High intakes of ultra-processed foods (UPFs) have been associated with adverse health outcomes including weight gain, cardiometabolic disease and cancer(1,2). However, the limitations of NOVA highlight the need for a novel classification which captures food properties and features of processing that drive these health outcomes. Proposed mechanisms underlying the associations between UPF and health include energy density and hyperpalatability(3,4). Within the US food system, the prevalence of both energy dense and hyperpalatable foods is increasing, and there is moderate overlap between foods identified by each definition(5). The aim of this study was to characterise the energy density and hyperpalatability of foods and habitual diets consumed by UK adults.

Logged dietary data, baseline characteristics and body composition were analysed from ZOE PREDICT 1 UK participants (n= 1001; NCT03479866)(6). Weighted logged dietary data, including more than 12,000 foods (beverages excluded), from participants with ≥2 free-living days (n=891) were included. Energy density (kcal/g) for all foods and participant diets was calculated. Hyperpalatable foods (HPF) were classified using a data-defined definition(7) and percentage energy intake (%EI) from HPFs were calculated. Differences in energy density and hyperpalatability were examined across top and bottom tertiles of energy intake.

In the PREDICT 1 cohort (n=891, 73% females, (mean±SD) age 46±12 years and BMI 26±4.9 kg/m2) 55% of foods were high in energy density (>2 kcal/g), 49% were defined as HPFs and 35% were both energy dense and hyperpalatable. Across participants’ habitual diets (beverages excluded), the average energy density was 2.1±1.0 kcal/g and proportion of energy consumed from HPFs was 48±16 %EI. There was a significant difference across tertiles of energy intake (low versus high) in the energy density of diets (kcal/g) and contribution of energy from HPFs (%EI) (P<0.05 for all).

This analysis provides valuable insights into the energy density and hyperpalatability of habitual diets in the PREDICT 1 cohort, demonstrating overlap between these features. Further work is required to understand the independent and combined role of these factors, and additional proposed mechanisms underlying the negative health effects associated with processing, such as food texture, energy-intake-rate, and additives and non-culinary ingredients. A novel classification should build on NOVA by encompassing the food properties and features of processing that drive adverse health outcomes. Policy makers must consider the role of these factors to ensure nutritionally beneficial and affordable foods are not penalised without appropriate scientific justification.