Personalised nutrition: Exciting opportunity or costly distraction?

In a webinar hosted by the European Nutrition Leadership Platform (ENLP) this week, Dr Nicola Guess, an academic dietitian and researcher specialising in the dietary prevention and management of type 2 diabetes, provided a critical analysis of the current evidence base for personalised nutrition interventions.

She explained there are three key ways to provide personalised nutrition, via: preferences, phenotype (observable health traits), or omics (data on the genome, microbiome, and metabolome).

Dr Guess asserted that there is plenty of evidence to show that if you personalise a diet to fit a person’s lifestyle and preferences then this will help them adhere to the diet and lead to better health outcomes.

She also noted the DASH study​ which showed diet can be personalised, based on phenotype, in order to lower blood pressure.

What’s more, she believes nutrition apps and AI can be extremely helpful in helping people to stick to nutrition plans and make healthy choices.

However, she questions the value in gathering omics data to provide diet advice.

Noting the huge investments in research aiming to show the benefit of omics approaches, Guess pointed out there are already many commercial products already on the market.

She said these companies generally base their marketing on the claim that everyone needs a diet as unique as they are.

“These services attain billions of data points and use machine learning and AI – it’s seen as cutting edge, and it's big news.”

Discussing the validity of basing nutrition recommendations on omics data, Dr Guess noted that you have to consider if the analysis of the person’s omics actually leads to personalised unique recommendations and if prescribing these diets actually attains better outcomes, compared to a standard healthy diet.

To find that out, she took viewers through the lesser discussed aspects of some of the personalise nutrition studies that have been published.

She argued that if you look at the algorithms created by a CGI-based personalised nutrition service, the algorithm will generally lead to a low carb diet with more protein and fat.

“We know high protein, low carb diets lower blood-sugar spikes in anyone.”

Looking at the algorithm for a service which takes urine samples, plasma and serum, saliva and more, in order to cluster people into health categories, she said that the algorithm created fairly standard healthy diet advice, but with increased fibre.

“This data doesn’t suggest the diets are personalised.”

She notes that Preventomics own study​ concluded that personalised dietary plans didn’t lead to greater benefits over a generic healthy diet in their 10 week clinical trial.

Type 1 error

She argued that omics driven services will gather billions of data points and, with this, they are bound to discover statistically significant correlations.

“But correlations don’t necessarily mean anything. None of it is necessarily causal.

“You can find statistical significance easily. But the question is whether it is clinically significant.

“Looking at glucose as a primary outcome, there is huge intra-person variation in glucose responses.”

She noted that a person's glucose response to a meal will be greatly impacted by the meal eaten previously, and will be greatly impacted by what exercise the person has done earlier that day, or the day before, and even the day before that.

“You can bring the same person back a second time and find completely different responses to the same foods.

“That’s what makes it so hard to use CGMs to work out what foods to avoid….

“If you want to control for intra-person variance you should give the same meal three times at the least. But as far as I’m aware these studies aren’t doing this.”

Equally she noted that saliva and blood samples only provide a snapshot picture of health which is unhelpful.

Control groups

She argued that another issue in personalised nutrition studies is when the control group on the standard healthy diet is given an inferior plan in terms of the amount of support and guidance they are given throughout the trial.

“Regular contact will help someone to make healthy choices and they will be more likely to get results.”

She suggested that if the personalisation group gets an app with regular suggestions of meals and prompts to make healthy choices, then the control group should be given that as well, but with the standard healthy diet advice.

“If you are going to encourage people to spend a lot of money on your app by suggesting your diet is going to be more efficacious than a standard healthy diet, that should be demonstrated with a careful control.”

Dangerous distraction

Not only does she question the validity of the methodologies being used and the marketing messages of costly services, but she is concerned there is a risk to general health.

She pointed out the huge issue of over consumption of ultra-processed foods, with ease of access to cheap junk food making it easy for people to make bad food choices.

“There’s a great wariness of nanny state - no one wants to regulate or tax food – so someone coming along and saying ‘hey it’s just because people aren’t eating the right foods for their own metabolism’ might be very attractive to people that don’t want to address the food elephant in the room.”

Answering an audience question regarding the potential for CGM’s to cause unhealthy choices and glucose phobia in the healthy population, Dr Guess asserted that high cholesterol and blood pressure are two hugely prevalent issues while glucose control is not a problem in the non-prediabetic/diabetic population, so she sees no benefits in focusing people’s attention on glucose.