There’s one kind of comment from readers that I always keep front of mind. Here’s one of them, “I take almost all of your tweaks and incorporate them into my own life… I send them on to family too.”
Comments like that are a huge compliment, but they also bring a great sense of responsibility. They’re the antithesis of a licence to knock out an issue of One Health Tweak a Week in an hour because I’m busy and tired.
If you’re to take health advice from someone, implement it into your own life, and share it with those you love, you need to know that the information you read has been carefully curated by someone who has thoroughly researched and understood the topic. Otherwise, you risk making important decisions based on an automated process where an AI scraped the web and reassembled the information into something that’s effectively a Google search written up in narrative form.
There was a time when it was hard to get detailed information about health and wellness; now we’re drowning in it.
Some of it comes from people genuinely invested in sharing the best-supported evidence, but for many, publishing is simply a business where the economics reward speed, confidence, and volume over scientific rigour. You’re left looking at something that seems authoritative and polished, but you have no idea if someone competent has properly interrogated the evidence behind it.
That lack of oversight can lead to material that’s superficial. Worse, it can be uninformed to the point of being potentially dangerous. I see that kind of content most days, and it makes me angry.
So this week, I wanted to share what it looks like to put out evidence-based health content each week.
I know some of you enjoy seeing how the sausage is made, but more broadly, I want to show you what careful health writing looks like behind the scenes, both so you can judge the advice I give you and so you’re better equipped to assess the much larger volume of health information you encounter elsewhere.
While this week’s newsletter may feel like a break from the usual format, I’ll be back next week with another simple, science-backed tweak to help you live healthier and longer.
Deciding what to write about
Over the last few years, I’ve collected a long list of potential subjects for the newsletter.
Very often, these come from your questions and comments. They’re particularly likely to catch my attention if they’re commonly asked questions, or if they’re ones to which I’m not entirely confident of the most up-to-date answer.
Poll questions in the newsletter have been especially helpful in understanding what you’re wondering about, struggling with, or considering changing. Those little clicks are always useful, so please keep them coming.
Some topics, like circadian rhythm and women’s health, especially around and after menopause, seem to resonate less with readers. Open rates and engagement fall, and I see that, but they affect too many people and offer too many opportunities to make life better to be set aside. So I’ll keep adding them into the mix alongside the always popular topics, like sleep and protein.
Digging into the research
Once the topic is nailed down, I start researching it.
Even though I might know the subject well, the field constantly evolves, so I can’t rely solely on what I remember. You’ll recognise this from topics like eggs or dietary fats, where popular sentiment can swing between ‘health food’ and ‘quietly killing you.’
I usually begin with a recent review to map the field, then work through the important original studies in full.
It can be tempting to read only the abstracts (summaries at the start of each paper), but they can be misleading or may miss key points relevant to the practical decisions you’re facing, which is why it’s important to read papers in full. I make brief notes as I read.
When I come upon some data that would illustrate a key point, I’ll create a graph (using Excel and Photoshop) and save it with my notes.
This phase will typically take me between six and eight hours, but on a topic with lots of complexity, like the recent issue on eggs or an upcoming one on creatine, it might take several times that. Friends telling me what they got up to at the weekend aren’t always welcome at that point!
I keep going until I understand the evidence well enough to explain where it agrees, where the conflicts are, and what it does and doesn’t justify doing.
Once this phase is completed, I’ll have a set of notes like those below for the recent issue on the benefits of substituting plant for animal protein at a few meals each week.
I don’t recommend settling down to read them over coffee. A quick skim will make the point.
You’ll see that I capture the populations studied, the main findings, contradictions and surprises, possible explanations, conflicts of interest and any graphs I’ve made to understand the results.
Tharrey et al. 2018 https://pmc.ncbi.nlm.nih.gov/articles/PMC6658814/ Researchers followed more than 81,300 American adults for an average of about 10 years. Those eating the most meat had a 61% higher risk of death from cardiovascular disease. Those eating the most nuts and seeds had a 40% lower risk. These figures are after adjustment for the usual demographic and socioeconomic factors.
Alt text: A graph illustrating how eating an equivalent amount of protein can lead to an increased risk of death from cardiovascular disease if it comes from meat or a reduced risk if it comes from nuts and seeds. Caption: Two people can eat identical amounts of protein, yet if one person gets it from meat, they may have an elevated risk of dying from cardiovascular disease, while the other, who gets it from nuts and seeds, may have a reduced risk (lower is better) | Data: Tharrey et al. 2018 {https://pmc.ncbi.nlm.nih.gov/articles/PMC6658814/} | Image: Author
Chen et al. 2020 https://pmc.ncbi.nlm.nih.gov/articles/PMC7250948/ Researchers followed almost 7800 Dutch adults for an average of 13 years. They found that those with the highest consumption of animal proteins had a 28% higher risk of death from cardiovascular disease. They also had an 18% higher risk of premature death from any cause. In contrast, those getting the most plant protein had a 10% lower risk of premature death from any cause. They then did a meta-analysis of 11 prospective studies involving more than 350,000 participants. This found that those eating the most animal protein had a 9% higher risk of dying from cardiovascular disease, whilst those eating the most plant protein had a 14% lower risk of dying from cardiovascular disease. They estimated that those consuming the most animal protein were getting about 75 g a day, while those consuming the least were eating about 38 g a day of pure protein. This suggests that cutting animal protein intake from 75 to 42 g a day, which would be equivalent to going from about 220g (7.8oz) to 100 g (3.5oz) of red meat a day, could cut the risk of dying from cardiovascular disease by about 9 %.
Huang et al. 2020 https://doi.org/10.1001/jamainternmed.2020.2790 Researchers followed more than 410,000 Americans with an average age of 62 for 16 years. They found that higher consumption of plant protein was associated with a lower risk of premature death. More usefully, they looked at the effects of switching some protein consumption from animal to plant sources. They found that switching just 18g (0.6oz) of pure plant protein from animal sources to plant protein was associated with a 10% lower risk of dying prematurely from any cause. Equivalent swaps might be:
From: 80g/2oz chicken breast, steak or hamburger patty, or 150g/5.3oz sausage, or 3 large eggs
To: 80g / 2.8 oz of peanuts, 230g / 8 oz of kidney beans, 76g / 2.7 oz of lentils, or 90g / 3.2 oz of tofu They estimated the reduction in risk for a variety of outcomes if specific animal-sourced proteins were switched to plant proteins (any animal protein, red meat, white meat, eggs, and dairy). These show that animal-to-plant swaps are not all equal. In this analysis, the greatest benefits come from switching from eggs to plant protein, with red meat coming second, dairy third and white meat (poultry and fish) fourth.
Alt text: A graph showing the reduction in death and disease associated with switching from various animal proteins to plant-based protein. Caption: Switching 18g/0.6oz of daily animal protein for plant protein can offer substantial health benefits, but the effect depends on what you swap out. Switching from eggs to plant protein has the greatest effect | Data: Huang et al. 2020 {https://doi.org/10.1001/jamainternmed.2020.2790} | Image: Author
Song et al. 2016 https://doi.org/10.1001/jamainternmed.2016.4182 A study of more than 131,000 American adults with an average age of 49, followed for up to 32 years, found that animal protein was not associated with all-cause mortality, but was associated with a 9% higher risk of death from cardiovascular disease. Those eating the most plant protein had a 10% lower all-cause mortality and a 12% lower risk of dying from cardiovascular disease. (A healthy lifestyle was defined as little or no smoking, alcohol < 14g in women or 28g/day in men, BMI of 18.5-25, and physical activity of ≥75 minutes vigorous or ≥150 minutes moderate activity per week - the benefits seemed greater in those with an unhealthy lifestyle). They estimated the reduction in risk for a variety of outcomes if 18g of specific animal-sourced proteins were switched to plant proteins. The greatest benefits came from switching from processed meat (34% lower all-cause mortality, 39% lower cardiovascular disease), followed by eggs (19% lower all-cause mortality, 12% lower cardiovascular disease), then red meat (12% lower all-cause mortality, 17% lower cardiovascular disease)
Budhathoki et al. 2019 https://doi.org/10.1001/jamainternmed.2019.2806 In this Japanese study, more than 70,600 adults were followed for 18 years. Those eating the most plant protein had a 13% lower risk of premature death from any cause and a 16% lower risk of death from cardiovascular disease. Switching 18g of protein from red meat to plant protein was linked to a 34% lower risk of death from any cause, and 18% lower when switching from poultry. Again, the greatest benefit came from switching from processed meats, where a switch to plant protein was associated with a 46% lower risk of premature death from any cause.
Meroño et al. 2022 https://dx.doi.org/10.1093/gerona/glab334 Italian researchers followed 1139 older Italian adults with an average age of 75 at baseline for 20 years. In contrast to what we’ve seen in other studies, animal protein in these older adults was associated with better long-term health outcomes. Those eating the most meat had a 21% lower risk of dying from any cause during the 20-year study and a 51% lower risk of dying from cardiovascular disease. The amount of plant protein consumed was not linked to any long-term outcomes. Sources of animal protein were 26% for dairy products, 26% for processed meat products, 20% for red meat, 7.7% for fish and seafood, 6.3% for chicken, 2.7% for eggs, and the rest for other meats. Sources of plant protein were 73% for cereals, 11.4% for vegetables, 9.0% for fruits and nuts, and 5.3% for legumes. This makes the results even more surprising, given that processed meats and dairy make up 50% of animal protein consumption. While more easily accessible protein may account for all-cause mortality through sarcopenia prevention, it’s hard to explain the reduction in cardiovascular disease. The authors suggested the lack of benefit from plant protein may be because most of the plant protein came from cereals, and little from vegetables, fruit, nuts and legumes, which are more strongly associated with better long-term health outcomes.
Carballo-Casla et al. 2024 https://doi.org/10.1001/jamanetworkopen.2024.26577 A Spanish study followed 8,543 adults aged 60+ from Spain and Sweden for up to 10 years. In this older group, those with the highest protein intake (1.6g/kg/day) had a 49% lower risk of dying from any cause over those ten years. When they compared animal and plant proteins, those eating the most animal protein had a 26% lower risk of dying, while those consuming the most plant protein had a 31% lower risk of dying. So, while consuming more protein, even if it’s from animals, reduces the risk of dying in later life, there’s still an advantage to getting a substantial amount of that protein from plants. The benefits of higher protein consumption (1.6g/kg/day) were greater for those aged 75 and older than for those aged 60-74, which suggests a continued need to focus on protein consumption. This study also looked at people with chronic kidney disease. They also had the lowest mortality at 1.6g/kg/day total protein, and benefited more from plant protein.
That’s a fairly short set of notes, so they’re useful as an example. For a more complex topic, they can be much longer.
Getting down into the weeds often reveals a far more interesting story than the one I’d originally set out to answer, or challenges widely held assumptions that turn out to be built on rocky ground.
One of the studies in the notes is a good example: the headline result gets quoted as if it shows meat is better than plants for older adults, but the details make clear that the comparison really wasn’t that simple - the plants they compared against were mostly refined carbs, like white bread and pasta. That kind of nuance is exactly what you miss if you rely on abstracts or AI summaries.
Deciding what I want to say
While I will have started out with an idea of what the message of the newsletter will be, sometimes it changes as I immerse myself in the research. So now comes the time to pull it all together into a coherent set of key points I want to include in the issue.
For the article on the benefits of basing some meals on plant rather than animal protein, it started like this: (again, a portion of my notes are here just for you to skim through, not to read in detail).
Core article thesis: Getting enough protein matters, especially as we age, but once you are getting enough, the next question is where it comes from. The best evidence suggests that keeping protein intake high while shifting some of it from animal sources - especially processed meat, red meat and heavy reliance on eggs - towards legumes, soy foods, nuts, seeds and whole grains better supports both longevity and healthier ageing.
While it’s important to get sufficient protein in later life to preserve strength and independence, there’s clear evidence that not all protein sources are equal. If your aim in boosting protein is to live longer and healthier, it would seem self-defeating to increase your consumption of protein sources that might work against those goals.
You don’t need to become a vegan to get the benefits of eating less meat:
Researchers in a study of more than 410,000 Americans with an average age of 62, followed for more than 16 years, estimated that switching just 18g (0.6oz) of pure plant protein from animal sources to plant protein was associated with a 10% lower risk of dying prematurely from any cause. {https://doi.org/10.1001/jamainternmed.2020.2790}
Equivalent swaps might be:
From: 80g/2oz chicken breast, steak or hamburger patty, or 150g/5.3oz sausage, or 3 large eggs
To: 80g / 2.8 oz of peanuts, 230g / 8 oz of kidney beans, 76g / 2.7 oz of lentils, or 90g / 3.2 oz of tofu
Because plant protein sources can be less protein-dense and sometimes less digestible than animal proteins, it’s not just a case of eating more plants; you need to plan around protein-rich plants (lentils, beans, tofu, tempeh, soya milk, soya yoghurt, edamame, nuts, seeds, peanut butter, hummus, whole grains, or a good plant-protein powder if needed). We’re also talking about whole plant proteins, not ultra-processed plant meat substitutes, as we discussed last week.
What happens if you’re older and losing muscle?
We’ve talked before about how retaining muscle is vital for strength and independence in later life, so how do you balance getting enough protein with cutting back on animal protein if your appetite isn’t what it used to be - since you generally have to eat a greater volume of plants to get the same amount of protein?
A study of Spanish and Swedish adults over 60 showed that those consuming the most protein (1.6g/kg/day) had a 49% lower risk of dying over 10 years. {https://doi.org/10.1001/jamanetworkopen.2024.26577} This confirms the importance of maintaining a high protein intake as we age. It found that while consuming lots of animal protein offered much higher survival benefits than not eating enough protein, the greatest benefits were seen in those eating the most plant protein. That wasn’t a binary choice; most people were eating both.
A similar study of Italian adults with an average age of 75 found that higher consumption of animal protein was associated with a reduced risk of death, but plant protein was not. This is an odd study as processed meats and dairy made up 50% of animal protein consumption, so you’d expect consumers of these products to be at higher risk. The authors suggested the lack of benefit from plant protein may be because most of the plant protein came from cereals, and little from vegetables, fruit, nuts and legumes, which are more strongly associated with better long-term health outcomes. {https://dx.doi.org/10.1093/gerona/glab334}
While there’s limited information from studies in this age group, my interpretation is that getting 1.2 to 1.6g/kg/day of protein is the first priority, and that getting a significant proportion of that protein from plants offers added survival benefits.
Since my outline for that article ran to more than 1500 words, I’ve just included an extract to give you an idea of what it looks like.
With the relief that I’ve finally condensed a stack of papers and innumerable mugs of coffee into an orderly summary, it’s time to write the first draft.
Writing the first draft
By this point, I’ve had enough writing, and I’m pretty desperate to get away from my desk. So, I take my notes, pace around and dictate a first draft.
Although I have comprehensive notes to work from, I’m still thinking things through as I dictate, so the first draft can be pretty rough, but it ensures that every draft begins with my reasoning, my voice, and my argument structure.
An added advantage is that it helps nudge me towards the 6,000-plus steps a day I committed to many newsletters ago!
While dictation software is a lot better than it used to be, it’s still far from perfect, and those helpful little ‘auto-corrections’ can add some eyebrow-raising changes at times. And that’s on top of me changing how I want to word things on the fly. So, the next step is to pass that rough draft to Claude to take out all the typos and repetition, and fix the punctuation.
After that come the rounds of editing, and that’s a collaboration between me, Grammarly and Claude. Grammarly picks up any remaining errors and also suggests sentence-level improvements. It’s invaluable, but also a little discouraging to feel the relief at having finished the first draft swept away by a tide of red and blue underlining.
Claude is more useful for answering questions like “Is this the best order to present this information?” “Can you suggest a better metaphor to explain this?” or “This paragraph feels long-winded. What’s a more concise way to say the same thing?”
Sometimes that’s the end of the writing. Other times, if I have a nagging feeling there’s something missing, I might pass it back to Claude and ask, “What questions might readers still have after reading this?” That’s often helpful, though, as your many questions show, it doesn’t capture every missed concern.
Putting it all together
With the text complete, it’s a case of copying it all to Substack and adding the formatting and graphics.
Next are the admin bits, like SEO titles and meta-descriptions. I outsource all of this to Claude or ChatGPT as they’re far better at that than I am, and I have no desire to learn any more SEO than I need to. I’m quite happy for ‘long tails’, ‘semantic clustering’, and ‘EEAT’ to remain mysteries.
If all has gone according to plan, I’ll have the issue ready by Tuesday or Wednesday, and I’ll schedule it to be published on Saturday.
Once it comes out, there are always lots of questions. I try to answer every one, and the answers are always from me. No AI involved.
A note on AI
I see other writers in my field publishing detailed, lengthy posts every day. Perhaps they work a lot faster than me, but I suspect AI plays a greater role in their process. Often, AI mannerisms and turns of phrase are very evident.
I know that if I ask an AI model to give me a review of a health topic, it will get much of it right. But it will get a few things wrong, and there’ll be aspects that are missing. There’s no way I could take that output and pass it on to you, confident that I was giving you the best, most evidence-based information.
Without that thorough understanding, I also wouldn’t be able to answer your questions. Indeed, many of my fellow writers publishing daily don’t answer questions, or only a select few.
Because I’ll know the topic backwards by the time I write, I can use AI as an intellectual sparring partner. I can ask questions like “What do readers find most confusing about this topic?”, or “What’s the strongest argument against the recommendation I make here?” This is genuinely useful and makes my articles stronger.
Substack recently introduced a collaboration with the AI ‘detector’ tool, Pangram. The introduction has been polarising.
Some have welcomed the supposed transparency it allows. Others are howling at the challenges it brings to their business model. Another group worry about the false flagging of highly structured writing (like scientific exposition), the undermining of the benefits AI assistance brings to people writing in languages that are not their first, or who live with conditions like dyslexia, and the potential for witch hunts.
I ran a recent issue of One Health Tweak a Week through Pangram. It claimed it was “mostly human written.” More specifically, 67% human-written, 33% AI-written, and 0% AI-assisted.
That’s an odd breakdown since I’d consider the newsletter to be AI-assisted. I was also a little affronted by that 66%, knowing the many hours I’d put in long before Claude tidied up my dictation and Grammarly politely pointed out my long-winded sentences with blue underlines.
Determined to get that figure closer to 100% human-written, I rewrote some of the paragraphs I remembered Grammarly had suggested rephrasing, deleting them first so I wouldn’t just paraphrase.
The result? The AI-written percentage went up by 8%. So, either I’m better at writing AI content than AI is, or Pangram is, at best, not very reliable.
Forbes magazine recently reported AI detection algorithms had asserted that the Declaration of Independence was 98% AI-generated. Either someone is keeping some pretty impressive time travel technology from us, or we should be sceptical of AI detectors, however reliable they claim to be.
If I were a painter and decorator, and not very good at my job, you’d not have me back to do the guest bedroom. You might write something acerbic on Trustpilot. It would be annoying and perhaps expensive, but it would be repairable.
Medicine taught me early on that a lack of diligence or poor judgment can have much more serious consequences, and they’re usually borne by someone else. That makes professional diligence less of a virtue and more of a basic obligation.
I hope that seeing the process I use to write One Health Tweak a Week gives you a better sense of what to look for when you’re deciding how much to trust other health information. A polished article and a confident tone can conceal some very flimsy foundations, but it becomes easier to spot the warning signs once you know what careful evidence-based writing should involve.
The conclusions I share in One Health Tweak a Week are the same ones I give to friends and family members.
An average issue takes about twelve hours of research and writing. That’s a little absurd for an article that takes five or six minutes to read, but it’s what allows me to stand behind what I write each week.
It’s also why you only hear from me once a week.
Of course, I’d love to be more efficient and regain a chunk of my weekends, but that’s not to be, unlike for those lazy Founding Fathers who outsourced their declaration drafting to ChatGPT!
Science-backed advice, simple steps, real results. Subscribe to One Health Tweak a Week today and take the guesswork out of healthy living.
It’s obvious you put a lot of effort and time selecting articles and crafting the final results. The only criticism I would make is the example you chose. I have a problem now believing studies based Human diets which is complicated from the tendency of the subjects to supply inaccurate info based on keeping food diaries. Maybe you could address. Is the size of the study big enough we can ignore that concern?
It’s obvious you put a lot of effort and time selecting articles and crafting the final results. The only criticism I would make is the example you chose. I have a problem now believing studies based Human diets which is complicated from the tendency of the subjects to supply inaccurate info based on keeping food diaries. Maybe you could address. Is the size of the study big enough we can ignore that concern?