Who consumes ultra-processed food? A systematic review of sociodemographic determinants of ultra-processed food consumption from nationally representative samples · Dicken SJ, Qamar S, Batterham RL

Introduction

The global prevalence of obesity and adiposity-related non-communicable diseases has greatly risen in recent decades, with 2 billion adults now living with overweight or obesity (1), compared with just over 100 million in 1975 (2). An increasing concern in relation to rising levels of obesity-related disease has been the nutrition transition during the same time period, away from minimally processed foods (MPF), and towards greater consumption of ultra-processed foods (UPF) (3). The most commonly used food processing classification, but not necessarily the most validated, is the NOVA classification (not an acronym). As defined by NOVA as foods made using extracts and components of whole foods, UPFs typically contain five or more ingredients, produced using industrial methods and ingredients not used in the home, such as bulking agents, emulsifiers and additives (4). The nutrition transition and rising sales of UPFs has been driven by economic development, with urbanisation, shifts in workplace dynamics, expansion of multinational corporations, and with rising income levels at the national and individual level (3,5). As such, UPF sales are increasing around the world, with sales in middle-income countries (MIC) rapidly rising to the levels sold in high-income countries (HIC) (6,7).

Higher intakes of UPF are associated with a range of adverse health outcomes including obesity, type 2 diabetes, cardiometabolic disease and all-cause mortality (8,9). As a result, health organisations and national dietary guidelines including the American Heart Association and Brazilian Dietary Guidelines recommend limiting UPF intake (10,11), and calls have been made for systemic, national and international-level change to support individuals to reduce their UPF intake (12). However, to understand the effectiveness of universal strategies to reduce UPF intake, or whether regional policy action or targeted UPF interventions may differentially benefit certain sociodemographic groups, it is crucial to determine which individuals are consuming large quantities of UPF, and therefore who is at greatest risk of future health problems.

Previous reviews and studies suggest that higher UPF intakes are associated with male gender, younger age, lower education level and higher incomes (3,5,13–15). However, previous studies are subject to a number of methodological limitations. National food balance sheets provide a minimalist tool to understand national-level trends in food consumption (16), but preclude the ability to consider sociodemographic differences in food intake. Household Budget and Expenditure Surveys based on food purchases can be used as a proxy to estimate household or individual intakes and allow for sociodemographic stratification, but still include biases such as food waste and food loss, and preclude the ability to determine food consumption by specific individuals in the household (16). Dietary assessment using food consumption tools (e.g. 24-hour dietary recalls, food diaries and food frequency questionnaires (FFQ)) are considered optimal measures to assess individual-level food consumption (16).

Many of the large-scale, health-based cohort studies that assess individual-level UPF intake and their relationship with health outcomes also consider sociodemographic associations with UPF consumption (8). Such studies use convenience sampling methods that may include biases such as a healthy volunteer selection bias, or lead to underrepresentation or exclusion of sociodemographic strata, resulting in such samples not being generalisable to the national population. Smaller samples from cities or specific regions may be representative of such regions, but are also unlikely to accurately represent the national population. In combination with food consumption assessments, nationally representative samples typically obtained with complex, multistage probabilistic sampling are needed to provide country-level distributions of UPF intake by sociodemographic factors.

No study has systematically reviewed and formally synthesised papers assessing the association between sociodemographic factors and individual-level UPF intake from nationally representative samples. Current reviews of sociodemographic predictors of individual-level UPF intake are not systematic (13,14), and do not focus on studies that are nationally representative (13), or consider only age and gender (15) or systematically review nationally representative samples without considering sociodemographic predictors of UPF intake (17).

The objective of this review was to systematically synthesise the evidence regarding sociodemographic characteristics associated with a high, individual-level UPF intake as defined by NOVA, in nationally representative samples.