PREDICT study provides personalised nutrition revelations
Dr Tim Spector, professor of genetic epidemiology and director of the TwinsUK Registry at King’s College London, and his team, joined the ASN's Nutrition Live Online 2020 conference last week and revealed the key preliminary findings from their PREDICT study - which he described as the largest ongoing programme to measure postprandial responses to food in nutritional science.
The ultimate aim of this collaborative, ongoing programme, is to develop a home test and app that will allow anyone to understand their personal nutritional response and choose foods that optimise their metabolism, improve long-term health and manage weight more easily.
The study, set to be published in Nature Medicine in the next few weeks, involved 1,000 individuals from the UK (unrelated, identical and non-identical twins) and 100 unrelated individuals from the US.
During a controlled clinic day, the team assessed postprandial (0-6h) metabolic responses to sequential mixed-nutrient dietary challenges. This was followed by a two week at-home phase.
Glycemic and lipaemic responses to multiple duplicate isocaloric meals of different macronutrient content and self-selected meals (>100,000), were tested at home using a continuous glucose monitor (CGM) and dry blood spots. Baseline factors included metabolomics, genomics, gut metagenomics and body composition.
Dietary data was collected using an app and dashboard combining weighed food records, photographs, bar coding and live nutritional support. Sleep and activity were monitored using wearable devices.
Throughout the study, over 32,000 muffins were consumed, 28,000 TAG readings taken, 132,000 meals were logged and over 2 million CGM glucose readings were taken.
Key Findings
Dr Sarah Berry, senior lecturer at King’s College London, said the participants' triacylglycerol, glucose and insulin results after meals showed that inter-individual responses to identical meals varied hugely.
The team found that variability within a person was much less variable – an important finding because its important to understand what’s the day-to-day variability for machine learning to be effective.
The data also revealed a huge variation in responses in twins, showing that genetics have little impact.
Dr Berry explained: “Less than 1% of lycemin and around 50% of glycemic response is genetic and therefore a large portion is modifiable which means lifestyle interventions have the potential to produce pronounced effects on metabolic health.
“In addition to looking at genetics we started to tease apart all the other factors which might affect responses… and we were able to quantify the contribution of the different determinants.”
For triacylglycerol, the biggest contributor was serum lipid markers, and other serum markers. Genetics was the lowest contributing factor and meal composition came seventh in the list of contributors, after microbiome and gender.
For glucose, the largest contributing factor was the meal composition, followed by genetics, meal context (factors such as the time of day or timing of exercise or sleep), serum glycemic markers and the microbiome.
The team found that meal context had a surprisingly large influence, with individuals showing a far higher glycemic response to the same meal when consumed at lunch, compared to when consumed at breakfast, but this variability was strongly influenced by age.
"We looked at the size of this effect and the variability between individuals to find out whether some are better off eating high carb meals in the morning and other in the evening.
"We found the time of day affect was particularly pronounced in young people but for people over the age of 60 there was far less influence, it may not matter so much what time of day you eat."
Microbiome
Dr Nicola Segata, associate professor at the University of Trento, Italy, discussed the findings from the microbiome tests.
He explained that the team took microbiome data from stool sample metagenomics, looked at 18 different measures of the person - age, gender, BMI, blood pressure etc - as well as their habitual diet and fasting and postpranial responses.
They then undertook shotgun metagenomics to cross reference the species in the microbiome with the other individual measures.
"We looked at what is the overall signature of a ‘healthy’ microbiome. Firmicutes CAG 95 bacterium was one of the reproducible markers of good health. There was a panel of 15 bacterial strains all found to be consistently associated with markers of good health and 15 consistently associated with markers of bad health."
The team were able to use the food frequency questionnaires to map out what affect foods had on the composition of the microbiome.
They mapped which bacterial components of the microbiome were linked with which foods and they categorised the bacterial components into 'good' or 'bad' cluster, associated with good or bad health. They also colour coded the foods according to whether they were considered healthy or unhealthy, plant based or animal based.
They found that meat (beef, poultry, sausages pork and ham) was most highly likely to be linked with the 'bad' cluster of bacterial strains while the 'healthy' plant-based foods consistently were linked with the healthy cluster of bacteria.
There were some exceptions to this rule, such as jam, dark chocolate, yogurt, shellfish, oily fish and eggs being linked with more good bacteria. Red wine was also found to provide several good bacterial species.
It also wasn't quite that clear cut, as the researchers also found the the quality of the food was a very important determinant, meaning a person could include meat in their diet and still have a healthy microbiome as long as their overall diet was healthy and diverse.
These results do suggest that there is some generic advice that can be given.
Dr Berry concluded: “So the general gut healthy dietary advice would be plenty of fibre, fermented foods etc and to consume a variety of unprocessed, healthy foods. But also it’s important to think about how we eat - to control responses to foods through timing of eating, meal ordering and how much exercise we are getting.”
Carb specific results
The researchers also carried out a 'PREDICT - carbs' study, involving 100 participants who were the most accurate at logging their meals during PREDICT. The study was undertaken remotely but was similar to PREDICT in that the same monitoring, testing, and logging of health data was used.
In a cross-over design, participants consumed different carb staples in meals throughout the day and data showed that glycemic response at lunch was hugely different depending on the meal consumed at breakfast even if the lunch meal was the same and availability of carbs was exactly the same across the breakfasts. This shows the meal patterns throughout the day have a huge impact on our responses to food.
The team also observed huge differences in postprandial responses, around two to three hours post meal, with some having a big dip in glucose while others did not.
"We found that only small variabilities in glucose dip created a really big differences in hunger and alertness and those with bigger dips had a shorter time before their next meal, more calories in their next meal and this bigger calorie intake carried on throughout the whole day. Therefore, we should be looking at the two to three hour glucose responses in relation to satiety and energy intake."
They also examined two inflammatory measures, IL-6 and NMR GlycA, and found huge variations in responses and discovered those with a GlycA response in the 90th percentile had a doubled ASCVD (Atherosclerotic Cardiovascular Disease) risk.
Ongoing and future studies from the team will include PREDICT 2, involving 1,000 US participants, and PREDICT 3, which will analyse around 10,000 US participants.
