GIFSQ ID: FTN-0102
Personalised Nutrition and GLP-1 Eating
Unit 1: Why One Diet Does Not Fit All
Note
Unit 1: Why One Diet Does Not Fit All
A question to start with
Two friends eat the same lunch: a bowl of rice, beans, and vegetables, with a glass of water. One feels full and steady for hours. The other feels hungry again within an hour, and her blood sugar, measured on a small sensor on her arm, rises much higher and falls much faster. The food was the same. The plates were the same. The bodies were not. This unit is about that difference: why the same food can do different things in different people, and why the idea of one diet for everyone, however good the diet is, misses the way bodies actually work.
For most of the last century, nutrition advice was given as one set of rules for everyone: eat this much, eat less of that, choose these foods, avoid those. The rules were based on averages, on what most people need most of the time, and they helped many people. But averages hide differences, and the differences are large. People differ in age, in size, in how active they are, in how their bodies handle sugar and fat, in the microbes in their gut, in their genes, in their sleep, in their stress, and in the foods they grew up eating. A diet that works well for one person can work poorly for another, even when both follow it carefully. Personalised nutrition starts from that fact, and asks a different question: not "what is the best diet?" but "what is the best diet for this person, at this time, in this life?"
Questions to think about
- Two people eat the same bowl of rice. Why might one person's blood sugar rise twice as high as the other's?
- A diet that helped your friend lose weight does nothing for you. Does that mean the diet is wrong, or that your body is different?
- If nutrition advice is personal, does that mean general advice, such as eating more vegetables, is useless?
Think about the rice question before you read on
Because the rise in blood sugar depends on more than the rice: it depends on how fast the gut empties, how much insulin the body makes and how well it works, what else was eaten with the rice, how active the person was, how they slept, and the microbes in their gut. The same rice meets a different body, and the body answers differently.
The old idea: one diet for everyone
The old idea was simple and it had a reason behind it. When nutrition science began, the main problems were deficiency: too little vitamin C, too little iron, too little protein, and the diseases those shortages caused. The answer was to find how much of each nutrient most people need, and to tell everyone to eat at least that much. The numbers were set for groups: men and women, children and adults, pregnant women and older people, and they were printed as recommended amounts. Those numbers saved lives, and they still matter: a person who eats far too little iron, or far too little vitamin C, will become ill, whoever they are.
The old idea also had a practical reason. Advice that fits everyone can be printed on a poster, taught in a school, and put on a food pack. Advice that fits one person needs a test, a conversation, and a plan, and it costs more to give. For a government feeding a whole country, or a school feeding a thousand children, one set of rules is the only practical answer. For a person who wants to feel better, to manage their weight, or to control their blood sugar, one set of rules may not be enough, and that gap is where personalised nutrition begins.
But the old idea had a limit, and the limit showed as the main problems changed. Deficiency is still a problem in many places, but in many others the main problems are now excess: too much energy, too much sugar, too much salt, and too little fibre, leading to weight gain, type 2 diabetes, and heart disease. Those problems are not the same in every person. Two people can eat the same amount of energy and one gains weight while the other does not. Two people can eat the same amount of sugar and one develops high blood sugar while the other does not. The average hides those differences, and a diet built on the average will work for the average person and will miss the people at the edges, who are often the people who need help most.
How people differ: the main sources of difference
People differ in ways that change how food works in the body. Age matters: a child needs more energy per kilogram than an adult, and an older person needs more protein per meal to keep muscle. Sex matters: men and women differ in size, in muscle, and in hormones, and those differences change energy needs and how fat is stored. Size matters: a taller, heavier person needs more energy than a smaller person, and a person with more muscle burns more energy at rest. Activity matters: a person who walks all day needs more energy than a person who sits all day, and a person who trains needs more protein and more carbohydrate at the right times.
Health matters most of all. A person with type 2 diabetes handles sugar differently from a person without it, because their insulin does not work as well. A person with high blood pressure must watch salt more closely. A person with kidney disease must watch protein and some minerals. A person with coeliac disease must avoid gluten entirely. A person taking a medicine may need to eat differently because the medicine changes appetite, or changes how the body handles a nutrient. A woman who is pregnant needs more iron, more folate, and more energy in later months. A person recovering from illness needs more protein and more energy to rebuild. Each of these is a difference that a single diet cannot cover, and each one is a reason to personalise.
Habits and life matter too. A person who works nights eats at different times and sleeps at different times, and both change how food is handled. A person who cooks at home can follow a plan that a person who eats out every day cannot. A person who does not like fish will not follow a plan built on fish, however good the plan is on paper. A person who cannot afford a food will not buy it, however good it is for them. A plan that ignores the life it must fit will fail, not because the science is wrong, but because the life is real. Personalised nutrition must fit the life, not only the body.
What personalised nutrition is, and what it is not
Personalised nutrition is the use of information about one person, their body, their health, their habits, and their life, to give food advice that fits them better than general advice does. The information can be simple: age, sex, height, weight, activity, health, medicines, and what the person eats now. It can be more detailed: blood tests for sugar, cholesterol, and vitamins, measurements of body fat and muscle, and records of what the person eats and how they feel after eating. It can be most detailed: genetic tests, gut microbe tests, and continuous glucose monitoring, which are covered in the units that follow. At each level, the aim is the same: to move from the average to the person, and to give advice that the person can follow and that works for them.
Personalised nutrition is not a promise that every person needs a different diet, or that general advice is wrong. General advice is the base: eat more vegetables and fruit, choose wholegrains, eat less sugar and salt, drink water, and keep portions in line with activity. Those rules help most people, most of the time, and personalised nutrition builds on them, it does not replace them. A person who eats no vegetables will not be helped by a genetic test. A person who drinks sugary drinks all day will not be helped by a microbiome test. The first step in personalising is often the simplest: to look at what the person eats now, and to change the one or two things that matter most, in a way that fits their life.
Personalised nutrition is also not a single test that tells the whole truth. No test, on its own, can say what a person should eat. A genetic test can say that a person is more likely to respond in a certain way, but it cannot say what will happen, because genes are only one part of the story. A gut microbe test can say which microbes are present, but it cannot say what they will do, because the microbes change with what is eaten. A glucose sensor can say how blood sugar rose after one meal, but it cannot say what will happen after a different meal on a different day. Each test adds a piece, and the pieces must be put together with the person's health, habits, and life, by someone who knows how to read them. A test without that reading is a number without a meaning.
Why the same food gives different blood sugar
Blood sugar is the clearest example of why one diet does not fit all, because it can be measured, and the measurement shows the difference. When a person eats carbohydrate, it is broken down to glucose, which enters the blood. The body releases insulin, which helps glucose move from the blood into cells, and the blood sugar rises and then falls. How high it rises, and how fast it falls, depends on the food, on the person, and on the moment. The same food can give a small rise in one person and a large rise in another, and the same person can give a different rise to the same food on two different days.
Several things change the rise. The food itself matters: how much carbohydrate it has, how fast it is digested, and what else is eaten with it, because fat, protein, and fibre all slow the rise. The person matters: how much insulin they make, how well their cells respond to it, how much muscle they have, and how active they are. The moment matters: a meal eaten after a poor night of sleep gives a higher rise than the same meal after a good night. A meal eaten after exercise gives a lower rise, because the muscles take up glucose more readily. A meal eaten when stressed gives a higher rise, because stress hormones raise blood sugar. The gut microbes matter too, because they change how carbohydrate is broken down and how the gut signals to the body. No single number, such as the glycaemic index of a food, can capture all of that, because the index is measured in a group, under set conditions, and the person in front of you is not the group.
That is why a food that is "good" for blood sugar in general may not be good for one person, and why a food that is "bad" in general may be fine for another. A person who measures their own blood sugar after meals, and who keeps a record of what they ate and what happened, learns their own pattern, and that pattern is more useful to them than any table of values. The pattern is personal, and it is the start of a personal eating plan, which Unit 6 describes in full.
The diagram shows the layers that shape a response to food. The steps point to the diagram: the food is only the first layer, and a plan that looks only at the food will miss the body, the gut, the moment, and the life around it.
The risks of getting personalisation wrong
Personalised nutrition carries risks when it is done badly. The first risk is over-testing: a person buys a set of tests, gets a set of numbers, and is told to eat in a way that is hard, expensive, and not based on good evidence. The second risk is over-promising: a test is sold as telling the person exactly what to eat, when it can only say what is more likely, and the person is disappointed when the promise fails. The third risk is under-eating: a plan that cuts out whole groups of foods, on the basis of a weak test, can leave a person short of nutrients, and can make eating a source of worry rather than pleasure. The fourth risk is cost: tests and plans that cost a lot of money can leave a person with less money for the food itself, and can make good eating feel like a luxury.
The defence against those risks is to keep personalisation grounded in what is known, to start with the simple information, to use tests only when they change the advice, and to judge a plan by whether it works and whether the person can keep it. A plan that is hard to follow will not be followed, however good its science is. A plan that is simple, that fits the person's life, and that is checked against real results, such as weight, blood sugar, or how the person feels, is a plan that has a chance. The units that follow show how each kind of information is used, and where its limits lie, so that the reader can judge a personalised claim for themselves.
Deriving glycaemic load, step by step
Glycaemic index, GI, says how fast a carbohydrate raises blood sugar, compared with pure glucose, which is given a GI of 100. A food with a GI of 70 raises blood sugar fast. A food with a GI of 40 raises it slowly. But GI on its own does not say how much carbohydrate is in a serving, and a food with a high GI eaten in a tiny serving may raise blood sugar less than a food with a low GI eaten in a large serving. Glycaemic load, GL, fixes that, by combining the GI with the amount of carbohydrate in the serving. The derivation is one multiplication and one division, and it is worth seeing worked through once, because it shows why serving size matters as much as the type of carbohydrate.
The formula is: GL equals GI times the grams of carbohydrate in the serving, divided by 100. A worked example, with round teaching numbers only, makes the arithmetic visible. Suppose a food has a GI of 70, and a serving contains 30 g of carbohydrate. The GL is 70 times 30, which is 2100, divided by 100, which is 21. Suppose the same food is eaten in a smaller serving, with 15 g of carbohydrate. The GL is 70 times 15, which is 1050, divided by 100, which is 10.5. The GI has not changed, because the food is the same, but the GL has halved, because the serving has halved, and the blood sugar rise will be smaller, because less carbohydrate was eaten. That is the lesson of GL: the same food can have a high load or a low load, depending on how much of it is eaten.
GL is usually read in three bands, as a guide, not as a rule for any one person: a GL below 10 is low, a GL from 11 to 19 is medium, and a GL of 20 or more is high, for one serving. Those bands are guides from the literature, and a person's own response, measured with a CGM, matters more than the band, because the rise also depends on what else was eaten, on the person's insulin sensitivity, and on the day. The maker and the reader should use GL as a tool for comparing servings, not as a verdict on a food: a food with a high GI can have a low GL if the serving is small, and a food with a low GI can have a high GL if the serving is large. The question to ask is always about the serving that will actually be eaten, not about the food in the abstract.
Practical note: Keep a food and feeling diary for three days: write down what you eat, when you eat it, how hungry you feel before and after, and how much energy you have an hour later. At the end of three days, look for your own pattern: which meals keep you full longest, which ones leave you hungry, and which ones give you steady energy. That pattern is the simplest form of personalised nutrition, and it needs no test at all.
Quick self-check
Question: Why can the same food give a different blood sugar rise in two people?
Because the rise depends on the food, the body, the gut, and the moment: how much insulin the body makes and how well it works, how much muscle there is, what else was eaten, how the person slept, whether they exercised, and the microbes in their gut. The same food meets a different body, and the body answers differently.
Question: Does personalised nutrition mean general advice, such as eating more vegetables, is useless?
No. General advice is the base that helps most people most of the time, and personalised nutrition builds on it. A person who eats no vegetables will not be helped by a genetic test, and the first step in personalising is often to change the one or two things that matter most, in a way that fits the person's life.
Try it yourself
Eat the same breakfast on two different days: once after a good night of sleep, and once after a short night, or once before a walk and once after sitting all morning. Write down how hungry you feel an hour later, and how much energy you have. Compare the two days: what changed, and what does that tell you about how sleep and activity shape your response to the same food? Keep the notes, and repeat the test with a different meal next week.
Key ideas to carry forward
- The same food can do different things in different people, because bodies differ in age, size, activity, health, genes, gut microbes, sleep, and stress.
- General nutrition advice is based on averages and helps most people most of the time, and it is the base that personalised nutrition builds on.
- Personalised nutrition uses information about one person to give advice that fits them better, from simple facts to blood tests, sensors, and genetic tests.
- Blood sugar is a clear example of difference: the same meal can give a small rise in one person and a large rise in another, and a different rise in the same person on different days.
- No single test can tell a person exactly what to eat: each test adds a piece, and the pieces must be read together with health, habits, and life.
- A plan that ignores the life it must fit will fail, however good its science is, because the person will not follow it.
- Personalisation done badly risks over-testing, over-promising, under-eating, and high cost, and the defence is to start simple and to judge by real results.
Glossary
Personalised nutrition
Food advice based on information about one person, their body, health, habits, and life, rather than on averages for everyone. Food tech example: a personal eating plan uses a person's blood sugar pattern to choose their breakfast.
Blood sugar
Glucose in the blood, which rises after eating carbohydrate and falls as insulin helps it move into cells. Food tech example: blood sugar rises after a bowl of rice and falls over the next hours.
Glucose
A simple sugar in the blood, made when carbohydrate is digested, and used by cells for energy. Food tech example: glucose from bread enters the blood after eating.
Insulin
A hormone made in the pancreas that helps glucose move from the blood into cells. Food tech example: insulin is released after a meal to lower blood sugar.
Type 2 diabetes
A condition in which the body does not make enough insulin or does not respond to it well, so blood sugar stays high. Food tech example: a person with type 2 diabetes watches carbohydrate at each meal to keep blood sugar steady.
Glycaemic index
A number that ranks foods by how fast they raise blood sugar, measured in a group under set conditions. Food tech example: white bread has a higher glycaemic index than wholegrain bread.
Microbiome
The community of microbes living in the gut, which helps digest food and signals to the body. Food tech example: the gut microbiome changes with what a person eats, and it shapes how they respond to fibre.
Folate
A B vitamin needed to make new cells, important in pregnancy and in growth. Food tech example: folate in green leaves and beans supports healthy growth.
Recommended amount
The amount of a nutrient judged to meet the needs of most people in a group, used as a guide for planning diets. Food tech example: the recommended amount of iron is higher for women than for men.
Wholegrain
Grain that keeps all its parts, bran, germ, and endosperm, and that is high in fibre. Food tech example: wholegrain bread is made from wholegrain flour.
Fibre
The part of plant foods that is not digested in the small gut, which slows digestion and feeds gut microbes. Food tech example: fibre in oats slows the rise in blood sugar after breakfast.
Energy
The power a food gives the body, measured in kilojoules or kilocalories, used for movement, warmth, and growth. Food tech example: a bowl of rice gives energy for the afternoon's work.
Hormone
A chemical messenger made in one part of the body that acts in another part, such as insulin. Food tech example: stress hormones raise blood sugar during a worrying day.
Pancreas
An organ that makes insulin and digestive juices, sitting behind the stomach. Food tech example: the pancreas releases insulin after a meal.
Food diary
A record of what a person eats and drinks, and when, used to see patterns and to plan changes. Food tech example: a three-day food diary shows which meals keep a person full longest.
Glycaemic index (GI)
A number that says how fast a carbohydrate raises blood sugar, compared with pure glucose, which is given a GI of 100. Food tech example: a food with a GI of 70 raises blood sugar fast, while a food with a GI of 40 raises it slowly.
Glycaemic load (GL)
A number that combines GI with the grams of carbohydrate in a serving: GL equals GI times carbohydrate, divided by 100. Food tech example: a food with a GI of 70 and 30 g of carbohydrate in a serving has a GL of 21.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.
- Ordovas, J. M., Ferguson, L. R. and Tai, E. S., Personalised nutrition and health, 2018.
Unit 2: Genes, Microbiome and Metabolism: What Shapes Response
Note
Unit 2: Genes, Microbiome and Metabolism: What Shapes Response
A question to start with
Two sisters eat the same meal, share the same home, and walk the same streets. One gains weight easily and feels hungry soon after eating. The other stays lean and feels full for hours. Their genes are similar, but not the same. Their gut microbes are different, shaped by the foods they chose, the medicines they took, and the years they lived apart. Their bodies handle fat and sugar in different ways. This unit is about the three systems that shape those differences: genes, the gut microbiome, and metabolism, and about what each one can and cannot tell a person about what to eat.
Genes are the instructions written in DNA, the long molecule in almost every cell, that tell the body how to make proteins. Metabolism is the set of chemical changes that turn food into energy, into new body parts, and into waste. The microbiome is the community of microbes, mainly bacteria, that live in the gut, in numbers so large that they outnumber the body's own cells. Each of these shapes how a person responds to food, and each of them is shaped, in turn, by what the person eats. The three are not separate. Genes set the starting point, microbes change the food before the body meets it, and metabolism is the result of the two working together, in a body that is also shaped by age, activity, sleep, and stress.
Questions to think about
- If two people share the same genes, why might they still respond differently to the same food?
- A gut microbe test says a person has few of one kind of bacteria. Does that mean they should eat a particular food?
- Why can a gene that raises the risk of a condition not tell a person whether they will get it?
Think about the sisters before you read on
Because genes are only one part of the story: the same genes can be read differently depending on what the person eats, how they live, and which microbes live in their gut. Two people with similar genes can have different microbes, different habits, and different histories, and each of those changes how the genes are used. A gene is a tendency, not a fate.
Genes: the starting point
Every person carries a set of genes, about twenty thousand of them, written in DNA. Genes make proteins, and proteins do the work of the body: they digest food, carry oxygen, build muscle, and signal hunger and fullness. Small differences in genes, called variants, change how those proteins work, and those changes can alter how a person responds to food. One variant may change how fast a person breaks down caffeine, so coffee keeps one person awake and leaves another unaffected. Another may change how a person tastes bitterness, so one person likes green vegetables and another finds them bitter. Another may change how the body handles fat, so one person gains weight more easily on a high-fat diet.
Some gene variants are linked to conditions that matter for eating. Variants in genes that control appetite can make a person feel hungry more often, or feel full more slowly. Variants in genes that control how the body handles fat can change blood cholesterol after eating. Variants in genes that control how the body handles sugar can change the risk of type 2 diabetes. Variants in the gene for lactase, the enzyme that breaks down milk sugar, decide whether an adult can digest milk: some adults keep making lactase for life, and some stop making it after childhood, and that difference is written in their genes. Each of these is a real link, found in studies of many people, and each one is a tendency, not a certainty.
The limit of genetic testing for nutrition is that most food responses are shaped by many genes at once, not by one gene, and each gene adds only a small part. A test that looks at one gene can say that a person is more likely to respond in a certain way, but it cannot say what will happen, because the other genes, the microbes, and the life around them also matter. A test that looks at many genes can give a better picture, but it still cannot predict a person's response to a single meal, because a meal is shaped by the food, the gut, and the moment as well as by the genes. Genetic testing is most useful when it finds a clear, single-gene condition, such as the lactase variant, or a rare condition where one gene has a large effect. For most everyday eating, genes are a background, not a prescription.
The microbiome: the gut community
The gut microbiome is the community of microbes that live in the intestines, mainly bacteria, but also other tiny living things. There are trillions of them, in hundreds of kinds, and they do work the body cannot do on its own. They break down fibre that the body cannot digest, and they make short-chain fatty acids from it, which feed the gut wall and signal to the body. They make some vitamins, such as vitamin K and some B vitamins. They train the immune system, and they compete with harmful germs for space and food. They also change how food is handled: they can take energy from food the body would otherwise pass, they can change how fast the gut moves, and they can signal hunger and fullness through the gut-brain link.
The microbiome is shaped by what a person eats, by where they live, by the medicines they take, and by their early life. A diet high in fibre, with many different plant foods, tends to support a rich and varied microbiome. A diet low in fibre, with few plant foods, tends to support a less varied one. Antibiotics can reduce the microbiome sharply, and it may take weeks or months to recover. Birth, breastfeeding, and early foods shape the first microbiome, and that early community can leave a mark that lasts. Stress, sleep, and exercise also shape it, though less strongly than food. The microbiome is not fixed. It changes within days when the diet changes, and it changes back when the diet changes back, which means it is a reflection of how a person eats, as much as a cause of how they respond.
Testing the microbiome is now sold directly to the public, and the buyer should know what the test can and cannot say. A test can say which kinds of microbes are present, and in what amounts, in a stool sample. It can say whether the community is rich and varied, or poor and narrow. It can compare the result with other people. What it cannot say, on its own, is what the person should eat, because the science linking a particular microbe to a particular food response is still young, and because the same microbe can do different things in different people. A test that says "eat more of this food" on the basis of a single microbe is going further than the evidence supports. The safer use of a microbiome test is as a general guide: a poor and narrow community is a sign to eat more fibre and more varied plant foods, and a rich community is a sign that the diet is already supporting the gut well.
Metabolism: the chemical work of the body
Metabolism is the set of chemical changes that turn food into energy, into new body parts, and into waste. It has two broad sides: breaking down, in which food is split into smaller parts and energy is released, and building up, in which small parts are joined into new proteins, fats, and other body parts. The rate at which the body burns energy at rest is called the basal metabolic rate, and it depends on size, muscle, age, and sex. A person with more muscle burns more energy at rest than a person with less muscle at the same weight. A younger person burns more than an older person. A taller person burns more than a shorter person. Activity adds to the resting rate, and the total is the energy the person needs each day.
Metabolism differs between people in ways that matter for eating. Some people burn carbohydrate quickly and feel hungry soon after a starchy meal. Some people burn fat more readily and feel steady on a meal with more fat. Some people make insulin quickly after eating, and their blood sugar falls fast, leaving them hungry. Some people make insulin slowly, and their blood sugar stays high longer. These differences are shaped by genes, by muscle, by activity, by sleep, and by the microbiome, and they change over time: a person who becomes more active, or who gains muscle, changes their metabolism, and a person who sleeps poorly for a week changes it too. Metabolism is not a fixed number. It is a moving picture, and a plan that treats it as fixed will soon be out of date.
The body also adapts to what it is given. A person who eats less energy for a long time burns less energy at rest, because the body becomes more efficient, and that is one reason weight loss slows after the first weeks. A person who eats more protein builds more muscle if they also exercise, and that muscle raises the resting rate. A person who eats at night, when the body is preparing to sleep, handles sugar less well than the same person eating the same food in the morning. These adaptations are normal, and they are the reason a personal eating plan must be checked and changed over time, rather than written once and followed for ever.
How the three systems work together
Genes, microbes, and metabolism work together in a loop, and the loop is the reason personalised nutrition is complex. Genes set how the body makes proteins, including the proteins that digest food and signal hunger. Microbes change the food before the body meets it, breaking down fibre, making new substances, and signalling to the gut wall. Metabolism is the result: the energy released, the sugar in the blood, the fat stored, and the hunger felt after eating. What is eaten then changes the microbes, and the microbes change the signals, and the signals change what the person wants to eat next. A person who eats a lot of fibre feeds microbes that make short-chain fatty acids, which signal fullness, so the person eats less. A person who eats little fibre starves those microbes, so the signal is weaker, and the person feels hungry sooner. The loop can run in either direction, and the food is both the cause and the result.
That loop is why a single test cannot tell the whole story. A genetic test shows the starting point. A microbiome test shows the community at one moment. A metabolic test, such as a blood sugar measurement, shows the result at one moment. Put together, with the person's health, habits, and life, they give a picture that is better than any one alone. But the picture is still a snapshot, and it must be updated as the person changes, as the diet changes, and as the body adapts. A personalised plan is therefore not a prescription written once. It is a process of testing, changing, measuring, and changing again, and the person who follows it learns their own pattern over time.
What to do with this knowledge
The practical use of genes, microbiome, and metabolism is to guide choices, not to dictate them. A person who knows they carry a variant that makes them break down caffeine slowly can choose to drink coffee earlier in the day. A person who knows they stop making lactase as an adult can choose lactose-free milk, or hard cheese, which has little lactose. A person who knows their microbiome is poor and narrow can choose to eat more varied plant foods, with more fibre, and to keep doing so for weeks, because the microbiome changes slowly. A person who knows their blood sugar rises high after rice can choose a smaller portion of rice, or rice with beans and vegetables, or rice after a walk, and can measure the result. Each of these is a small, testable change, and the test is whether it works for that person, in their life, over time.
The buyer of a genetic or microbiome test should ask three questions before paying. First, what will the test change? If the answer is the same for every result, the test is not needed. Second, how strong is the evidence? If the link is based on a small study, or on a study in animals, the advice is a guess, not a fact. Third, who will read the result? A result read by a qualified professional, with the person's health and history in view, is worth more than a result read by an app alone. A test that is sold with a promise to tell the person exactly what to eat, without those three answers, is selling certainty that the science does not yet support.
The diagram shows the loop. The steps point to the diagram: the loop is why a single test cannot tell the whole story, and why a personal plan must be checked and changed over time.
Practical note: For one week, eat a different plant food each day that you do not usually eat: a bean, a lentil, a root vegetable, a leafy green, a nut, a seed, or a wholegrain you have not tried. Keep a short note of how you feel after each one, and count how many different plant foods you ate in the week. A higher count is a simple way to feed a richer microbiome, and it needs no test at all.
Quick self-check
Question: Why can a gene variant not tell a person exactly what to eat?
Because most food responses are shaped by many genes at once, each adding only a small part, and by the microbes, the life, and the moment as well. A variant can say a person is more likely to respond in a certain way, but it cannot predict what will happen after a single meal.
Question: Why does the microbiome change when the diet changes?
Because microbes feed on what the person eats, especially fibre, and different microbes thrive on different foods. A diet high in varied plant foods feeds a rich and varied community, while a diet low in fibre starves many kinds, so the community changes within days when the diet changes.
Try it yourself
Count the different plant foods you eat in one week: vegetables, fruits, beans, lentils, nuts, seeds, and wholegrains, counting each kind once. Write the number down. Then, for the next week, add five new kinds you do not usually eat, and count again. Note how you feel after meals in the second week compared with the first. That count is a simple measure of how varied your diet is, and variety is one of the best ways to support a rich gut microbiome.
Key ideas to carry forward
- Genes are instructions in DNA that make proteins, and small variants change how the body handles caffeine, lactose, fat, and appetite.
- Most food responses are shaped by many genes at once, so a gene variant is a tendency, not a fate, and it cannot predict a single meal.
- The gut microbiome is a community of microbes that breaks down fibre, makes vitamins and short-chain fatty acids, and signals hunger and fullness.
- The microbiome is shaped by food, medicines, and early life, and it changes within days when the diet changes.
- Metabolism is the chemical work of turning food into energy, body parts, and waste, and its rate differs by size, muscle, age, activity, and sleep.
- Genes, microbes, and metabolism work in a loop, with food both shaping the loop and being shaped by it.
- A genetic or microbiome test is useful only if it changes the advice, if the evidence is strong, and if the result is read by someone qualified.
Glossary
Gene
A part of DNA that carries instructions for making one protein. Food tech example: the gene for lactase carries instructions for making the enzyme that breaks down milk sugar.
DNA
The long molecule in cells that carries genetic instructions, written in a code of four bases. Food tech example: DNA in a cheek cell carries the instructions for how a person handles caffeine.
Variant
A small difference in a gene between people, which can change how a protein works. Food tech example: a variant in the lactase gene decides whether an adult keeps digesting milk.
Microbiome
The community of microbes living in the gut, which helps digest food and signals to the body. Food tech example: the gut microbiome changes with what a person eats, and it shapes how they respond to fibre.
Metabolism
The set of chemical changes that turn food into energy, into new body parts, and into waste. Food tech example: metabolism turns the carbohydrate in rice into glucose for energy.
Basal metabolic rate
The rate at which the body burns energy at rest, depending on size, muscle, age, and sex. Food tech example: a person with more muscle has a higher basal metabolic rate.
Short-chain fatty acid
A substance made by gut microbes from fibre, which feeds the gut wall and signals fullness. Food tech example: short-chain fatty acids made from oat fibre help the gut wall stay healthy.
Lactase
The enzyme that breaks down lactose, the sugar in milk, in the gut. Food tech example: a person who stops making lactase as an adult may feel discomfort after drinking milk.
Caffeine
A substance in coffee, tea, and some drinks that keeps a person awake, broken down at different speeds in different people. Food tech example: a person who breaks down caffeine slowly stays awake longer after coffee.
Insulin
A hormone made in the pancreas that helps glucose move from the blood into cells. Food tech example: insulin is released after a meal to lower blood sugar.
Enzyme
A protein that speeds up a chemical change in the body, such as breaking down a food part. Food tech example: lactase is the enzyme that breaks down lactose, the sugar in milk.
Vitamin K
A vitamin needed for blood clotting and bone health, made in part by gut microbes. Food tech example: vitamin K in green leaves supports blood clotting.
Gut-brain link
The signalling between the gut and the brain, through nerves and hormones, which shapes hunger and fullness. Food tech example: signals from the gut tell the brain that a meal was filling.
Antibiotic
A medicine that kills bacteria, which can sharply reduce the gut microbiome for weeks. Food tech example: an antibiotic course can leave the gut microbiome poorer until it recovers.
Protein
A body part made of amino acids, used to build muscle, enzymes, and other body parts, and found in foods such as beans, eggs, and meat. Food tech example: protein in lentils helps build and repair muscle.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Ordovas, J. M., Ferguson, L. R. and Tai, E. S., Personalised nutrition and health, 2018.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.
- Sonnenburg, J. L. and Sonnenburg, E. D., The Good Gut, 2015.
Unit 3: Measuring the Body: CGM, Wearables and Tests
Note
Unit 3: Measuring the Body: CGM, Wearables and Tests
A question to start with
A woman wears a small white disc on the back of her arm. It is a continuous glucose monitor, a sensor that reads her blood sugar every few minutes, day and night, and sends the numbers to her phone. She eats a bowl of oats and watches the line rise gently. She eats white rice and watches it rise steeply, then fall fast, leaving her hungry. She eats the same rice after a walk, and the line rises less. In a week, she has learned more about her own body than a year of general advice taught her. This unit is about the tools that make that learning possible: glucose sensors, wearables, and blood tests, and about how to use them well and how to avoid being misled by them.
Measuring the body has changed fast. A generation ago, blood sugar was measured by pricking a finger, a few times a day, and the numbers were written in a book. Now a sensor on the arm measures it every few minutes, and a watch on the wrist counts steps, measures heart rate, and tracks sleep. Blood tests that once needed a hospital can now be ordered online and done at home. Each of these tools gives numbers, and numbers feel certain, but each tool has limits, and a number without its limit is a trap. This unit teaches the tools, their limits, and the habit of reading them together, not one at a time.
Questions to think about
- A glucose sensor shows a steep rise after a meal. Does that mean the meal was bad, or that the rise was normal for that person at that time?
- A watch says you slept badly. Should you change your diet because of one night of data?
- A blood test shows a vitamin is low. What should you check before taking a supplement?
Think about the steep rise before you read on
It means the rise happened, and it must be read with the rest of the picture: what was eaten, how much, what else was eaten with it, whether the person walked, slept, or was stressed, and what the rise usually looks like for that person. A single rise, on its own, does not judge a meal. A pattern over days, with the same meal eaten in the same way, does.
Continuous glucose monitoring: what it is and what it shows
A continuous glucose monitor, or CGM, is a small sensor worn on the skin, usually on the back of the arm or on the belly, that measures glucose in the fluid just under the skin every few minutes, and sends the numbers to a phone or a reader. It is worn for one or two weeks at a time, depending on the type, and it shows a line, a trace, of glucose through the day and night: the rise after each meal, the fall after, the level during sleep, and the effect of exercise, stress, and illness. It does not measure blood glucose directly. It measures the fluid under the skin, which follows blood glucose with a short delay, of some minutes, so the sensor lags behind a finger prick when glucose is changing fast.
The CGM shows three things that a finger prick cannot show easily. First, it shows the pattern: how high the glucose rises after each meal, how long it stays high, and how fast it falls. Second, it shows the night: whether glucose falls low during sleep, or stays high through the night. Third, it shows the effect of ordinary life: a walk after a meal flattens the rise, a poor night of sleep makes the rise higher, a stressful day makes it higher, and an illness makes it higher still. Those patterns are the useful part of a CGM, because they show the person's own response, in their own life, over days, not in a clinic for an hour.
The CGM also has limits that the buyer must know. The sensor can be wrong, especially on the first day, when it is settling, and when glucose is changing fast, when the lag matters most. The sensor can be knocked off, can irritate the skin, and can give a low reading if the person lies on it during sleep, a false low called a compression low. The sensor measures glucose, not health: a flat line is not a promise of health, and a steep rise is not a verdict on a meal. And the sensor can create worry: a person who watches the line all day can become anxious about every rise, and can start to avoid foods that are fine, or to eat in a way that flattens the line but leaves them hungry and unhappy. A CGM is a tool for learning, and like any tool, it needs a purpose, a time limit, and a person to help read it.
Wearables: steps, heart rate, and sleep
A wearable, such as a watch or a band, measures movement, heart rate, and sleep, and some measure blood oxygen and skin temperature. The step count shows how active a person is, and it is the most reliable number a wearable gives, though it can be fooled by a bumpy ride or by waving an arm. Heart rate shows how hard the body is working, at rest and during exercise, and a resting heart rate that falls over weeks is a sign of improving fitness. Sleep tracking shows how long a person was in bed, how much they moved, and an estimate of light and deep sleep, though the estimate is rough, because a wearable cannot measure the brain waves that define sleep stages. It can say that a person slept badly, but it cannot say why, or whether the sleep was truly poor.
Wearables help personalised nutrition in three ways. First, they show activity, and activity changes how food is handled: a walk after a meal lowers the glucose rise, and a day with many steps burns more energy. Second, they show sleep, and sleep changes how food is handled: a short night raises hunger, raises the glucose rise, and makes a person choose sweeter foods. Third, they show patterns over weeks, and patterns are more useful than single days: a person who sees that they sleep better on days they walk more, and eat better on days they sleep better, has learned something about their own loop. The wearable does not give that lesson. The person gives it to themselves, by looking at the pattern.
The limits of wearables are the same as the limits of any measurement: they estimate, they err, and they can mislead. A step count can be high on a day when the person did little, because the watch counted a car ride. A sleep score can be low on a night when the person slept well, because the watch read movement as waking. A heart rate can be high because of stress, not exercise. The habit to learn is to look at trends over weeks, not at single days, and to check the wearable against how the person feels: if the watch says the sleep was poor and the person feels rested, the feeling matters more. A wearable is a guide, not a judge.
Blood tests: what they can and cannot say
Blood tests measure what is in the blood at one moment: sugar, cholesterol, vitamins, minerals, and markers of how organs are working. A fasting glucose test measures sugar after an overnight fast. A glycated haemoglobin test, called HbA1c, measures the average blood sugar over the past two to three months, because glucose sticks to haemoglobin in red blood cells, and the amount that sticks shows the average. A cholesterol test measures the fats in the blood, including the types that carry cholesterol to and from tissues. A vitamin D test measures the vitamin made in skin by sunlight and taken in food. An iron test measures iron stores and whether the blood can carry enough oxygen. Each test is a snapshot, and the snapshot must be read with the person's health, medicines, and history.
Blood tests are useful when they change the advice. A high HbA1c in a person with diabetes changes the eating plan, because it shows the average sugar has been high for months. A low iron test in a person who feels tired changes the advice, because it points to a cause that food can help. A low vitamin D test in a person who spends little time outdoors changes the advice, because it points to a need for food or for a supplement. A blood test that is normal, and that does not change the advice, is a test that did not need to be done, and the money spent on it could have been spent on food.
Blood tests also have limits. A single test can be high or low for a short-term reason: a meal before the test, an illness, a medicine, or a hard exercise session can all change the numbers. A test that is slightly outside the normal range may mean nothing, because the normal range is set for a group, and a healthy person can sit just outside it. A test for a vitamin can be low because the person has not eaten that vitamin lately, or because the body is not absorbing it, and the fix is different in each case. A blood test must be read by someone who knows the person's history, and who can say whether the result needs a change in food, a change in medicine, or simply a repeat test later.
Putting the measurements together
No single measurement tells the whole story, and the skill is to put them together. A CGM that shows a steep rise after rice, a wearable that shows a short night before it, and a food diary that shows the rice was eaten alone, without beans or vegetables, together point to a simple change: eat the rice with beans and vegetables, after a better night, or after a walk. A blood test that shows a high HbA1c, a wearable that shows little activity, and a diary that shows sugary drinks every day together point to a different change: cut the drinks, walk more, and check the HbA1c again in three months. Each measurement adds a piece, and the pieces must be read together, with the person's life in view.
The habit to build is a short, regular review: once a week, look at the CGM pattern, the wearable trends, the food diary, and how the person feels, and ask three questions. What worked this week? What did not work? What one change will I try next week? That review, done by the person with a professional where needed, is the heart of personalised nutrition, and it needs no new device. A person who reviews weekly, and who changes one thing at a time, will learn more in a month than a person who buys every test and never looks at the pattern.
The diagram shows the measurement habit. The steps point to the diagram: the question at the start and the stopping at the end are the two steps that prevent measuring from becoming a habit without a purpose.
Privacy, cost, and when to stop measuring
Measuring the body creates data about the body, and that data is personal. A CGM trace, a wearable record, and a blood test result together say a great deal about a person's health, habits, and life. That data should be kept by the person, shared only with people the person chooses, such as a doctor or a dietitian, and not sold or shared without clear consent. A buyer should read the terms of any app or service before using it, and should ask where the data is stored, who can see it, and how it can be deleted. A service that does not answer those questions clearly should not be trusted with health data.
Cost matters too. A CGM, a wearable, and a set of blood tests can cost a lot of money, and the money spent on measuring is money not spent on food. A person should measure when the measurement will change the advice, and should stop when the question is answered. A CGM worn for two weeks to learn a personal pattern is a good use. A CGM worn for a year, watched every day, with no change in eating, is a cost without a benefit. The best measure of a tool is not how much data it gives, but how much it changes what the person does, and whether that change lasts.
Deriving CGM numbers: time in range and glucose variability, step by step
A continuous glucose monitor gives a trace, a line that rises and falls through the day, and the trace must be turned into numbers before it can be used. Two numbers matter most. The first is time in range, the share of the day that blood sugar spends inside a target range set by the person's care team. The second is glucose variability, how much the blood sugar swings up and down, measured by the coefficient of variation, CV, which is the standard deviation divided by the mean, times 100. Both are derived from the trace, step by step, and both are worth seeing worked through once, with round teaching numbers only, so the reader can follow the arithmetic and see what each number means.
Time in range is a count. Suppose a CGM takes a reading every 15 minutes, which is 96 readings in a day. Suppose 67 of those readings fall inside the target range. The time in range is 67 divided by 96, times 100, which is about 70 percent. That means the person's blood sugar was inside the range for about 70 percent of the day, and outside it, above or below, for about 30 percent. The target range itself, and the time in range a person should aim for, must be set by their doctor, not by this unit, and the numbers here are teaching numbers, not targets. What matters is the method: count the readings inside the range, divide by the total number of readings, and multiply by 100. A higher time in range means steadier blood sugar, and a lower time in range means the person spent more of the day too high or too low.
Glucose variability is a measure of swing. The mean is the average of all the readings. The standard deviation, SD, is a measure of how far the readings spread around the mean: a small SD means the readings sit close to the mean, and a large SD means they swing widely. The coefficient of variation, CV, is the SD divided by the mean, times 100, which gives the swing as a percentage of the mean, so that two people with different means can be compared. A worked example, with round teaching numbers only, shows the arithmetic. Suppose a person's mean glucose is 8 units, and their SD is 2 units. The CV is 2 divided by 8, times 100, which is 25 percent. Suppose another person's mean is also 8, but their SD is 4. Their CV is 4 divided by 8, times 100, which is 50 percent. The two people have the same mean, but the second person's blood sugar swings twice as much, and that swing, the ups and downs, is what the CV captures, and what a plan tries to smooth.
Both numbers must be read together. A person can have a good mean and a bad CV, with blood sugar swinging from high to low around a fine average, and a person can have a good time in range and a bad CV, if the swings stay inside the range but are still large. A person can also have a good CV and a bad mean, with blood sugar steady but steady too high. The three numbers, mean, time in range, and CV, each say something the others do not, and the care team reads them together, with the food diary beside them, to see which meals cause the swings, and which meals keep the trace steady. A single number, on its own, is not a verdict. It is one piece of the picture.
Practical note: If you use a wearable, look at its weekly trend, not its daily score, for one month. Write down one line each Sunday: steps, sleep, and how you felt. At the end of the month, look for the link between the three: do you eat better on days you walk more, and do you sleep better on days you eat better? That link, if it is there, is your own pattern, and it is worth more than any single day's score.
Quick self-check
Question: Why is a pattern over days more useful than a single glucose reading?
Because a single reading is shaped by the moment: what was eaten, how much, sleep, stress, activity, and illness. A pattern over days, with the same meal eaten in the same way, shows the person's own response, and it can be tested by changing one thing and watching the pattern change.
Question: Why should a blood test be read with the person's history, not on its own?
Because a single test can be high or low for a short-term reason, such as a meal, an illness, or a medicine, and because the normal range is set for a group. A person who knows the history can say whether the result needs a change in food, a change in medicine, or simply a repeat test later.
Try it yourself
Keep a food and feeling diary for one week, without any device: write down what you eat, when you eat it, how active you were, how you slept, and how hungry and energetic you felt an hour after each meal. At the end of the week, look for your own pattern: which meals kept you full, which days you slept best, and whether activity and sleep lined up with better eating. That diary is the base that any device should build on, and it costs nothing.
Key ideas to carry forward
- A continuous glucose monitor shows the pattern of glucose through the day and night, with a short delay, and it is most useful for learning personal responses over days.
- A CGM can be wrong on the first day, can lag when glucose changes fast, and can create worry if watched all day without a purpose.
- Wearables measure steps, heart rate, and sleep, and they are most useful for trends over weeks, not for single days.
- Blood tests are snapshots that must be read with health, medicines, and history, and they are useful when they change the advice.
- No single measurement tells the whole story: the skill is to put the CGM, the wearable, the food diary, and how the person feels together.
- Review weekly, change one thing at a time, and measure again to see if the change worked.
- Health data is personal: share it only with people the person chooses, and stop measuring when the question is answered.
Glossary
Continuous glucose monitor
A small sensor worn on the skin that measures glucose in the fluid under the skin every few minutes and sends the numbers to a phone. Food tech example: a CGM shows how blood sugar rises after a bowl of rice.
Wearable
A device worn on the body, such as a watch or band, that measures steps, heart rate, and sleep. Food tech example: a wearable shows that a person walks more on days they eat better.
HbA1c
A blood test that measures average blood sugar over the past two to three months, by how much glucose has stuck to haemoglobin. Food tech example: an HbA1c test shows whether blood sugar has been high for months.
Haemoglobin
The protein in red blood cells that carries oxygen, and that glucose sticks to over time. Food tech example: glucose sticks to haemoglobin, so HbA1c shows the average blood sugar.
Compression low
A false low glucose reading from a CGM when the person lies on the sensor during sleep. Food tech example: a CGM shows a low during the night because the sensor was pressed by the body.
Resting heart rate
The heart rate when the body is at rest, which falls over weeks as fitness improves. Food tech example: a resting heart rate that falls over a month is a sign of improving fitness.
Fasting glucose
Blood sugar measured after an overnight fast, before eating. Food tech example: a fasting glucose test is taken in the morning before breakfast.
Cholesterol
A fat-like substance in the blood, carried by different types of particles, and linked to heart health. Food tech example: a cholesterol test measures the fats in the blood.
Vitamin D
A vitamin made in skin by sunlight and taken in food, needed for bone health and other work. Food tech example: vitamin D is low in a person who spends little time outdoors.
Iron
A mineral needed to carry oxygen in the blood, found in meat, beans, and green leaves. Food tech example: iron in lentils supports the blood's ability to carry oxygen.
Food diary
A record of what a person eats and drinks, and when, used to see patterns and to plan changes. Food tech example: a three-day food diary shows which meals keep a person full longest.
Trend
A pattern over time, such as steps rising over weeks, which is more useful than a single day's number. Food tech example: a trend of better sleep over a month is more useful than one night's score.
Snapshot
A measurement at one moment, such as a blood test, which must be read with the history around it. Food tech example: a blood test is a snapshot of the blood on that day.
Consent
Clear agreement to share personal data, given by the person the data is about. Food tech example: a person gives consent for their CGM data to be shared with their doctor.
Pattern
A repeated shape in data, such as a glucose rise that happens after the same meal each time. Food tech example: a pattern of high rises after rice points to a change in portion or pairing.
Time in range
The share of the day that blood sugar spends inside a target range set by the person's care team. Food tech example: a time in range of 70 percent means blood sugar was inside the range for about 70 percent of the day.
Coefficient of variation (CV)
A measure of glucose variability: the standard deviation divided by the mean, times 100. Food tech example: a mean of 8 and a standard deviation of 2 gives a CV of 25 percent.
Standard deviation
A measure of how far readings spread around their mean: a small value means the readings sit close to the mean. Food tech example: a small standard deviation in glucose readings means blood sugar swings little through the day.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.
- American Diabetes Association, Standards of Care in Diabetes, 2024.
Unit 4: AI Meal Planning: How It Works, What It Can and Cannot Do
Note
Unit 4: AI Meal Planning: How It Works, What It Can and Cannot Do
A question to start with
A person opens an app, types "I am vegetarian, I do not like mushrooms, I need lunch for four people, and I have 30 minutes", and in seconds the app gives a plan: a chickpea curry, a shopping list, and the nutrition for each serving. It feels like magic, and it feels personal. But where did the plan come from, who checked it, and what would happen if the person followed it every day for a year? This unit is about the machine behind the magic: how AI meal planning works, what it does well, what it does badly, and how to use it without letting it do the thinking for you.
AI meal planning is the use of computer programs, often called artificial intelligence, to suggest meals, recipes, and shopping lists based on information about a person. The programs are trained on large sets of data: recipes, nutrition tables, food prices, and, in some cases, records of what people ate and how they responded. They use that training to guess what will fit a person's needs, likes, and limits, and to put together a plan that meets a set of rules, such as a calorie target, a protein target, or a list of foods to avoid. The plan looks personal, because it uses personal information, but it is built from patterns in data, and patterns are not the same as knowledge about one person.
Questions to think about
- An app gives a meal plan in seconds. What did it use to make the plan, and what did it not know about you?
- An AI plan says a meal has 25 g of protein. How would you know if that number is right?
- If an AI can plan meals, why would a person still need a dietitian?
Think about the protein number before you read on
By checking it against a trusted source, such as a nutrition table or the food's label, and by asking whether the portion matches what the app assumed. AI can misread a recipe, use the wrong ingredient, or assume a portion that is different from the one you will eat, so the number must be checked before it is relied on, especially for health.
How AI meal planning works, in plain words
AI meal planning works in three steps: it takes in information, it matches that information against patterns, and it gives out a plan. The information can be what the person types: age, sex, height, weight, activity, health, likes, dislikes, allergies, budget, and time. It can be what the person has done before: meals they liked, meals they skipped, and foods they bought. It can be what a device measured: steps, sleep, or glucose from a sensor. The program puts that information together and looks for meals that fit the rules: enough protein, not too much sugar, no allergens, within the budget, and ready in the time given.
The matching is done by a model, a set of rules learned from data. The model may be simple: a set of if-then rules written by a person, such as "if the person is vegetarian, do not suggest meat". Or it may be complex: a machine learning model trained on thousands of recipes and nutrition tables, which has learned which foods tend to go together, which meals meet which targets, and which plans people tend to like. The model does not understand food. It does not taste, it does not know what a meal feels like to eat, and it does not know the person's life. It finds patterns in numbers, and it gives out the pattern that fits best. That is both its strength and its limit: it can search thousands of options in seconds, and it can miss the thing that matters most to the person in front of it.
The plan that comes out is usually a set of meals for a day or a week, with recipes, portions, shopping lists, and nutrition numbers. Some apps also adjust the plan as the person uses it: if the person skips a meal, the app suggests another; if the person says they are still hungry, the app adds a snack; if the person's glucose rises after a meal, the app suggests a different meal next time. That adjustment is called feedback, and it is the part of AI planning that feels most personal, because the plan changes with the person. But the adjustment is still based on patterns, and it still needs a person to judge whether the change is right.
What AI meal planning does well
AI meal planning does several things well, and they are worth using. It is fast: it can give a plan for a week in seconds, where a person would take an hour. It is broad: it can search thousands of recipes, including foods the person has never tried, and it can find meals that meet a set of rules, such as a protein target and a budget, at the same time. It is consistent: it does not forget a rule, it does not get tired, and it applies the same limits to every meal. It is good at counting: it can add up energy, protein, fibre, sugar, and salt across a day, and it can show whether the day meets the targets. And it is good at lists: it can turn a week's meals into a shopping list, grouped by shop section, so the person buys only what is needed.
AI planning is also good at variety, when it is set up well. A person who plans their own meals tends to repeat the same five or six dishes, because those are the ones they know. An AI can suggest new dishes, new vegetables, and new ways to use a food, and that variety can make eating more interesting and can bring in nutrients the person was missing. It can also adapt to a change quickly: if the person says they have less time this week, the app can suggest quicker meals; if the person says they have a smaller budget, it can suggest cheaper ones. That speed of change is hard for a person to match on their own.
What AI meal planning cannot do
AI meal planning cannot do several things, and those are the things that matter most for health. It cannot diagnose: it cannot say whether a person has diabetes, coeliac disease, or a food allergy, and it cannot say what to eat for a medical condition. It cannot judge safety: it cannot say whether a meal is safe for a person with kidney disease, or whether a supplement is safe with a medicine, and a plan that is safe for most people may be unsafe for one person. It cannot taste: it cannot say whether a meal will taste good, whether the texture will be right, or whether the person will enjoy it, and a plan that is not enjoyed will not be followed. It cannot know the person's life: it does not know that the person cooks for a family, that they eat out on Fridays, that they cannot afford a food, or that they do not like the smell of a food, unless the person tells it, and even then it may not weigh that information as a person would.
AI can also be wrong, and its errors are not always obvious. It can misread a recipe, using the wrong ingredient or the wrong amount, and give a nutrition number that is wrong. It can invent a recipe that looks right but does not work in a kitchen, because it has never cooked. It can suggest a food that contains an allergen the person must avoid, because the allergen was hidden in a compound ingredient. It can give advice that is out of date, because it was trained on data from years ago. It can repeat a bias in its training data, suggesting foods that fit one culture and not another, or suggesting expensive foods to people who cannot afford them. Each of these errors is a reason to check the plan, not to follow it blindly, and the check must be done by a person who knows food and health.
The data problem: what the AI was trained on
The quality of an AI plan depends on the data it was trained on, and that data is rarely shown to the user. A model trained on recipes from one country will suggest foods from that country, and may not know the foods of another. A model trained on nutrition tables that are out of date will give numbers that are out of date. A model trained on records of what people ate, without knowing whether those people were healthy, will learn to suggest what people ate, not what they should eat. A model trained on a small set of people will not know how to plan for a person who is different from them. The buyer of an AI service should ask what data it was trained on, how recent it is, and whether it was checked by dietitians or doctors. A service that cannot answer those questions is asking the buyer to trust a black box, and a black box should not be trusted with health.
Privacy is part of the data problem. An AI meal planner that knows a person's health, weight, likes, and habits holds a great deal of personal information. That information should be kept by the person, shared only with people the person chooses, and not used to sell the person more food, more tests, or more services. A service that gives a free plan and then sells the data, or that shows adverts for foods based on the person's health, is not a health service. It is a marketing service, and the buyer should treat it as one. The terms of the service, which few people read, say what happens to the data, and they should be read before the person types in their health information.
Using AI well: a habit for the reader
AI meal planning is best used as a helper, not as a decider. The person decides the aim: what they want to change, and why. The AI suggests options: meals, recipes, and lists that fit the aim. The person checks the options: do they fit the health, the allergies, the budget, the time, and the taste? The person chooses, cooks, and eats, and then judges: did it work, did it taste good, was it enough, and would I eat it again? That judgement is fed back to the AI, or simply kept by the person, and the next plan is better for it. The AI does the searching and the counting. The person does the judging, the tasting, and the deciding. That division keeps the strengths of the AI and the strengths of the person together, and it keeps the person in charge of their own eating.
A practical habit is to use the AI for the first draft, and to check the draft before cooking. Check the allergens first, by reading the recipe and the ingredients, not by trusting the app's filter. Check the portions, by weighing the main ingredients the first time, so the app's numbers match the plate. Check the nutrition, by comparing one or two numbers with a trusted table or a label. Check the cost, by pricing the shopping list. And check the time, by cooking the meal once and timing it. If the draft passes those checks, cook it. If it fails, change it, and tell the app what was wrong, so the next draft is better. That habit takes a few minutes, and it is the difference between using AI and being used by it.
The diagram shows the habit. The steps point to the diagram: the person decides the aim and judges the result, the AI does the searching and counting in the middle, and a professional is called in when health is at stake.
When to ask a professional
AI meal planning is not a substitute for a dietitian or a doctor, and the line between them is clear. A dietitian is needed when a person has a medical condition, such as diabetes, kidney disease, coeliac disease, or a food allergy, when a person takes medicine that changes how food works, when a person is pregnant, breastfeeding, or recovering from illness, when a child or an older person needs a special diet, or when a person has tried to change their eating and it has not worked. A dietitian can diagnose, can read tests with the person's history, can judge safety, and can give advice that carries professional responsibility. An AI can do none of those, and a service that claims it can is overclaiming.
The best use of AI and a professional together is to let the AI do the routine work, and the professional do the judgement. The person can use the AI to find meals, to count nutrients, and to make shopping lists, and can take the plan to the dietitian, who checks it, changes it where needed, and sets the targets. That way the person gets the speed of the AI and the judgement of the professional, and the professional's time is spent on the decisions that need a person, not on the counting that a machine can do. A health system that uses AI in that way will help more people, and will help them better, than a system that uses AI alone.
Practical note: Ask an AI meal planner for one day of meals that fit a simple aim, such as more vegetables and less sugar, with a budget and a time limit. Then check the plan yourself: read each recipe for allergens, weigh the main ingredients, compare one nutrition number with a trusted table, and price the shopping list. Write down what the AI got right and what it got wrong. That check is the skill this unit teaches, and it takes less time than cooking the meals.
Quick self-check
Question: Why can an AI meal plan be wrong about protein in a meal?
Because it can misread a recipe, use the wrong ingredient or amount, or assume a portion that is different from the one you will eat. The number must be checked against a trusted table or label, and the portion must be weighed the first time, before the number is relied on.
Question: When should a person ask a dietitian instead of using an AI plan?
When they have a medical condition, take medicine that changes how food works, are pregnant, breastfeeding, or recovering from illness, need a special diet for a child or an older person, or have tried to change their eating and it has not worked. A dietitian can diagnose, judge safety, and give advice with professional responsibility, which an AI cannot.
Try it yourself
Use any AI meal planner you can access, or a recipe site with a filter, to get a plan for one day that fits your likes, your budget, and your time. Then check it with the habit in this unit: allergens, portions, nutrition, cost, and time. Cook one meal from the plan and judge it: did it taste good, was it enough, and would you eat it again? Write down your verdict, and what you would change in the plan next time.
Key ideas to carry forward
- AI meal planning takes in personal information, matches it against patterns learned from data, and gives out meals, recipes, and shopping lists.
- AI is fast, broad, consistent, and good at counting and at lists, and it can suggest variety a person might not find alone.
- AI cannot diagnose, cannot judge safety for a medical condition, cannot taste, and cannot know a person's life unless it is told.
- AI can be wrong: it can misread a recipe, invent a meal that does not work, miss an allergen, or repeat a bias in its training data.
- The quality of an AI plan depends on the data it was trained on, and the buyer should ask what that data is and whether it was checked by professionals.
- Use AI as a helper, not a decider: decide the aim yourself, let the AI suggest, check the draft, cook, judge, and feed the judgement back.
- Ask a dietitian or doctor when health, medicine, or a condition is involved, and let the AI do the routine work while the professional does the judgement.
Glossary
Artificial intelligence
A computer program that learns patterns from data and uses them to make suggestions or decisions. Food tech example: an AI meal planner learns from recipes and nutrition tables to suggest meals.
Machine learning
A kind of artificial intelligence that learns from examples, rather than from rules written by a person. Food tech example: a machine learning model learns which foods tend to go together from thousands of recipes.
Model
A set of rules or patterns learned from data, used by a program to make a suggestion. Food tech example: a meal planning model has learned which meals meet a protein target within a budget.
Training data
The data a model learns from, such as recipes, nutrition tables, and records of what people ate. Food tech example: a model trained on recipes from one country suggests foods from that country.
Feedback
Information about what happened after a suggestion, used to make the next suggestion better. Food tech example: telling an app that a meal was too small is feedback that changes the next plan.
Dietitian
A qualified professional who gives food and nutrition advice for health, including for medical conditions. Food tech example: a dietitian plans meals for a person with diabetes.
Portion
The amount of a food that is served or eaten at one time. Food tech example: a portion of rice in a meal plan is the amount the person will eat.
Shopping list
A list of foods to buy to make a set of meals, grouped by shop section. Food tech example: an AI plan turns a week's meals into a shopping list.
Bias
A slant in data or in a model, such as suggesting foods from one culture and not another. Food tech example: a model trained on expensive recipes suggests expensive foods to people who cannot afford them.
Black box
A system whose inner working cannot be seen, so the user cannot tell how it reached its answer. Food tech example: an AI that gives a meal plan without saying what data it used is a black box.
Privacy
The right of a person to keep their personal data private, and to choose who can see it. Food tech example: health data given to a meal planning app should not be sold without clear consent.
Consent
Clear agreement to share personal data, given by the person the data is about. Food tech example: a person gives consent for their health data to be shared with their doctor.
Nutrition table
A table giving the nutrients in a food, such as energy, protein, sugar, and salt, used to check a meal plan. Food tech example: a nutrition table shows the protein in 100 g of lentils.
Allergen
A food part, usually a protein, that causes an immune response in some people. Food tech example: peanut protein in a biscuit is an allergen for a person with peanut allergy.
Calorie target
A daily energy target, in kilocalories, set for a person's needs and aims. Food tech example: a meal plan is built to meet a daily calorie target.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Ordovas, J. M., Ferguson, L. R. and Tai, E. S., Personalised nutrition and health, 2018.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.
- Topol, E., Deep Medicine, 2019.
Unit 5: Eating with GLP-1 Medicines: Protein, Fibre and Portions
Note
Unit 5: Eating with GLP-1 Medicines: Protein, Fibre and Portions
A question to start with
A woman starts a GLP-1 medicine for her weight and blood sugar. In the first weeks, she feels full after a few bites, she has no wish for a large meal, and she sometimes feels sick if she eats too fast. She eats less, and she loses weight, but she also worries: is she eating enough protein to keep her muscle, is she eating enough fibre to keep her gut working, and is she drinking enough? Her doctor says the medicine is working, and her dietitian says the food must change with it. This unit is about that change: how GLP-1 medicines work, how they change appetite and eating, and how to eat well while taking them, with enough protein, enough fibre, and portions that fit the new appetite.
GLP-1 medicines are a group of medicines that copy a hormone the body makes after eating, called glucagon-like peptide-1, or GLP-1. The hormone signals fullness to the brain, slows how fast the stomach empties, and helps the pancreas release insulin when blood sugar is high. The medicines copy that hormone, in a stronger and longer-lasting form, so the signals last longer. They are used to treat type 2 diabetes and, in some cases, to support weight loss, always under a doctor's care. They are not a diet, and they are not a substitute for food. They change how the body responds to food, and the food must change with them, or the person can end up eating too little of the nutrients they need.
Questions to think about
- A medicine makes you feel full after a few bites. Why might you still need to plan protein carefully?
- Why does fibre matter more, not less, when a person is eating smaller amounts?
- A person on a GLP-1 medicine feels sick after a large, fatty meal. What is happening, and what should change?
Think about the sick feeling before you read on
Because the medicine slows how fast the stomach empties, so a large, fatty meal sits in the stomach longer, and the fullness signal stays strong. Fat slows the stomach further, so a large fatty meal can cause nausea, discomfort, and vomiting. The fix is smaller portions, eaten slowly, with less fat at one meal, and with protein and fibre spread through the day rather than loaded into one large meal.
How GLP-1 medicines work
GLP-1 is a hormone made in the gut after eating. It tells the brain that food has arrived, it slows the stomach so food leaves it more slowly, and it helps the pancreas release insulin when blood sugar is high, while also lowering a hormone that raises blood sugar. The effect is a lower blood sugar after meals, a longer feeling of fullness, and a smaller appetite. GLP-1 medicines copy that hormone, but they last much longer in the body than the natural hormone, which is broken down within minutes. Some are taken as a weekly injection, some as a daily injection, and some as a daily tablet, and the dose is raised slowly over weeks, so the body can adapt and the side effects stay manageable.
The main effects on eating are three. First, appetite falls: the person feels hungry less often, and feels full sooner when they eat. Second, the stomach empties more slowly, so food stays in the stomach longer, and the person feels full for longer after a meal. Third, cravings change: many people find they want less sweet and fatty food, and they find it easier to stop eating when they are full. Those effects help with weight loss and with blood sugar, because the person eats less energy without feeling as hungry as they would on a diet alone. But the same effects create new risks: eating too little protein, eating too little fibre, drinking too little, and losing muscle as well as fat, if the food is not planned.
The medicines also have side effects, and the food can make them better or worse. Nausea is the most common, especially when the dose is raised, or when a meal is large, fatty, or eaten quickly. Vomiting, diarrhoea, and constipation can also happen, and they are more likely when the diet is low in fibre, low in fluid, or high in fat. Heartburn can happen, because the stomach empties slowly. These side effects are usually mild and they often settle as the body adapts, but they are a sign that the food needs to change: smaller meals, eaten slowly, with less fat, more fibre, and enough fluid, and with the dose raised only as the doctor directs. A person who has severe vomiting, who cannot keep fluid down, or who has severe pain in the belly should seek medical care at once, and should not try to manage it with food alone.
Protein: keeping muscle while losing weight
When a person loses weight, they lose fat and they can also lose muscle, and muscle matters. Muscle keeps the body strong, keeps the resting energy burn higher, and helps the body handle sugar. Losing muscle makes it harder to keep the weight off, because the body burns less energy at rest, and it makes the person weaker and more tired. Protein is the nutrient that protects muscle, but only if it is eaten in enough amount, spread through the day, and paired with activity that uses the muscle, such as walking, climbing stairs, and resistance exercise. A person on a GLP-1 medicine who eats very little, and who eats little protein, will lose more muscle than a person who eats enough protein and stays active.
How much protein is enough depends on the person, their size, their age, and their activity, and the right amount should be set with a doctor or dietitian, not from a single number in this unit. In general, a person losing weight needs more protein per kilogram than a person who is not losing weight, because the body is using protein to repair and to keep muscle while it is also using fat for energy. The protein should be spread across the day, with some at each meal, rather than loaded into one meal, because the body can only use a certain amount at one time to build muscle, and the rest is used for energy. A small amount at breakfast, a moderate amount at lunch, and a moderate amount at dinner, with a protein food at a snack if needed, is a pattern that works for many people.
Good protein foods are those that give a lot of protein for the energy they carry, and that are easy to eat in small amounts. Eggs, fish, chicken, beans, lentils, tofu, yoghurt, and cheese are all good choices, and the choice depends on what the person likes, can afford, and can digest. A person who feels full quickly may find a small amount of a protein-rich food easier than a large plate of food, and may do better with foods that are soft and easy to eat, such as eggs, yoghurt, and fish, rather than tough meat. A person who is vegetarian can meet their protein needs with beans, lentils, tofu, eggs, and dairy, but they need to plan, because plant proteins are often lower in protein per bite, and they need to be eaten in larger amounts or combined through the day.
Fibre: keeping the gut working
Fibre is the part of plant foods that is not digested in the small gut, and it matters more, not less, when a person is eating less. Fibre slows digestion, so blood sugar rises more gently. Fibre feeds the gut microbes, which make short-chain fatty acids that keep the gut wall healthy. Fibre adds bulk to the stool, so the gut keeps moving, and it helps prevent constipation, which is a common side effect of GLP-1 medicines, because the gut moves more slowly and the person may be eating less and drinking less. A person who eats too little fibre while taking a GLP-1 medicine is more likely to be constipated, to feel bloated, and to have gut discomfort.
Fibre should be raised slowly, with enough fluid, because a sudden rise in fibre can cause wind and bloating, especially when the gut is already moving slowly. The best sources are foods the person already likes: oats, beans, lentils, vegetables, fruit, nuts, seeds, and wholegrains. A spoonful of oats at breakfast, a serving of beans at lunch, and a serving of vegetables at dinner, spread through the day, is a pattern that many people can follow. Fibre supplements can help, but they are not a substitute for fibre from food, because food brings other nutrients with it, and because the variety of fibres from different foods feeds a richer microbiome. Fluid matters with fibre: water helps fibre do its work, and a person who eats more fibre without drinking more will feel worse, not better.
Portions, timing, and what to avoid
Portions change on a GLP-1 medicine, because appetite changes. The old portion, the plate that was right before, is now too large, and the person who tries to finish it will feel sick. The new portion is smaller, and it is built around protein and fibre first, with carbohydrate and fat fitted around them. A plate that is half vegetables, a quarter protein food, and a quarter starchy food, in a smaller plate than before, is a pattern that works for many people. Eating slowly matters, because the fullness signal takes time to arrive, and a person who eats quickly will eat too much before the signal comes. Putting the fork down between bites, chewing well, and stopping when full, rather than when the plate is empty, are habits that matter more on this medicine than before.
Timing matters too. Many people do better with three small meals and one or two small snacks, rather than one or two large meals, because a large meal is more likely to cause nausea. A small snack with protein, such as yoghurt or a boiled egg, can keep energy steady between meals. A meal eaten late at night, or a large meal eaten just before bed, can cause heartburn and poor sleep, because the stomach is still full when the person lies down. A light meal in the evening, eaten a few hours before bed, is often better.
Foods that are best limited are those that are high in fat, high in sugar, or hard to digest in a slow stomach. Fried foods, fatty meats, rich sauces, and large amounts of cheese are more likely to cause nausea. Sugary drinks and sweets can cause a fast rise in blood sugar and then a fall, and they give energy without protein or fibre, so they use up the small appetite without feeding the body. Spicy foods and alcohol can worsen heartburn and nausea, and alcohol also adds energy without nutrients. None of these foods is forbidden for everyone, but a person who notices that a food makes them feel unwell should eat less of it, or eat it in a smaller amount, and should tell their doctor or dietitian, so the plan can be adjusted.
The diagram shows the eating pattern. The steps point to the diagram: protein first and fibre slowly are the two steps that protect muscle and gut, and the last step, telling the doctor, is the one that keeps the medicine safe.
What to watch, and when to seek help
A person on a GLP-1 medicine should watch their weight, their strength, and how they feel, and should tell their doctor if the weight is falling too fast, if they feel weak, if they cannot eat enough to meet their needs, or if they have vomiting that does not settle. The doctor may check blood tests, may change the dose, or may refer the person to a dietitian. The dietitian can set the protein and fibre targets, can plan meals that fit the new appetite, and can help the person find foods they can eat and enjoy. The person should not stop the medicine on their own, and should not change the dose on their own, because the medicine is prescribed for a reason, and stopping it suddenly can bring back the appetite and the blood sugar it was controlling.
The long-term picture matters too. A GLP-1 medicine helps while it is taken, and the weight and blood sugar often rise again if it is stopped and the eating does not change. The food habits learned while taking it, smaller portions, protein at each meal, fibre, water, and activity, are the habits that keep the benefit after the medicine is stopped or the dose is lowered. The medicine is a tool that makes the habits easier to learn, because the appetite is lower and the cravings are weaker, but the habits are what last. A person who learns the habits while the medicine helps, and who keeps them after, has used the medicine well.
When appetite is very low: small meals, fluids, and what not to force
GLP-1 medicines lower appetite, and for some people the appetite falls so low that eating enough becomes the problem, rather than eating too much. A person who feels full after a few mouthfuls, who feels sick at the smell of food, or who skips meals because eating feels like a chore, is a person at risk of eating too little protein, too little fibre, and too little fluid, and of losing muscle as well as fat. The plan for a very low appetite is therefore different from the plan for a normal appetite: it is built on small, frequent meals, on foods that are easy to eat, and on fluids, and it is built with the prescriber, because a medicine dose that leaves a person unable to eat may need to be reviewed by the doctor who prescribed it.
Small, frequent meals are the base. Three small meals and two small snacks, each no bigger than a teacup, are easier to face than three large plates, and they give the person five chances a day to take in protein, rather than three. Each small meal should carry protein first, because protein is the nutrient the body needs most when food is scarce: a spoonful of yoghurt, an egg, a little cheese, a few spoonfuls of lentils, or a small piece of fish. Fluids should be taken between meals, not with them, because a large glass of water with a meal fills the stomach and leaves no room for food, while sipping water, soup, or a diluted drink between meals keeps the person hydrated without taking the place of food. Foods that are cold, or at room temperature, are often easier than hot foods, because hot food smells stronger, and the smell is what makes a nauseous person turn away. Foods that are dry, such as toast or crackers, can be easier than foods that are rich or fatty, which sit heavily in a slow stomach.
What not to force matters as much as what to eat. A person who feels sick should not be forced to eat a large meal, or to eat a food that makes them feel worse, because forcing food can make the nausea worse, and can make the person avoid eating altogether. A person who has vomited, who cannot keep fluids down, who feels dizzy, or who has not eaten for a day should seek medical help, and should tell their prescriber, because those are signs that the medicine, the dose, or the eating plan needs to be reviewed. The goal is not to eat a perfect diet on a day when appetite is very low. It is to eat a little protein, a little fibre, and enough fluid, and to keep doing that, day after day, until appetite returns, and then to rebuild the plan from there, with the care team.
Practical note: For one week, keep a short record of each meal: what you ate, how much, how full you felt after, and whether you felt sick. At the end of the week, look for the pattern: which meals left you comfortably full, which ones caused nausea, and which ones gave you protein and fibre. That record is the base for a conversation with your doctor or dietitian, and it is more useful than any general plan.
Quick self-check
Question: Why does protein matter more when a person is losing weight on a GLP-1 medicine?
Because weight loss takes fat and muscle, and protein protects muscle, but only if it is eaten in enough amount, spread through the day, and paired with activity. A person who eats very little protein while losing weight will lose more muscle, feel weaker, and burn less energy at rest, making it harder to keep the weight off.
Question: Why should fibre be raised slowly, with enough fluid, on a GLP-1 medicine?
Because the gut already moves more slowly on the medicine, and a sudden rise in fibre can cause wind, bloating, and discomfort. Raising fibre slowly, from foods such as oats, beans, and vegetables, with enough water, lets the gut adapt and keeps it moving without discomfort.
Try it yourself
Plan one day of small meals for a person on a GLP-1 medicine, using foods you have at home: three small meals and one snack, each with a protein food, a fibre food, and a small portion of starchy food, on a small plate. Write the plan down, with the protein and fibre in each meal. Then eat one of those meals yourself, slowly, stopping when full, and note how it feels compared with your usual portion. That feeling is the change the medicine brings.
Key ideas to carry forward
- GLP-1 medicines copy a gut hormone that signals fullness, slows the stomach, and helps insulin work, so appetite falls and blood sugar improves.
- The main effects on eating are a smaller appetite, a longer feeling of fullness, and weaker cravings, which help with weight loss and blood sugar.
- Protein protects muscle during weight loss, and it must be eaten in enough amount, spread through the day, and paired with activity.
- Fibre matters more when eating less, because it slows digestion, feeds gut microbes, and keeps the gut moving, and it must be raised slowly with enough fluid.
- Portions must be smaller, eaten slowly, with protein and fibre first, and with less fat at one meal, to avoid nausea.
- Nausea, vomiting, and constipation are common side effects, and they are eased by smaller meals, less fat, more fibre, and enough fluid.
- The habits learned while taking the medicine, smaller portions, protein, fibre, water, and activity, are what keep the benefit after the medicine is stopped or lowered.
Glossary
GLP-1
Glucagon-like peptide-1, a hormone made in the gut after eating that signals fullness, slows the stomach, and helps insulin work. Food tech example: GLP-1 medicines copy this hormone in a longer-lasting form.
Insulin
A hormone made in the pancreas that helps glucose move from the blood into cells. Food tech example: insulin is released after a meal to lower blood sugar.
Pancreas
An organ that makes insulin and digestive juices, sitting behind the stomach. Food tech example: the pancreas releases insulin after a meal.
Protein
A body part made of amino acids, used to build muscle, enzymes, and other body parts, and found in foods such as beans, eggs, and meat. Food tech example: protein in lentils helps build and repair muscle.
Fibre
The part of plant foods that is not digested in the small gut, which slows digestion and feeds gut microbes. Food tech example: fibre in oats slows the rise in blood sugar after breakfast.
Nausea
A feeling of sickness in the stomach, often before vomiting, common when a GLP-1 dose is raised or a meal is large and fatty. Food tech example: nausea after a large fried meal on a GLP-1 medicine is a sign to eat smaller and less fatty meals.
Short-chain fatty acid
A substance made by gut microbes from fibre, which feeds the gut wall and signals fullness. Food tech example: short-chain fatty acids made from oat fibre help the gut wall stay healthy.
Microbiome
The community of microbes living in the gut, which helps digest food and signals to the body. Food tech example: the gut microbiome changes with what a person eats, and it shapes how they respond to fibre.
Type 2 diabetes
A condition in which the body does not make enough insulin or does not respond to it well, so blood sugar stays high. Food tech example: a person with type 2 diabetes watches carbohydrate at each meal to keep blood sugar steady.
Blood sugar
Glucose in the blood, which rises after eating carbohydrate and falls as insulin helps it move into cells. Food tech example: blood sugar rises after a bowl of rice and falls over the next hours.
Muscle
Body tissue that moves the body and burns energy, which is protected by protein and activity during weight loss. Food tech example: muscle is kept by eating protein and by walking and resistance exercise.
Resistance exercise
Exercise that works muscles against a force, such as lifting, pushing, or climbing stairs, which helps keep muscle. Food tech example: climbing stairs is a form of resistance exercise that helps keep leg muscle.
Heartburn
A burning feeling in the chest when acid from the stomach rises, more likely when the stomach empties slowly. Food tech example: a large meal late at night can cause heartburn on a GLP-1 medicine.
Constipation
Difficulty passing stool, common when the gut moves slowly and fibre and fluid are low. Food tech example: constipation on a GLP-1 medicine is eased by more fibre and more water.
Dietitian
A qualified professional who gives food and nutrition advice for health, including for medical conditions. Food tech example: a dietitian sets protein and fibre targets for a person on a GLP-1 medicine.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- American Diabetes Association, Standards of Care in Diabetes, 2024.
- Wilding, J. P. H., Batterham, R. L. and Calanna, S., Once-weekly semaglutide in adults with overweight or obesity, 2021.
- Ordovas, J. M., Ferguson, L. R. and Tai, E. S., Personalised nutrition and health, 2018.
Unit 6: Designing a Personal Eating Plan
Note
Unit 6: Designing a Personal Eating Plan
A question to start with
A man wants to eat better. He has read ten articles, each with a different plan, and each plan claims to be the best. One says to cut carbohydrate. One says to cut fat. One says to eat six small meals. One says to fast for sixteen hours. He tries one for a week, feels hungry, and stops. He tries another, feels tired, and stops. He is not lazy. He is lost, because no plan was built for him: his work, his family, his budget, his health, and his taste were never in the plan. This unit is about building a plan that is built for one person, step by step, from what is known about them, and about keeping it simple enough to follow for months, not days.
A personal eating plan is a written plan for what a person will eat, when, and how much, built from their body, their health, their habits, and their life. It is not a diet from a book. It is a plan that starts with the person, sets a clear aim, chooses a small number of changes, and checks whether those changes work. A good plan is simple: it names the aim in one line, it names three or four changes, it names how the changes will be checked, and it names when the plan will be reviewed. A plan that tries to change everything at once will fail, because the person cannot follow it, and a plan that changes nothing will also fail, because it does not help. The skill is to find the middle: a few changes that matter most, in a form the person can keep.
Questions to think about
- A plan says to eat fish three times a week, but the person does not like fish and cannot afford it. Is the plan good or bad?
- Why is it better to change three things and keep them for three months than to change ten things for one week?
- How do you know if a personal eating plan is working, and when should it be changed?
Think about the fish plan before you read on
It is a bad plan for that person, however good fish is in general. A plan must fit the person's taste, budget, and life, or it will not be followed. The protein and the nutrients fish gives can come from other foods the person likes and can afford, such as beans, eggs, or chicken, and the plan should be built around those instead.
Step one: start with the person, not the diet
A plan starts with information, and the information is gathered before any food is chosen. The person's age, sex, height, weight, and activity give the energy needs. Their health gives the limits: diabetes, high blood pressure, kidney disease, coeliac disease, allergy, or medicines that change how food works. Their habits give the starting point: what they eat now, when they eat, how they cook, whether they eat out, and what they drink. Their life gives the shape: who they cook for, how much time they have, how much money they can spend, what foods they like and dislike, and what their culture and family eat. Each of these is written down, in a short record, so the plan can be built from it and so it can be checked later.
The record is made by asking, not by guessing. The person is asked what they eat on a normal day, what they eat on a weekend, what they eat when they are busy, and what they eat when they celebrate. They are asked what they like, what they dislike, what they cannot eat, and what they wish they ate more of. They are asked about their health, their medicines, their sleep, and their activity. They are asked about their aim: do they want to lose weight, to control blood sugar, to feel more energetic, to eat better for their family, or simply to feel less confused about food? The aim must be theirs, not the planner's, because a person will follow a plan that serves their aim, and will not follow a plan that serves someone else's.
The record also includes measurements, where they are useful: weight, waist, blood pressure, blood sugar, or a blood test, taken at the start, so the plan can be judged later. A plan with no starting measurement cannot be judged, because no one knows whether it worked. A plan with a measurement that is checked once and never again is also hard to judge. The measurements are taken at the start, and again after a set time, such as four weeks or three months, so the change can be seen.
Step two: set a clear aim
An aim must be clear, small, and measurable. "Eat better" is not an aim. "Eat vegetables at lunch and dinner, five days a week, for a month" is an aim. "Lose weight" is not an aim. "Lose 3 kg in three months, by eating smaller portions at dinner and walking for 30 minutes a day" is an aim. "Control blood sugar" is not an aim. "Keep blood sugar after lunch below the level my doctor set, on most days, by eating rice with beans and vegetables and walking after lunch" is an aim. Each aim names what will change, by how much, and by when, and each aim is set with the person, not for them, so they own it.
A plan should have one main aim, and at most two or three smaller aims. A plan with ten aims is a plan with no aim, because the person cannot hold ten aims in mind at once, and cannot tell which one matters most. The main aim is the one that matters most to the person, and the smaller aims support it. A person whose main aim is to lose weight may have a smaller aim to eat vegetables at two meals a day, and another to drink water instead of sugary drinks. A person whose main aim is to control blood sugar may have a smaller aim to walk after lunch, and another to eat protein at breakfast. Each aim is written down, with the date it will be checked, so the person and the planner both know what success looks like.
Step three: choose a few changes that matter most
From the record and the aim, the planner chooses three or four changes that matter most, and that the person can keep. The changes are chosen by asking: what in the current eating matters most for the aim, and what is easiest to change? A person who drinks three sugary drinks a day and wants to lose weight will gain more from cutting those drinks than from any other single change. A person who eats no vegetables and wants to feel better will gain more from adding vegetables than from changing their fat. A person who skips breakfast and then overeats at lunch will gain more from eating a protein breakfast than from counting every calorie. The biggest change is not always the hardest one, and the planner's skill is to find the change that gives the most benefit for the least effort.
The changes are written as actions, not as rules. "Eat vegetables at lunch" is an action. "Eat healthily" is a rule. "Drink water instead of a sugary drink at lunch" is an action. "Cut sugar" is a rule. Each action names what to do, when to do it, and what to do it with, so the person knows exactly what to do on Monday morning. Each action is small enough to do every day, and large enough to matter. A change that is too large will be dropped in a week. A change that is too small will not help. The test is whether the person can see themselves doing it for three months, not for three days.
The plan also names what will stay the same. A person who loves rice does not need to give up rice. They need a smaller portion of rice, with beans and vegetables, eaten after a walk. A person who loves bread does not need to give up bread. They need wholegrain bread, with protein, and a portion that fits their aim. A plan that takes away everything a person loves will fail, because the person will go back to what they love, and will feel they have failed. A plan that keeps what they love, in a smaller amount or in a better form, will last, because the person does not feel punished by it.
Step four: build the plate and the week
Once the changes are chosen, the plan is built into meals. A simple plate pattern helps: half the plate vegetables and salad, a quarter protein food, and a quarter starchy food, with water to drink, and with fruit for a snack or for dessert. That pattern works for many people, for lunch and dinner, and it can be adapted: the vegetables can be cooked or raw, the protein can be beans, eggs, fish, or chicken, and the starchy food can be rice, bread, potato, or pasta, in a portion that fits the aim. Breakfast follows a similar pattern: a protein food, a wholegrain or fruit, and a drink, in a portion that keeps the person full until lunch. Snacks are planned, not left to chance: a piece of fruit, a handful of nuts, or a yoghurt, rather than a biscuit from a machine.
The week is planned as a set of meals that repeat, with variety. A person does not need seven different dinners. They need three or four dinners they like, that they can cook, that fit the plan, and that they can repeat with small changes. A shopping list is made from the week's meals, so the person buys only what is needed, and the kitchen holds the foods the plan needs, not the foods the plan is trying to avoid. A person who has no vegetables at home will not eat vegetables, however good their aim is. A person who has sugary drinks at home will drink them, however good their plan is. The kitchen is part of the plan, and the shopping list is the tool that shapes it.
Portions are set by a simple method, not by weighing every gram. A hand is a useful guide: a palm for protein, a fist for starchy food, two handfuls for vegetables, and a thumb for fat, such as oil or butter. That guide is rough, but it is quick, it needs no scales, and it can be used anywhere, at home, at work, or in a restaurant. A person who needs exact portions, such as a person with diabetes who counts carbohydrate, will need to weigh food at first, until they learn the amounts by eye. Most people do not need that precision, and a plan that asks for it will be dropped.
Step five: check, review, and change
A plan is checked after a set time, usually four weeks, by looking at the measurements, the diary, and how the person feels. Did the weight change? Did the blood sugar change? Did the person feel more energetic? Did they follow the plan, and if not, which part was hard? The answers decide what happens next. If the plan worked and the person liked it, it is kept, and a new small change is added. If the plan did not work, the reason is found: was the aim too large, was the change too hard, was the portion wrong, or was life in the way? The plan is then changed, not the person blamed. A person who did not follow a plan has not failed. The plan has failed them, and the planner's job is to make a better one.
Review is built into the plan from the start, with a date written down, so it is not forgotten. The review is done with the person, not to them, and it asks what worked, what did not, and what to try next. A plan that is reviewed and changed is a plan that lives. A plan that is written once and never looked at again is a plan that dies, and the person who was given it is left with a piece of paper and a feeling of failure. The best plans are the ones that are reviewed, changed, and kept, over months and years, as the person's life, health, and taste change.
The diagram shows the plan in order. The steps point to the diagram: the review at the end is the step that keeps the plan alive, and the small number of changes in the middle is the step that keeps it possible to follow.
Practical note: Write your own one-page plan today: your aim in one line, three changes you will make, the plate pattern you will use, and the date four weeks from now when you will review it. Keep the page where you can see it, and tick each change every day you do it. At the review date, look at the ticks and at how you feel, and write the next version. That page is your personal eating plan, and it is worth more than any plan written for someone else.
Quick self-check
Question: Why is a clear, small aim better than a large, vague one?
Because a clear aim names what will change, by how much, and by when, so the person knows what success looks like and can check it. A vague aim such as "eat better" cannot be checked, so the person cannot tell whether the plan worked, and cannot tell what to change next.
Question: Why should a plan name what will stay the same, not only what will change?
Because a plan that takes away everything a person loves will fail, and the person will go back to what they love and feel they have failed. A plan that keeps what they love, in a smaller amount or a better form, will last, because the person does not feel punished by it.
Try it yourself
Write a one-page plan for yourself, using the steps in this unit: your aim in one line, three changes, your plate pattern, your shopping list for one week, and your review date four weeks from now. Then follow it for one week, keeping a short diary of what you ate and how you felt. At the end of the week, review it: which change was easiest, which was hardest, and what would you change in the plan for the next week? That review is the start of a plan that lives.
Key ideas to carry forward
- A personal eating plan starts with the person: body, health, habits, life, and aim, gathered by asking, not by guessing.
- An aim must be clear, small, and measurable, with one main aim and at most two or three smaller aims.
- Choose three or four changes that matter most and are easiest to keep, written as actions, not as rules.
- A plan must fit taste, budget, time, and culture, or it will not be followed, however good it is on paper.
- Build the plate and the week with a simple pattern, half vegetables, a quarter protein, a quarter starchy food, and a shopping list that fits it.
- Portions can be set by hand: a palm for protein, a fist for starchy food, two handfuls for vegetables, and a thumb for fat.
- Check the plan after four weeks, review what worked and what did not, change the plan rather than blaming the person, and set the next review date.
Glossary
Personal eating plan
A written plan for what a person will eat, when, and how much, built from their body, health, habits, and life. Food tech example: a personal eating plan sets three changes for a person who wants to lose weight.
Aim
A clear, small, measurable goal, such as eating vegetables at two meals a day for a month. Food tech example: an aim to lose 3 kg in three months is clear and measurable.
Plate pattern
A simple guide for a meal: half the plate vegetables, a quarter protein food, and a quarter starchy food. Food tech example: a plate with half vegetables, a quarter beans, and a quarter rice follows the pattern.
Food diary
A record of what a person eats and drinks, and when, used to see patterns and to plan changes. Food tech example: a three-day food diary shows which meals keep a person full longest.
Portion
The amount of a food that is served or eaten at one time. Food tech example: a portion of rice in a meal plan is the amount the person will eat.
Shopping list
A list of foods to buy to make a set of meals, grouped by shop section. Food tech example: a shopping list from a week's meals means the kitchen holds the foods the plan needs.
Review
A check of a plan after a set time, asking what worked, what did not, and what to change next. Food tech example: a four-week review of a plan shows the weight fell and the person liked the changes.
Wholegrain
Grain that keeps all its parts, bran, germ, and endosperm, and that is high in fibre. Food tech example: wholegrain bread is made from wholegrain flour.
Protein food
A food rich in protein, such as beans, eggs, fish, chicken, or yoghurt. Food tech example: beans are a protein food that fits a vegetarian plan.
Starchy food
A food rich in starch, such as rice, bread, potato, or pasta, which gives energy. Food tech example: rice is the starchy food on a plate with beans and vegetables.
Blood pressure
The force of blood against the walls of arteries, which is raised by too much salt and lowered by healthy eating and activity. Food tech example: blood pressure is checked at the start of a plan and again at review.
Energy
The power a food gives the body, measured in kilojoules or kilocalories, used for movement, warmth, and growth. Food tech example: a bowl of rice gives energy for the afternoon's work.
Fibre
The part of plant foods that is not digested in the small gut, which slows digestion and feeds gut microbes. Food tech example: fibre in oats slows the rise in blood sugar after breakfast.
Dietitian
A qualified professional who gives food and nutrition advice for health, including for medical conditions. Food tech example: a dietitian helps build a personal eating plan for a person with diabetes.
Waist
A measurement around the waist, used with weight to judge health risk and to track change. Food tech example: waist is measured at the start of a plan and again at review.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Ordovas, J. M., Ferguson, L. R. and Tai, E. S., Personalised nutrition and health, 2018.
- Rolls, B. J., The Ultimate Volumetrics Diet, 2012.
- American Diabetes Association, Standards of Care in Diabetes, 2024.
Unit 7: Evidence and Limits: What Is Proven
Note
Unit 7: Evidence and Limits: What Is Proven
A question to start with
A headline says "This diet cuts diabetes risk by half". A second headline says "This food doubles your risk of heart disease". A third says "Scientists find the perfect diet for your genes". Each headline is based on a real study, and each one is also, in some way, wrong: the risk was halved in a small group, over a short time, in people who were not like you; the food was linked to risk in a study that could not prove cause; the gene test was based on a link that is real but small. This unit is about reading evidence like a food scientist: what counts as proof, what counts as a clue, and what counts as a sales pitch dressed as science.
Evidence in nutrition comes in layers, and the layers are not equal. At the bottom are stories and opinions: a person says a food helped them, a writer says a diet works, a company says its product is proven. Above them are studies that observe people: researchers watch what people eat and what happens to them over years, and they find links. Above those are studies that test people: researchers give one group a diet and another group a different diet, and they measure what happens. At the top are reviews that put many studies together, and ask whether the pattern holds across them. A claim that is based on a story is not the same as a claim based on a test, and a claim based on one test is not the same as a claim based on many tests that agree. The reader who knows the layers can judge a headline in a minute.
Questions to think about
- A study finds that people who eat more vegetables live longer. Does that prove vegetables make people live longer?
- A diet is tested for eight weeks and works. Does that prove it works for a year?
- An app says its plan is "scientifically proven". What should you ask before believing it?
Think about the vegetables question before you read on
No, it does not prove it on its own. People who eat more vegetables also tend to exercise more, smoke less, and earn more, and any of those could explain the longer life. An observational study finds a link, and the link is a clue, but only a test that changes the diet and measures the result can show cause, and even then the test must be long, large, and well done.
The layers of evidence
The weakest layer is the story: one person, or a few people, who ate a food and felt better. Stories are useful as a start, because they suggest what to test, but they prove nothing, because the person may have changed other things at the same time, may have felt better for a reason that has nothing to do with the food, or may be remembering the past in a kinder light. A diet that works for one person may not work for another, and a story that is told well can feel like proof when it is only a story. The food scientist listens to stories, and then asks for a test.
The next layer is the observational study. Researchers follow a large group of people for years, record what they eat, and record what happens to their health. They find links: people who eat more fibre have less heart disease, people who drink more sugary drinks gain more weight, people who eat more processed meat have more bowel cancer. Those links are real, and they are the base of much nutrition advice, but they cannot prove cause, because the people who eat one way differ from the people who eat another way in many other things. Researchers try to allow for those differences, by a method called adjustment, but they cannot allow for everything, and a link that survives adjustment is still a link, not a proof. Observational studies are best for finding risks that are large, that grow with the amount eaten, and that appear in many different groups.
The next layer is the randomised trial. Researchers take a group of people, divide them by chance into two or more groups, give each group a different diet, and measure what happens. Because the groups were divided by chance, they should be similar in everything except the diet, so a difference in the result can be put down to the diet. That is the strongest single test, and it is the test that proves a diet works, or does not work, for the people in the trial, over the time of the trial. But trials have limits too: they are often short, often small, often in people who are not like the person reading the result, and often hard to follow, because people do not eat what they are told to eat for months on end. A trial that lasts eight weeks proves what happens in eight weeks, not what happens in a year.
The top layer is the systematic review and the meta-analysis, which put many studies together and ask whether the pattern holds. A review that finds the same result in many trials, in many places, in many kinds of people, is stronger than any single trial. A review that finds different results in different trials, or that finds the result only in small trials, is a sign that the evidence is weak or that the effect depends on something the trials did not measure. The best reviews also ask whether studies that found no effect were published, because studies that find nothing are less likely to be published, and a review that misses them will overstate the effect. That problem is called publication bias, and it is one reason a single glowing review should be read with care.
What is proven, in plain terms
Some things in nutrition are proven well enough to act on, and they are the base of the advice in this subject. Eating more vegetables and fruit, in varied kinds, is linked to better health in many studies, and the link is strong, consistent, and plausible, because vegetables and fruit bring fibre, vitamins, minerals, and plant substances with little energy. Eating more wholegrains instead of refined grains is linked to better health, for the same reasons. Eating less sugar, especially in drinks, is linked to less weight gain and better teeth, and the link is strong enough to act on. Eating less salt is linked to lower blood pressure, and the link has been shown in trials. Eating enough protein, spread through the day, with activity, helps keep muscle, especially in older people, and that has been shown in trials. Eating fibre, from varied plant foods, supports the gut and slows the rise in blood sugar, and that has been shown in trials. These are not new ideas, and they are not exciting, but they are the ideas with the strongest evidence behind them, and they are the ones a personal plan should be built on.
Some things are partly proven, and they need care. The idea that a person's genes can guide their diet is partly proven: some gene variants change how a person handles caffeine, lactose, or fat, and those links are real, but most food responses are shaped by many genes, and the tests that claim to guide a whole diet from genes go further than the evidence supports. The idea that the gut microbiome can guide diet is partly proven: fibre and varied plant foods support a richer microbiome, and a richer microbiome is linked to better health, but the claim that a single microbe test can say what one person should eat is not yet proven. The idea that a CGM can guide eating is partly proven: it shows a person's own glucose pattern, and that pattern can guide choices, but the claim that flattening every glucose rise makes a person healthier, even when their average is normal, is not yet proven. Each of these is a tool with a real use and a real limit, and the limit is where the sales pitch begins.
Some things are not proven, and they should be treated as such. A diet that claims to be perfect for one person's genes, based on a single test, is not proven. A supplement that claims to fix the microbiome, based on a small study, is not proven. A food that claims to burn fat, to cleanse the body, or to boost immunity, based on a story or a single study, is not proven. A plan that claims to work for everyone, based on a trial in a small group for a short time, is not proven for everyone. The reader who asks "what is the evidence, how strong is it, and who was studied?" will not be misled by those claims, and will spend their money on food, not on promises.
How to read a claim: five questions
A claim about food can be judged with five questions, and they take a minute to ask. First, what is the claim, exactly? "Cuts risk by half" means something different from "lowers blood sugar by a small amount", and a claim that is vague, such as "supports wellbeing", means almost nothing. Second, what is the evidence? Is it a story, an observational study, a trial, or a review, and how many people were studied, for how long, and in what way? Third, who was studied? A trial in young, healthy men does not prove a diet for older women, and a trial in people with diabetes does not prove a diet for people without it. Fourth, how large is the effect? A risk that falls from 2 in 100 to 1 in 100 is a halving, but it is also a change of 1 in 100, and the headline that says "halved" is truer than the one that says "changed by 1 in 100", but the second is more useful. Fifth, who gains if you believe it? A claim from a company that sells the food, the test, or the plan has a reason to make the claim sound stronger than it is, and the reader should look for evidence from a source that does not gain from the answer.
Those five questions are the habit this unit teaches, and they apply to every claim in this subject, including the claims in the earlier units. A claim about a gene, a microbe, a sensor, or a medicine should be read with the same five questions, and the answer should be kept in mind when the plan is built. A plan built on a strong claim will stand. A plan built on a weak claim will fall, and the person who built it will feel they have failed, when the claim failed them.
The limits of personalised nutrition today
Personalised nutrition is a young field, and its limits are honest limits, not failures. The science of genes and food is young, and most links are small. The science of the microbiome and food is young, and most links are not yet clear enough to guide a single person's diet. The science of glucose and food is stronger, because glucose can be measured, but the link between flattening every rise and long-term health is not yet proven. The science of AI and food is young, and most apps have not been tested in trials to show they help people eat better or stay healthier. Each of these will grow stronger with time, with larger trials, with longer follow-up, and with better measures. Until then, the honest position is to use what is proven, to use the new tools as guides rather than as deciders, and to keep checking whether the plan works for the person in front of you.
The future of personalised nutrition will be built on better trials: trials that are larger, longer, and in people like the person who will use the result; trials that measure what matters, such as weight, blood sugar, and how the person feels, not only a number on a sensor; and trials that test the whole plan, not only one part of it. When those trials exist, the advice will be stronger, and the plans will be better. Until then, the best plan is the one built on the proven base, personalised by the person's own pattern, checked by measurement, and reviewed with a professional where health is at stake. That plan is not exciting, and it does not make a good headline, but it is the plan that works, and it is the plan this subject has been building, unit by unit.
The diagram shows the five questions. The steps point to the diagram: the fourth question, how large the effect is in real numbers, is the one that headlines most often hide, and it is the one that matters most for deciding what to do.
How to read a personalised nutrition study: five checks before you believe it
A claim about personalised nutrition usually rests on a study, and a study must be read before it is believed. Five checks, done in order, will catch most of the weak claims. The first is: who was studied? A study of young, healthy people, eating one set of foods, in one country, may not apply to an older person, with diabetes, eating different foods, in another country. The reader should ask whether the people in the study are like the person the claim is being made for, and if they are not, the claim does not travel. The second is: how many people were studied, and for how long? A study of 20 people for two weeks can show a short-term change, but it cannot show a lasting effect, and it cannot show a rare harm. A study of 200 people for a year can say more, and a study of 2,000 people for two years can say more still. Small, short studies are a start, not a verdict.
The third check is: was there a comparison group? A study that gives everyone a personal diet, and finds they improve, has not shown that the personal diet caused the improvement, because the people may have improved from attention, from eating more carefully, or from time passing. A study that gives one group a personal diet and another group a general diet, chosen at random, and finds the personal group does better, has shown something real, because the only difference between the groups was the diet. That design is called a randomised controlled trial, and it is the strongest design for a diet claim, though it is also the hardest and most costly to run. The fourth check is: what was measured, and was it health, or only a number? A study that measures a change in blood sugar, in weight, or in symptoms has measured health. A study that measures only a change in microbes, or only a change in a blood marker, has measured a number, and the number may or may not mean better health. The reader should ask whether the outcome measured is the outcome they care about.
The fifth check is: who paid for the study, and who benefits from its result? A study paid for by a company that sells the test, the app, or the supplement being tested is not automatically wrong, but it must be read with more care, and its result must be checked against studies paid for by others. A study that is published in a journal where other scientists have reviewed it is stronger than a claim on a company's website with no study behind it, and a claim with no study at all is not a claim. It is an advertisement. A reader who runs these five checks, who was studied, how many and for how long, was there a comparison group, what was measured, and who paid, will be misled far less often than a reader who takes a headline at face value.
Practical note: Take a headline about food from this week, from a newspaper, a website, or a social feed, and judge it with the five questions in the diagram. Write down your answers in five lines. If you cannot answer one of the questions from the headline alone, find the study behind it, or treat the claim as not proven until you can. That habit, done once a week, will change how you read food news for good.
Quick self-check
Question: Why can an observational study find a link without proving cause?
Because people who eat one way differ from people who eat another way in many other things, such as exercise, smoking, and income, and any of those could explain the link. Researchers try to allow for those differences, but they cannot allow for everything, so a link is a clue, not a proof. Only a trial that changes the diet can show cause.
Question: Why is a risk that falls from 2 in 100 to 1 in 100 both a halving and a small change?
Because the relative change is a half, but the absolute change is 1 in 100, which is the number that matters for deciding what to do. A headline that says "halved" is true, but it hides how small the change is in real numbers, so the reader should always ask for the absolute numbers.
Try it yourself
Find two headlines about food from this week: one that excites you and one that worries you. Judge each one with the five questions, and write down your verdict for each: proven, partly proven, or not proven. Then write one line for each saying what you will do differently, if anything, because of it. Keep the two verdicts with your notes, and repeat the exercise next week with two new headlines. Over a month, you will have a picture of how often food news is proven, and how often it is not.
Key ideas to carry forward
- Evidence in nutrition comes in layers: stories, observational studies, randomised trials, and reviews, and the layers are not equal.
- An observational study finds a link, but it cannot prove cause, because people who eat one way differ in many other ways too.
- A randomised trial is the strongest single test, because groups divided by chance differ only in the diet, but trials are often short, small, and in people unlike the reader.
- A systematic review that puts many studies together is stronger than any single trial, if it includes studies that found no effect.
- Proven advice includes eating more vegetables, fruit, and wholegrains, eating less sugar in drinks and less salt, and eating enough protein with activity to keep muscle.
- Gene, microbiome, CGM, and AI claims are partly proven: each has a real use and a real limit, and the limit is where the sales pitch begins.
- Judge every claim with five questions: what is it, what is the evidence, who was studied, how large is the effect, and who gains if you believe it.
Glossary
Observational study
A study that watches what people eat and what happens to their health over time, finding links without changing the diet. Food tech example: an observational study finds people who eat more fibre have less heart disease.
Randomised trial
A study that divides people by chance into groups, gives each group a different diet, and measures what happens. Food tech example: a randomised trial gives one group a high-fibre diet and another a low-fibre diet and measures blood sugar.
Systematic review
A review that puts many studies together and asks whether the pattern holds across them. Food tech example: a systematic review of trials finds wholegrains lower blood sugar in many groups.
Meta-analysis
A method that combines the numbers from many studies into one overall result. Food tech example: a meta-analysis combines ten trials of fibre and blood sugar into one number.
Publication bias
The tendency for studies that find an effect to be published more than studies that find none, which can overstate an effect in a review. Food tech example: publication bias makes a supplement look better than it is.
Relative risk
A risk given as a proportion, such as halved or doubled, which can hide how large the change is in real numbers. Food tech example: a risk halved from 2 in 100 to 1 in 100 is a relative change of half.
Absolute risk
A risk given as a real number, such as 1 in 100, which shows how large a change is for a person. Food tech example: an absolute risk of 1 in 100 means one person in a hundred is affected.
Adjustment
A method in an observational study that tries to allow for differences between groups, such as age or smoking. Food tech example: adjustment for smoking lets researchers compare eaters more fairly.
Placebo
A treatment with no active effect, used in a trial to compare with the real treatment. Food tech example: a placebo pill is given to one group in a trial of a supplement.
Confounding
A mix-up in a study, where a third factor explains a link that looks like cause. Food tech example: exercise confounds the link between vegetables and long life, because vegetable eaters also exercise more.
Evidence
Information from studies that supports or weakens a claim about food and health. Food tech example: trials give stronger evidence for fibre than stories do.
Claim
A statement about what a food or diet does, which must be judged against evidence. Food tech example: a claim that a food boosts immunity must be backed by evidence.
Peer review
Checking of a study by other scientists before it is published, which catches errors but does not prove the result. Food tech example: a study that passes peer review has been checked, but it can still be wrong.
Follow-up
The time people in a study are followed, to see what happens to them over months or years. Food tech example: a trial with a one-year follow-up shows whether a diet works for a year.
Wholegrain
Grain that keeps all its parts, bran, germ, and endosperm, and that is high in fibre. Food tech example: wholegrain bread is made from wholegrain flour.
Randomised controlled trial
A study in which people are put into groups at random, one group getting the new diet and another getting a comparison diet, so the difference in outcome can be tied to the diet. Food tech example: a randomised trial of a personal diet against a general diet shows whether the personal diet truly does better.
Comparison group
A group in a study that gets a general or standard treatment, so the new treatment can be judged against it. Food tech example: a personal diet is judged against a comparison group that follows a general diet.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Ioannidis, J. P. A., The challenge of reforming nutritional epidemiologic research, 2018.
- Ordovas, J. M., Ferguson, L. R. and Tai, E. S., Personalised nutrition and health, 2018.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
Unit 8: Personalised Nutrition for Diabetes, Gut Disorders and Sport
Note
Unit 8: Personalised Nutrition for Diabetes, Gut Disorders and Sport
A question to start with
Three people sit down to the same meal: a bowl of rice, lentils, and vegetables. The first has type 2 diabetes, and their blood sugar rises high after the rice and stays high for hours. The second has irritable bowel syndrome, and the lentils leave them bloated and in pain. The third is a runner training for a long race, and the same meal is too small, too low in energy, and too low in salt for the training they did that morning. One meal, three bodies, three different needs. Personalised nutrition is often talked about in general terms, about genes and microbes and wearables, but its real test is here, in specific conditions, where the general advice must be turned into a plan for one person with one condition and one life. This unit takes three conditions, diabetes, gut disorders, and sport, and shows how personalisation works in each, what is proven, what is not, and where a professional must be involved.
This unit is not medical advice for any one reader. It explains the ideas a food technologist and a nutrition professional use when they design a plan for a person with a condition, and it marks clearly where the plan must be made with a doctor or a registered dietitian. A reader with diabetes, a gut disorder, or a training goal should use this unit to ask better questions of their own care team, not to replace that team. The numbers in the worked examples are round teaching numbers, clearly marked as examples, so the arithmetic can be followed without mistaking them for a prescription for any one person.
Questions to think about
- Two people with type 2 diabetes eat the same bowl of rice. One has a high blood sugar rise, the other a modest one. What could explain the difference?
- A food that is healthy for most people, such as beans, causes pain in a person with a gut disorder. Is the food unhealthy, or is the match between food and person wrong?
- A runner eats the same meal as a person who sits at a desk all day. Why might the runner need more salt, more carbohydrate, and more water, and why would the same meal be too much for the desk worker?
Think about the beans question before you read on
The match is wrong, not the food. Beans are rich in fibre and in fermentable carbohydrates, which gut microbes turn into gas. In most people that gas is a small price for the fibre, the protein, and the nutrients. In a person whose gut is sensitive, the same gas causes pain and bloating, because their gut feels the stretch more, or their microbes make more gas, or their gut moves the gas more slowly. The food has not changed. The person has, and the plan must change with them.
Diabetes: blood sugar, carbohydrate, and the personal response
Type 2 diabetes is a condition in which the body does not use insulin well, or does not make enough of it, so blood sugar rises higher than it should after eating, and stays high for longer. Insulin is the hormone that moves sugar from the blood into cells, and when it does not work well, the sugar stays in the blood. Over time, high blood sugar damages blood vessels, nerves, eyes, and kidneys, which is why controlling the rise after meals matters. The rise after a meal is called the postprandial rise, and it is the part of diabetes that food can change most directly, because the food is what puts the sugar into the blood in the first place.
Carbohydrate is the nutrient that raises blood sugar most, because carbohydrate is broken down into sugar in the gut. The amount of carbohydrate matters, and the type matters too: a carbohydrate that is broken down quickly, such as white rice or a sugary drink, raises blood sugar fast and high, while a carbohydrate that is broken down slowly, such as whole grains with fibre, raises it more slowly and less high. But the amount and the type are not the whole story. The same carbohydrate gives a different rise in different people, and in the same person on different days, because the rise depends on insulin sensitivity, on the microbes in the gut, on sleep, on stress, on exercise, and on what else was eaten with the carbohydrate. A meal of rice eaten alone gives a higher rise than the same rice eaten with vegetables, protein, and fat, because the vegetables, protein, and fat slow the emptying of the stomach and the breakdown of the carbohydrate.
Personalised nutrition for diabetes therefore starts with measurement, not with a diet sheet. A continuous glucose monitor, a CGM, worn for a week or two, shows the rise after each meal, in real time, and the person can see which meals give a high rise and which give a modest one. A food diary kept beside the CGM shows what was eaten, and the two together show the pattern: this breakfast gives a high rise, that breakfast does not, and the difference is the food, not the person. The plan that follows is built on those patterns: keep the meals that work, change the meals that do not, and change them in small ways first, such as eating vegetables before the rice, or walking for ten minutes after the meal, or swapping half the rice for lentils, and then measuring again to see whether the change worked. That loop, measure, change, measure again, is the heart of personalisation for diabetes, and it is more useful than any single diet, because it is built on the person's own data.
A worked example shows the arithmetic of carbohydrate counting, with round teaching numbers only. Suppose a person eats a meal with 60 g of carbohydrate, and their CGM shows a rise of 3 units of blood sugar after that meal. Suppose they eat the same meal another day, but with half the rice replaced by vegetables, so the carbohydrate falls to 30 g, and the rise falls to 1.5 units. The halving of carbohydrate has halved the rise, in this example, because the rise is roughly in proportion to the carbohydrate eaten, when the rest of the meal stays the same. That proportion is the reason carbohydrate counting works as a tool: it lets the person predict the rise from the amount eaten, and adjust the amount to keep the rise in the range their care team has set. The range itself, and the target for any one person, must be set by their doctor, not by this unit, and the numbers here are teaching numbers, not targets.
The limits must be stated plainly. A CGM shows the rise, but it does not show why the rise happened, and it does not tell the person what to eat. A diet that works for one person with diabetes may not work for another, and a diet that is low in carbohydrate may be high in fat, which brings its own risks. Medication matters: a person who takes insulin or a medicine that lowers blood sugar must not change their food without talking to their doctor, because less carbohydrate can mean less medicine is needed, and the dose must be changed by the prescriber. And personalised nutrition does not cure diabetes. It helps manage it, alongside medication, activity, sleep, and weight, and the plan must be reviewed as the person's condition, medication, and life change.
Gut disorders: when the gut sets the rules
Gut disorders cover a group of conditions in which the gut is sensitive, inflamed, or does not move food in the usual way. Irritable bowel syndrome, IBS, is the most common, and it causes pain, bloating, and changes in bowel habit, without damage that can be seen on a test. Inflammatory bowel disease, which includes ulcerative colitis and Crohn's disease, causes inflammation and damage in the gut, and needs medical care. Coeliac disease, covered in earlier units, is an immune response to gluten that damages the gut. Each of these is different, and each needs its own care, but they share one feature: the gut sets the rules for what can be eaten, and the rules are personal, because the same food can be fine for one person with IBS and painful for another.
In IBS, the foods most often involved are the fermentable carbohydrates, known as FODMAPs, a name that stands for fermentable oligosaccharides, disaccharides, monosaccharides, and polyols. These are short chains of sugar found in wheat, onion, garlic, beans, lentils, apples, pears, and milk, among others. They are not digested in the small intestine, so they pass to the large intestine, where gut microbes ferment them and make gas. In most people that gas is a small matter. In a person with IBS, the gut feels the gas more, because the gut is more sensitive to stretch, and the pain and bloating follow. The personalised approach in IBS is therefore not to cut out all FODMAPs for life, but to test them, one group at a time, under the care of a dietitian: cut them out for a short period, then reintroduce them one group at a time, and keep the ones that are tolerated, and limit only the ones that cause symptoms. That reintroduction is the personalisation, because the list of foods that cause symptoms is different for each person.
The low FODMAP approach has a cost that must be stated. Cutting out wheat, onion, garlic, beans, and many fruits for weeks narrows the diet, and a narrow diet can be low in fibre, low in variety, and hard to keep, and it can change the gut microbes in ways that are not wanted. That is why the approach is done for a short time, with a dietitian, and with reintroduction built in from the start, and why a person should not cut out whole groups of foods on their own, without a plan to bring them back. The dietitian also checks the rest of the diet: is the person eating enough, is the diet varied, is the person drinking enough water, and is stress, sleep, or eating too fast making the symptoms worse, because the gut responds to all of these, not only to food.
In inflammatory bowel disease, the plan is different, because the gut is inflamed and damaged, and the food must be gentle on a gut that is healing. During a flare, when symptoms are strong, the plan may be low in fibre, low in fat, and low in foods that are hard to digest, with small, frequent meals, and with attention to fluids and to nutrients that are lost, such as iron and vitamin B12. Between flares, the plan broadens, and the person can eat a wider diet, with fibre reintroduced slowly, and with attention to the foods that they know cause them trouble. In both phases, the plan is made with the medical team, because the medicine, the surgery, and the nutrition must work together, and because a diet that is too narrow can cause malnutrition in a person whose gut already absorbs poorly.
Sport: energy, carbohydrate, protein, and the training day
Sport is the condition where personalised nutrition is most visible, because the need changes from day to day, and even from hour to hour. A runner who trains for two hours in the morning needs more energy, more carbohydrate, and more water than the same runner on a rest day, and a runner who trains in the heat needs more salt than one who trains in the cold. A person who lifts weights needs more protein, spread across the day, to build and repair muscle. A person who plays a team sport needs carbohydrate before the game, water during it, and carbohydrate and protein after it. The plan is therefore not one diet, but a set of plans for training days, rest days, and competition days, and the person learns to match the food to the day.
Carbohydrate is the fuel for hard exercise, because it is stored in the muscle and the liver as glycogen, a chain of glucose molecules, and it is burned quickly when the muscle works hard. When glycogen runs low, the person feels tired, heavy, and slow, a state often called hitting the wall. The personalised question is how much carbohydrate the person needs, and when: a large meal with carbohydrate three to four hours before a long session, a small snack with carbohydrate one hour before, a drink with carbohydrate during a session longer than an hour, and a meal with carbohydrate and protein within a couple of hours after, to refill glycogen and repair muscle. The amounts depend on the person's weight, the length and hardness of the session, and the time until the next session, and they are worked out with a sports dietitian, not guessed from a general table.
Protein is the nutrient that builds and repairs muscle, and the personalised question is how much, and when. A person who trains hard needs more protein than a person who does not, and the protein should be spread across the day, in meals and snacks, rather than eaten in one large meal, because the muscle can only use a certain amount at one time for building, and the rest is burned for energy. A worked example, with round teaching numbers only, shows the arithmetic. Suppose a person weighs 70 kg, and their sports dietitian sets a protein target of 1.6 g of protein per kilogram of body weight per day. The daily target is 70 times 1.6, which is 112 g of protein per day. Spread across four meals, that is 28 g per meal. The person can then check each meal against that number: does this breakfast give 28 g, and if not, what could be added to reach it? That check, done with real foods and real labels, is the personalisation: the number is general, and the foods that meet it are personal, chosen from what the person likes, can afford, and can digest before training.
Water and salt are the parts of sports nutrition most often forgotten, and the most dangerous to forget. Sweat is water and salt, and a person who sweats for two hours loses both, and must replace both. Water alone, drunk in large amounts without salt, can dilute the blood, a condition that is rare but severe, and the reason sports drinks contain salt as well as carbohydrate. The personalised question is how much the person sweats, which is measured by weighing before and after a session, and how salty their sweat is, which varies from person to person. A person who loses a lot of sweat, or whose sweat is very salty, needs more salt than a person who sweats little, and the plan must be tested in training, never for the first time on the day of a competition.
The diagram shows the personalisation loop for a condition. The loop is the same for diabetes, gut disorders, and sport, and the measurements inside it are different for each, which is why the plan must be made with a professional who knows the condition.
Where a professional must be involved
Three lines mark where a food technologist or a general reader must hand over to a professional. The first is diagnosis: no one should diagnose diabetes, IBS, or any gut disorder from a wearable, an app, or a food diary. Diagnosis is done by a doctor, with tests, and the plan starts after the diagnosis, not before it. The second is medication: a person who takes insulin, a diabetes medicine, or a medicine for gut inflammation must not change their food in a way that changes their need for medicine without talking to the prescriber, because the dose must be changed by the prescriber. The third is a narrow diet: a person who cuts out whole groups of foods, for weeks or for life, needs a dietitian to check that the diet is enough, that it is varied, and that it does not cause a deficiency. A personalised plan that is made without those three lines is not personalised. It is a guess, and a guess with a condition behind it can cause harm.
Practical note: If you know a person with diabetes, a gut disorder, or a training goal, ask them one question: what did you measure, and what did it change? The answer will tell you more about personalised nutrition than any diet sheet, because it will show the loop, measure, change, measure again, that this unit is built on. Do not offer them a diet. Offer them the question, and let their care team build the plan.
Quick self-check
Question: Why does the same bowl of rice give a different blood sugar rise in two people with type 2 diabetes?
Because the rise depends on insulin sensitivity, gut microbes, sleep, stress, exercise, and what else was eaten with the rice, and those differ from person to person, and from day to day in the same person. The rise is personal, so the plan must be built on the person's own measurements.
Question: A person weighs 70 kg and their protein target is 1.6 g per kilogram per day. What is their daily protein target?
70 times 1.6 gives 112 g of protein per day. Spread across four meals, that is 28 g per meal, which is a teaching example, not a prescription for any one person.
Try it yourself
Choose one of the three conditions in this unit, and write a one-page plan for a made-up person, with the condition, the goal, the measurement you would use, and the first small change you would test. For diabetes, the measurement is a CGM and a food diary. For a gut disorder, the measurement is a symptom diary and a reintroduction list. For sport, the measurement is the training load and the weight before and after a session. Then write the second change you would test if the first one worked, and the change you would test if the first one did not work. That plan, with its measurements and its next steps, is personalised nutrition in miniature, and it is more useful than a diet sheet, because it is built to be tested.
Key ideas to carry forward
- One meal gives different results in different people, and in the same person on different days, because insulin sensitivity, gut microbes, sleep, stress, exercise, and the rest of the meal all shape the response.
- In type 2 diabetes, carbohydrate raises blood sugar, and the personalised work is to measure the rise after each meal, change one thing at a time, and measure again.
- Carbohydrate counting lets a person predict the rise from the amount eaten, and the target range must be set by their doctor, not by a general table.
- In gut disorders, the foods most often involved are the fermentable carbohydrates, FODMAPs, and the personalised work is a short test with a dietitian, followed by reintroduction one group at a time.
- A narrow diet, kept for weeks without a plan to bring foods back, can be low in fibre and variety, and must be done with a dietitian.
- In sport, the plan changes by the day: carbohydrate is the fuel for hard exercise, protein is spread across the day to build and repair muscle, and water and salt must both be replaced.
- Diagnosis, medication changes, and narrow diets are the three lines where a professional must be involved, and a plan made without them is a guess.
Glossary
Type 2 diabetes
A condition in which the body does not use insulin well, or does not make enough of it, so blood sugar rises higher than it should after eating. Food tech example: a person with type 2 diabetes watches the blood sugar rise after a meal of rice.
Insulin
The hormone that moves sugar from the blood into cells. Food tech example: insulin moves sugar from the blood into muscle cells after a meal.
Postprandial rise
The rise in blood sugar after a meal. Food tech example: the postprandial rise after rice is higher when the rice is eaten alone than when it is eaten with vegetables and protein.
Carbohydrate counting
Counting the grams of carbohydrate in a meal to predict the blood sugar rise. Food tech example: a person counts 60 g of carbohydrate in a meal and predicts the rise from that amount.
Glycogen
A chain of glucose molecules stored in muscle and liver, used as fuel for hard exercise. Food tech example: glycogen in muscle is burned when a runner runs hard, and refilled by carbohydrate after the run.
Irritable bowel syndrome (IBS)
A common gut disorder causing pain, bloating, and changes in bowel habit, without damage that can be seen on a test. Food tech example: a person with IBS feels pain and bloating after eating beans, because their gut feels the gas more.
Inflammatory bowel disease
A group of conditions, including ulcerative colitis and Crohn's disease, in which the gut is inflamed and damaged. Food tech example: a person with inflammatory bowel disease needs medical care and a diet that is gentle on a healing gut.
FODMAPs
Fermentable carbohydrates found in wheat, onion, garlic, beans, and some fruits, which gut microbes turn into gas. Food tech example: FODMAPs in beans are fermented by gut microbes, making gas that causes bloating in a sensitive gut.
Reintroduction
Bringing foods back into the diet one group at a time, after a short period of cutting them out, to find which ones cause symptoms. Food tech example: reintroduction after a low FODMAP test shows that onion causes symptoms but lentils do not.
Flare
A period when the symptoms of a gut disorder are strong. Food tech example: during a flare, a person with inflammatory bowel disease eats small, frequent meals that are low in fibre and fat.
Protein target
The amount of protein a person aims to eat per day, often set in grams per kilogram of body weight. Food tech example: a protein target of 1.6 g per kilogram per day for a 70 kg person gives 112 g per day.
Sweat loss
The water and salt lost in sweat during exercise, measured by weighing before and after a session. Food tech example: sweat loss of 1 kg in a two-hour run means the runner lost about 1 litre of water and salt.
Sports dietitian
A dietitian who specialises in food for sport, who sets carbohydrate, protein, and fluid plans for training and competition. Food tech example: a sports dietitian sets a runner's carbohydrate plan for a long race.
Symptom diary
A record of what was eaten and what symptoms followed, used to find which foods cause trouble. Food tech example: a symptom diary shows that garlic causes bloating, while rice does not.
Registered dietitian
A qualified professional who can diagnose nutrition problems and make a diet plan for a person with a condition. Food tech example: a registered dietitian makes a low FODMAP plan with reintroduction built in.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- American Diabetes Association, Standards of Care in Diabetes, 2024.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.
Unit 9: The Microbiome and Personalised Nutrition in Depth
Note
Unit 9: The Microbiome and Personalised Nutrition in Depth
A question to start with
Two people eat the same bowl of beans. One feels fine, and their blood sugar rises a little. The other feels bloated, and their blood sugar rises a lot. The difference is not in the beans. It is in the gut, in the trillions of microbes that live there, the community called the microbiome. Those microbes break down foods the human gut cannot, make vitamins and small molecules that talk to the body, and shape how the body handles sugar, fat, and even appetite. This unit goes deeper into that community than Unit 2 did: what it is, how it is measured, how food shapes it, how it shapes the response to food, and what is proven, what is promising, and what is still guesswork.
The microbiome is the most talked-about part of personalised nutrition, and the most oversold. A test that names the microbes in a stool sample can tell a story that sounds precise, and a diet built on that story can sound scientific, but the science is younger than the marketing, and the gap between the two is where a careful reader must stand. This unit explains the science as it stands, with its limits stated plainly, so the reader can tell a real finding from a sales claim, and can use the microbiome wisely without expecting it to do more than it can.
Questions to think about
- A test says your gut is low in a microbe linked to good health. Does that mean you should eat a food that feeds that microbe, and will the microbe grow if you do?
- Two people eat the same fibre, and one makes more gas than the other. What could explain the difference, and is the gas a sign of harm or of the fibre working?
- A company sells a probiotic and says it will fix your microbiome. What would you need to see before you believed that claim?
Think about the gas question before you read on
Gas is made when gut microbes ferment fibre that the human gut cannot digest. Different people carry different microbes, so the same fibre is fermented by different microbes, at different speeds, making different amounts of gas. Gas is a sign that the fibre is being fermented, which is also how the microbes make short-chain fatty acids, the small molecules that feed the gut lining. Gas is therefore not a sign of harm on its own. It is a sign that the fibre is working, and whether it causes pain depends on how sensitive the gut is to the gas, not only on how much gas is made.
What the microbiome is, in plain words
The microbiome is the community of microbes that live in the gut, mostly in the large intestine. It includes bacteria, which are the most studied, and also other microbes such as archaea, fungi, and viruses, though the bacteria are the ones most tests measure. The number of microbes is very large, in the trillions, and the number of kinds, called species, is in the hundreds in any one person, and in the thousands across all people. Each person carries a different mix, shaped by birth, by early feeding, by diet, by medicines, especially antibiotics, by illness, by age, and by where they live. That mix is personal, and it is also changeable, because diet can shift it in days, and antibiotics can shift it in hours, though the shift after antibiotics is often followed by a slow return towards the earlier mix.
The microbes do work the human body cannot do alone. They break down fibre and other carbohydrates that human enzymes cannot break down, turning them into short-chain fatty acids, small molecules that feed the cells of the gut lining, that signal to the immune system, and that affect how the body handles sugar and fat. They make some vitamins, such as vitamin K and some B vitamins. They train the immune system, teaching it which microbes to tolerate and which to attack. They compete with harmful microbes for space and food, making it harder for those microbes to grow. And they make small molecules that travel in the blood and talk to the brain, the liver, and the fat tissue, which is why the gut is sometimes called a second brain, though that phrase is a metaphor, not a fact, and should not be taken to mean the gut thinks.
Two words are used constantly in microbiome talk, and they are often mixed up. Diversity is the number of different kinds of microbes in a gut, and how evenly they are spread. A diverse gut, with many kinds present in fair numbers, is generally linked to better health, though diversity on its own is not a diagnosis. Composition is which kinds are present, and in what amounts. Two people can have the same diversity and a different composition, and two people can have the same composition on one day and a different one after a week of different food. Both words matter, and a test that reports only one of them tells only half the story.
How the microbiome is measured, and what a test can and cannot say
The most common test is a stool test, in which a sample of stool is sent to a laboratory, the DNA of the microbes in it is read, and the kinds present are named, with their relative amounts. The reading is done by sequencing, a method that reads the DNA letters of a short piece of DNA that is shared by bacteria, or by reading all the DNA in the sample, which is called shotgun sequencing, and which can name more kinds and can also say what the microbes are able to do, such as which fibres they can break down. The result is a list of kinds, with percentages, and often a score, a diversity number, and a set of diet suggestions.
A test can say which kinds are present in that stool sample, on that day, and in what relative amounts. It cannot say, on its own, what those microbes are doing, whether they are alive or dead, whether they are in the small intestine or the large intestine, or whether the person is healthy or ill. A stool sample is a sample of the end of the gut, not of the whole gut, and the microbes in the small intestine, where much digestion happens, are different from those in the stool. A test taken after a week of travel, illness, or antibiotics will differ from a test taken at home, and a test taken after a high-fibre week will differ from one taken after a low-fibre week. The test is a snapshot, not a portrait, and a diet built on one snapshot is built on a single day.
The suggestions that come with a test must be read with care. A suggestion to eat more fibre, to eat a wider range of plants, or to eat fermented foods is usually sound advice for most people, and it does not need a test to be true. A suggestion to eat one specific food because it feeds one specific microbe, or to avoid one specific food because it starves one microbe, is a much stronger claim, and it needs stronger evidence, because the link between one food, one microbe, and one health outcome is rarely proven. A test that sells a supplement alongside its suggestions has a conflict of interest, and the reader should ask whether the suggestion serves the person or the sale. A good test states its limits, names its method, and gives suggestions that are general, food-based, and safe. A test that promises a personal diet that will fix the gut, the weight, or the blood sugar, on the basis of one stool sample, promises more than the science can deliver.
How food shapes the microbiome, and how the microbiome shapes the response to food
Food shapes the microbiome every day, because the microbes eat what the person eats. Fibre is the main food for the gut microbes, and a diet rich in many kinds of fibre, from vegetables, fruits, legumes, whole grains, nuts, and seeds, feeds many kinds of microbes, and supports diversity. A diet low in fibre starves them, and diversity falls. A diet rich in one kind of fibre feeds the microbes that can break down that fibre, and they grow, while others shrink. That is why a change in diet can shift the microbiome in days: the microbes that can eat the new food grow quickly, and the ones that cannot shrink. The shift is fast, and it is also reversible: when the old diet returns, the old mix returns, which is why a short diet change gives a short microbiome change, and a lasting change needs a lasting diet change.
The microbiome also shapes the response to food, and this is the part that matters most for personalised nutrition. The same food can give a different blood sugar rise in two people, and part of that difference is the microbes: the microbes break down the food differently, make different small molecules from it, and signal differently to the body. A study that fed the same foods to many people, and measured their blood sugar, found that the rise could be predicted, in part, from the person's microbiome, along with their genes, their weight, and their habits. That finding is real, and it is also limited: the prediction was better than guessing, but it was not exact, and it was made in one group of people, eating one set of foods, and it may not hold in another group, eating other foods. The microbiome is one piece of the prediction, not the whole answer.
Fermented foods, such as yoghurt, kefir, and fermented vegetables, add live microbes to the gut, and they also add the small molecules those microbes made during fermentation. Eating them can raise the diversity of the gut, in some studies, and can change the microbes present, though the microbes from food do not usually settle in the gut for long. They pass through, do their work, and leave, which is why a fermented food must be eaten often to keep its effect. Probiotics, which are live microbes sold as a supplement or added to a food, work in the same way: they pass through, they may do a job while they are there, and they do not usually colonise the gut for life. That is why a probiotic is judged by the strain, the dose, and the condition it is used for, not by the word probiotic alone, and why a probiotic that helps one condition may do nothing for another.
Short-chain fatty acids: the small molecules that do the talking
Short-chain fatty acids, called SCFAs, are the small molecules that gut microbes make when they ferment fibre, and they are the clearest link between fibre, microbes, and health. The three main ones are acetate, propionate, and butyrate. Butyrate is the main fuel for the cells of the gut lining, and it helps keep the gut wall strong, so that microbes and their products do not leak into the blood. Propionate travels to the liver, where it affects how the liver handles sugar and fat. Acetate travels more widely, and affects appetite and fat storage. Together, they signal to the immune system, calming it, and to the gut itself, slowing or speeding its movement, and to the brain, through nerves and through the blood, affecting appetite and mood. That chain, from fibre to microbes to SCFAs to the body, is the best understood part of microbiome science, and it is the reason fibre is the most solid advice in this unit.
The chain also shows why a single fibre is not enough. Different fibres are fermented by different microbes, making different SCFAs, in different amounts. A diet with many kinds of fibre, from many plants, feeds many microbes, and makes a wider set of SCFAs, than a diet with one fibre, however large the dose. That is why the advice is to eat a wide range of plants, not to take a single fibre supplement, and why a diet that is rich in plants, with variety, is the most reliable way to support the microbiome. The advice is simple, and it does not need a test to be true, which is both its strength and its frustration: the most useful microbiome advice is the advice that was already known, to eat many plants, and the test adds little to it.
A worked example, with round teaching numbers only, shows how a maker thinks about fibre variety. Suppose a person eats 30 g of fibre per day, all from one source, such as wheat bran. Suppose another person eats the same 30 g, but from six sources: vegetables, fruit, lentils, whole grains, nuts, and seeds, with 5 g from each. The total is the same, 30 g, but the second person feeds a wider set of microbes, because each fibre is broken down by different microbes, and the second person makes a wider set of SCFAs. The lesson is that the variety of fibre matters as well as the amount, and that a label that says only "high fibre" does not say whether the fibre is varied. A food technologist who designs a food for gut health should therefore use more than one fibre, from more than one source, and should name them, so the buyer can see the variety.
The diagram shows the chain from fibre to the body. Each step leads to the next, and the chain is only as strong as its first step: a diet with one fibre feeds only the microbes that can eat that fibre, and makes only a narrow set of short-chain fatty acids.
What is proven, what is promising, and what is still guesswork
Proven, in the sense of many studies agreeing, is this: a diet rich in many kinds of fibre supports a diverse microbiome, and a diverse microbiome is linked to better health. Fibre is fermented into SCFAs, and SCFAs feed the gut lining and calm the immune system. Antibiotics can disturb the microbiome, and the disturbance is followed by a slow return. Fermented foods can raise diversity in some people. Those findings are solid, and they support the advice to eat many plants, to eat fermented foods, and to use antibiotics only when a doctor prescribes them.
Promising, in the sense of early studies that need more work, is this: the microbiome can help predict the blood sugar rise after a meal, and a diet built on that prediction can lower the rise in some people. Certain microbes are linked to a better response to certain diets, and to a lower risk of certain conditions, though the link is not yet a cause. Probiotics can help in specific conditions, such as certain kinds of diarrhoea, when the right strain is used at the right dose, though the same probiotic may do nothing for another condition. Those findings are worth watching, and they are the reason the field is growing, but they are not yet a basis for a personal diet for every reader.
Still guesswork is this: that a stool test can name the one diet that will fix a person's weight, blood sugar, or mood, and that a supplement can fix a microbiome that a diet cannot. The test can name the microbes, but it cannot yet say what diet will change them in the way that matters, or whether changing them will change the health outcome. The supplement can add microbes, but those microbes pass through, and they do not usually settle, and the claim that they will fix the gut is a claim about a single product, not about the diet, the sleep, the stress, and the medicines that also shape the gut. A reader who is offered a test or a supplement with a promise that sounds too neat should ask for the studies behind it, and should ask whether the studies were done in people like them, eating foods like theirs, and whether the outcome measured was health, or only a change in the microbes.
Practical note: The most useful microbiome test you can do at home costs nothing: count the different plants you eat in a week, including vegetables, fruits, legumes, whole grains, nuts, seeds, herbs, and spices. If the number is below ten, add one new plant a week, and keep a note of how you feel. That count, kept over weeks, is a better guide to your microbiome than a single stool test, because it measures the food that feeds the microbes, and it is the food, not the test, that changes them.
Quick self-check
Question: Why is a stool test a snapshot, not a portrait, of the microbiome?
Because it measures the microbes in one stool sample, on one day, from the end of the gut, and the mix changes with diet, travel, illness, and antibiotics. A test taken after a high-fibre week differs from one taken after a low-fibre week, so one test cannot stand for the whole person.
Question: Why does eating many kinds of fibre matter more than eating a large amount of one fibre?
Because different fibres are fermented by different microbes, making different short-chain fatty acids. Many kinds of fibre feed many microbes and make a wider set of SCFAs, while one fibre feeds only the microbes that can eat it.
Try it yourself
Count the different plants you eat in one week, including vegetables, fruits, legumes, whole grains, nuts, seeds, herbs, and spices, and write the number down. Then, for the next week, add three new plants you do not usually eat, and count again. Write one line saying whether you felt any different, and one line saying whether the change was easy or hard to keep. That count, and that note, are the start of a personalised microbiome plan, because they measure the food that feeds your microbes, and they show whether a change is one you can live with.
Key ideas to carry forward
- The microbiome is the community of microbes in the gut, mostly bacteria in the large intestine, and each person carries a different mix, shaped by birth, diet, medicines, illness, and age.
- Diversity is the number of kinds present, composition is which kinds they are, and both matter, though neither on its own is a diagnosis.
- A stool test names the microbes in one sample, on one day, from the end of the gut, and it is a snapshot, not a portrait, and it cannot say what the microbes are doing.
- Fibre feeds the microbes, and many kinds of fibre feed many microbes, which make short-chain fatty acids that feed the gut lining, calm the immune system, and signal sugar and appetite.
- Fermented foods and probiotics add live microbes that pass through and do a job while they are there, and they do not usually settle in the gut for life.
- The microbiome can help predict the blood sugar rise after a meal, in part, but the prediction is not exact, and it is one piece of the answer, not the whole answer.
- Proven is the value of many plants and many fibres; promising is prediction and specific probiotics; still guesswork is a test or a supplement that claims to fix the gut on its own.
Glossary
Microbiome
The community of microbes that live in the gut, mostly bacteria in the large intestine. Food tech example: the microbiome in a person's gut breaks down fibre that human enzymes cannot break down.
Species
A kind of microbe, such as one kind of bacterium in the gut. Food tech example: a gut carries hundreds of bacterial species, each with its own job.
Diversity
The number of different kinds of microbes in a gut, and how evenly they are spread. Food tech example: a diet rich in many plants supports a diverse microbiome.
Composition
Which kinds of microbes are present in a gut, and in what amounts. Food tech example: two people can have the same diversity and a different composition of microbes.
Stool test
A test that reads the DNA of microbes in a stool sample to name the kinds present. Food tech example: a stool test names the bacteria in a sample and gives their relative amounts.
Sequencing
A method that reads the DNA letters of microbes to name them. Food tech example: sequencing a stool sample reads the DNA of the bacteria in it.
Shotgun sequencing
A sequencing method that reads all the DNA in a sample, naming more kinds and saying what the microbes can do. Food tech example: shotgun sequencing can say which fibres the microbes in a sample can break down.
Short-chain fatty acids (SCFAs)
Small molecules made when gut microbes ferment fibre, including acetate, propionate, and butyrate. Food tech example: butyrate, made from fibre, is the main fuel for the cells of the gut lining.
Butyrate
A short-chain fatty acid that feeds the cells of the gut lining and helps keep the gut wall strong. Food tech example: butyrate from fermented fibre feeds the gut lining.
Propionate
A short-chain fatty acid that travels to the liver and affects how the liver handles sugar and fat. Food tech example: propionate made from fibre signals to the liver.
Acetate
A short-chain fatty acid that travels widely in the body and affects appetite and fat storage. Food tech example: acetate made by gut microbes signals appetite.
Fermented food
A food made by microbes, such as yoghurt, kefir, or fermented vegetables, which carries live microbes and their products. Food tech example: yoghurt is a fermented food that adds live microbes to the gut.
Probiotic
Live microbes sold as a supplement or added to a food, judged by the strain, the dose, and the condition. Food tech example: a probiotic strain that helps one kind of diarrhoea may do nothing for another condition.
Prebiotic
A fibre or other substance that feeds the microbes already in the gut. Food tech example: inulin in a food is a prebiotic that feeds gut microbes.
Archaea
A group of microbes in the gut, distinct from bacteria, that are less studied than bacteria. Food tech example: archaea in the gut make gas from the products of bacterial fermentation.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.
- Valdes, A. M., Walter, J. and Segal, E., Role of the gut microbiota in nutrition and health, 2018.
Unit 10: Sticking to a Plan: Behaviour Change, Privacy and Ethics
Note
Unit 10: Sticking to a Plan: Behaviour Change, Privacy and Ethics
A question to start with
A person downloads a meal planning app, wears a glucose monitor, and follows a personal eating plan for three weeks. The plan works: their blood sugar is steadier, they feel better, and they lose a little weight. In week four, they go on holiday, eat out every day, and stop logging their food. In week five, they are back home, but the app is silent, the monitor is in a drawer, and the plan is forgotten. The food was right, the measurements were right, and the plan still failed, because the plan was built for a person who never gets tired, never goes on holiday, and never eats out. This unit is about the part of personalised nutrition that is not about food at all: how people change what they do, how they keep a change going when life gets in the way, and how the data that makes personalisation possible, the person's own health data, is kept private and used fairly.
Behaviour change is the reason most eating plans fail, and it is also the reason the best plans succeed. A plan that asks for a perfect diet, tracked perfectly, every day, will fail, because no one lives a perfect life. A plan that asks for one small change, that fits the person's life, and that survives a bad week, will last, because it was built for a real person, not for an ideal one. The second half of the unit is about the price of personalisation: the data. A personal plan runs on data about the body, the genes, the microbes, and the meals, and that data is among the most private information a person has. Who holds it, who can see it, who can sell it, and who can use it to judge the person, are questions that must be answered before the data is collected, not after it has leaked.
Questions to think about
- A plan asks a person to log every meal, every day, for a year. Why might that plan fail even if the meals in it are perfect?
- An app offers a free personal diet in return for your glucose data, your genes, and your food diary. What are you paying with, if not money?
- A person follows a plan for three weeks, then stops for a week, then starts again. Has the plan failed, or is that how real change works?
Think about the three weeks question before you read on
That is how real change works. A break, a holiday, or a bad week is not a failure. It is a normal part of changing a habit, and the plan that survives is the one that expects breaks and has a way back: a smaller step to restart with, a meal that is easy to make, and a person to talk to, rather than a guilt message from an app. A plan that calls a break a failure teaches the person to give up. A plan that calls it a pause teaches them to restart.
Why plans fail: the gap between knowing and doing
Knowing what to eat and eating it are different things, and the gap between them is where most plans fail. The gap has reasons that are ordinary and human. The first is that a new behaviour costs effort, and effort runs out: a person who cooks a new meal every night, logs every bite, and wears a monitor every day will be tired by week three, and tired people fall back to what is easy. The second is that life gets in the way: a holiday, an illness, a busy week at work, a family celebration, or a late night all break the routine, and a plan with no way back from a break stays broken. The third is that the environment fights the plan: a kitchen full of snacks, a workplace with a vending machine, a shop that sells a large pack for less than a small one, and friends who eat out every Friday all pull the person back to the old way. The fourth is that the plan asks for too much at once: five new foods, three new habits, and a daily log, all starting on Monday, is a plan that asks for a new life, not a new meal.
The cure for each of these is to make the change smaller, easier, and more forgiving. A small change, such as eating vegetables before the main dish, or walking for ten minutes after lunch, or swapping one sugary drink a day for water, is a change a person can keep when they are tired, busy, or away from home. An easy change is one that fits the person's life: a breakfast that can be made in five minutes, a lunch that can be bought near work, a dinner that the family will eat too. A forgiving change is one that survives a break: the plan says what to do after a holiday, after a celebration, or after a week when nothing went right, and the answer is a small step, not a restart from zero. The plan that lasts is the one that was built for the person's worst week, not for their best week, because the worst week is the one that decides whether the change survives.
Habits are the way a change stops costing effort. A habit is a behaviour that is tied to a cue, such as a time, a place, or an action that already happens, and that is repeated until it needs no decision. A person who eats vegetables before the main dish every time they sit down to dinner, or who walks after lunch every day, or who drinks water before coffee every morning, has tied the new behaviour to a cue that already exists, and the behaviour happens without a daily argument with themselves. The cue must be stable, the behaviour must be small, and the reward must be felt soon: the vegetables taste good, the walk clears the head, the water is refreshing. A habit that is tied to a cue that moves, such as "when I feel motivated", will not form, because motivation moves, and the cue must stay still.
The tools that help, and the tools that hurt
Self-monitoring, the act of watching what you do, is one of the strongest tools for behaviour change, and it is also one of the easiest to overdo. A food diary, a step count, a glucose trace, or a weight log all give feedback, and feedback lets the person see whether a change is working, which is motivating. But a diary that takes twenty minutes a day, a monitor that beeps at every rise, or a scale that is checked three times a day will tire the person, and a tired person stops monitoring, and then stops the plan. The rule is to monitor the smallest amount that still gives a useful answer: a food diary for one week to find the pattern, then a diary for one day a week to check it, a glucose monitor for two weeks to learn the rise after meals, then a check when a new food is tried. Monitoring is a tool for learning, not a job for life, and a plan that needs lifelong logging is a plan that has mistaken the tool for the goal.
Goals work best when they are specific, small, and set by the person, not by the app. A goal that says "eat better" is not a goal, because no one can tell whether they have done it. A goal that says "eat vegetables before the main dish at dinner, five days a week" is a goal, because it can be counted, and the person can see whether they did it. A goal that is set by the person, from their own life, is kept longer than a goal set by an app, because the person owns it. A goal that is reviewed, with the person, every week or two, and changed when it is too hard or too easy, is a goal that stays alive, while a goal that is set once and never reviewed dies quietly in the app.
Support is the tool most often left out, and the one that matters most when a plan is hard. A person who is changing what they eat needs someone to talk to: a family member who cooks with them, a friend who walks with them, a dietitian who reviews the plan, or a group that meets, in person or online, and shares what worked and what did not. Support is not advice. It is a person who listens, who does not judge a bad week, and who helps the person restart with a small step. An app that sends a message saying "you missed your goal" is not support. It is a judgement, and it teaches the person to delete the app. A plan that has a person in it, not only an app, is a plan that survives the weeks when the app is silent.
Privacy: whose data is it, and who can see it
Personalised nutrition runs on data, and the data is private. A glucose trace shows when a person eats, when they sleep, and when they are stressed. A food diary shows what a person eats, when, and with whom. A gene test shows a person's inherited risks, and those risks are shared with their family, who did not agree to be tested. A microbiome test shows the microbes in a person's gut, and those microbes change with illness and with medicine. A wearable shows where a person goes, how much they move, and how they sleep. Put together, these data can tell a story about a person that the person did not intend to tell, and that story, once it leaves the person's hands, cannot be taken back.
The first question to ask of any app, test, or service is: who holds the data, and where is it stored? A service that stores data on its own servers, in another country, under another country's laws, is a service whose data the person cannot control, however good its privacy words sound. The second question is: who can see it? A service that shares data with advertisers, with insurers, with employers, or with research partners, without the person's clear agreement, is a service that treats the person's data as its own. The third question is: can the data be deleted? A service that lets the person download their data and delete it, and that deletes it from its backups, is a service that respects the person's ownership. A service that keeps the data after the person leaves, or that sells it in a form that can be traced back to the person, is a service that does not. The fourth question is: what happens if the service is sold, or fails? The data must not pass to a new owner without the person's agreement, and the person must be told, in plain words, what will happen.
Consent is the answer to these questions, but consent must be real, not a long document that no one reads. Real consent is specific: it says what data is collected, for what purpose, for how long, and with whom it is shared, in plain words, on one page. Real consent is freely given: the person can say no, and can still use the service, or can use a smaller version of it, without being punished for saying no. Real consent can be taken back: the person can withdraw it, and the data is deleted, and the service stops using it. A consent that is buried in a long document, that is required for the service to work at all, or that cannot be withdrawn, is not consent. It is a condition, and the person who accepts it has not agreed. They have given in, because there was no other way to get the service.
Ethics: fairness, bias, and the limits of personalisation
Ethics asks who benefits from personalised nutrition, and who is left out. The first question is fairness of access. A personal plan that needs a glucose monitor, a gene test, a microbiome test, and a paid app is a plan that only a person with money can buy, and a person without money, who may need it most, cannot. A service that is free, but that takes the person's data in return, is a service that charges the poor with their privacy and the rich with money, which is not fair either. The answer is not to stop personalised nutrition, but to make its most useful parts, the food advice, the measurement of blood sugar, the support, available to everyone, and to say plainly who can afford the rest, and who cannot.
The second question is bias in the data. An AI meal planner trained mostly on foods eaten by one group of people, in one country, will give poor advice to a person who eats different foods, and a gene test built from studies of one group will predict poorly for another group. A microbiome test built from one population will misread another. The bias is not in the person. It is in the data the system was trained on, and the maker must say whose data it was trained on, and must test the system on other groups, before selling it to them. A personalised service that works well for one group and badly for another is not personalised. It is personalised for the group it was built on, and the rest are an afterthought.
The third question is the line between help and judgement. A service that tells a person their blood sugar is high, and suggests a smaller portion, is helping. A service that tells a person they are failing, that shames them for a bad week, or that reports their eating to an insurer or an employer who then judges them, has crossed the line from help to judgement. A service that uses a person's data to set an insurance price, or to decide whether they get a job, has crossed a line that should not be crossed, and the person must be told, before they give their data, whether that use is possible. The rule is simple: data given for health must be used for health, and not for judging, pricing, or punishing the person who gave it.
The diagram shows the behaviour change steps in order. The step most often missing is the fourth one, the plan for the break, and it is the step that decides whether a change survives a holiday, an illness, or a busy week.
A checklist before you share your health data
Before a person gives their health data to any app, test, or service, they can run through a short checklist, and the service that passes it is the service worth trusting. First, what data is collected, and why is each piece needed? A service that collects a gene test to plan a meal should be able to say, in plain words, how the genes change the meal. Second, who holds the data, and where is it stored? Third, who can see it, and is any of it shared with advertisers, insurers, or employers? Fourth, can the data be downloaded and deleted, and is it deleted from backups too? Fifth, can consent be taken back, and what happens then? Sixth, is the service tested on people like the person who will use it, eating foods like theirs? Seventh, does the service have a person to talk to, or only an app? A service that answers all seven, in plain words, on one page, has earned the person's trust. A service that cannot answer them, or that hides the answers in a long document, has not, and the person should not give it their data.
Practical note: Open the privacy page of an app you use, and try to answer the seven questions above from that page alone. If you can answer them in ten minutes, the service has done its job. If you cannot, write down the questions it did not answer, and ask the service, or choose a service that does answer them. That exercise, done once, teaches more about data privacy than a long document ever will.
Quick self-check
Question: Why does a plan that asks for perfect logging every day often fail?
Because logging costs effort, and effort runs out, and a tired person stops logging and then stops the plan. Monitoring is a tool for learning, and the rule is to monitor the smallest amount that still gives a useful answer, not to log for life.
Question: What makes consent real, rather than a long document no one reads?
Real consent is specific, saying what data is collected, for what purpose, for how long, and with whom it is shared, in plain words. It is freely given, so the person can say no and still use the service, and it can be taken back, with the data deleted when it is withdrawn.
Try it yourself
Choose one small eating change you would like to keep, such as eating vegetables before the main dish, or drinking water before coffee, or walking for ten minutes after lunch. Write down the cue you will tie it to, the smallest version of it you could do on a bad day, and the step you will take to restart it after a break. Then try it for one week, and write one line each day saying whether you did it, and one line saying what helped or what got in the way. At the end of the week, read your seven lines and ask: was the change small enough, was the cue steady enough, and was the restart step clear enough? That week of notes is a behaviour change experiment, and it is the same experiment a dietitian runs with a client, on a smaller scale.
Key ideas to carry forward
- Knowing what to eat and eating it are different things, and the gap is filled by small changes, steady cues, and a plan that survives a bad week.
- A change that is small, easy, and forgiving lasts longer than a change that asks for a new life on Monday.
- Habits form when a behaviour is tied to a cue that already happens, repeated until it needs no decision, and rewarded soon.
- Self-monitoring is a tool for learning, not a job for life, and the rule is to monitor the smallest amount that still gives a useful answer.
- Goals work best when they are specific, small, set by the person, and reviewed often.
- Support is a person who listens and helps restart, not an app that judges a missed goal.
- Health data is private, and a service must say who holds it, who can see it, whether it can be deleted, and what happens if the service is sold.
- Real consent is specific, freely given, and can be taken back, with the data deleted when it is withdrawn.
- Personalised nutrition must be fair: it must work for people who eat different foods, and it must not be used to judge, price, or punish the person who gave their data.
Glossary
Behaviour change
Changing what a person does, such as what they eat, and keeping the change going over time. Food tech example: behaviour change is eating vegetables before the main dish at dinner, five days a week, for months.
Habit
A behaviour tied to a cue, repeated until it needs no decision. Food tech example: drinking water before coffee every morning is a habit tied to the cue of waking up.
Cue
A time, place, or action that already happens, used to trigger a new behaviour. Food tech example: sitting down to dinner is the cue for eating vegetables before the main dish.
Self-monitoring
Watching what you do, such as keeping a food diary or a step count, to see whether a change is working. Food tech example: a food diary kept for one week shows which meals give a high blood sugar rise.
Relapse
A return to an old behaviour after a break, such as eating out every day on holiday. Food tech example: a relapse after a holiday is a normal part of change, and the plan should have a small step to restart.
Consent
Agreement to share data, which must be specific, freely given, and able to be taken back. Food tech example: consent to share glucose data says what data is shared, for what purpose, and for how long.
Privacy
The control a person has over who can see their personal information, such as health data. Food tech example: privacy means a person's glucose trace is seen only by those they agree to share it with.
Health data
Information about a person's body and health, such as glucose readings, genes, microbes, and food diaries. Food tech example: a food diary and a glucose trace together are health data that can tell a story about a person.
Bias
A slant in data or in a system, such as an AI trained mostly on one group's foods, which makes it work badly for another group. Food tech example: a meal planner trained on one country's foods gives poor advice to a person who eats different foods.
Fairness
The question of who benefits from a service and who is left out, such as who can afford a personal plan. Food tech example: fairness asks whether a personal plan that needs costly tests is available to people without money.
Wearable
A device worn on the body, such as a watch or a monitor, that measures steps, heart rate, sleep, or glucose. Food tech example: a wearable glucose monitor shows the blood sugar rise after each meal.
Adherence
Sticking to a plan over time, such as following an eating plan for weeks and months. Food tech example: adherence to a plan is higher when the change is small and fits the person's life.
Support
Help from a person, such as family, a friend, or a dietitian, who listens and helps restart after a break. Food tech example: support is a dietitian who reviews a plan with a client every two weeks.
Data deletion
Removing a person's data from a service, including from its backups, when the person asks for it. Food tech example: data deletion means a glucose trace is removed from the app's servers after the person leaves.
Insurance pricing
Setting the price of insurance from a person's data, which is a use of health data that must be disclosed before the data is given. Food tech example: a service must say whether eating data could be used for insurance pricing.
References
- Gibney, M. J., Lanham-New, S. A. and Cassidy, A., Introduction to Human Nutrition, 2nd edition, 2009.
- Mann, J. and Truswell, A. S., Essentials of Human Nutrition, 4th edition, 2012.
- Michie, S., van Stralen, M. M. and West, R., The behaviour change wheel, 2011.
- Berry, S. E., Valdes, A. M. and Drew, D. A., Human postprandial responses to food, 2020.
- Zeevi, D., Korem, T. and Zmora, N., Personalized nutrition by prediction of glycemic responses, 2015.