perfectly with the 2018/19 outbreak—an index case bringing the virus onto the ship, and then infecting a number of people over the course of a week who are only now showing symptoms.

The most plausible explanation based on the data we have so far is that this outbreak will follow the same trend as it did in 2018/19. That means that there will probably be more people from the cruise ship who start getting sick over the next few weeks, but there will be very few additional cases after that. As long as the 23 people who disembarked the ship in St Helena don’t start going to parties if they get a fever, the risk of there being further transmission is extremely low.

It could happen. I know I’ll be keeping an eye on the situation, because it’s always good to be up to date. That being said, I would be very surprised to see a significant epidemic or even a large number of cases from this outbreak. It’s entirely possible that fewer than 30 people will eventually test positive to the disease.

Subscribe now

Ultra-processed foods are the demon of the day. Where once we were all concerned with sugar, or fat, or how many food miles our steak had seen—it’s always better to be eating a poorly-travelled cow—the buzzword of the day is processing. The simple act of processing food is, it seems, enough to make food almost inedibly dangerous.

The most recent slew of headlines in this ongoing campaign of terror is about dementia. Apparently, eating even a single extra packet of potato chips—or crisps, for my British readers—per day is enough to significantly increase your risk of dementia. As CNN put it, “Even small amounts of ultraprocessed foods increase dementia risk”.

It’s the sort of story that feeds very nicely into our general ideas about ultra-processed foods. We all know that they’re unhealthy, and so being told that they cause yet another disease is just par for the course. Obviously, things that are this bad for your health would possible cause dementia as well as a wide range of other conditions.

But the reality is that, like most news about ultra-processed foods—or UPFs—this new story is not nearly as simple as it might seem. It’s possible that you could lower your dementia risk by eating fewer french fries, but the new data certainly doesn’t prove that.

Let’s look at the science.

New Study

The new paper that has everyone so riled up is a cross-sectional observational study looking at about two thousand 40-70 year olds in Australia. The authors asked people what foods they ate using a food frequency questionnaire, and then compared the proportion of UPFs that people reported to their long-term risk of dementia calculated using a calculated figure called the cardiovascular risk factors, aging, and incidence of dementia (CAIDE, for some reason) score. They also looked at scores that measured attention span and memory.

The authors found that there was no association between UPF intake and the CAIDE score. There was, however, some association between how many UPFs people reported eating and their attention span, and there was also an association between UPFs and a modified CAIDE score which only looked at certain aspects of the original score. They then argued that this meant that UPF intake was associated with long-term dementia risk, because a higher modified CAIDE score would potentially be predictive of people getting dementia at some point in the future.

This is all extremely silly.

There are lots of problems here. For one thing, the research was cross-sectional. That means that we only have a single point in time where these people were measured on their eating habits and dementia risk. Without looking at these people over a period of years, we have no idea if the eating habits would really translate into dementia cases or if it’s all just meaningless noise.

The outcome measure is also a surrogate. CAIDE scores do predict dementia, to some extent, but they are not the same as a diagnosis of the disease. Worse, the main CAIDE score was not associated with UPF intake, and the modified CAIDE score has not been tested properly. It’s entirely possible that the modified CAIDE score does not have any association with dementia at all.

Worse still, the authors ran a problematic statistical model. Part of the CAIDE score calculation includes age and physical activity, but the authors also added those things into their regression. This means that their variables were collinear, and it creates some issues for our interpretation of their estimates. In simple terms, it means that the statistical model may be spitting out somewhat meaningless figures rather than useful numbers.

This is all in addition to the usual problems that any observational study of food intake already has. People don’t answer food frequency questionnaires correctly, meaning it’s really hard to know whether the study accurately measures UPF intake. There are all sorts of confounding factors that may get in the way of a relationship between UPFs and dementia and make the study unhelpful. There may be issues like reverse causality—whereby people who have early dementia eat more UPFs—which this sort of paper simply cannot account for.

And on top of all of that, the results from this study are singularly unimpressive. If we take the findings at face value, they suggest that:

1. UPF intake is not associated with the main CAIDE score, and therefore not necessarily tied to dementia risk at all.

2. Every 10% more UPFs you eat in absolute terms increases your modified CAIDE score by 0.24.

Let’s say we ignore point 1 here. What does point 2 mean?

Well, take a person eating 2500 calories per day, who has the median intake of UPFs in this study. That means they’re getting about 20% of their calories from UPFs, which is roughly 500 per day. If this person increased that to 30% of their calories, or 750 per day, they would have increase their modified CAIDE score from 1.54 to 1.78.

This is, indeed, about one medium-sized packet of potato chips. But we have no idea what this really means. There’s no straight line between a modified CAIDE score and the long-term risk of dementia. It’s hard to take any serious meaning away from an 0.24 increase in a score that measures “physical activity, history of hypercholesterolemia, history of hypertension, and body mass index”.

Even at face value, the results tell us very little about dementia risk and UPF consumption.

Share

Eating For Health

Food is a universal obsession, and being healthy is one of the most important aspects of eating for many people. That’s not going to change any time soon—we all want to know which things to eat to avoid diseases like diabetes and dementia.

But most of the diet science that gets media attention is not particularly useful for your life. This is especially true for almost all of the UPF noise. It’s true that eating large amounts of potato chips is probably not a great way to improve your wellbeing, but it’s not as if we didn’t all already know that without the label of ultra-processed to add to the mix.

This study provides essentially no information on what you should and should not eat. The authors didn’t even measure dementia, they just used a modified surrogate score which may have no relationship to dementia risk at all. It’s just a very vague observational study that proves mostly that we publish a lot of somewhat meaningless research in academic journals.

It’s possible that certain foods increase your risk of dementia. We know that obesity is part of the dementia equation, and eating a cake a day will probably make you fatter and that could result in a higher long-term risk of cognitive decline. But we still don’t really know if it matters whether you make the cake yourself, or if you buy one from the supermarket which is the question that UPFs are, in theory, intended to answer.

UPFs are the bogeyman of today, but they remain poorly-defined and not that useful when it comes to health. Cutting back on high energy foods may reduce your risk of dementia, but beyond that there’s currently no strong evidence that processing itself is making your brain more likely to decay as you age.

Subscribe now

Ivermectin is one of the strangest holdovers from the pandemic years. It’s a remarkable medication—both very safe and incredibly effective at treating a group of nasty parasitic diseases—but the furor about the drug has been more about contrarianism than treating river blindness.

People who disliked government intervention against COVID-19 seized on the idea that ivermectin could save us all, based largely on fraudulent research. Years later, and despite extensive evidence that the drug has no benefit for COVID-19, there’s still a bizarre fixation on ivermectin from people who mistrust the medical establishment.

After proving that much of the ivermectin for COVID-19 literature was fake back in 2021, it’s really odd for me to be writing about the drug again five years later.

The newest viral sensation is the claim the ivermectin can cure cancer, which has been garnering massive attention on X. This comes from arguably the worst piece of research that I’ve ever seen. While ivermectin is an amazing drug, there’s no reason to believe that it can cure or treat cancer.

Ivermectin Investigation

The new piece of research that has seen thousands of reposts online is the preprint of a survey study of patients attending appointments for cancer with The Wellness Company. The Wellness Company is a fascinating online service that sells everything from scammy supplements such as their spike detox—71% of tested users reported decreased spike levels!—to untested tirzepatide (Mounjaro) tablets that are not FDA-approved. It’s a strange amalgam of unregulated medications, supplements, and outright quackery jumbled into a single website.

One of the services that The Wellness Company offers is prescribing a combination of ivermectin and mebendazole—another anti-parasitic medication—for various diseases including cancer. While there is some very preliminary evidence that these medications may have some impacts on cancer cells in petri dishes, there is essentially no data showing that they have any impact on cancer in real people suffering from disease. There are many thousands of things that impact cancer cells in a lab—very few of them treat cancer in human beings.

Read more

I’ve been writing a weekly newsletter/blog post about science in the media since 2016. That’s a lot of writing. I’ve written over 500 pieces about a wide swathe of scientific subjects over the last decade, covering everything from whether pasta will prevent high cholesterol to the question of how many people would die of COVID-19 during the pandemic years.

Of the many things I’ve written about, cancer seems to come up the most often. Partly that’s because “cancer” is not one disease, and therefore there are just so many things to get wrong in media headlines about all the different types of cancer out there. But it’s also because cancer is a big unknown, and it remains one of the scariest words that many of us will hear in our lifetimes. More than a third of people will eventually be diagnosed with cancer, which means that the chance that either you or a loved one will experience some form of cancer in your lifetime is close to 100%.

We care about cancer. A lot. Which brings me to one of the most absurd headlines I’ve ever come across. There have been a few versions of the story, but here’s one from MSN dot com: “Eating fruits, vegetables and whole grains may increase chance of early onset lung cancer”.

For anyone who knows literally anything about cancer, this is very surprising. Fruit, vegetables, and whole grains are generally considered to be excellent ways to avoid getting cancer in the future. The stories all blame this on pesticides, but it is hard to understand how barely measurable residues of pesticides from food that you eat could increase your risk of lung cancer, which generally is most impacted by the air you breathe in.

This is, thankfully, because the story is utter nonsense. There is no link between eating fruit, vegetables, and whole grains and a higher risk of cancer. If anything, these foods may reduce your risk of cancer.

The Science

The new study that has everyone riled up about the potential for killer cauliflower is a conference abstract of a food-based survey of 187 young people who were diagnosed with lung cancer. The authors took a database of people aged <50 years who have lung cancer, and asked them about their food consumption using a food frequency questionnaire. Food frequency questionnaires are long survey instruments that capture lots of information about what people report that they eat.

The authors then computed a Healthy Eating Index (HEI) score for these people, and compared those HEI scores to the general population. These scores give you a reasonable understanding of how healthy people’s diets are, based mostly on comparing the diet to the American dietary guidelines. The authors found that their sample of young people who had lung cancer had somewhat healthier eating than the population at large, with HEI scores of 63-67 compared to 58 for the average American. They speculated that this could indicate an association between healthy foods such as fruit, vegetables, and whole grains, and getting lung cancer at a young age.

This is all absolutely absurd. Truly ridiculous.

Firstly, this is a conference presentation. There is little information on all sorts of important things about the study—for example, how the HEI scores were calculated—because we don’t have the full study to read. It is rarely a good idea to rely on results being presented at a conference without a full paper published, because the findings of papers often change dramatically between the first presentation and the final publication.

The study also very straightforwardly does not assess the association between healthy eating and lung cancer. The authors surveyed people after they had been diagnosed with cancer, not before. Therefore, we have no idea what these people ate prior to getting the cancer. It’s quite likely that the lung cancer diagnoses impacted their diets—many people start eating more healthily when they’re told they’ve got cancer!

The authors also don’t really have a control group. They compared young people who participated in this study to the average person in the United States, but that is unlikely to be a reasonable comparison. The institute running the study is based in Los Angeles, California, and it’s not a huge stretch to imagine that a sample of people primarily from L.A. might have healthier eating scores generally than the U.S. average regardless of their cancer status.

The authors also didn’t control for age, socioeconomic status, or any other factors. All of this means that the comparison of HEI scores between the cancer patients and the general population is a bit meaningless. Yes, the scores are different, but we have absolutely no idea why.

The speculation about pesticides is also truly absurd. There have been some concerns about various pesticides and their impact on human health for decades, but that has literally nothing to do with the study. The authors did not measure exposure to pesticides—or at least, the abstract that they published does not report this measurement—so the entire pesticides thing appears to come from the assumption that people who have higher HEI scores will have more pesticide exposure. But that’s not necessarily true—people who eat more fresh produce may have lower pesticide exposure than those who eat less produce depending on the specific fruits and vegetables they like to eat and the pesticide residues that you see in more highly processed food.

In other words, this is an entirely unsupported, unfounded, and likely incorrect assumption that the authors are making to explain an association that may not even exist. Not a very useful thing to base your life decisions on.

Healthy Eating

Generally speaking, a higher intake of fresh fruit, vegetables, and whole grains is associated with a lower risk from all kinds of cancer. How much of this is causal—e.g. those foods actually preventing cancer—is hard to say, but it’s probably going to be at least a bit.

This new conference abstract provides us with no information that would impact that statement. All we can say is that young people with lung cancer seem to eat reasonably healthy food. It’s technically possible that this is because healthy food gives you lung cancer, but there are far more plausible explanations than that.

This study should not change anything about your diet. Frankly, conference abstracts such as this probably shouldn’t be reported on at all, but this one is particularly useless as a guide to your personal health decisions. If you like to eat fresh produce, please don’t stop because of these ridiculous headlines.

Subscribe now

Anyone who has spent any time looking into losing weight online has come across the simple fact: most diets fail. It’s something that I’ve personally said before, and it’s something that you will see over and over again if you are looking for a way to lose weight. It seems to be pretty generally accepted that if you go on a diet, you will probably regain all of the weight that you lose long-term. You might make it to 6 months lighter, but eventually you will be as heavy or heavier than you were at the start.

But recently I’ve been wondering; is that true?

If you look at the science of the question, the water is surprisingly muddy. While it’s probably reasonable to say that many diets fail, there is evidence that many also succeed. The data appears to show that the average person will keep at least some of the weight that they’ve lost off for two years or more.

Let’s look at the science.

Read more

Water fluoridation is one of the greatest success stories of public health. Based on a simple observation—people who have weird-looking teeth also seem to lose less of them—we managed to massively enhance dental care for literally billions of people. It’s a simple, low-cost intervention that saves millions of teeth a year in the United States alone.

As with many public health victories, fluoridation is also very controversial. People who distrust even the concept of public health believe that anything the government adds into drinking water must be part of some sort of conspiracy, or at least in some way bad for us all. This has been seemingly supported for a long time by a series of studies showing that IQ is negatively correlated with the amount of fluoride in drinking water—in other words, that adding fluoride into water may knock a few IQ points off of entire generations of people.

This data has, however, always been sub-par. Most of the studies on fluoride in water have been conducted in China or India, and look at natural rather than artificial fluoridation. This is a big problem because natural sources of fluoride tend to provide a much higher level of exposure than the absolutely minuscule amount that’s added to drinking water for dental health.

Also, some of the research is just awful. One of the most famous meta studies—a paper that aggregated all of the research on fluoridation together—published in 2025 had serious numeric and statistical errors.

Which brings us to the most recent fluoride publication. We now have quite strong evidence that adding fluoride to water probably does not cause people to lose IQ points.

Read more

Gender-affirming care for children has become one of the defining issues of our time. This has never made sense to me, because it’s a tiny pool of patients and the care itself is no more or less harmful than many of the medical treatments we offer. There’s some question about the evidence-base, and the data has been slow in coming, but if people paid this much attention to literally any other medical issue that children face we’d never talk about anything else.

The newest study that’s rocking the social media commentators is a paper purporting to show that children who were given gender-affirming care in Finland had worse mental health than kids who didn’t get the same care. In this case, gender-affirming care was mostly comprised of hormones and puberty blockers—medications designed to delay or reverse puberty so that children could live as their chosen gender.

The study itself is, however, not very good. There are obvious issues with the analysis which make it largely useless as a scientific investigation. Let’s look at the data.

The Study

The new paper is titled “Psychiatric Morbidity Among Adolescents and Young Adults Who Contacted Specialised Gender Identity Services in Finland in 1996–2019: A Register Study”. It’s from the same Finnish team that I wrote about in 2024, with some similar issues in the study as well.

The paper is a simple epidemiological exercise. The authors looked at the Finnish registry of patients, which includes medical information on most services provided in the country. They compared all children who had been treated at the country’s gender identity clinics before they were largely shut down with a control group who were roughly the same age and came from the same municipalities. The primary outcome of the study was the number of psychiatric appointments—how many times kids saw a psychiatrist* for any problem—2+ years after their first referral to the gender clinic or their inclusion in the cohort for the control group.

The study showed that teens who attended the gender clinics had far more psychiatric appointments than those who did not. This was especially true for those who were seeking a medical transition from male to female, who had very few appointments before their referral but a very large number after.

The authors say that this shows that “severe psychiatric morbidity”—very bad mental health issues—was higher in children with gender dysphoria who attended this service and had medical assistance with transition. They then argue that this casts doubt on the idea that gender-affirming care can improve mental health, and imply that such care could actually cause mental health to get worse.

The first obvious issue with this paper is very simple: the outcome variable. We use proxies in epidemiological studies because it can be hard to assess, say, mental health hospitalizations, and so instead of using that as an outcome researchers look for other things that can indicate some deterioration of mental health.

The problem here is that the outcome variable is useless. The Finnish gender service—the intervention that the authors are testing—was a psychiatric clinic. All four of the authors of the paper appear to be psychiatrists who saw patients as part of this service. The authors do note that “gender identity assessments” were excluded from this, but not the plethora of other potential consultations at the clinic nor referrals from these psychiatrists for external assessment. There was a 2-year lag, but even so it would be absurd to argue that having more psychiatric appointments 2+ years after referral to a psychiatrist was an indicator of issues with mental health.

You could argue the exact opposite—perhaps still going to see a psychiatrist after being referred is a good thing!

The control group is also problematic. The authors picked 8 children—4 of each sex—for every child referred to the clinic. However, they only matched these kids on year of age and broad region of birth. That means that the control group was almost by definition VERY different from the intervention cohort, and so the comparisons made in the study are likely heavily biased.

You could attempt to control for some of these issues, but the authors did not. They even note in their discussion that things such as socio-economic status could impact these results, which is true but rather understating the issue. Everything from place of birth, parental marriage status, and ethnicity can impact mental health. The authors only controlled for age, municipality, and—in their statistical analysis—year of referral and number of specialist consultations that happened prior to being referred to the gender clinic.

One way to do this study in a useful way would be to match the children in the intervention cohort carefully with a proper control group. For example, you could look at all children referred to a specialist eating disorder clinic, and then assess whether the two groups had similar levels of psychiatric treatment 2+ years later. But just arbitrarily picking children of the same general age doesn’t give us any useful information about the mental health of the kids who were referred to the gender clinic.

There are also weaknesses in the study design and analysis. The authors used a Cox regression, which assumes that the proportional hazards—the ratio of risk for the intervention teens when compared to the control—of the groups were steady throughout the period that they examined. They did not, however, examine whether this assumption held for their analysis using the standard methodology such as a Kaplan-Meier curve. If you look at the rate of specialist consults in the groups, you can even see that there are changes to the proportional hazards over time. This is bad because it means that their statistical analysis may be giving misleading results.

They also said that they included 8 controls for each of their 2,083 gender-referred children, but the control group only had 16,643 kids in it—they’re missing 21 patients from the control group. All in all, just not a very robust statistical analysis.

Searching For Meaning

This study is somewhat worthless as evidence. The simple reading is that the authors have shown that kids who get referred to a psychiatrist go to more psychiatric appointments than kids who are not referred. It’s a bit like showing that 10 year olds who get given free bicycles are more likely to cycle than children who don’t get free bikes—it’s the expected consequence of the initial intervention, and tells us very little about the mental health of the children in question.

Which begs the question, why do this study? The authors report that they compiled this dataset in June 2022. The Finnish national registry has data on all sorts of useful outcomes, including things like mental health hospitalizations, emergency department presentations, and similar. In their previous study, the authors already showed that suicide deaths had been reduced in the children who got gender-affirming care when compared to those who were referred but didn’t get such care, so we know that this analysis could’ve been better.

Instead, they’ve conducted a largely pointless analysis that tells us nothing. It’s quite confusing.

In a rational world where people divided their attention based on facts rather than politics, this study would’ve been read by perhaps 100 people, and mostly ignored. Instead, it’s going viral on Twitter/X.

This is an inconclusive, limited analysis that shows very little about the population in question and does not answer any important questions about gender-affirming care. But since it’s about gender clinics, every person online feels like it deserves their personal attention and comment. It’s all very depressing.

*Note: It has been brought to my attention that this could be confusing for American readers. In Finland, as in Canada, the UK, and much of Europe, a psychiatrist appointment requires referral from a general physician. This means that some mild mental health services are provided at the genera practice level, and specialists require at least one assessment before the initial appointment. You can read more about this sort of system here: https://thl.fi/en/topics/mental-health/mental-health-services

There are a few topics that health writers on the internet tend to stay away from, because of the nature of the debate. There’s nothing quite like writing what you see as a fairly innocuous piece about breastfeeding or ketogenic diets and getting bizarre hate mail the next day accusing you of what amounts to mass murder. Of these topics, vaping is arguably the most controversial.

I expect to get a fair few emails about this piece.

The reason for this is fairly understandable. The online debate around vaping has two broad sides—the public health argument, which tends to be that vaping is similar to smoking and should be discouraged, and a broadly libertarian viewpoint suggesting that vaping is the key to saving hundreds of millions of lives.Both sides believe strongly that their perspective is lifesaving, and thus both sides generally have the furious surety of the indignantly righteous. It makes for…loud arguments.

The recent news has not helped this discussion. According to a recent paper, it seems that vaping advocates have been wrong all along. Not only does vaping cause cancer, but it can “no longer be be caricatured as safer than smoking” according to the published research. This has led to an explosion of headlines arguing that vaping is a terrible blight on society which kills people.

As so often is the case, the truth is somewhat less impressive than the headlines suggest. Vaping probably does indeed cause cancer, but whether that’s a problem depends largely on perspective.

The Study

The new piece of research that has everyone so worked up is a qualitative assessment of the carcinogenicity of e-cigarettes published in the journal Carcinogenesis. This type of paper is essentially an academic op-ed—the authors have identified studies that they think are pertinent, and written up what they think about the research landscape and what they believe it shows.

This sort of paper can be useful. These are eminent professors, and knowing what they think about the cancer-causing potential of vaping is not useless. They’ve also included lots of references which I for one enjoyed reading as I went through the paper.

But it’s by definition not evidence. The authors have not done an experiment, nor have they aggregated together data from previous experiments in a usable format. It’s just a series of concepts tied together into a story about how vaping may be related to cancer. With no protocol or real methodology, all we can really say about the paper is that it reflects the beliefs of the authors and how they read the literature.

If you look at the arguments made in the paper, they are somewhat unconvincing. For example, the authors make the claim that:

“E-cigarettes are a source of exposure to metals from the heating coils, soldered joints, and other metallic components [26, 27]. Reviewing metal/metalloid exposure from e-cigarettes in 2020, Zhao et al. [28] identified 24 studies, four of which described levels in biofluids of users. Most levels were similar or higher than levels found in smokers. A subsequent review referred to the serious health effects associated with such metal exposure [13]. A 2025 study found elevated nickel concentrations were observed among e-cigarette users compared to nonsmokers [29].”

Zhao et al is a systematic review that identified four studies looking at whether people who vape have higher levels of metals in their blood/urine than people who don’t vape. The authors also looked at the composition of vape vapour vs cigarette smoke. This review does state in the discussion that levels of metal contamination between people who vape is similar to that of cigarette smokers, but if you look at the included research the reality is rather less impressive. While there are a few metals that sometimes appear in the blood of vapers as much or more than smokers, in most cases the blood levels of metals are similar between vapers, smokers, and non-smokers. Here’s an example from the one of the four papers included in Zhao et al:

This basically shows a very mixed bag. Vapers have slightly more of some metals in their blood than smokers, but they are less likely to have mercury, thorium, thallium, vanadium, cadmium, or lead than smokers. In some cases, vapers have less metals in their blood than non-smokers. There’s definitely some potential issues caused by smoking—for example, of the 10 people in this study who had mercury in their blood, 1 vaped, 1 didn’t vape or smoke, and 8 smoked—but much less of a signal for vaping.

Another paper looked at the US NHANES data. This is a study that the US does every two years asking ~10,000 people nationally to fill in a number of surveys about their behaviours and give a range of blood and urine samples. The authors looked at heavy metal levels between people who’d said that they smoked cigars, cigarettes, or e-cigarettes exclusively over the last 5 days. They found a handful of statistically significant differences between the groups, including that e-cigarette users had higher manganese levels than cigar smokers (but not cigarette smokers).

Again, this tells us almost nothing about heavy metal contamination from vaping. There was no control group of non-smokers, and there are innumerable possible weaknesses in such a paper. It gives us a vague idea of what might be a problem, but tells us very little about whether vaping actually exposes people to harm.

So while it may be true that heavy metal levels in vaper’s blood is similar to that of smoking, the paragraph in the original paper is quite misleading. We have no idea what the amount of heavy metal exposure from vaping is, nor how this compares to smoking. At best, the evidence is very vague.

If you look at the arguments made in the paper, they follow a similar trend. They’ve got a section on volatile organic compounds, for example. In this, they single out, among other things, acetaldehyde, benzene, and acrylamide as substances found in the urine of people who vape. All three of these compounds are likely to or definitely cause cancer.

However, the fact that they have been reported in the urine of some vapers tells us very little of use. In fact, all three of these substances—acetaldehyde, benzene, and acrylamide—are produced at lower levels in vapes than in conventional cigarettes. Benzene in particular does not appear to be a product of vaping per se, as there are some vaping systems that do not produce any benzene contamination. All three substances are also present in things like toast and fruit juice, so finding any amount in people’s urine doesn’t tell us much about their cancer risk.

The authors of the paper imply that vaping itself exposes you to very high levels of these chemicals. A more nuanced reading of the evidence would, in my opinion, suggest instead that vaping does expose you to some risk but that this is a) substantially lower than smoking and b) may be mitigated by better regulation on vaping devices.

Which brings us to the real issue here. It’s not whether vaping may cause some cancers.

Is Vaping Safer Than Smoking?

We’ve known for at least a decade that vaping will probably give some of the people who do it cancer. While vapes may produce less carcinogens than cigarettes, they do still produce some, and breathing any cancer-causing chemicals into your lungs long-term is a bad idea.

This paper adds very little to that equation in terms of evidence. Yes, there are reasons to suspect that vaping causes cancer. The real question—the thing that we’ve always cared much more about—is how much cancer rather than a binary yes/no.

This is true of pretty much everything that causes cancer. The substance I mentioned above—acrylamide—is produced when you make toast. We don’t really care about this because the amount of acrylamide that you are exposed to in food is very low, and so even if it does cause cancer it would be a tiny number every year worldwide.

What people are really interested in with vaping is not whether it may cause the occasional cancer, but whether it is safer than smoking. And on this important point, the new paper provides no new evidence. Clearly, the authors believe that vaping is as dangerous as smoking—they say so in their paper—but they have not identified any data showing that this is the case.

And if you look at all of the information out there, there is some quite strong evidence that vaping is better for you than smoking. Obviously both are bad, but if you currently smoke, you can probably reduce your risk of bad stuff significantly by switching to a vape. We currently don’t know exactly how much—and it’s possible that the reduction in risk is quite small, especially if you continue to smoke as well—but it’s very likely that replacing cigarettes with vapes is good for your long-term health.

To sum up the study, I think there was an excellent quote by Professor Lion Shahab from UCL:

“While it is clear that e-cigarettes expose users to harmful chemicals, which may lead to later disease, I would urge against sensationalisation of evidence. No-one would argue that e-cigarettes are entirely risk-free. They should be used as a harm reduction product to help those who smoke to quit and reduce their risk of developing smoking-related diseases. They should not be used by someone who has never smoked. However, this review does not offer a ‘smoking gun’ that e-cigarettes cause oral or lung cancer, nor does it make an attempt at quantifying this risk, which is unsurprising because the evidence is simply not there to allow for such an estimation.”

Yes, vaping is harmful. If you’ve never picked up a vape or a cigarette, don’t start now. It is, however, probably less harmful than smoking. Switching from smoking to vaping is probably good for you, and this new paper provides no evidence that would cause us to change that assessment.

After visiting Seoul and Busan in 2019 for a scientific conference, I fell in love with Korean food. I am a huge fan of bibimbap, I buy my own gochujang and gochugaru to make various Korean dishes, and I have tried making my own Korean pickles—including kimchi—with mixed success.

I am, therefore, primed to see any good news about kimchi as positive. I eat enough of the stuff. So when the headlines came out that you could potentially mitigate any negative impacts of microplastic consumption by eating kimchi, it seemed like a dream come true. It’s not just tasty; it might be remarkably—and uniquely—good for your health!

Except, unfortunately, at this point the pedantry kicked in. I am nothing if not consistent when it comes to looking at the evidence behind things that seem too good to be true, and this was no different.

Sadly, the reality is not nearly as amazing as the hype. Kimchi is delicious, but the health benefits are probably somewhat oversold. The story behind the headlines, however, is one of the funniest examples of scientific shenanigans that you’ll hear all year. It turns out that there is a secretive government department dedicated to trying to get you to eat more kimchi.

Let’s look at the science.

Read more