Scientific Updates /
Nova UPF classification: too heterogeneous for NCD risk associations
05 December 2023
Background
A continuing increase in non-communicable diseases (NCDs) in mid- to high-income countries has been associated with the concurrent increase in ultra-processed food (UPF) consumption. Additionally, it is now common for individuals to live with multimorbidity, i.e., with at least two chronic diseases such as cancer, cardiovascular disease (CVD), and/or type 2 diabetes (T2D), while UPF consumption can contribute to as much as 60% of energy intake.
Aim
The study aimed to investigate the association of UPF consumption with the risk of specific NCDs – cancer at any site, CVD, and T2D – both as single outcomes and co-existing as multimorbidity. The study explored subgroups of UPF as well as total UPF consumption.(1)
Method
Participants
As the foundation for their research, the investigators utilised data from the well-established prospective cohort study – the European Prospective Investigation into Cancer and Nutrition (EPIC). EPIC examines the association of diet, lifestyle, genetics, and environmental risk factors with the incidence of cancer and other diseases across 10 European countries, involving a total of 520,000 participants (70% female) recruited between 1992 and 2000. Participants aged 35-74 years with complete datasets and a willingness to participate in this study were included.
Diet assessment and estimation of UPF consumption
Dietary assessments, conducted using food frequency questionnaires (FFQ) at recruitment into the EPIC study in the 1990s, were used to evaluate intakes. The FFQ reflected participants' eating habits over the last 12 months and included more than 11,000 food items.
UPF food classification
Foods and drinks classified as UPF by the Nova system were the focus of this study. Typically, UPF (group 4 of the Nova system) includes mass-produced ready-to-consume pre-packaged products such as carbonated soft drinks, sweet or savoury packaged snacks, confectionery, mass-produced packaged breads and rolls (whole grain or white), margarines and other spreads (including those based on unsaturated oils), biscuits, pastries, cakes, breakfast cereals (including fortified unsweetened and whole grain varieties), flavoured milk drinks and yogurts, processed meats, and ready meals.
For more on the Nova food classification system, please click here.
UPF were identified and calculated as a composite variable. The authors modified the UPF calculation to better reflect the degree of processing of ingredients at the time of data collection in the 1990s.
UPF assessment
Correlations were made between UPF consumption per 1 standard deviation (SD) increments of energy-adjusted g/day and NCD and multimorbidity outcomes. One standard deviation corresponded to approximately 260g UPF daily.
Assessment of UPF subgroups
In addition to total UPF intakes, the team further categorised UPF foods into 9 sub-groups and compared each to different NCDs and multimorbidity outcomes. The 9 UPF subgroups were:
Ultra-processed breads and cereals
Sauces, spreads, and condiments
Sweets and desserts
Savory snacks
Plant-based alternatives (to meat and dairy)
Animal-based products
Ready-to-eat/heat mixed dishes
Artificially and sugar-sweetened beverages
Other UPF
Accommodating for confounding factors
The authors accommodated for socio-demographics, lifestyle, physical activity, body weight, education level, menopausal status, and use of HRT.
Disease outcome measures
Incidents of cancer at any site, CVD, and T2D, as well as mortality, were considered. Additionally, the study examined the presence of multimorbidity involving any combination of the mentioned diseases.
Results
266,666 participants from the original EPIC dataset were included, with 60% being women, and the mean follow-up period was 11.2 years.
Mean and (SD) UPF intakes
Men: 413g/day (292), accounting for 34% of total calories
Women: 326g/day (242), constituting 32% of total calories
NCD incidence after 11.2 years follow up
21,917 cases of cancer
10,939 cases of CVD
11,322 cases of T2D
4,461 individuals developed multimorbidity
Despite more women than men in the baseline measures, multimorbidity was more common in men than women (61% vs. 39%).
The most common multimorbidity was cancer among participants with CVD, followed by cancer with T2D, and the least common was T2D with CVD.
Total UPF intake associated with NCDs and multimorbidity
Positive associations were confirmed between higher consumption of UPF per 1 SD (equivalent to 260g per day) with each of the three NCDs – indicated as Hazards Ratio per 1 SD (HR1SD). The positive association remained after adjusting for body weight, except for T2D, which was attenuated.
Cancer: HR1SD 1.01; 95% CI: 1.00–1.03
CVD: HR1SD 1.06; 95% CI: 1.04–1.08
T2D: HR1SD 1.11; 95% CI: 1.10–1.13
A similar positive association was demonstrated between higher UPF consumption and the risk of multimorbidity – HR1SD 1.09; 95%CI: 1.03–1.09. However, the transition from one NCD to multimorbidity was not as clear, with some transitions showing a null HR association.
Not all UPF foods were associated with multimorbidity
When each of the 9 UPF subgroups were compared to NCDs and multimorbidity outcomes, associations were not demonstrated for all UPFs.
UPF breads and cereals demonstrated an inverse association with multimorbidity. Interestingly, breads and cereals reduced the risk of developing multimorbidity with HR1SD 0.97; 95%CI: 0.94-1.00.
UPF subgroups demonstrating no association with multimorbidity development:
Plant-based alternatives to meat and dairy
Ready-to-eat/heat mixed dishes
Sweets and desserts
Savoury snacks
Other unspecified UPF
UPF subgroups demonstrating a positive associations with the development of multimorbidity
Consumption of:
Animal-based products: HR1SD 1.09; 95%CI: 1.05-1.12
Artificially and sugar-sweetened beverages: HR1SD 1.09; 95%CI: 1.06-1.12
Sauces, spreads, and condiments association was borderline at HR1SD 1.03; 95%CI: 1.00-1.06
Discussion
The authors highlight the significant heterogeneity within the Nova UPF group, hindering clear associations with NCD outcomes. Variation across the 9 UPF subgroups underscores the importance of considering nutritional quality and energy density as well as the degree of processing.
It's unsurprising that overall UPF consumption is linked to increased NCD risk and multimorbidity transition, given the prevalent salt, fat, and sugar content, high energy density, and lack of positive nutrients, notably fibre and vitamins. Adjusting for carbonated drink intake showed no difference, likely due to the inclusion of both artificially and sugar-sweetened varieties.
Regarding specific UPF subgroups, positive associations were noted with animal-based foods (contributing to saturated fat) and sweetened beverages (linked to higher free sugar intake). The association with artificially sweetened beverages, though interesting, aligns with established patterns among individuals with pre-existing cardiometabolic events.
Notably, some UPF classified foods, such as cereals and breads, showed an inverse association, likely due to the inclusion of wholegrain varieties with positive nutrient profiles and lower energy density. Other UPF categories, like plant-based alternatives, ready-to-eat dishes, and unspecified UPF, displayed no clear association, suggesting variable nutritional profiles.
This study found sweets and desserts to have no association with multimorbidity.
The authors acknowledge the absence of established mechanisms for Nova UPF's increased NCD risk, attributing established mechanisms to nutritional quality and energy density and the fact that diets with higher UPF intakes tend to exhibit poorer overall diet quality.
Conclusion
For a more precise understanding of the association between the Nova UPF food group and NCDs, it is crucial to further categorise foods, considering their diverse nutritional profiles. While an overall positive association exists between UPF intake and NCDs, this research underscores that it is UPFs high in energy density, saturated fats, sugars, and salt that are more likely to be linked to NCDs. Conversely, UPFs with a healthier nutrient profile, such as wholegrain breads and cereals, exhibited an inverse association, while others like plant-based dairy and meat alternatives showed no correlation. This study emphasises the significance of evaluating the overall quality of the entire diet, transcending the focus on specific foods and food groups.
Save article as PDF
Expert opinion
BNF warns against focusing solely on UPF classification
Original research
Labelling soya foods as ultra-processed is unhelpful