Abstract
Background & Aims:
Higher intake of ultra-processed foods (UPF) has been linked with higher risks of cancer, cardiovascular disease, and diabetes, as well as all-cause mortality. However, studies on UPF and cause-specific mortality remain limited, especially among disadvantaged populations. We aimed to examine associations of UPF intake with all-cause and cause-specific mortality among low-income Americans.
Methods:
In the Southern Community Cohort Study (SCCS), a prospective cohort of mostly low-income Black and White Americans, we included 77,060 participants who completed a food frequency questionnaire (FFQ) at baseline (2002–2009) and had at least 1 year follow-up. All 89 items in the FFQ were categorized using the Nova classification. UPF intake was calculated as % of daily foods intake by weight (grams). Cox regression was used to estimate HR (95% CI) for the association of UPF intake (quartile or per 10% increase) with total and cause-specific mortality (cancer, coronary heart disease [CHD], stroke, and diabetes) after adjusting for sociodemographics, lifestyles, and disease history.
Results:
Of 77,060 participants, 46,175 (59.9%) were women, 49,857 were Black (64.7%), and mean age was 52.4 (SD: 8.8) years at baseline. The mean intake of UPF was 41.0% (SD: 15.7%). UPF intake was inversely associated with Healthy Eating Index and intakes of fiber, minerals, and vitamins but positively associated with intakes of sugars and fats (all PFDR<0.0001). During an average follow-up of 12.2 years, we documented 17,895 total deaths, including 4,267 from cancer, 2,208 from CHD, 867 from stroke, and 997 from diabetes. In the fully adjusted model, higher UPF intake was not associated with all-cause, cancer, CHD, or stroke mortality but showed a significant association with increased diabetes mortality (HR [95% CI] = 1.32 [1.07, 1.62] for the highest versus lowest quartiles [>51.1% vs. <29.3%] and 1.09 [1.04, 1.15] per 10% increase). The adverse UPF-diabetes mortality association was noted regardless of sex, race, income, neighborhood deprivation, lifestyles, and cardiometabolic disease history, while particularly evident in participants with no more than high school education or a history of hypercholesterolemia (HR [95% CI] per 10% increase=1.12 [1.05, 1.18] and 1.14 [1.07, 1.22], respectively; both Pinteraction<0.05).
Conclusions:
Among predominantly low-income Black and White American adults, UPF intake was associated with increased diabetes mortality, especially for individuals with limited education or hypercholesterolemia. Our findings suggest the potential impact of increasing access and intake of un/minimally processed food to replace UPF on reducing diabetes-related mortality among populations facing socioeconomic and health disparities.
Keywords: ultra-processed foods, diabetes mortality, diet quality, prospective cohort study, multi-ethnic populations
Introduction
With the emergence of industrial food system, food consumption patterns have changed dramatically worldwide. Ultra-processed foods (UPF), characterized as palatable and ready-to-eat food with long shelf life, have become dominant in terms of daily energy contribution in some developed countries and are rapidly altering traditional dietary patterns based on freshly prepared meals or minimal-processed food in developing countries [1]. According to a systematic review [2], worldwide consumption of UPF ranges from 10% to 70% of total daily energy intake, with the highest contribution (50–70%) being seen in the United States (US), which was even higher among individuals who are non-Hispanic Black or with a low socioeconomic status (SES) [3].
After multiple physical, biological, and/or chemical processes for industrial production, UPF usually contain low fiber and vitamins but high saturated fat, salt, sugar, and additives, such as dyes, flavor enhancers, emulsifiers, and humectants [1]. Those attributes of UPF have led to concerns about potential harm to human health. Indeed, higher intakes of UPF have been associated with higher risks of cancer[4, 5], cardiovascular disease (CVD) [6, 7], and diabetes [8–11], as well as all-cause [7, 12–15] and CVD mortality [16, 17] in prospective cohort studies conducted in UK, France, Spain, Italy, and the US. However, results on UPF and cause-specific mortality remain limited [18], and findings for CVD mortality have been inconsistent [7, 12, 15–17]. Furthermore, few studies have focused on disadvantaged populations (e.g., racial/ethnic minorities and individuals with low SES), despite their high UPF intake and disproportionately high morbidity and mortality burdens [3, 19, 20].
The Southern Community Cohort Study (SCCS) is a large, multi-racial/ethnic, prospective cohort study of predominantly low-income Black and White Americans [21]. Using data from the SCCS, we investigated the associations of UPF intake with all-cause and major cause-specific mortality (i.e., cancer, coronary heart disease [CHD], stroke, and diabetes). We hypothesized that UPF intake was associated with increased mortality among SCCS participants, particularly those with low SES (including education, income, and neighborhood deprivation).
Materials & Methods
Study population
The SCCS is a prospective cohort study; its study design and procedures have been described previously [21]. In brief, 84,735 Americans aged 40–79 years were enrolled from 12 southeastern states from March 2002 to September 2009. There were two types of enrollment strategies: 86% of the participants were enrolled in person from community health centers (CHC) that serve primarily low-income, uninsured populations, while the remaining 14% were enrolled by mail from the general population in the same areas. Structured questionnaires were used to collect information on sociodemographics, diet, lifestyles, anthropometrics, and medical history. About two-thirds of the SCCS participants were self-reported Black/African Americans, and over 50% had annual household income <$15,000, making SCCS a unique large population-based study examining race and SES-related health disparities in the US. Institutional Review Boards of Vanderbilt University Medical Center and Meharry Medical College approved the SCCS study, and all participants provided written informed consent.
In this analysis, we excluded participants who left more than ten items blank in the food frequency questionnaire (FFQ) or reported total energy intake <600 or >8,000 kcal/day (N=6,756) or died within the first year after enrollment (N=919). A total of 77,060 participants were included in the main analyses.
Diet assessment
Diet information was collected at baseline using an 89-item FFQ designed to capture food commonly consumed in the southeastern US [22]. The frequency of consumption was asked in nine categories ranging from “never” to “2 or more times per day” based on usual intake over the past 12 months. To estimate the intake amount of each food item, we applied the race- and sex-specific portion sizes from the NHANES and US Department of Agriculture (USDA) Continuing Survey of Food Intakes by Individuals, restricted to participants aged 30–84 years, reported as non-Hispanic White or Black, and living in South census regions [23, 24]. The intake of each food (grams/day) was estimated by multiplying the average portion size and the frequency. Total energy and nutrient intakes were estimated using the USDA Food Composition Databases. The validity of the FFQ was assessed by the agreement between assigned quintiles of nutrient intake and 24-hour dietary recalls; the Kappa values ranged from 0.82 to 0.96 for macronutrients and from 0.73 to 0.95 for micronutrients [23]. Healthy Eating Index (HEI-2010) was calculated by linking the FFQ data with MyPyramid Equivalents Database and scoring (0 to 100 points) based on the 2010 Dietary Guidelines for Americans [25, 26].
All items in the FFQ were categorized into four groups (un/minimally processed, processed culinary ingredient, processed foods, and ultra-processed foods) based on the Nova, a valid tool for food classification according to the extent and purpose of food processing [1] (see details in Supplemental Table 1). UPF comprise foods and drinks that are industrial formulations typically with five or more ingredients and often contain additives with cosmetic function (e.g., flavorings, colorings, and emulsifiers) [1]. Examples include carbonated and sugar-sweetened drinks, sweet or savory packaged snacks, mass-produced packaged bread, cookies and cakes, breakfast cereals, energy bars, and many ready-to-heat products such as frozen pizza, pasta, pies, chicken nuggets, fish sticks, sausages, burgers, and instant soups, noodles, and desserts [1]. We classified 46 food items as UPF, of which 32 food items were assigned 100% weight and 14 were assigned 20%, 50%, or 70% correction weights considering the chance a food/food group is likely to be UPF in the SCCS population. These UPF were further categorized into six subgroups (i.e., grains and fried potatoes; protein food; condiments, fats, and oils; sweets and snacks; beverages; mixed dishes) based on What We Eat in America (WWEIA) Food categories provided by the USDA [27].
Mortality ascertainment
Vital status and date and underlying cause of death (International Classification of Diseases 9th and 10th Revision codes) were ascertained via annual linkage to the National Death Index and/or Social Security Administration mortality files up to Dec 31, 2019. The outcomes of interest were total deaths and deaths from cancer (ICD-9 codes 140–208 and ICD-10 codes C00-C97), CHD (410–414 and I20–I25), stroke (430–438 and I60–I69), and diabetes (249–250 and E10-E14).
Statistical analysis
The primary exposure was the proportion of total UPF in the diet, as % of weight (grams) per day. Baseline characteristics were summarized as mean and SD for continuous variables and frequency and percentage for categorical variables and compared across quartiles of UPF intake using ANOVA or chi-square test. Spearman partial correlations between UPF and intakes of selected nutrients were estimated, adjusting for enrollment source, enrollment age, sex, and total energy intake and correction for multiple testing.
HR (95% CI) for the associations between UPF intake (coded as quartiles or per 10% increase) and mortality outcomes were estimated using Cox proportional hazards regression models with age as the time scale. Entry time was enrollment age; exit time was age at death or Dec 31, 2019, whichever came first. The proportional hazards assumption was evaluated by Schoenfeld residuals and log-log (survival) vs. log-time plots, which suggested no violation. Three models were fitted with sequential adjustment: 1) enrollment source (CHC, general population), enrollment age (continuous), sex (men, women), race (White, Black, others), educational attainment (less than high school, graduated high school, some college, graduated college or higher), annual household income (<$15,000, $15,000 – 24,999, $25,000 – 49,999, ≥ $50,000), neighborhood deprivation index (continuous), and marital status (currently married, separated/divorced, widowed, single); 2) variables in model 1 and smoking status (never, former, current), smoking pack-years (continuous), alcohol drinking (never, moderate, heavy), total physical activity (metabolic equivalent of task [MET] hours/day, continuous), and sitting hours (continuous); (3) variables in model 2 and body mass index (BMI, continuous), history of cancer (yes, no), CHD (yes, no), stroke (yes, no), diabetes (yes, no), hypertension (yes, no), and hypercholesterolemia (yes, no), and HEI-2010 (continuous). Inclusion of HEI allowed us to evaluate the UPF-mortality association independent of the established diet quality measurement. Sex-specific median or mode was assigned to missing data (<2%) for each study covariate included in the analysis.
Stratification and interaction analyses were conducted to examine potential effect modification by sociodemographics, lifestyles, obesity, and history of diseases. P value for interaction (UPF intake × stratifying variable) was obtained from the Wald test. Restricted cubic splines were used to evaluate potential non-linear association with four knots at 12.5th, 37.5th, 62.5th, and 87.5th percentiles and 12.5th percentile as the referent, where UPF intake was ~20% of daily food intake by weight [grams]). Sensitivity analyses were conducted by 1) reassigning correction weights of the 14 UPF items (70% as 100%; 20% or 50% as 0%) and 2) excluding participants with a history of cancer, CHD, or stroke at baseline (N=14,229). Two-sided P <0.05 was considered statistically significant. All analyses were performed in SAS (version 9.4; SAS Institute, Cary, NC, USA).
Results
Baseline characteristics of study participants
Of 77,060 included study participants, 46,175 (59.9%) were women, 49,857 were Black (64.7%), and the mean (SD) age at enrollment was 52.4 (8.8) years (Table 1). The mean (SD) of total UPF intake was 41.0% (15.7%) of daily food intake by weight (grams). Compared to those in the lowest UPF quartile (<29.3%), participants in the highest quartile (>51.1%) were younger and more likely to be men, Black, enrolled from CHC, and not currently married and have lower education and income levels but higher neighborhood deprivation index; meanwhile, those with high UPF intake were more likely to be current smokers and heavy drinkers and have higher sitting time, total energy intake, and BMI, and lower HEI-2010 scores but were less likely to have existing chronic diseases, suggesting possible dietary change after disease diagnosis.
Table 1.
Baseline characteristics of study participants across UPF intake categories in the SCCS (2002–2009)
| Baseline characteristics | All | Intake of UPFa | |||
|---|---|---|---|---|---|
| Quartile 1 (<29.3%) | Quartile 2 (29.4–39.7%) | Quartile 3 (39.8–51.0%) | Quartile 4 (51.1–97.5%) | ||
| No. of participants | 77060 (100.0) | 19253 (25.0) | 19260 (25.0) | 19293 (25.0) | 19254 (25.0) |
| Men, n (%) | 30885 (40.1) | 6091 (31.6) | 7919 (41.1) | 8896 (46.1) | 7979 (41.4) |
| Women, n (%) | 46175 (59.9) | 13162 (68.4) | 11341 (58.9) | 10397 (53.9) | 11275 (58.6) |
| Enrollment age, years, mean (SD) | 52.4 (8.8) | 55.1 (9.4) | 53.2 (8.9) | 51.3 (8.3) | 49.9 (7.6) |
| CHC enrollment, n (%) | 66130 (85.8) | 15424 (80.1) | 16380 (85.1) | 17044 (88.3) | 17282 (89.8) |
| Race, n (%) | |||||
| White Americans | 23764 (30.8) | 7828 (40.7) | 5890 (30.6) | 4830 (25.0) | 5216 (27.1) |
| Black Americans | 49857 (64.7) | 10168 (52.8) | 12529 (65.1) | 13741 (71.2) | 13419 (69.7) |
| Others | 3439 (4.5) | 1257 (6.5) | 841 (4.4) | 722 (3.7) | 619 (3.2) |
| Educational attainment, n (%) | |||||
| <12 years | 21527 (27.9) | 4341 (22.6) | 5438 (28.2) | 5849 (30.3) | 5899 (30.6) |
| Graduated high school | 24996 (32.4) | 5306 (27.6) | 6126 (31.8) | 6617 (34.3) | 6947 (36.1) |
| Some college | 19763 (25.7) | 5502 (28.6) | 4950 (25.7) | 4714 (24.4) | 4597 (23.9) |
| Graduated college or higher | 10391 (13.5) | 3966 (20.6) | 2658 (13.8) | 2015 (10.4) | 1752 (9.1) |
| Unknown | 383 (0.5) | 138 (0.7) | 88 (0.5) | 98 (0.5) | 59 (0.3) |
| Annual household income, n (%) | |||||
| <$15,000 | 41143 (53.4) | 8772 (45.6) | 10223 (53.1) | 10935 (56.7) | 11213 (58.2) |
| $15,000 – $24,999 | 16124 (20.9) | 3832 (19.9) | 3963 (20.6) | 4189 (21.7) | 4140 (21.5) |
| $25,000 – $49,999 | 11005 (14.3) | 3334 (17.3) | 2774 (14.4) | 2508 (13.0) | 2389 (12.4) |
| ≥ $50,000 | 7705 (10.0) | 2939 (15.3) | 2038 (10.6) | 1437 (7.5) | 1291 (6.7) |
| Unknown | 1083 (1.4) | 376 (2.0) | 262 (1.4) | 224 (1.2) | 221 (1.2) |
| Neighborhood Deprivation index, mean (SD) | 0.70 (1.20) | 0.40 (1.13) | 0.68 (1.17) | 0.84 (1.22) | 0.88 (1.22) |
| Marital status, n (%) | |||||
| Currently married | 27587 (35.8) | 7688 (39.9) | 7095 (36.8) | 6491 (33.6) | 6313 (32.8) |
| Separated/divorced | 25495 (33.1) | 6193 (32.2) | 6317 (32.8) | 6518 (33.8) | 6467 (33.6) |
| Widowed | 7436 (9.7) | 2399 (12.5) | 2041 (10.6) | 1600 (8.3) | 1396 (7.3) |
| Single | 15923 (20.7) | 2740 (14.2) | 3644 (18.9) | 4566 (23.7) | 4973 (25.8) |
| Unknown | 619 (0.8) | 233 (1.2) | 163 (0.9) | 118 (0.6) | 105 (0.6) |
| Smoking status, n (%) | |||||
| Never | 27894 (36.2) | 7963 (41.4) | 7055 (36.6) | 6391 (33.1) | 6485 (33.7) |
| Former | 17901 (23.2) | 5534 (28.7) | 4663 (24.2) | 4106 (21.3) | 3598 (18.7) |
| Current | 30735 (39.9) | 5580 (29.0) | 7403 (38.4) | 8674 (45.0) | 9078 (47.2) |
| Unknown | 530 (0.7) | 176 (0.9) | 139 (0.7) | 122 (0.6) | 93 (0.5) |
| Smoking pack-years among ever-smokers, mean (SD) | |||||
| 23.3 (22.9) | 23.5 (24.1) | 23.4 (22.9) | 23.0 (22.3) | 23.4 (22.6) | |
| Alcohol drinking b , n (%) | |||||
| None | 35876 (46.6) | 9636 (50.1) | 9137 (47.4) | 8477 (43.9) | 8626 (44.8) |
| Moderate | 27583 (35.8) | 7097 (36.9) | 6862 (35.6) | 6825 (35.4) | 6799 (35.3) |
| Heavy | 12884 (16.7) | 2296 (11.9) | 3050 (15.8) | 3826 (19.8) | 3712 (19.3) |
| Unknown | 717 (0.9) | 224 (1.2) | 211 (1.1) | 165 (0.9) | 117 (0.6) |
| Total physical activity, MET-hours/Day, mean (SD) | 22.6 (18.8) | 21.2 (17.2) | 22.5 (18.7) | 23.6 (19.9) | 23.0 (19.4) |
| Total sitting time, hours/day, mean (SD) | 9.1 (4.6) | 8.7 (4.4) | 9.1 (4.6) | 9.3 (4.7) | 9.5 (4.8) |
| HEI-2010, mean (SD) | 57.9 (12.1) | 67.1 (11.2) | 60.0 (10.1) | 55.1 (9.5) | 49.3 (9.7) |
| Total energy intake, kcal/day, mean | |||||
| (SD) | 2546.0 (1439.4) | 2028.2 (1064.2) | 2559.4 (1431.7) | 2874.4 (1578.6) | 2721.2 (1485.3) |
| BMI, kg/m 2 , mean (SD) | 30.4 (7.5) | 30.1 (7.2) | 30.4 (7.4) | 30.1 (7.6) | 30.8 (7.9) |
| Prevalent medical conditions, n (%) | |||||
| Hypertension | 42157 (54.7) | 10698 (55.6) | 10760 (55.9) | 10397 (53.9) | 10302 (53.5) |
| Hypercholesterolemia | 26213 (34.0) | 7524 (39.1) | 6822 (35.4) | 6029 (31.3) | 5838 (30.3) |
| Cancer | 6117 (7.9) | 2089 (10.9) | 1575 (8.2) | 1257 (6.5) | 1196 (6.2) |
| CHD | 5385 (7.0) | 1462 (7.6) | 1452 (7.5) | 1295 (6.7) | 1176 (6.1) |
| Stroke | 4956 (6.4) | 1348 (7.0) | 1309 (6.8) | 1177 (6.1) | 1122 (5.8) |
| Diabetes | 18831 (24.4) | 4887 (25.4) | 5042 (26.2) | 4573 (23.7) | 4329 (22.5) |
Abbreviations: BMI, body mass index; CHC, Community Health Center; CHD, Coronary heart disease; HEI-2010, Healthy Eating Index; MET, metabolic equivalent of task; SCCS, Southern Community Cohort Study; UPF, ultra-processed foods.
Intake of UPF was estimated as the proportion of UPF in daily food intake (% of grams). All P values were less than 0.001 except for smoking pack-years (P=0.25).
Heavy drinking was defined as alcohol consumption of >2 drinks per day in men or >1 drink per day in women; moderate drinking was defined as alcohol consumption of >0 to ≤2 drinks per day in men or >0 to ≤1 drink per day in women.
Correlations between UPF and diet quality and nutrients intake
As shown in Table 2 (all PFDR<0.0001), UPF intake, modeled as a continuous variable, was inversely correlated to HEI-2010 (r = −0.53) and intakes of fiber (r = −0.44), minerals such as potassium (r = −0.63), magnesium (r = −0.57), copper, calcium, phosphorus, iron, zinc, and selenium, vitamins such as vitamin A (r = −0.41), Bs, particularly folate (r = −0.58) and B6 (r = −0.41), C, E, K, and carotenoid (β-cryptoxanthin, lutein, zeaxanthin, lycopene). On the other hand, UPF intake was positively correlated to intakes of sugars and fats, particularly saturated stearic acid (C18:0) and palmitoleic acid (C16:1); both r=0.18.
Table 2.
Spearman partial correlation between nutrients and intake of UPF
| Nutrient variables | Spearman partial correlation coefficients with intake of UPFa |
|---|---|
| Potassium | −0.63 |
| Food folate | −0.58 |
| Magnesium | −0.57 |
| Healthy Eating Index | −0.53 |
| Dietary fiber | −0.44 |
| Vitamin A | −0.41 |
| Vitamin B6 | −0.41 |
| Copper | −0.41 |
| Carotene | −0.40 |
| Beta − cryptoxanthin | −0.40 |
| Alpha−tocopherol | −0.34 |
| Vitamin C | −0.34 |
| Lutein & zeaxanthin | −0.32 |
| Vitamin B1 | −0.31 |
| Calcium | −0.31 |
| Phosphorus | −0.31 |
| Vitamin K | −0.30 |
| Vitamin B2 | −0.29 |
| Niacin | −0.27 |
| Lycopene | −0.23 |
| Iron | −0.22 |
| Retinol | −0.21 |
| Zinc | −0.21 |
| Protein | −0.16 |
| Vitamin B12 | −0.16 |
| Fatty acid 22:6/ DHA | −0.13 |
| Selenium | −0.13 |
| Total sugars | 0.06 |
| Total fat | 0.08 |
| Monounsaturated fat | 0.10 |
| Fatty acid 18:1/ oleic acid | 0.10 |
| Saturated fat | 0.12 |
| Fatty acid 16:0/ palmitic acid | 0.12 |
| Fatty acid 18:0/ stearic acid | 0.18 |
| Fatty acid 16:1/ palmitoleic acid | 0.18 |
Abbreviations: UPF, ultra-processed foods.
Intake of UPF was estimated as the proportion of UPF in daily food intake (% of grams) and used as a continuous variable in correlation analysis. Adjusted for enrollment age, sex, enrollment source, and total daily calories. False discovery rates were calculated at all nutrients level to account for multiple testing, and all PFDR <0.0001.
Associations between UPF intake and mortality
During an average follow-up of 12.2 years (SD: 3.4 years), we documented 17,895 total deaths, including 4,267, 2,208, 867, and 997 deaths from cancer, CHD, Stroke, and diabetes (Table 3). After adjustment for sociodemographics (Model 1) and lifestyle factors (Model 2), higher UPF intake was associated with higher all-cause, stroke, and diabetes mortality (HR [95% CI] for the highest versus lowest quartiles [Q4 vs. Q1] = 1.06 [1.01, 1.11], 1.26 [1.03, 1.54], and 1.26 [1.05, 1.51] in Model 2, respectively), but not with cancer or CHD mortality. While after further adjustment for BMI, history of chronic diseases, and HEI-2010 (Model 3), the associations with all-cause and stroke mortality were attenuated, higher UPF intake was still significantly associated with diabetes mortality. In the fully adjusted Model 3, the highest quartile of UPF intake showed a 32% (95%CI: 7%−62%) higher diabetes mortality compared to the lowest quartile, and per 10% increase in UPF intake was associated with a 9% (95%CI: 4%−15%) higher diabetes mortality. The UPF-diabetes mortality association appeared linear (P-linearity = 0.0003, P-nonlinearity = 0.56, and P-overall =0.003; Supplementary Figure 1).
Table 3.
Associations between intake of UPFa and all-cause and cause-specific mortality in the SCCS
| Cases | Person-years | Model 1: HR (95% CI)b | Model 2: HR (95% CI)c | Model 3: HR (95% CI)d | |
|---|---|---|---|---|---|
| All-cause mortality | |||||
| Quartile 1 | 4,304 | 233,517 | 1 | 1 | 1 |
| Quartile 2 | 4,642 | 233,246 | 1.08 (1.03, 1.12) | 1.06 (1.01, 1.10) | 1.01 (0.97, 1.06) |
| Quartile 3 | 4,630 | 236,346 | 1.10 (1.06, 1.15) | 1.06 (1.02, 1.11) | 1.01 (0.97, 1.06) |
| Quartile 4 | 4,319 | 237,141 | 1.11 (1.07, 1.16) | 1.06 (1.01, 1.11) | 0.99 (0.94, 1.04) |
| Per 10% increase | 17,895 | 940,250 | 1.02 (1.01, 1.03) | 1.01 (1.00, 1.02) | 0.99 (0.98, 1.00) |
| Cancer mortality | |||||
| Quartile 1 | 1,085 | 233,517 | 1 | 1 | 1 |
| Quartile 2 | 1,084 | 233,246 | 1.01 (0.93, 1.10) | 0.98 (0.90, 1.07) | 0.96 (0.88, 1.05) |
| Quartile 3 | 1,106 | 236,346 | 1.08 (0.99, 1.18) | 1.02 (0.94, 1.11) | 0.99 (0.90, 1.08) |
| Quartile 4 | 992 | 237,141 | 1.07 (0.98, 1.17) | 1.00 (0.91, 1.09) | 0.95 (0.86, 1.05) |
| Per 10% increase | 4,267 | 940,250 | 1.01 (0.99, 1.03) | 0.99 (0.97, 1.01) | 0.98 (0.96, 1.00) |
| CHD mortality | |||||
| Quartile 1 | 560 | 233,517 | 1 | 1 | 1 |
| Quartile 2 | 586 | 233,246 | 1.06 (0.94, 1.19) | 1.04 (0.93, 1.17) | 0.98 (0.87, 1.10) |
| Quartile 3 | 570 | 236,346 | 1.08 (0.96, 1.22) | 1.05 (0.93, 1.19) | 0.97 (0.85, 1.10) |
| Quartile 4 | 492 | 237,141 | 1.03 (0.91, 1.17) | 0.99 (0.87, 1.12) | 0.88 (0.76, 1.01) |
| Per 10% increase | 2,208 | 940,250 | 1.01 (0.98, 1.04) | 0.99 (0.97, 1.02) | 0.96 (0.93, 1.00) |
| Stroke mortality | |||||
| Quartile 1 | 195 | 233,517 | 1 | 1 | 1 |
| Quartile 2 | 235 | 233,246 | 1.20 (0.99, 1.46) | 1.18 (0.98, 1.43) | 1.12 (0.92, 1.36) |
| Quartile 3 | 223 | 236,346 | 1.21 (0.99, 1.47) | 1.16 (0.95, 1.42) | 1.09 (0.88, 1.35) |
| Quartile 4 | 214 | 237,141 | 1.31 (1.07, 1.61) | 1.26 (1.03, 1.54) | 1.15 (0.91, 1.44) |
| Per 10% increase | 867 | 940,250 | 1.07 (1.03, 1.12) | 1.06 (1.02, 1.11) | 1.04 (0.99, 1.10) |
| Diabetes mortality | |||||
| Quartile 1 | 226 | 233,517 | 1 | 1 | 1 |
| Quartile 2 | 250 | 233,246 | 1.08 (0.90, 1.29) | 1.07 (0.89, 1.28) | 1.04 (0.86, 1.25) |
| Quartile 3 | 254 | 236,346 | 1.11 (0.93, 1.34) | 1.12 (0.93, 1.35) | 1.13 (0.94, 1.38) |
| Quartile 4 | 267 | 237,141 | 1.28 (1.07, 1.54) | 1.26 (1.05, 1.51) | 1.32 (1.07, 1.62) |
| Per 10% increase | 997 | 940,250 | 1.08 (1.03, 1.12) | 1.07 (1.03, 1.12) | 1.09 (1.04, 1.15) |
Abbreviations: CHD, coronary heart disease; SCCS, Southern Community Cohort Study; UPF, ultra-processed foods.
Intake of UPF was estimated as the proportion of UPF in daily food intake (% of grams).
Adjusted for enrollment source, enrollment age, sex, race, education, income, deprivation index, and marital status.
Adjusted for covariates in Model 1 plus smoking status, pack-years, alcohol drinking, physical activity, and sitting hours.
Adjusted for covariates in Model 2 plus BMI, history of cancer, CHD, stroke, diabetes, hypertension, and hypercholesterolemia, and HEI-2010.
Among UPF food groups, beverages, such as sodas and fruit-flavored drinks, contributed the most (mean [SD]:19.4% [16.0%] in the diet), followed by grains and fried potatoes (6.4% [4.1%]) and protein food (6.0% [4.4%]), especially animal-based protein (5.9% [4.4%]). Per 10% increase in UPF intake from beverages and animal-based protein food was associated with, respectively, a 10% (95% CI: 5%−15%) and 23% (95% CI: 4%−45%) higher diabetes mortality (Supplementary Table 2).
Similar adverse associations of UPF intake with diabetes mortality were found among men and women, Blacks and Whites, and participants stratified by household income, neighborhood deprivation index, marital status, lifestyle factors, and history of major chronic diseases (Fig. 1). However, we observed potential modifying effects by educational attainment and history of hypercholesterolemia: the adverse association was more evident among participants with high school or less education (HR [95% CI] per 10% increase =1.12 [1.05, 1.18]) or a history of hypercholesterolemia (1.14 [1.07, 1.22]), compared to those with more than high school education (1.03 [0.94, 1.12], Pinteraction =0.04) or without hypercholesterolemia (1.03 [0.96, 1.11], Pinteraction =0.03), suggesting subpopulations who may be particularly vulnerable to UPF exposure that may increase their risk of death due to diabetes.
Fig. 1. Associations between intake of UPF (Q4 vs. Q1/ per 10% increase) and diabetes mortality in subgroups.
Intake of UPF was estimated as the proportion of UPF in daily food intake (% of grams). Adjusted for enrollment source, enrollment age, sex, race, education, income, deprivation index, marital status, smoking status, pack-years, alcohol drinking, physical activity, sitting hours, BMI, HEI-2010, and history of cancer, CHD, stroke, diabetes, hypertension, and hypercholesterolemia (Model 3).
In sensitivity analyses, we found similar adverse associations between UPF intake and diabetes mortality after reassigning the correction weights of 14 UPF items (HR [95% CI] per 10% increase=1.09 [1.05, 1.14] in Model 3) and after excluding participants with cancer or CVD at baseline (HR [95% CI] per 10% increase=1.06 [1.00, 1.12] in Model 3).
Discussion
In this large prospective cohort study among ~77,000 predominantly low-income Americans with an average follow-up of ~12 years, higher UPF intake was associated with higher diabetes mortality, although not with all-cause, cancer, CHD, or stroke mortality after adjustment for potential confounders. The association between UPF and diabetes mortality appeared linear (HR~1.32 in the highest vs. lowest quartile and ~1.09 per 10% increase of UPF intake) and independent of established risk factors for metabolic disorders and premature deaths, including HEI-2010, a diet quality measure based on the Dietary Guidelines for Americans. Furthermore, the adverse UPF-diabetes mortality association was observed regardless of sex, race, household income, neighborhood deprivation index, lifestyle factors (e.g., smoking), obesity, and history of cardiometabolic diseases, while particularly evident among individuals with no more than high school education or a history of hypercholesterolemia.
To our knowledge, this was the first prospective study that examined the relations of UPF with all-cause and cause-specific mortality in a multi-racial cohort of mostly low-income Americans. We noted higher UPF intake was associated with higher diabetes mortality, which has not been investigated by other studies but agreed well with previous findings on an increased risk of incident diabetes [8–11]. A meta-analysis showed that moderate or high intake of UPF increased diabetes risk by 12% and 31%, respectively, compared to the non-/low consumption of UPF [28]. The associations of UPF with diabetes incidence and mortality could be partly explained by the nutrient attributes of UPF [29]. Consistent with findings from a nationally representative US population [30, 31], we found that UPF was correlated with higher intakes of total energy, saturated fat, and sugar, while lower intakes of minerals, vitamins, fiber, and polyunsaturated fat such as DHA, as well as a lower overall dietary quality (i.e., HEI-2010). Furthermore, our additional analysis suggested that ultra-processed animal-based foods (e.g., bacon), which are high in saturated fat and sodium, and beverages, mostly sugar-sweetened soft drinks, showed significant associations with increased diabetes mortality. There are established adverse effects of saturated fat, processed meats, and sugar-sweetened beverages and protective effects of dietary fiber, potassium, magnesium, B vitamins, polyunsaturated fat, fresh fruits and vegetables, and healthy dietary patterns on diabetes prevention and management [29, 32, 33]. The correlated poor dietary quality and nutritional profiles underline the harmful health effect of UPF.
Meanwhile, the persistent, significant association between UPF intake and diabetes mortality after adjusting for overall dietary quality indicates that non-nutrient attributes of UPF also play a role. UPF normally contain a variety of additives (e.g., artificial sweeteners) [34], compounds generated during processing (e.g., acrylamide and acrolein) [35], and contaminants from package materials (e.g., bisphenol A and phthalates) [36]. Although there are regulatory safety thresholds for many chemicals in food, the long-term health effects of their low-level, chronic exposures have not been well understood [37]. Of note, evidence is emerging for the adverse effects of artificial sweeteners [38–40], neo-formed acrylamide [41], bisphenol A [42], and phthalates [43] on the risk of diabetes or other cardiometabolic diseases. Those food additives might alter the gut microbiota composition, function, microbial metabolites, and microbiome-host interactions [44], which can activate various potential mechanisms leading to insulin resistance and diabetes [45]. Since the level of non-nutrient ingredients in UPF is often very low, their long-term collective impact on health is worthy of investigation. Our findings provide clues for future prospective studies with systematic molecular approaches, such as metabolomic profiling of UPF or biospecimens of individuals with high consumption of UPF.
Importantly, the adverse association of UPF with diabetes mortality was observed in men and women, Blacks and Whites, and individuals from different levels of household income and neighborhood deprivation, suggesting that the harmful effect of high UPF intake is likely universal. On the other hand, the adverse association seemed particularly evident among participants with limited education or a history of hypercholesterolemia. UPF consumption was higher among our participants with high school education or below than those with higher education (mean [SD] =42.6% [15.5%] vs. 38.6% [15.7%]), in line with a previously reported inverse association between educational attainment and UPF consumption [3]. People with higher education may consume less UPF and have greater knowledge about health and also better healthcare access and treatment adherence (uncontrolled potential confounders), which might attenuate the effect of UPF on disease and related mortality. However, our participants with a history of hypercholesterolemia had a lower intake of UPF than those without (mean [SD] =39.6% [15.7%] vs. 41.8% [15.7%]), suggesting potential dietary changes after disease diagnosis. A stronger association observed among participants with a history of hypercholesterolemia suggests a potential synergistic effect between UPF intake and hypercholesterolemia on diabetes. Prior studies have demonstrated adverse associations of UPF with blood lipid profiles [46, 47]; meanwhile, statin use is associated with poor glycemic control and diabetes progression [48]. The fact that our participants with hypercholesterolemia had slightly lower UPF intake than those without suggests that the observed association may be diluted. The health effect of UPF among individuals with existing metabolic diseases warrants further investigation.
We did not find a significant association between UPF and all-cause mortality after adjusting for lifestyle factors and disease status, although a significant association (HR=1.22–1.62 for the highest quartile vs. lowest quartile and 1.14 for per 10% increase) has been reported in several cohort studies, including the Seguimiento Universidad de Navarra (SUN) cohort in Spain [12], the French NutriNet-Sante study [13], the UK Biobank [7], as well as the NHANES and the Adventist Health Study in the US [14, 15]. We also found null associations between UPF and cancer, CHD, or stroke mortality in multivariable-adjusted models. Current evidence on cause-specific mortality remains inconsistent. For example, the Moli-sani study in Italy [17] and the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (PLCO) in the US [16] reported a positive association of UPF intake with CVD mortality, while the SUN cohort [12] and the NHANES III [15] found null association; nevertheless, an increased CVD mortality was observed in a meta-analysis (pooled HR=1.50 for the highest vs. lowest intake categories) [18]. Meanwhile, the Moli-sani study and SUN cohort also reported a null association between UPF intake and cancer mortality [12, 17], like ours. Prospective studies with large sample sizes, especially from racially, ethnically, and geographically diverse populations, are still needed to examine the relations between UPF intake and cause-specific mortality.
Our findings have major implications. Food processing is an essential component of industrialized food systems, and consumption of UPF is increasing worldwide [49]. The US is ranked top of all countries in terms of UPF sales per capita and contributions to diet [49, 50]. Our findings provide novel evidence of an adverse association between UPF and diabetes mortality in low-income Black and White Americans, independent of overall dietary quality. To date, research evidence has supported the need to consider the harmful health effect of high UPF consumption and replace UPF with unprocessed or minimally processed food. Indeed, over 45 countries, including from Europe, Asia, Africa, and South America, have taken action, including increased taxes on the UPF, front-of-package warning labels, school food policies, and media campaigns [51]. Furthermore, the adverse UPF-diabetes mortality association, particularly among participants with limited education or a history of hypercholesterolemia, highlights the need to explain the harmful effects of UPF in plain language and increase accessibility and affordability of unprocessed or minimally processed food among socioeconomically marginalized communities.
Our study has several strengths. It is based on a large-scale, prospective cohort with a long follow-up among low-income Americans who have been underrepresented in other US cohorts. We applied the Nova system, a widely accepted UPF classification system, to classify foods in FFQ, making our results comparable with other studies. However, our study also has several limitations. First, although the 89-item FFQ was designed and validated to collect dietary information in our targeted study population, it was not specifically designed to capture the degree of processing, which might lead to potential Nova food item misclassifications and thus inaccurate UPF estimates. These likely non-differential UPF estimate errors may bias the studied associations towards the null. Also, the diet information was self-reported by participants, who might underreport UPF intake due to social desirability bias. Second, we followed the previous practice of defining UPF intake using weight [6, 10] instead of energy proportion in diet, considering no energy contribution from some UPF (e.g., artificially sweetened beverages); this method, however, ignores the energy densities of different UPF. Nevertheless, the results were similar with or without adjusting for total energy intake (data not shown). Third, the observational nature of our study precludes ruling out residual confounding bias and establishing causality.
Conclusion
In conclusion, our study linked the higher intake of UPF with increased diabetes mortality among predominantly low-income Black and White Americans, independent of sociodemographics, lifestyles, and overall diet quality defined by recent Dietary Guidelines for Americans. The adverse association between UPF and diabetes mortality seemed particularly evident among individuals with limited education or a history of hypercholesterolemia. Our findings suggest the need to consider UPF in nutritional education and public health policies and the potential impact of increasing access and intake of un/minimally processed food to replace UPF to reduce diabetes-related mortality.
Supplementary Material
Acknowledgments
We thank all SCCS participants and staff for their contribution to the study.
Funding
The Southern Community Cohort Study (SCCS) is supported by the National Cancer Institute of the National Institutes of Health under Award Number U01CA202979 to William Blot, Martha Shrubsole, and Wei Zheng. SCCS data collection was performed by the Survey and Biospecimen Shared Resource, which is supported in part by the Vanderbilt-Ingram Cancer Center (P30 CA68485). Danxia Yu is supported by R01HL149779. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Abbreviations:
- BMI
body mass index
- CHC
community health center
- CHD
coronary heart disease
- CVD
cardiovascular disease
- FFQ
food frequency questionnaire
- HEI
Healthy Eating Index
- ICD
International Classification of Diseases
- MET
metabolic equivalent of task
- NHANES
National Health and Nutrition Examination Survey
- PLCO
Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial
- SCCS
Southern Community Cohort Study
- SES
socioeconomic status
- SUN
Seguimiento Universidad de Navarra cohort
- UPF
ultra-processed foods
- US
United States
- USDA
United States Department of Agriculture
- WWEIA
What We Eat in America
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest
The authors declare that there are no conflicts of interest.
Data availability
Data described in the manuscript, code book, and analytic code will be available upon research study application and approval by the cohort committees.
Reference
- [1].Monteiro CA, Cannon G, Levy R, Moubarac J-C, Jaime P, Martins AP, et al. NOVA. The star shines bright. 2016;7:28–38. [Google Scholar]
- [2].Marino M, Puppo F, Del Bo C, Vinelli V, Riso P, Porrini M, et al. A Systematic Review of Worldwide Consumption of Ultra-Processed Foods: Findings and Criticisms. Nutrients. 2021;13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Baraldi LG, Steele EM, Canella DS, Monteiro CA. Consumption of ultra-processed foods and associated sociodemographic factors in the USA between 2007 and 2012: evidence from a nationally representative cross-sectional study. Bmj Open. 2018;8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Fiolet T, Srour B, Sellem L, Kesse-Guyo E, Alles B, Mejean C, et al. Consumption of ultra-processed foods and cancer risk: results from NutriNet-Sante prospective cohort. Bmj-Brit Med J. 2018;360. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Chang K, Millett C, Rauber F, Levy RB, Huybrechts I, Kliemann N, et al. Ultra-processed food consumption, cancer risk, and cancer mortality: a prospective cohort study of the UK Biobank. 2022;400:S31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Srour B, Fezeu LK, Kesse-Guyot E, Alles B, Mejean C, Andrianasolo RM, et al. Ultra-processed food intake and risk of cardiovascular disease: prospective cohort study (NutriNet-Sante). BMJ. 2019;365:l1451. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Chen X, Chu J, Hu W, Sun N, He Q, Liu S, et al. Associations of ultra-processed food consumption with cardiovascular disease and all-cause mortality: UK Biobank. Eur J Public Health. 2022;32:779–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Levy RB, Rauber F, Chang K, Louzada MLD, Monteiro CA, Millett C, et al. Ultra-processed food consumption and type 2 diabetes incidence: A prospective cohort study. Clin Nutr. 2021;40:3608–14. [DOI] [PubMed] [Google Scholar]
- [9].Llavero-Valero M, Martin JES, Martinez-Gonzalez MA, Basterra-Gortari FJ, de la Fuente-Arrillaga C, Bes-Rastrollo M. Ultra-processed foods and type-2 diabetes risk in the SUN project: A prospective cohort study. Clin Nutr. 2021;40:2817–24. [DOI] [PubMed] [Google Scholar]
- [10].Srour B, Fezeu LK, Kesse-Guyot E, Alles B, Debras C, Druesne-Pecollo N, et al. Ultraprocessed Food Consumption and Risk of Type 2 Diabetes Among Participants of the NutriNet-Sante Prospective Cohort. Jama Intern Med. 2020;180:283–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Chen Z, Khandpur N, Desjardins C, Wang L, Monteiro CA, Rossato SL, et al. Ultra-Processed Food Consumption and Risk of Type 2 Diabetes: Three Large Prospective U.S. Cohort Studies. Diabetes Care. 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Rico-Campa A, Martinez-Gonzalez MA, Alvarez-Alvarez I, Mendonca RD, de la Fuente-Arrillaga C, Gomez-Donoso C, et al. Association between consumption of ultra-processed foods and all cause mortality: SUN prospective cohort study. Bmj-Brit Med J. 2019;365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Schnabel L, Kesse-Guyot E, Alles B, Touvier M, Srour B, Hercberg S, et al. Association Between Ultraprocessed Food Consumption and Risk of Mortality Among Middle-aged Adults in France. Jama Intern Med. 2019;179:490–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Orlich MJ, Sabate J, Mashchak A, Fresan U, Jaceldo-Siegl K, Miles F, et al. Ultra-processed food intake and animal-based food intake and mortality in the Adventist Health Study-2. Am J Clin Nutr. 2022;115:1589–601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Kim H, Hu EA, Rebholz CM. Ultra-processed food intake and mortality in the USA: results from the Third National Health and Nutrition Examination Survey (NHANES III, 1988–1994). Public Health Nutr. 2019;22:1777–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Zhong GC, Gu HT, Peng Y, Wang K, Wu YQ, Hu TY, et al. Association of ultra-processed food consumption with cardiovascular mortality in the US population: long-term results from a large prospective multicenter study. Int J Behav Nutr Phys Act. 2021;18:21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Bonaccio M, Di Castelnuovo A, Costanzo S, De Curtis A, Persichillo M, Sofi F, et al. Ultra-processed food consumption is associated with increased risk of all-cause and cardiovascular mortality in the Moli-sani Study. American Journal of Clinical Nutrition. 2021;113:446–55. [DOI] [PubMed] [Google Scholar]
- [18].Suksatan W, Moradi S, Naeini F, Bagheri R, Mohammadi H, Talebi S, et al. Ultra-Processed Food Consumption and Adult Mortality Risk: A Systematic Review and Dose-Response Meta-Analysis of 207,291 Participants. Nutrients. 2021;14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [19].Murphy SL, Kochanek KD, Xu J, Arias E. Mortality in the United States, 2020. NCHS Data Brief. 2021:1–8. [PubMed] [Google Scholar]
- [20].Stringhini S, Carmeli C, Jokela M. Socioeconomic status and the 25 × 25 risk factors as determinants of premature mortality: a multicohort study and meta-analysis of 1.7 million men and women (vol 389, pg 1229, 2017). Lancet. 2017;389:1194-. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Signorello LB, Hargreaves MK, Blot WJ. The Southern Community Cohort Study: investigating health disparities. J Health Care Poor Underserved. 2010;21:26–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Buchowski MS, Schlundt DG, Hargreaves MK, Hankin JH, Signorello LB, Blot WJ. Development of a culturally sensitive food frequency questionnaire for use in the Southern Community Cohort Study. Cell Mol Biol (Noisy-le-grand). 2003;49:1295–304. [PubMed] [Google Scholar]
- [23].Signorello LB, Munro HM, Buchowski MS, Schlundt DG, Cohen SS, Hargreaves MK, et al. Estimating nutrient intake from a food frequency questionnaire: incorporating the elements of race and geographic region. Am J Epidemiol. 2009;170:104–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Schlundt DG, Buchowski MS, Hargreaves MK, Hankin JH, Signorello LB, Blot WJ. Separate estimates of portion size were not essential for energy and nutrient estimation: results from the Southern Community Cohort food-frequency questionnaire pilot study. Public Health Nutr. 2007;10:245–51. [DOI] [PubMed] [Google Scholar]
- [25].Bowman SA, Friday JE, Moshfegh AJJUDoA. MyPyramid Equivalents Database, 2.0 for USDA survey foods, 2003–2004: documentation and user guide. 2008.
- [26].Guenther PM, Casavale KO, Reedy J, Kirkpatrick SI, Hiza HA, Kuczynski KJ, et al. Update of the Healthy Eating Index: HEI-2010. J Acad Nutr Diet. 2013;113:569–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].AGRICULTURE USDO. WWEIA Food Categories for use with WWEIA, NHANES 2017 - March 2020 Prepandemic. 2017.
- [28].Delpino FM, Figueiredo LM, Bielemann RM, da Silva BGC, Dos Santos FS, Mintem GC, et al. Ultra-processed food and risk of type 2 diabetes: a systematic review and meta-analysis of longitudinal studies. Int J Epidemiol. 2022;51:1120–41. [DOI] [PubMed] [Google Scholar]
- [29].Evert AB, Dennison M, Gardner CD, Garvey WT, Lau KHK, MacLeod J, et al. Nutrition Therapy for Adults With Diabetes or Prediabetes: A Consensus Report. Diabetes Care. 2019;42:731–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Steele EM, Popkin BM, Swinburn B, Monteiro CA. The share of ultra-processed foods and the overall nutritional quality of diets in the US: evidence from a nationally representative cross-sectional study. Popul Health Metr. 2017;15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Liu J, Steele EM, Li Y, Karageorgou D, Micha R, Monteiro CA, et al. Consumption of Ultraprocessed Foods and Diet Quality Among U.S. Children and Adults. American journal of preventive medicine. 2022;62:252–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Micha R, Wallace SK, Mozaffarian D. Red and processed meat consumption and risk of incident coronary heart disease, stroke, and diabetes mellitus: a systematic review and meta-analysis. Circulation. 2010;121:2271–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Malik VS, Popkin BM, Bray GA, Despres JP, Willett WC, Hu FB. Sugar-sweetened beverages and risk of metabolic syndrome and type 2 diabetes: a meta-analysis. Diabetes Care. 2010;33:2477–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Chazelas E, Deschasaux M, Srour B, Kesse-Guyot E, Julia C, Alles B, et al. Food additives: distribution and co-occurrence in 126,000 food products of the French market. Sci Rep. 2020;10:3980. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Morales FJ, Mesias M, Delgado-Andrade C. Association between Heat-Induced Chemical Markers and Ultra-Processed Foods: A Case Study on Breakfast Cereals. Nutrients. 2020;12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [36].Buckley JP, Kim H, Wong E, Rebholz CM. Ultra-processed food consumption and exposure to phthalates and bisphenols in the US National Health and Nutrition Examination Survey, 2013–2014. Environ Int. 2019;131:105057. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].Maffini MV, Neltner TG, Vogel S. We are what we eat: Regulatory gaps in the United States that put our health at risk. Plos Biol. 2017;15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [38].Azad MB, Abou-Setta AM, Chauhan BF, Rabbani R, Lys J, Copstein L, et al. Nonnutritive sweeteners and cardiometabolic health: a systematic review and meta-analysis of randomized controlled trials and prospective cohort studies. CMAJ. 2017;189:E929–E39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [39].Johnson RK, Lichtenstein AH, Anderson CAM, Carson JA, Despres JP, Hu FB, et al. Low-Calorie Sweetened Beverages and Cardiometabolic Health: A Science Advisory From the American Heart Association. Circulation. 2018;138:e126–e40. [DOI] [PubMed] [Google Scholar]
- [40].Witkowski M, Nemet I, Alamri H, Wilcox J, Gupta N, Nimer N, et al. The artificial sweetener erythritol and cardiovascular event risk. Nat Med. 2023. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [41].Zhang Y, Huang M, Zhuang P, Jiao J, Chen X, Wang J, et al. Exposure to acrylamide and the risk of cardiovascular diseases in the National Health and Nutrition Examination Survey 2003–2006. Environ Int. 2018;117:154–63. [DOI] [PubMed] [Google Scholar]
- [42].Ranciere F, Lyons JG, Loh VH, Botton J, Galloway T, Wang T, et al. Bisphenol A and the risk of cardiometabolic disorders: a systematic review with meta-analysis of the epidemiological evidence. Environ Health. 2015;14:46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [43].Sun Q, Cornelis MC, Townsend MK, Tobias DK, Eliassen AH, Franke AA, et al. Association of urinary concentrations of bisphenol A and phthalate metabolites with risk of type 2 diabetes: a prospective investigation in the Nurses’ Health Study (NHS) and NHSII cohorts. Environ Health Perspect. 2014;122:616–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [44].Zinocker MK, Lindseth IA. The Western Diet-Microbiome-Host Interaction and Its Role in Metabolic Disease. Nutrients. 2018;10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [45].Li WZ, Stirling K, Yang JJ, Zhang L. Gut microbiota and diabetes: From correlation to causality and mechanism. World J Diabetes. 2020;11:293–308. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [46].Rauber F, Campagnolo PD, Hoffman DJ, Vitolo MR. Consumption of ultra-processed food products and its effects on children’s lipid profiles: a longitudinal study. Nutr Metab Cardiovasc Dis. 2015;25:116–22. [DOI] [PubMed] [Google Scholar]
- [47].Pagliai G, Dinu M, Madarena MP, Bonaccio M, Iacoviello L, Sofi F. Consumption of ultra-processed foods and health status: a systematic review and meta-analysis. Br J Nutr. 2021;125:308–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [48].Mansi IA, Chansard M, Lingvay I, Zhang S, Halm EA, Alvarez CA. Association of Statin Therapy Initiation With Diabetes Progression: A Retrospective Matched-Cohort Study. Jama Intern Med. 2021;181:1562–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [49].Baker P, Machado P, Santos T, Sievert K, Backholer K, Hadjikakou M, et al. Ultra-processed foods and the nutrition transition: Global, regional and national trends, food systems transformations and political economy drivers. Obes Rev. 2020;21:e13126. [DOI] [PubMed] [Google Scholar]
- [50].Martinez Steele E, Baraldi LG, Louzada ML, Moubarac JC, Mozaffarian D, Monteiro CA. Ultra-processed foods and added sugars in the US diet: evidence from a nationally representative cross-sectional study. BMJ Open. 2016;6:e009892. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [51].Popkin BM, Barquera S, Corvalan C, Hofman KJ, Monteiro C, Ng SW, et al. Towards unified and impactful policies to reduce ultra-processed food consumption and promote healthier eating. Lancet Diabetes Endocrinol. 2021;9:462–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data described in the manuscript, code book, and analytic code will be available upon research study application and approval by the cohort committees.