Association of ultra-processed food consumption with cardiovascular risk factors among patients with type-2 diabetes mellitus

Type-2 diabetes mellitus (T2DM) is highly prevalent and linked to multiple comorbidities, including hypertension, dyslipidemia, and micro- and macrovascular complications. Diet quality and broader lifestyle factors contribute to both T2DM and cardiovascular disease (CVD) risk. The NOVA classification defines ultra-processed foods (UPFs) as multi-ingredient industrial formulations (e.g., soft drinks, packaged snacks, sweets, processed meats, ready-to-eat meals). Prior studies associate higher UPF intake with adverse cardiometabolic outcomes, such as dyslipidemia and obesity, in general populations and children, but evidence is scarce in T2DM populations. Novel lipid- and anthropometry-derived indices (CRI-1/CRI-2, AIP, LAP; ABSI, BRI, AVI) can better capture CVD risk. This study investigates the association between UPF consumption and novel CVD risk factors among T2DM patients, addressing a gap in an understudied high-risk group.

Observational evidence links higher UPF intake to increased risks of diabetes, CVD, hypertension, obesity, and poorer overall diet quality. Prospective studies in older adults and children report that higher UPF consumption is associated with incident dyslipidemia and increases in total and LDL cholesterol. Meta-analyses show consistent associations of UPFs with adverse nutritional profiles and chronic disease risk. Large cohorts (e.g., NutriNet-Santé, UK Biobank) report higher CVD risk per unit increase in UPF energy share. However, few studies specifically assess UPFs versus detailed lipid and novel CVD indices in T2DM populations, motivating the current analysis.

Design and ethics: Cross-sectional study conducted per the Declaration of Helsinki and approved by Tehran University of Medical Sciences Ethics Committee (IR.TUMS.MEDICINE.REC.1400.185). Written informed consent obtained. Setting and participants: 490 adults aged 35–80 with T2DM were recruited from the Endocrine and Metabolism Research Institute’s diabetes research center (May 2021–September 2022). Inclusion/exclusion: Excluded if on insulin; pregnant/breastfeeding; on estrogen therapy; with autoimmune disease, acute GI disorder, acute renal disease, liver cancer; implausible energy intake (<800 or >4200 kcal/day). Measures: Dietary assessment via validated 168-item semi-quantitative FFQ administered by trained nutritionists; household measures converted to grams and analyzed with Nutritionist IV (Iran-modified). UPFs were classified using NOVA; FFQ items considered UPF included biscuits, crackers, cakes, canned foods, burgers, sausages, flavored milks, ice creams, industrial juices, compotes, sauces, jams, soft drinks, sweets, chips, packaged bread, ready-to-eat meals, pizza, etc. UPFs were also grouped as sweetened beverages, sweets, salty snacks, and ultra-processed meats/fast foods. Anthropometry: Weight (SECA scale, 0.1 kg), height (stadiometer, 0.5 cm), waist and hip circumferences (tape). BMI calculated as kg/m². Novel anthropometric indices calculated as previously described: ABSI, BRI, AVI using WC, BMI, height, and hip measures. Biochemistry and derived CVD risk indices: Laboratory values (FBS, LDL-C, HDL-C, total cholesterol [TC], triglycerides [TG]) obtained from clinical tests. Derived indices computed using standard formulae: CRI-1 (TC/HDL-C), CRI-2 (LDL-C/HDL-C), AIP (log[TG/HDL-C]), LAP ((WC−65)×TG for men; (WC−58)×TG for women), and a cholesterol index (LDL−HDL multiplied by TG; modified if TG>400). Other variables: Demographics, medication and supplement use, socioeconomic status (composite 1–10 score), smoking, marital status, education, employment, family size, transport, housing; physical activity via short IPAQ. Major adverse cardiac events (MACE) were considered but not analyzed due to absence of CV death in the sample. Statistical analysis: Participants were categorized into tertiles of daily UPF intake (g/day). Baseline comparisons across tertiles used ANOVA (continuous) and chi-square (categorical). Dietary intakes across tertiles were assessed via ANCOVA (energy-adjusted). Associations between UPF and outcomes were examined using linear regression in two ways: per 20-g increase in UPF intake and by tertiles of UPF intake. Binary logistic regression was used by dichotomizing ABSI, BRI, AVI, AIP and other indices using literature-based cut points (e.g., ABSI 0.08, BRI 5.20, AVI 17.30, AIP 0.11). Models: crude; Model 1 adjusted for energy intake; Model 2 further adjusted for socioeconomic status, age, sex, and smoking (covariates chosen based on baseline differences). Trends across tertiles treated as ordinal. SPSS v26; two-sided p<0.05.

Sample characteristics: Mean age 62.6±9.8 years; 59% female overall. Significant differences across UPF tertiles in age, sex distribution, welfare, education, and smoking (all p<0.001). Medication use for blood sugar (~99%) and lipids (~98%) was common and similar across tertiles. UPF intake levels: Tertile 1: 18.67 g/d (3.76% energy); Tertile 2: 43.81 g/d (7.20%); Tertile 3: 117.37 g/d (12.94%). Overall mean UPF ~59.83 g/d (7.96% energy). Biochemical and anthropometric differences across tertiles (ANOVA): TC higher with greater UPF (p=0.03); CRI-1 p=0.03; CRI-2 p=0.053. Anthropometrics: weight (p<0.001), BMI (p=0.003), WC (p=0.03), ABSI (p=0.03), AVI (p=0.03) differed; BRI and HC did not. Linear regression per 20-g UPF increase (Table 4, fully adjusted Model 2 unless noted): • TC: B=1.214 (SE 0.537), p=0.024; 95% CI 0.159–2.269. • HDL: Crude B=−0.394 (0.138), p=0.004; Model 1 B=−0.371 (0.155), p=0.017; attenuated in Model 2 B=−0.237 (0.154), p=0.125. • No significant associations for BMI, WC, HC, FBS, TG, LDL. Linear regression per 20-g UPF increase for novel indices (Table 5): • Crude only: CRI-1 B=0.032 (0.012), p=0.007; CRI-2 B=0.022 (0.009), p=0.014; AIP B=0.006 (0.003), p=0.034. Associations attenuated after adjustment. Linear regression across UPF tertiles (Table 6, crude): • CRI-1: B=0.126 (0.050), p=0.012 (NS after energy adjustment). • CRI-2: B=0.091 (0.039), p=0.019 (NS after energy adjustment). • AVI: B=0.565 (0.216), p=0.009; energy-adjusted B=0.736 (0.247), p=0.003. • Cholesterol index and AIP showed borderline p-values in crude (p≈0.056 and 0.053 respectively). Overall interpretation: Higher UPF intake is associated with higher TC and lower HDL (partially adjusted), and with adverse CVD risk indices in crude models; tertile analyses revealed significant positive associations particularly with AVI, suggesting a dose–response in some anthropometric CVD risk markers.

The study addresses whether UPF intake is linked to adverse cardiovascular risk among T2DM patients. Findings indicate that greater UPF consumption is associated with higher total cholesterol and lower HDL (attenuating after full adjustment), and higher values of novel CVD risk indices in crude analyses; tertile comparisons reveal significant associations with AVI and crude associations with CRI-1/CRI-2, suggesting a dose–response pattern. These results align with cohorts in non-diabetic populations showing increased CVD risk and dyslipidemia with UPF intake. Despite relatively modest UPF energy shares compared with Western populations, associations with CVD-related indices were detectable in this Iranian T2DM cohort. Lack of strong associations with LDL and TG, and attenuation after multivariable adjustment, may reflect high prevalence of lipid- and glucose-lowering medication use and potential confounding by age, sex, socioeconomic status, and smoking. Mechanistically, UPFs are energy-dense, high in refined carbohydrates and saturated/trans fats, and low in fiber and micronutrients, contributing to insulin resistance, visceral adiposity, systemic inflammation, oxidative stress, gut microbiota alterations, and endothelial dysfunction—all pathways promoting CVD. The divergence between linear and logistic/tertile analyses suggests possible non-linear relationships or sensitivity to distributional assumptions and outliers, underscoring the importance of multiple analytic approaches.

In T2DM patients, higher UPF consumption is associated with an adverse lipid profile (notably higher total cholesterol and lower HDL in partially adjusted models) and with higher values of certain anthropometric and lipid-derived CVD risk indices, particularly when comparing highest versus lowest UPF intake groups. Reducing UPF intake may help mitigate cardiovascular risk in this high-risk population. Future prospective and longitudinal studies are needed to clarify temporality and causality, evaluate dose–response, and assess the impact of medication use and other confounders.

Cross-sectional design precludes causal inference. Many participants were on lipid- and glucose-lowering medications, likely attenuating associations for lipid profile measures. UPF intake was relatively low in this older T2DM cohort, potentially limiting power to detect stronger associations. Dietary intake was assessed by FFQ, which is subject to recall and measurement error. Residual confounding cannot be excluded despite adjustments.