Contribution of Nova-classified ultraprocessed foods to energy and nutrient intakes in very old people: the Newcastle 85+ Study

Abstract

Objectives

To estimate the contribution of ultraprocessed foods (UPF) to total energy intake and to macronutrient and micronutrient intakes among very old people aged 85 years in the Newcastle 85+ Study.

Design

Cross-sectional observational analysis of baseline dietary and demographic data from the Newcastle 85+ cohort.

Setting

Community-dwelling and institutionalised adults in Newcastle on Tyne and North Tyneside, UK, recruited through general practice registered between June 2006 and October 2007.

Participants

Eight hundred participants (62% female) aged 85 years at baseline, with two complete non-consecutive 24-hour dietary recalls.

Primary outcome

The primary outcome was the contribution of UPF (Nova group 4) to total energy intake, macronutrient intakes, expressed as percentage of total energy for carbohydrate, protein, total fat, saturated fat and added sugars, as grams per day for fibre and to micronutrient intakes (vitamins A, B₆, B₁₂, C, D, E and folate) and minerals (calcium, potassium, magnesium, zinc, selenium, phosphorus, iron and sodium). All evaluated across sex, education and socioeconomic status, adjusted tertiles of UPF intake.

Results

Among the 800 participants included in the analysis, UPF contributed 56% of total energy intake, surpassing that from unprocessed foods (27%). Total energy intake did not differ across tertiles of UPF consumption (lowest vs highest tertile: 1759.5 kcal/day (95% CI 1684.6 to 1834.4) vs 1740.0 kcal/day (1667.3 to 1812.7)). Higher UPF intake was associated with a higher proportion of energy from carbohydrates and added sugars, and a lower proportion from protein and saturated fat. Intakes of several micronutrients were lower in the highest versus the lowest UPF tertile, including vitamin C (59.9 mg/day (49.8 to 70.0) vs 94.0 mg/day (83.7 to 104.4)) and potassium (2455.9 mg/day (2334.1 to 2577.6) vs 2786.3 mg/day (2660.8 to 2911.8)). By contrast, calcium from fortified foods increased across tertiles (6.3 mg/day (3.7 to 8.9) to 15.4 mg/day (12.9 to 17.9)).

Conclusions

This study highlights the potential role of UPF in the diets of very old people: higher UPF intake was not associated with higher energy intakes often observed in younger populations. Some UPF, particularly fortified products, may contribute to meeting micronutrient requirements in very old people where dietary inadequacies are common. Further research is needed to confirm these findings and to inform dietary guidance for very old people.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Dietary intake was assessed using two non-consecutive interviewer-administered 24-hour recalls, improving estimation of usual intake compared with a single recall.

• The Nova classification of dietary data adhered strictly to a standardised protocol with recipe disaggregation and sensitivity analyses to assess classification uncertainty.

• The study relied on dietary data collected in 2006–2007, which may not reflect current ultraprocessed food availability or formulation.

• The analysis was cross-sectional, limiting causal inference between ultraprocessed food intake and nutrient profiles.

• Participants were recruited through general practice registers, which may limit generalisability to very old people not engaged with primary care.

Introduction

In the United Kingdom (UK), the number of adults aged 65 and over has increased by 52% between 1981 and 2021. This demographic shift is even more pronounced among those aged 75 and over, whose population grew by 84%, and those aged 85 and over, whose numbers rose by 186% during the same timeframe.¹ As life expectancy continues to rise, ensuring optimal nutritional intake among very old people (aged 85 and over) is increasingly vital to mitigate the prevalence of chronic diseases.

In response to growing dietary concerns, Monteiro et al² developed the Nova food classification system, which categorises foods according to the nature, purpose and extent of their processing.² The Nova classification was presented within a Food and Agriculture Organisation framework on food processing and has since been widely applied in public health and nutrition research.³ The system classifies foods into four groups: (1) unprocessed or minimally processed foods, (2) processed culinary ingredients, (3) processed foods, (4) ultraprocessed foods (UPF).² As a conceptual framework, Nova has been used in subsequent research to explore differences in dietary patterns and nutritional profiles across levels of food processing, supporting investigations into their potential health implications.³ The impact of food processing on health has received considerable attention due to its association with diet-related non-communicable diseases.4,8 UPF have been linked to obesity,9,13 cardiovascular diseases,14,16 metabolic disorders17,20 and numerous other comorbid conditions across the life course.

In a recent meta-analysis of nationally representative data, the authors observed that higher intake of UPF (as defined by Nova) among adults aged 18 and over were associated with poorer diet quality and increased energy intake.²¹ Specifically, UPF commonly consumed included sweets (eg, confectionery) and sugar-sweetened beverages (SSB). Thus, elevated UPF consumption was linked to higher intake of free sugars, total fats and saturated fats, and concurrently associated with lower intake levels of dietary fibre, protein and essential minerals (including potassium, zinc and magnesium), and key vitamins (A, C, D, E, B12, niacin).²¹

However, the interpretation that UPF consumption is uniformly detrimental to health has been increasingly challenged. Recent studies have highlighted substantial heterogeneity in the associations between UPF and health outcomes, depending on the level of exposure and the specific UPF subgroups consumed. Visioli et al argued that adverse associations observed in epidemiological studies may not reflect a causal effect of ultraprocessing as such, but instead be driven by residual confounding, broader dietary patterns or specific UPF subgroups rather than the overall UPF category.²² Consistent with this interpretation, a multinational cohort study has shown that only selected UPF subgroups, mostly SSB and ultraprocessed fats or sauces, are consistently associated with adverse outcomes, whereas other UPF categories, including breads and breakfast cereals, frequently show null or inverse associations with cardiometabolic risk and mortality.²³ Similarly, a study on females aged 60 and above demonstrated that with every increase in servings of subgroups of UPF, mainly from SSB and fats per day, frailty risk increased; conversely, these associations were not found with whole grain subgroups.²⁴ Evidence indicates that the health effects of UPF differ by physical form, with ultraprocessed drinks, particularly SSB, consistently associated with higher all-cause mortality, whereas UPF show null or heterogeneous associations after multivariable adjustment. These findings suggest that adverse health associations attributed to UPF may be driven predominantly by liquid rather than solid ultraprocessed products.²⁵

In addition, evidence further suggests that associations between UPF intake and adverse health outcomes are exposure-dependent and context-specific. In countries with lower average UPF consumption, increased health risks have been observed even at nominal levels of UPF intake. In contrast, in populations with chronically high UPF consumption, associations appear marginal or absent across much of the exposure distribution. Dose–response analyses in high-UPF settings indicate limited marginal risk beyond certain intake thresholds, supporting the hypothesis that UPF intake may act as a proxy for broader lifestyle or dietary patterns rather than an independent causal factor.²² Consistent with this interpretation, a large population-based cohort study reported that dietary quality, assessed using the Alternative Healthy Eating Index, showed a stronger and more consistent inverse association with all-cause mortality than UPF intake, with no consistent associations observed between UPF consumption and mortality within strata of dietary quality.²⁶

While individuals of all ages face considerable health risks associated with UPF, older adults appear to be particularly vulnerable due to age-associated physiological changes and their increased susceptibility to chronic diseases.²⁷ Research among Portuguese younger adults (aged 18 and above) and older adults (aged 65 and above) indicated that high UPF intake was associated with inadequate intake of certain vitamins (B6 and C) and minerals (folate, potassium, zinc and magnesium) among younger adults, though no such association was observed among the older cohort.²⁸ In addition, a systematic review concluded that high UPF was associated with frailty, obesity, dyslipidaemia, renal function decline and poor cognitive function in older adults aged 60 and above.²⁹ However, their findings showed that high UPF intake was associated with unintentional weight loss, possibly indicating that older adults consuming the highest amount of UPF may also be consuming the least total energy.²⁹ These studies might indicate different outcomes among older adults aged 60 and above when compared with adults aged 18–60 years.

Despite these findings, many studies have not employed a standardised protocol for classifying foods using the Nova system, introducing potential subjectivity and misclassification errors.³⁰ Such inconsistencies may lead to a type I error (false positive), wherein associations between UPF intake and adverse health outcomes are incorrectly inferred due to classification inaccuracies.³¹ For instance, misclassification could result in overstating the health risks attributed to UPF, when they may instead be driven by other dietary components, such as nutrient imbalances. Conversely, type II errors (false negative) may occur when a true association goes undetected, possibly due to underestimation of UPF consumption, especially in cases where mixed dishes contain hidden UPF ingredients.³² Therefore, to reduce the likelihood of both type I and II errors, it is imperative to employ a clear, validated and consistent classification protocol that accurately captures all UPF and enables the identification of true dietary associations. Martinez-Steele et al³³ published a comprehensive protocol grounded on best practices for applying the Nova food classification system, offering a standardised approach for researchers.

Applying this protocol is particularly important in understudied populations such as very old people (aged 85 and above), where dietary patterns may differ markedly from younger populations. Although evidence around food processing remains scarce for the very old, existing literature suggests that this age group derives much of their energy from cereals and cereal products (CCP), such as packaged breads and breakfast cereals, which are foods that fall within the Nova UPF category. Notably, CCPs have been identified as the highest contributors to fibre, folate, iron and selenium intake, and the second-highest sources of vitamin D, calcium and potassium in this population.³⁴

Despite growing evidence on the health risks associated with high UPF consumption, research specifically focusing on very old people remains limited. Accordingly, we examined whether, among very old people aged 85 years and over, UPF contribute a substantial proportion of total energy intake and whether high UPF intake (expressed as percentage total energy) is associated with differences in macronutrient distribution and micronutrient intakes. We hypothesised that higher UPF intake would be associated with a less favourable micronutrient profile, consistent with findings in younger adult populations. However, we also hypothesise that intakes of certain nutrients may be higher, reflecting the prominent contribution of cereal and cereal products such as packaged breads and breakfast cereals, which fall within the Nova UPF category, to fibre and key micronutrient intakes in this population, while noting that intakes of some nutrients may be higher where commonly consumed UPF (eg, breads and breakfast cereals) are fortified. This study aimed to address this gap by estimating the contribution of UPF consumption to nutrient intake (energy, micronutrients and macronutrients) among individuals aged 85 years and over. The study will also examine a nuanced characterisation of the contribution of specific food groups within Nova categories to total energy and nutrient intakes.

Methods

Study design and setting

This study is a cross-sectional analysis of the Newcastle 85+ cohort study. It adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist³⁵ to ensure rigorous and transparent reporting. The protocol was published as a peer-reviewed protocol paper.³⁶

Participants and recruitment

The Newcastle 85+ Study is a prospective observational cohort that included participants living in Newcastle on Tyne and North Tyneside in Northeast England, UK. Participants were recruited from general practitioner (GP) registers in Northeast England and aimed to recruit very old people, aged 85 and above (n=1470; men=496, women=974) who turned 85 in 2006. For the present analysis, only baseline data collected between June 2006 and October 2007 were used³⁷ (online supplemental figure S1).

All 64 general practices in Newcastle and North Tyneside were contacted; 53 (83%) agreed to participate and mailed invitations to their eligible patients. GPs excluded 11 patients due to terminal illness or posing a safety risk to the visiting nurse. Of the 1459 invited, 50 participants could not be contacted and were withdrawn (due to death (n=17), change of address (n=24) and no response (n=9)) (online supplemental figure S2). Leaving 1409 invitees reached, of these, some patients declined the invitation (n=358), and 1042 participants consented. Patients with undetectable GP records were excluded (n=188), leaving 851 participants at baseline, phase 1 interview (online supplemental figure S2). For the present secondary analysis, participants were eligible if they completed two non-consecutive 24-hour dietary recalls at baseline; participants with only one 24-hour recall were therefore excluded from this analytic sample (n=2). Thus, 800 participants with two non-consecutive 24-hour recalls were included. Because the study uses the entire Newcastle 85+ cohort available for the dietary substudy, a conventional a priori power or sample-size calculation was not applicable. Further details on the interviews are described elsewhere³⁷ and summarised in online supplemental figure S2.

Baseline and outcome measures

Baseline measurements were obtained by an interview (phase 1), with a detailed health assessment of each participant conducted by a trained research nurse. Core sociodemographic and health variables collected at baseline and used as covariates in this analysis were sex (male/female), educational attainment (coded as 0–9, 10–11, or above 12 years of full-time education), socioeconomic position (based on the three-category National Statistics Socio-Economic Classification (NS-SEC-3: managerial/professional; intermediate; routine/manual)), physical activity (low (score 0–1)/moderate (score 2–6)/high (score 7–18)), medication burden: total prescribed drugs abstracted from GP record and grouped into 0–2, 3–5, or ≥6 medications, psychological status: 15-item Geriatric Depression Scale (GDS-15), categorised as none (0–4), mild (5–9), severe (≥10), residential status: living in the community versus institutionalised (care-home or long-stay ward), smoking (never, former, current) and alcohol use (current drinker vs non-drinker) from interview, multimorbidity: disease count (0–1, 2, ≥3 clinical-diagnosed conditions) compiled from GP record review and anthropometry (weight measured using a digital scale; standing height for body mass index (BMI) calculation and waist to hip ratio and waist circumference using a non-stretch tape at the mid-axillary line and widest point of the buttock) BMI categories followed WHO cut-points: <18.5 kg/m² (underweight), 18.5–24.9 (normal), 25.0–29.9 (overweight), ≥30.0 (obese).

Dietary data collection

Diet was assessed by interviewer administered 24-hour multiple pass recall, collected 1 week apart on non-consecutive days. To address measurement bias in dietary assessment, all 24-hour recalls were conducted by trained research nurses. Additionally, participants recorded their food intake on the previous day including the type, quantity and eating occasion, the trained research nurse used Photographic Atlas of Food Portion Sizes to help estimate portion sizes of participant’s dietary intake. Dietary intake data were coded at the individual food-code level using the McCance and Widdowson food composition tables, which include ingredient-level definitions for a substantial proportion of composite and ‘homemade’ food items, thereby enabling Nova classification based on documented ingredients rather than assumptions where such information was available.³⁸ Data were entered into a Microsoft Access-based database, and each food has a distinct food code with over 2000 food codes based on the McCance and Widdowson food composition tables.³⁸ Food codes were then grouped into 118 distinct food groups developed by the Human Nutrition Research Centre at Newcastle University. Each food represented the average weight measured in grams of food consumed by participants across both 24-hour recalls.³⁹ Only baseline (phase 1) contained dietary intake; no further recalls were obtained at follow-up phases.³⁶

Categorisation of dietary intake using Nova

To enhance the accuracy and consistency of food classification, this study adopts the protocol by the creators of the Nova system for categorising foods into one of its four groups.³³ A structured three-step approach was employed to improve both efficiency and transparency of classification, thereby producing more reliable estimates of UPF intake.³³

In brief, first, all unique food codes (N=2127) were compiled and sorted into single-ingredient (N=670) or multi-ingredient items (N=1456). Composite dishes were reviewed and, where relevant, recipes were disaggregated into constituent ingredients (N=47 recipes) to support ingredient-level Nova assignment. Disaggregation was undertaken using McCance and Widdowson’s Food Composition Tables (sixth edition).³⁸ Cooking yield factors were applied to account for changes in weight due to moisture loss or gain during preparation, as described in the book. Where conversion data were unavailable, estimates were derived from the US Department of Agriculture (USDA) Foods Database.⁴⁰ To enable comparison across recipes, all ingredient weights were standardised to 100 g portions of cooked food. Recipes directly reported by the cohort were prioritised for full disaggregation. Recipes from external sources (eg, BBC Good Food, Mary Berry)^{41 42} were used to aid interpretation of likely ingredients; however, these sources were used solely to inform classification decisions rather than to replace reported intake data. A step-by-step visual summary of this classification process is presented in figure 1.

Figure 1. The classification of the Newcastle 85+dietary data strategy using Nova food classification³³.

Second, unambiguous foods were directly assigned to Nova groups 1–4 based on processing criteria, while industrially produced items (eg, clearly identifiable branded products and commercially packaged items) were classified using brand and ingredient list information. Third, items that could not be classified with certainty from the available description (eg, where it was unclear whether an item was homemade vs commercially manufactured) were flagged as uncertain foods. Across the full dietary dataset, 71 food codes (3.3% of all codes) fell into this category and were the only items treated as uncertain by the team. All Nova coding was performed initially by FS and subsequently reviewed by AWW and AMF. Any disagreements were resolved by TRH through discussion and consensus. These 56 flagged items were then reclassified to their most plausible alternative Nova group for sensitivity analysis, with UPF energy contributions and exposure thresholds recalculated to evaluate the impact of potential classification uncertainty.

Statistical analysis

All analyses were conducted in R V.4.3.2.⁴³ Baseline characteristics (sex, anthropometry, education, socioeconomic factors, physical activity, medication, geriatric depression scale, residential status, smoking, alcohol, multimorbidity) were summarised using gtsummary package (V.2.2.0). Continuous variables were reported as mean±SD, while categorical variables were presented as counts and percentages. Additionally, Pearson χ², Fisher’s exact or Kruskal-Wallis were used to compare UPF tertiles.

The primary exposure variable was the percentage of total energy intake (%TE) of UPF, derived from Nova group 4 food items, averaged across two non-consecutive 24-hour recalls. Participants were ranked by this proportion and categorised into tertiles using dense ranking to avoid ties (T1 <49.9%; T2 49.9–62.7%; T3 >62.7%). These tertiles were entered as categorical independent variables in linear regression models, with type III sums of squares estimated using the car R package (V.3.1–2) and adjusted marginal means were obtained using the emmeans R package (V.1.10.1).

Dietary intake variables

A total of 33 dietary outcomes were analysed. These included energy and macronutrient intakes: total energy in calories (kcal), carbohydrate, protein, total fat, saturated fat, added sugars (all expressed both as a % of energy and in g/day) and fibre. Micronutrients included riboflavin, folate, vitamins A (retinol and β-carotene), B₆, B₁₂, C, D, E, thiamin, niacin, biotin and minerals included calcium (natural and fortified), phosphorus, magnesium, potassium, iron, zinc, selenium and sodium (online supplemental table S1).

Regression modelling strategy

UPF intake, expressed as tertiles of %TE, was treated as the primary independent (exposure) variable in all models, while dietary intakes (energy, macronutrients and micronutrients) were specified as dependent (outcome) variables.

For every dietary outcome, three regression models were fitted, each controlling for an increasing set of potential confounders: the unadjusted model (online supplemental table S2), and the fully adjusted model adjusting for sex, educational attainment and socioeconomic factors (a three-level socioeconomic classification).

These potential confounders were included as they could possibly be correlated with UPF food intake and health outcomes.^{44 45} The validity of our findings could be strengthened when adjusting our models with these variables.

For each model, omnibus p values were derived using type III sums of squares (car::Anova). Adjusted marginal means (EMMeans) were completed for each tertile using emmeans, and Bonferroni-corrected pairwise comparisons (T1 vs T2, T1 vs T3, T2 vs T3) were conducted. Primary interpretations are based on the fully adjusted model, results from the unadjusted model are provided in online supplemental table S2.

Sensitivity analysis

To test robustness to food-code misclassification, all items flagged as uncertain were reassigned to their plausible alternative Nova group; percentage of UPF was recomputed, and the fully adjusted models were rerun.

Patient and public involvement

Patients or the public were not involved in the design, conduct, reporting or dissemination plans of our research.

Results

A total of 800 participants (62% female) were included for complete-case analysis (table 1). A clear gradient in UPF intake was demonstrated. Mean UPF energy contribution rose from 40.6% (±10%) in the lowest tertile (T1) to 55.5% (±12%) in the middle (T2) and 71.1% (±9%) in the highest (T3) (p<0.001) (table 1). In descriptive analyses, despite large differences in the proportion of energy derived from UPF, there were no statistically significant differences across UPF tertiles in sex distribution, BMI, waist circumference, educational attainment, socioeconomic classification, smoking status, alcohol consumption, medication burden, depressive symptoms or institutionalisation (table 1). Multimorbidity was the only health parameter that differed across tertiles of UPF. The proportion of participants with three or more diagnosed conditions increased from 35% in T1 to 41% in T2 and 44% in T3 (p=0.031) (table 1).

Table 1. Baseline sociodemographic, anthropometric and lifestyle characteristics of the Newcastle 85+ cohort across tertiles of dietary energy derived from ultra-processed foods.

Characteristic¹	Low UPF (n=254)	Medium UPF (n=254)	High UPF (n=292)	Overall (n=800)	P value²
% energy from UPF, mean (SD)	40.6 (7.2)	55.5 (3.6)	71.1 (7.1)	56.0 (14.0)
Sex, n (%)					0.30
Male	106 (42)	95 (37)	104 (36)	305 (38)
Female	148 (58)	159 (63)	188 (64)	495 (62)
Weight (kg), mean (SD)	64 (13)	64 (13)	65 (13)	64 (13)	0.60
BMI, n (%)					0.20
Underweight	14 (6.5)	14 (6.2)	19 (7.1)	47 (6.6)
Normal	124 (57)	115 (51)	125 (47)	364 (51)
Overweight	64 (30)	76 (34)	91 (34)	231 (33)
Obese	14 (6.5)	21 (9.3)	33 (12)	68 (9.6)
Waist circumference (cm), mean (SD)	91 (12)	90 (12)	92 (12)	91 (12)	0.40
Waist–hip ratio, mean (SD)	0.88 (0.08)	0.88 (0.07)	0.88 (0.09)	0.88 (0.08)	0.60
Education (years), n (%)					0.60
0–9	151 (61)	167 (67)	187 (65)	505 (64)
10–11	64 (26)	58 (23)	64 (22)	186 (24)
≥12	34 (14)	25 (10)	37 (13)	96 (12)
NS-SEC, n (%)					0.40
Managerial/professional	90 (37)	75 (31)	95 (34)	260 (34)
Intermediate	34 (14)	42 (17)	34 (12)	110 (14)
Routine/manual	118 (49)	125 (52)	150 (54)	393 (52)
Physical activity, n (%)					0.068
Low	62 (25)	53 (21)	62 (21)	177 (22)
Medium	94 (37)	108 (43)	144 (49)	346 (43)
High	96 (38)	91 (36)	86 (29)	273 (34)
Medications, n (%)					0.60
0–2	48 (19)	36 (14)	49 (17)	133 (17)
3–5	69 (27)	76 (30)	76 (26)	221 (28)
≥6	136 (54)	140 (56)	167 (57)	443 (56)
Geriatric Depression Scale, n (%)					0.50
None	188 (81)	184 (78)	214 (78)	586 (79)
Mild	28 (12)	28 (12)	42 (15)	98 (13)
Severe	17 (7.3)	24 (10)	19 (6.9)	60 (8.1)
Institutionalised, n (%)	24 (9.4)	26 (10)	22 (7.5)	72 (9.0)	0.50
Smoking, n (%)					0.40
Never	82 (32)	96 (38)	104 (36)	282 (35)
Current	20 (7.9)	11 (4.3)	15 (5.1)	46 (5.8)
Former	152 (60)	147 (58)	173 (59)	472 (59)
Alcohol, n (%)					0.15
Non-drinker	88 (35)	98 (39)	125 (43)	311 (39)
Drinker	166 (65)	156 (61)	167 (57)	489 (61)
Disease count, n (%)					0.031
0–1	92 (37)	70 (28)	71 (24)	233 (29)
2	70 (28)	77 (31)	91 (31)	238 (30)
≥3	88 (35)	103 (41)	129 (44)	320 (40)

1n (%); mean (SD); 2Pearson’s Chi-squared test; Kruskal-Wallis rank sum test; Fisher’s exact test.

BMI, body mass index; NS-SEC, The National Statistic Socioeconomic Classification; UPF, ultra-processed foods.

UPF (Nova 4) were the dominant source of dietary energy, supplying 950±375 kcal/day (56.3% of total energy), whereas unprocessed or minimally processed foods (Nova 1) provided 466±245 kcal/day (27.6%) (figure 2; online supplemental table S1). Processed culinary ingredients (Nova 2) and processed foods (Nova 3) contributed 7% and 9% of total energy intake, respectively (figure 2; online supplemental table S1).

Figure 2. Contribution of Nova groups to total energy intake. Bar chart showing mean daily energy intake (kcal) by Nova food‐processing group (1–4) among 800 participants aged≥85 years (mean±SD). Nova 4 (ultra-processed foods) contributed 56% of total energy, Nova 1 (unprocessed/minimally processed) 27%, Nova 2 (processed culinary ingredients) 7% and Nova 3 (processed foods) 9%. Data are derived from two non-consecutive 24-hour recalls.

Carbohydrate energy was predominantly derived from Nova 4 items (≈65%) (figure 3), with white bread (16.4%), wholemeal bread (10.7%) and cakes (10.4%) as the top contributors (figure 3; online supplemental table S1). Nova 1 contributed just over 27% of carbohydrate energy, chiefly from fresh fruit (22.7%) and potato (19.4%) (online supplemental table S1). Added (free) sugars were overwhelmingly sourced from Nova 4 (≈85%), primarily other cordials and squashes (24.4%), chocolate (16.0%) and cakes (13.0%) (online supplemental table S1). Additionally, dietary fibre intake exhibited a bimodal distribution: Nova 4 contributed 51% of total fibre, primarily from wholemeal and white breads and fibre-enriched breakfast cereals, while Nova 1 contributed 43% (fruits, vegetables, pulses, oats) (online supplemental table S1).

Figure 3. Macronutrient energy contributions by Nova group. Stacked bar chart displaying the percentage of energy derived from carbohydrates, proteins, total fat and saturated fat for each Nova group. Carbohydrate energy from Nova 4 was ≈65%, with proteins and saturated fats predominantly from Nova 1 and Nova 2, respectively. Values represent group means; sample size n=800. Online supplemental table S1 provides detailed means±SD.

Protein energy was primarily supplied by Nova 1 (54%; 24.8±8.9% kcal/day), with Nova 4 accounting for 26% (12.0±3.4% kcal/day) (figure 3; online supplemental table S1). Total fat intake was largely attributable to processed culinary ingredients (Nova 2; 39%), followed by Nova 4 (20%), while saturated fat originated mainly from butter and hard margarines in Nova 2 (43%), cheese in Nova 3 (25%) and UPF baked goods in Nova 4 (15%) (figure 3; online supplemental table S1).

Nova 1 food items were the main sources of β-carotene (79%), vitamin C (81%) and riboflavin (50%) (figure 4; online supplemental table S1). Vitamins D and B₁₂ were supplied approximately one-third each by Nova 3 and Nova 4, and contributed across Nova 1, Nova 3 and Nova 4 (figure 4; online supplemental table S1). Minerals and vitamin E were chiefly derived from Nova 4: selenium (56%), iron (59%), phosphorus (44%) and vitamin E (61%), reflecting enrichment of staple items such as white bread, wholemeal bread/rolls, cakes and fortified cereals (figure 5; online supplemental table S1). Zinc, calcium and folate intakes were more evenly distributed between Nova 1 and Nova 4 (≈45–51% vs ≈41%) (figure 5; online supplemental table S1).

Figure 4. Vitamin contributions by Nova group. Stacked bar chart illustrating the proportion (%) of total intake of β-carotene, vitamin C, riboflavin, vitamin D and vitamin B₁₂ sourced from each NOVA group. Nova 1 supplies the majority of β-carotene, vitamin C and riboflavin, whereas vitamin D and B₁₂ are more evenly distributed across Nova 1–4. Data based on two 24-hour recalls; n=800; see online supplemental table S1 for full values.

Figure 5. Mineral and vitamin E contributions by Nova group. Stacked bar chart showing the percentage contributions of selenium, iron, phosphorus, zinc, calcium and vitamin E by Nova group. Nova 4 is the main source of selenium, iron, phosphorus and vitamin E, while zinc and calcium are more evenly distributed between Nova 1 and Nova 4. Values represent mean contributions; n=800; detailed data in online supplemental file table S1.

In analyses, total energy intake exhibited a non-significant U-shaped pattern across UPF tertiles (T1: 1740 kcal/day; T2: 1749 kcal/day; T3: 1727 kcal/day; all p>0.05). Carbohydrate‐derived energy increased linearly from 45.6% to 50.5% across tertiles (p<0.001), while protein declined from 16.1% to 14.5% (p<0.001) and saturated fat from 15.1% to 12.2% (p<0.001). Micronutrient intakes of folate, vitamin C, potassium, magnesium, zinc and vitamin B₁₂ were significantly lower in the highest UPF group (all p<0.001), whereas fortified calcium increased from 6.3 to 15.4 mg/day (p<0.001). No significant differences were observed for fibre, iron, sodium or total calcium (all p>0.05).

After adjustment for sex, education and socioeconomic status, mean daily energy intake did not differ significantly across UPF tertiles (T1: 1759.5 kcal/day; T2: 1755.5 kcal/day; T3: 1740.0 kcal/day; all p>0.6). In the fully adjusted analyses, the energy share from carbohydrates increased across UPF tertiles, rising from 45.6% in T1 to 50.5% in T3 (p<0.001), with added sugars doubling from 2.2% to 3.4% of total energy (p<0.001). Energy contributions from protein and saturated fat declined from 16.1% to 14.5% and from 15.1% to 12.2%, respectively (both p<0.001), and overall fat intake was significantly lower in T3 compared with T1 (p=0.016). Mean vitamin C intake fell from 94.0 mg/day to 59.9 mg/day (p<0.001), accompanied by significant reductions in folate, potassium, magnesium, zinc and vitamin B₁₂ (all p<0.001). Calcium derived from fortified sources increased from 6.3 mg/day to 15.4 mg/day across tertiles (p<0.001). No significant differences were observed for non‐starch polysaccharide fibre, total calcium, iron, thiamine, riboflavin, vitamin D₃ or sodium (all p>0.05) (table 2).

Table 2. Model-adjusted means ± standard error energy and nutrient intakes by tertiles of ultra-processed foods in the Newcastle 85+ Study’s.

Nutrient	First tertile (≤ 49.91) mean±SE (95% CI)	Second tertile (≥ 49.91 and ≤ 62.73) mean±SE (95% CI)	Third tertile (≥ 62.73) mean±SE (95% CI)	1 vs 2 Δ (SE, t-ratio)	1 vs 3 Δ (SE, t-ratio)	2 vs 3 Δ (SE, t-ratio)
Energy intake (kcal)	1759.5±38.2 (1684.6–1834.4)a	1755.5±38.6 (1679.8–1831.2)a	1740.0±37.0 (1667.3–1812.7)a	4.06 (42.12, 0.10)	19.55 (40.70, 0.48)	15.48 (40.66, 0.38)
Energy intake (kJ)	7380.9±159.9 (7067.0–7694.8)a	7367.1±161.6 (7049.8–7684.3)a	7304.2±155.1 (6999.7–7608.7)a	13.86 (176.50, 0.08)	76.69 (170.56, 0.45)	62.83 (170.38, 0.37)
Carbohydrates (%)	45.6±0.6 (44.4–46.7)a	48.4±0.6 (47.2–49.5)b	50.5±0.6 (49.4–51.6)c	−2.83 (0.64, −4.40)*	−4.96 (0.62, −7.99)*	−2.13 (0.62, −3.44)*
Added sugars (%)	2.2±0.2 (1.7–2.7)a	2.7±0.2 (2.2–3.2)a	3.4±0.2 (2.9–3.8)b	−0.50 (0.27, −1.86)	−1.17 (0.26, −4.48)*	−0.67 (0.26, −2.56)*
Protein (%)	16.1±0.3 (15.5–16.6)a	15.6±0.3 (15.0–16.1)a	14.5±0.3 (13.9–15.0)b	0.48 (0.31, 1.53)	1.60 (0.30, 5.32)*	1.12 (0.30, 3.74)*
Total fat (%)	37.1±0.6 (36.0–38.2)a	35.7±0.6 (34.6–36.8)ab	35.4±0.5 (34.4–36.5)b	1.35 (0.62, 2.19)	1.66 (0.60, 2.79)*	0.31 (0.60, 0.52)
Saturated fat (%)	15.1±0.4 (14.4–15.8)a	13.6±0.4 (12.8–14.3)b	12.2±0.4 (11.5–12.9)c	1.53 (0.40, 3.82)*	2.87 (0.39, 7.42)*	1.34 (0.39, 3.47)*
β-carotene (µg)	234.1±52.0 (132.1–336.1)a	258.0±52.5 (154.9–361.1)a	208.2±50.4 (109.3–307.2)a	−23.86 (57.37, −0.42)	25.88 (55.44, 0.47)	49.74 (55.38, 0.90)
Biotin (µg)	32.4±0.9 (30.6–34.2)a	32.0±0.9 (30.2–33.9)a	28.8±0.9 (27.0–30.6)b	0.40 (1.03, 0.39)	3.63 (0.99, 3.67)*	3.24 (0.99, 3.27)*
Calcium, T (mg)	759.2±21.6 (716.9–801.6)a	780.7±21.8 (737.9–823.5)a	751.3±20.9 (710.2–792.4)a	−21.46 (23.81, −0.90)	7.91 (23.01, 0.34)	29.37 (22.98, 1.28)
Fibre (g)	11.7±0.4 (10.9–12.5)a	11.7±0.4 (10.9–12.5)a	10.8±0.4 (10.0–11.6)a	0.01 (0.45, 0.03)	0.91 (0.43, 2.10)	0.90 (0.43, 2.08)
Iron (mg)	10.2±0.3 (9.6–10.8)a	10.2±0.3 (9.6–10.8)a	9.8±0.3 (9.2–10.4)a	−0.04 (0.34, −0.13)	0.39 (0.33, 1.19)	0.43 (0.33, 1.33)
Folate (µg)	236.2±6.4 (223.7–248.7)a	230.5±6.5 (217.9–243.2)a	208.2±6.2 (196.0–220.4)b	5.66 (7.05, 0.80)	27.99 (6.81, 4.11)*	22.32 (6.80, 3.28)*
Calcium, FF (mg)	6.3±1.3 (3.7–8.9)a	9.9±1.3 (7.2–12.5)b	15.4±1.3 (12.9–17.9)c	−3.58 (1.46, −2.44)*	−9.08 (1.42, −6.42)*	−5.51 (1.41, −3.90)*
Potassium (mg)	2786.3±63.9 (2660.8–2911.8)a	2640.8±64.6 (2513.9–2767.6)a	2455.9±62.0 (2334.1–2577.6)b	145.56 (70.56, 2.06)	330.45 (68.19, 4.85)*	184.89 (68.12, 2.71)*
Magnesium (mg)	243.5±5.9 (231.9–255.1)a	239.1±6.0 (227.4–250.8)a	223.9±5.7 (212.7–235.1)b	4.38 (6.50, 0.67)	19.62 (6.28, 3.12)*	15.24 (6.28, 2.43)*
Sodium (mg)	2609.2±80.8 (2450.6–2767.7)a	2701.5±81.6 (2541.3–2861.8)a	2753.2±78.4 (2599.4–2907.0)a	−92.37 (89.17, −1.04)	−144.03 (86.17, −1.67)	−51.67 (86.08, −0.60)
Niacin (mg)	16.9±0.5 (15.8–17.9)a	16.7±0.5 (15.6–17.7)a	15.7±0.5 (14.7–16.8)a	0.20 (0.58, 0.34)	1.12 (0.56, 1.99)	0.92 (0.56, 1.64)
Calcium, N (mg)	158.6±12.2 (134.6–182.6)a	148.0±12.3 (123.7–172.2)ab	116.9±11.8 (93.6–140.1)b	10.63 (13.48, 0.79)	41.73 (13.02, 3.20)*	31.10 (13.01, 2.39)
Phosphorus (mg)	1164.7±26.6 (1112.4–1216.9)a	1147.8±26.9 (1095.0–1200.6)ab	1086.6±25.8 (1035.9–1137.3)b	16.85 (29.38, 0.57)	78.05 (28.39, 2.75)*	61.20 (28.36, 2.16)
Retinol (µg)	1160.4±194.2 (779.3–1541.6)a	918.5±196.2 (533.3–1303.7)a	393.9±188.3 (24.2–763.7)b	274.34 (211.17, 1.30)	821.01 (204.18, 4.02)*	546.67 (204.18, 2.68)*
Riboflavin (mg)	1.7±0.1 (1.6–1.8)a	1.7±0.1 (1.6–1.8)a	1.5±0.1 (1.4–1.6)a	0.00 (0.06, 0.03)	0.14 (0.06, 2.38)	0.14 (0.06, 2.35)
Selenium (µg)	49.4±2.3 (44.9–53.9)a	48.0±2.3 (43.5–52.5)ab	43.4±2.2 (39.1–47.8)b	1.42 (2.52, 0.57)	5.99 (2.43, 2.46)*	4.56 (2.43, 1.88)
Thiamine (mg)	1.3±0.1 (1.2–1.5)a	1.5±0.1 (1.3–1.6)a	1.4±0.1 (1.3–1.5)a	−0.13 (0.07, −1.72)	−0.05 (0.07, −0.67)	0.08 (0.07, 1.11)
Vitamin B₁₂ (µg)	6.1±0.6 (4.8–7.3)a	5.0±0.6 (3.8–6.3)a	3.4±0.6 (2.2–4.6)b	1.01 (0.69, 1.46)	2.62 (0.67, 3.91)*	1.61 (0.67, 2.41)*
Vitamin B₆ (mg)	3.2±0.9 (1.5–4.8)a	3.8±0.9 (2.1–5.4)a	2.3±0.8 (0.7–4.0)a	−0.58 (0.94, −0.62)	0.83 (0.91, 0.91)	1.41 (0.91, 1.55)
Vitamin C (mg)	94.0±5.3 (83.7–104.4)a	75.9±5.3 (65.5–86.4)b	59.9±5.1 (49.8–70.0)c	18.09 (5.84, 3.10)*	34.13 (5.64, 6.05)*	16.04 (5.63, 2.85)*
Vitamin D (µg)	3.2±0.2 (2.8–3.7)a	2.9±0.2 (2.5–3.4)a	2.7±0.2 (2.2–3.1)a	0.30 (0.26, 1.14)	0.57 (0.26, 2.23)	0.27 (0.26, 1.06)
Vitamin E (mg)	5.4±0.3 (4.8–6.1)a	5.6±0.3 (4.9–6.3)a	6.8±0.3 (6.1–7.4)b	−0.18 (0.37, −0.50)	−1.35 (0.36, −3.77)*	−1.16 (0.36, −3.26)*
Water (g)	2068.5±44.0 (1982.1–2154.8)a	1946.5±44.5 (1859.3–2033.8)b	1863.8±42.7 (1780.1–1947.6)b	121.94 (48.56, 2.51)*	204.64 (46.92, 4.36)*	82.71 (46.87, 1.76)
Zinc (mg)	9.0±0.3 (8.4–9.5)a	8.4±0.3 (7.9–8.9)a	7.5±0.3 (6.9–8.0)b	0.58 (0.30, 1.91)	1.53 (0.29, 5.23)*	0.95 (0.29, 3.26)*

^*Percentage of total energy intake. Values are presented as intake per day, model-adjusted means±SE error (SE) with 95% CIs. Superscript letters (a, b, c) indicate post hoc pairwise comparisons between ultra-processd food tertiles; values sharing at least one letter are not significantly different, whereas values with no shared letter are significantly different at p<0.05. Pairwise differences are reported as adjusted mean differences (Δ), followed by SE and t-ratio in parentheses (Δ [SE, t-ratio]). Asterisks (*) denote statistical significance at p<0.05. All models were adjusted for sex, socioeconomic status (National Statistics Socio-economic Classification, NS-SEC), and education level.

FF, fortified foods; N, natural; T, total.

In the sensitivity analysis, 71 of 2127 food items (3.3%) with ambiguous Nova classifications were reassigned to their most probable groups, resulting in an increase in the mean UPF energy contribution from 54.8% to 56.5%, alongside decreases in energy from Nova 1 (–1.3%), Nova 2 (–0.2%) and Nova 3 (–0.6%) and an increase for Nova 4 (+2.1%) (online supplemental table S3). The UPF tertile thresholds also shifted downward from 49.9% and 62.7% to 43.6% and 56.6%, respectively (online supplemental table S3), although only 12% of participants were reassigned to different tertiles and the overall demographic and dietary profiles remained unchanged.

Discussion

In this population-based study of very old people aged 85 and over, higher consumption of UPF, as a %TE, was characterised by shifts in macronutrient sources and reductions in several essential micronutrients. Total energy intake did not differ across UPF tertiles after adjustment for key demographic and socioeconomic factors. This finding is consistent with prior UK evidence showing that associations between UPF intake and energy or free sugar consumption are attenuated or absent in older age groups (≥65).⁴⁶ Nevertheless, macronutrient composition and micronutrient density were significantly compromised in participants with higher UPF intakes. Specifically, high UPF intake was associated with higher energy intake from carbohydrates, added sugars and lower energy intake from protein and saturated fat. Intakes of several essential micronutrients, including vitamin C, folate, magnesium, potassium, zinc and vitamin B₁₂, were significantly lower in the highest UPF tertile, while intakes of calcium (from fortified sources) and vitamin E were higher with increasing UPF, reflecting common industrial fortification practices. These associations remained robust after controlling for key demographic and socioeconomic factors and were consistent in sensitivity analyses accounting for classification uncertainty.

The absence of marked sociodemographic, anthropometric and health-related differences across UPF tertiles suggests that high UPF consumption in this very old cohort was not strongly socially or behaviourally patterned, contrasting with findings from younger populations.1247,49 This lack of demographic and socioeconomic patterning is important for interpreting the dietary findings, as it suggests that the observed differences in macronutrient composition and micronutrient density are unlikely to be driven by underlying social or lifestyle confounding. Still, the higher prevalence of multimorbidity observed among participants with higher UPF intake should be interpreted cautiously, as this finding was derived from baseline descriptive characteristics rather than from primary analytical models and does not imply a causal relationship. In very old people, greater disease burden may plausibly influence dietary choices through reduced mobility, functional limitations or increased reliance on convenient and ready-made foods, including UPF.50,52 Conversely, long-term exposure to poorer-quality diets earlier in life may also contribute to later-life morbidity.^{53 54} However, the temporal direction of this association cannot be determined within the present analysis. Moreover, this cautious interpretation is consistent with evidence from systematic reviews and meta-analyses linking UPF intake to adverse health outcomes, including all-cause mortality, predominantly in mixed-age populations (aged 18 and above), where long-term exposure and temporal sequencing can be more clearly examined.⁵⁵

Population ageing in the UK is progressing rapidly, and very old people (aged 85 and over) represent a priority group for understanding the role of dietary patterns in supporting healthy ageing. Applying the Nova classification system, UPF accounted for 54% of total energy intake among Newcastle 85+ Study participants, exceeding the contribution from unprocessed or minimally processed foods (27%). In this cohort, UPF intake was largely derived from solid staple foods such as breads and fibre-enriched breakfast cereals and contributed nearly half of the total dietary fibre, while Nova group 1 items were the main contributors to protein and plant-based fibre. Although high-UPF diets typically displace nutrient-dense foods,⁵⁶ our data highlight a dual role for UPF in very old people; in addition to contributing excess added sugars, they also supplied higher levels of certain fortified nutrients, notably calcium (6.3 mg/day to 15.4 mg/day) and vitamin E (5.4 mg/day to 6.8 mg/day).

Importantly, this interpretation should be considered, considering emerging evidence that associations between UPF intake and health outcomes could vary by physical form. Large prospective cohort studies have shown that liquid UPF, particularly SSB, are consistently associated with higher all-cause mortality, whereas solid UPF displays null, inverse or heterogeneous associations after multivariable adjustment.²⁵ This distinction is especially relevant in the Newcastle 85+ cohort, in which UPF intake was largely derived from solid foods such as breads and breakfast cereals rather than liquid UPF, providing important context for the lack of association with total energy intake and for the observed contribution of fortified nutrients.

Notably, the increase observed for micronutrients with higher UPF intake was limited to calcium derived from fortified sources, while total calcium intake did not differ across UPF tertiles. This indicates that the observed association reflects industrial fortification rather than differences in ingredient composition. Our food composition data explicitly distinguish between natural and fortified calcium, allowing this distinction to be examined directly; however, for most other micronutrients, the Nova classification and available food composition data do not allow separation of naturally occurring nutrients from fortification. This limitation reflects a known challenge in applying Nova-based classifications to nutrient-specific analyses.⁵⁷

This is especially important for very old people, where physiological limitations (eg, poor dentition, reduced appetite) may increase reliance on convenient, ready-made fortified UPF items.⁵⁰ Still, the concurrent declines in protein, saturated fat and several micronutrients raise concerns. Protein is critical for the preservation of muscle mass and strength in older adults,⁵⁸ and inadequate intake may contribute to the development of frailty and sarcopenia.59,61 However, the relationship with frailty may be bi-directional, as previous studies have shown that frail older adults and those with functional limitations are more likely to rely on ready-made or convenience meals, reflecting barriers to food shopping and preparation rather than dietary preference alone.^{51 62} Additionally, deficiencies in vitamins C and A, folate, magnesium and potassium could exacerbate age-related health risks, including impaired cognition, bone demineralisation and reduced immune function.63,67

Our sensitivity analysis, in which 3.3% of food items were reclassified, increased the mean UPF contribution by 1.7 percentage points and altered tertile cut-offs modestly. Only 12% of participants changed tertile, and most nutrient estimates differed by less than 2% between models. As recommended by Martinez-Steele et al,³³ this demonstrates that our conclusions about the nutrient profile of UPF intake are not unduly driven by classification uncertainties.

Comparison with younger adult cohorts

In the USA, adults aged ≥20 years in the highest quintile of UPF consumption consumed approximately 500 kcal/day more than those in the lowest quintile, along with significantly lower intakes of fibre, potassium, magnesium, zinc, iron, calcium and folate.⁶⁸ Similar nutrient inadequacies have been reported in the UK for adults aged 19–64,⁴ in French adults aged 18–75¹⁴ and in Brazilian adults aged ≥20 years.⁶⁹

However, findings from this cohort of very old people are in contrast to the trends in younger adults. After adjusting for sex, education and socioeconomic status, total energy intake did not differ significantly across UPF consumption tertiles. These findings suggest that the energy-surplus hypothesis used to explain UPF-related obesity in younger populations³⁰ may have limited applicability in advanced age, particularly among very old people, where physiological changes, including reduced appetite, anorexia, dental problems and restricted access to food, might already limit energy intake, highlighting the need for age-specific dietary guidance and research.

Importantly, the observed nutrient profile associated with higher UPF intake in this very old cohort highlights a distinct role for fortification. While higher UPF consumption was accompanied by lower overall micronutrient density for several nutrients, intakes of fortified calcium, vitamin E and biotin increased, most likely reflecting industrial fortification of commonly consumed UPF such as breads and ready-to-eat cereals. This finding suggests that, in advanced age, certain fortified UPF may partially offset micronutrient inadequacies that would otherwise arise from reduced dietary variety or limited intake of unprocessed foods.

These results parallel prior NHANES analyses, which found that removing UPF-sourced fortified cereals from the diet would sharply increase the prevalence of nutrient inadequacies across multiple vitamins and minerals.⁷⁰ Usual intake data suggested that, for US adults aged 19–99, the removal of vitamins and minerals provided by fortified ready-to-eat cereals (classified as UPF) would considerably increase the proportion of the population falling below the estimated average requirement for niacin, iron, thiamine, vitamin A, vitamin B6 and zinc, with little impact on those exceeding the tolerable upper intake levels.⁷⁰ These results imply that many adults, especially very old people, may be unable to meet micronutrient requirements without the support of UPF. Given that older adults aged 63–90 possibly rely on ready-made and shelf-stable foods,⁵⁰ fortified UPF may play a valuable role in maintaining nutritional adequacy where unprocessed alternatives are impractical or inaccessible. This nuanced perspective challenges simplistic narratives around UPF avoidance⁷¹ and suggests that, under certain conditions, selected UPF may act as nutritional safeguards rather than dietary hazards.⁷²

Implications for dietary guidance

UPFs are heterogeneous in nutritional quality. Although high UPF intake is generally linked to poorer nutrient profiles in adults,⁷³ select fortified UPFs, particularly low-sugar breakfast cereals and breads, can supply essential vitamins and minerals without the negative health associations seen for other UPFs. In a multinational cohort study, ultraprocessed breads and cereals were not associated with increased risk of cancer or cardiometabolic multimorbidity (HR 0.97, 95% CI 0.94 to 1.00).²³ Our study shows that the majority of energy in very old people population is from breads and cereals; therefore, for adults aged ≥85 years, dietary guidance should prioritise nutrient-dense whole foods while cautiously incorporating fortified UPFs to maintain intakes of nutrients at risk (vitamin D, B₁₂, calcium).^{74 75} Discretionary UPF with minimal nutritional benefit (eg, SSBs, confectionery) should be limited. Collaboration among healthcare providers, caregivers and industry is essential to develop convenient, fortified UPFs tailored to very old people, balancing ease of consumption, palatability and nutritional adequacy.

Strengths and limitations

Strengths of this study include the use of two non-consecutive 24-hour recalls, strict application of the Nova classification protocol,³³ and comprehensive covariate adjustment for sociodemographic factors. Although repeated recalls improve estimation of usual intake, 24-hour recalls represent short-term snapshots of diet and may not fully reflect habitual dietary patterns.⁷⁶ Our findings were robust to sensitivity analyses addressing classification uncertainty.

However, limitations must be acknowledged. Dietary data were collected in 2005–2006, and since then, the UK food supply has undergone substantial changes due to reformulation, product turnover and evolving fortification practices. Therefore, the absolute level and main contributors of UPF intake in this cohort may not accurately reflect those of very old people today. Self-reporting may be particularly vulnerable to error in very old people,⁷⁷ and the cohort’s recruitment through GP practices may have introduced selection bias. These factors may limit generalisability to contemporary older populations or those with different healthcare access patterns. Additionally, the association between UPF intake and multimorbidity observed in baseline characteristics is descriptive and cross-sectional; therefore, temporality cannot be established, and reverse causation cannot be excluded, particularly in older adults, where disease burden may influence food choice and reliance on convenient foods. Finally, the analyses were based on cross-sectional baseline data and the statistical models were correlational in nature; therefore, the observed associations cannot be interpreted as evidence of causality.

Generalisability

The Newcastle 85+ cohort comprises community-dwelling adults born in 1921 in Northeast England, recruited via GP registers in 2006–2007. As such, our findings may not extend to very old people in other geographic regions, healthcare systems, or birth cohorts with different dietary environments and UPF exposures. Furthermore, changes in food processing and fortification practices since 2006–2007 could limit applicability to present-day diets.

In conclusion, our findings highlighted that in adults aged ≥85 years, high UPF intake is associated with lower intakes of key micronutrients, despite stable energy intake and some benefits from fortification. Given the growing population of the oldest old, future research should further investigate the nutritional implications of UPF consumption in this age group. Further research on UPF intake and contribution to nutritional intake is urgently needed in this age group to confirm these findings, particularly given the drive to reduce intakes of UPF in the population.

Data availability statement

The data that support the findings of this study are from the Newcastle 85+ Study and are not publicly available due to ethical and governance restrictions. Data may be obtained from the Newcastle 85+ Study data custodians upon reasonable request and subject to appropriate approvals.