1 INTRODUCTION

Food processing has changed dietary patterns across the globe, with processed and ultra-processed foods contributing to an ever-increasing share of the energy intake of populations.1, 2 Classifying foods in accordance with their level of processing improved the understanding of these new patterns and enabled the systematic assessment of how they might contribute to diet-related conditions and noncommunicable diseases.3 A diet rich in ultra-processed foods is poorer in quality, with less protein, fibre, vitamins and micronutrients and more added sugars.4-7 In terms of health outcomes, they cause weight gain8 and are associated with cancer,9 metabolic syndrome,10 hypertension11 and all-cause mortality.12, 13

Diet is strongly associated with dental caries. The intake of added sugars is a particularly important risk factor for caries, with evidence pointing to the significance of controlling both amount and, to a lesser extent, frequency of its intake.14, 15 Studies suggest a moderate effect of refined starches and a negligible effect of whole starches and whole fruits on the development of dental caries.16 Ultra-processed food intake is very strongly associated with free sugars intake.4, 6, 7 Even though the most consumed ultra-processed foods are not those typically associated with caries, studies show that a reduction in the proportional daily energy contribution (%E) of these foods would entail a reduction in sugar intake and an increase in the intake of less cariogenic foods.4, 6 In this sense, restricting ultra-processed intake could be a good strategy to help meet the World Health Organization (WHO) recommendations on sugar intake.17

New nutritional guidelines take processing into account, particularly by using the NOVA (not an acronym) classification and focusing on reducing the dietary contribution of ultra-processed foods.18 Vast literature on a variety of foods and drinks and their association with dental caries is available.14, 16, 19, 20 Nevertheless, the relationship between different levels of food processing with dental caries has not been systematically investigated. Thus, with this study we aimed to assess whether proportional contribution of unprocessed or minimally processed, processed or ultra-processed foods to daily energy intake is associated with dental caries in USA adults.

2 METHODS 2.1 Population and sample

This secondary analysis used cross-sectional data from two cycles of the continuous National Health and Nutrition Examination Survey (NHANES): 2011-2012 and 2013-2014. These data were obtained using a complex, multistage probability sampling design, and they are representative of noninstitutionalized civilian residents in the United States of America in the midpoint of the two cycles included in the analysis.21 We have restricted our study population to adults aged 20-59 years old with a complete oral examination record (a comparison with participants with incomplete examination is presented in Supplement S1) and two 24-hour dietary recalls, in accordance with age cut-points proposed by the NHANES analytic guidelines.21 A flow chart detailing our studied population is presented in Figure S1.

2.2 Data collection

Participants were first interviewed at home, completing a screener questionnaire, then a series of structured questionnaires which were applied according to sample eligibility. Data collected at this time included sociodemographic background and health history. Participants were then referred to attend a mobile examination centre (MEC), where physical examinations, laboratory tests and a dietary interview were performed. Three to 10 days after the MEC examinations, participants were contacted by telephone and invited to do a follow-up dietary interview. This dietary interview aimed to obtain detailed information on dietary intake from selected participants. These data were used to estimate consumption of types and amounts of foods and beverages, including their energetic and nutritional components. Portion size estimation is further explained in the NHANES dietary interview manuals.22

2.3 Food classification

Food items recorded in these dietary recalls were classified according to NOVA, a system of food classification based on the extent and purpose of industrial food processing.23 This classification divides foods into four groups: unprocessed or minimally processed foods; processed culinary ingredients; processed foods; and ultra-processed foods. Unprocessed or minimally processed foods, also referred to as group 1 (G1) foods, are edible parts of plants, animals, fungi or algae, either in their natural state or after being altered by processes such as removal of inedible or unwanted parts, drying, crushing, grinding, fractioning, filtering, roasting, boiling, pasteurization, refrigeration, freezing or nonalcoholic fermentation. These processes do not add substances such as fats, oils, salt or sugar into the original food. Processed culinary ingredients, or group 2 (G2) foods, are obtained from unprocessed or minimally processed foods through pressing, refining, grinding, milling or spray drying. These ingredients are not usually consumed alone, as they are normally added to G1 foods to cook or season them. Processed foods, or group 3 (G3) foods, are products made by adding G2 ingredients to G1 foods. Most G3 foods have few ingredients, and the processes involved in making them include various preservation or cooking methods. Fermented alcoholic beverages are classified as processed foods. Ultra-processed foods, or group 4 (G4) foods, are industrialized products which normally have five or more ingredients. While these often include ingredients used in processed foods, they also contain substances not commonly used in domestic cooking and additives which have the purpose to imitate taste, smell, or texture of G1 or G3 foods. G1 foods are commonly absent, and if present, represent a small proportion of these G4 formulations. Ultra-processed foods are manufactured with the use of industrial processes that have no domestic equivalents, such as extrusion and moulding and preprocessing for frying.

We have used NHANES provided Food Codes energy values. For hand-made recipes, NOVA was applied to underlying ingredients (Standard Reference Code), while energy values were calculated using variables from the Food and Nutrient Database for Dietary Studies and the US Department of Agriculture National Nutrient Database for Standard Reference. This classification process has been previously described in further depth.4, 24 Energy intakes for each NOVA group were calculated, which was then used to determine the percentage of contribution of each group to total daily energy intake (% E). Participants who did not respond to both dietary recalls were excluded from our analysis.

2.4 Oral examination

We used two measures of dental caries as outcomes in this study: dental caries experience, represented by the DMFT index and untreated dental caries, which was defined as the presence of at least one surface with a surface condition code 0-4 (ie a carious lesion), or the presence of at least one untreated carious root tip. DMFT was calculated as the total count of codes E, J, K, M, P, Q, R, T, X and Z in the ‘Coronal Caries: Tooth Count’ segment of the dental examination. Edentulous participants were included in the dental caries experience analyses and assigned a DMFT of 28 (since the crown surface examination protocol excluded third molars). When analysing the prevalence of untreated caries, edentulous participants were excluded. Oral health examinations were conducted by licensed dentists in a room at the MEC using light, compressed air and a portable dental chair. Caries scoring criteria used in the dental examination, along with quality assurance and training/calibration details, are further described in Supplement S1. These processes are described in-depth in the NHANES plan and operations manual.25

2.5 Covariates

Covariates included in our analyses were age, gender, race/ethnicity, level of education, income, access to oral health services and total energy intake. Age was categorized as 20-29, 30-39, 40-49 and 50-59 for bivariable analysis, while it was used as a continuous variable in our models. Education was categorized as ‘Less than 9th Grade’, ‘9-11th Grade’, ‘High School Grad/GED or Equivalent’, ‘Some College or Associate Degree’ or ‘College Graduate or Above’. Race/ethnicity was categorized as follows: non-Hispanic white; Mexican American; other Hispanic; non-Hispanic black; non-Hispanic Asian; and Other race (including multi-racial). Income was assessed using the ratio of family income to localized poverty threshold levels. Access to oral health services was evaluated with the use of two variables. The first originated from the question ‘When did you last visit a dentist?’, which had its answers categorized into ‘Within last year’, ‘More than 1 year but less than 5 years’ and ‘More than 5 years, never or does not know’. The second originated from the question ‘During the past 12 months, was there a time when you needed dental care but could not get it at the time?’, with ‘Yes’ or ‘No’ as response options. Body mass index (BMI) was categorized into underweight for values <18.5, normal for values ≥18.5 and <25, overweight for values ≥25 and <30, and obese for values ≥30. A smoking status variable was constructed, where participants who answered ‘no’ to the question ‘Have you smoked at least 100 cigarettes in your entire life?’ were categorized as ‘never smokers’. If they answered ‘yes’ to this question and ‘no’ to the question ‘Do you now smoke cigarettes?’, they were categorized as ‘ex-smokers’, while, if they had answered ‘yes’ to both questions, they were categorized as ‘current smokers’. In all these variables, ‘Refuse’ and ‘Don't know’ were set as missing values.

2.6 Statistical analysis

NHANES dietary recall day 2 sampling weights were used in all analyses. Daily energy contribution of NOVA groups 1, 3 and 4 was cut into quartiles. Group 2 had to be cut into tertiles as more than 25% of the sample did not recall consuming foods in this group. Proportions were presented as weighted percentages, and means were presented alongside their standard errors. Between-group differences were assessed with the chi-square test for categorical variables and the Mann-Whitney-Wilcoxon or Kruskal-Wallis test (as appropriate) for continuous variables. Logistic regression models were estimated to assess controlled associations between quantiles of energy contribution of each NOVA group and the outcome of prevalence of untreated dental caries. Exponentiated coefficients from these models were presented as odds ratios (OR) with their respective 95% confidence intervals. We used Poisson regression to evaluate associations of NOVA groups with DMFT and number of decayed surfaces, with both outcomes modelled as count variables. DMFT models were offset by the total number of valid examined teeth. The coefficients of these models were also exponentiated and presented as rate ratios (RR) with respective 95% confidence intervals. All models were controlled for age, gender, race/ethnicity, income-to-poverty threshold ratio, level of education, access to oral health services, smoking status, BMI and total energy intake. Participants with missing data (509 in total) were excluded from these models. Analyses were conducted using R 3.6.1 (R Core Team, Vienna, Austria) and the package survey 3.36.26

3 RESULTS

We analysed data from 5720 individuals from continuous NHANES cycles 2011-2012 and 2013-2014. From these, 123 (2.2%) were edentulous and were excluded from the analysis of the prevalence of untreated dental caries. Mean DMFT was 9.7 (± 0.2), and the prevalence of untreated dental caries was 26.0%. The mean % E for each group were: 28.6% (± 0.5) for G1 foods, 4.3% (± 0.1) for G2 foods, 10.1% (± 0.2) for G3 foods and 56.9 (± 0.5) for G4 foods. The mean age was 39.6 (± 0.4) years, while women accounted for 50.8% of the sample. Mean daily energy intake was 2170 kcal (± 17.7).

Table 1 shows the outcome of the bivariable analysis for mean DMFT. Older, less educated and female participants had higher DMFT scores, on average. Differences in race/ethnicity were also significant, while poverty was not. Not being able to get dental care last year was also associated with a higher DMFT, as was a higher body mass index and being a current smoker.

TABLE 1. Mean DMFT score by sociodemographic and health characteristics in the NHANES cycles 2011-2014 No. participants DMFT (± SE) P-value Total 9.7 (0.2) Age 20-29 1432 5.3 (0.2) 30-39 1432 8.1 (0.3) 40-49 1442 11.1 (0.2) 50-59 1414 14.2 (0.3) <.001a Sex Male 2744 9.3 (0.2) Female 2976 10.2 (0.3) <.001b Race/ethnicity Non-Hispanic white 2245 10.2 (0.2) Mexican American 709 8.5 (0.4) Other Hispanic 520 9.6 (0.3) Non-Hispanic black 1301 9.3 (0.4) Non-Hispanic Asian 724 8.0 (0.3) Other race—including multi-racial 221 9.5 (0.6) <.001a Family Income-to-Poverty Threshold Ratio >1 4009 9.6 (0.2) ≤1 1317 10.5 (0.7) .512b Education Less Than 9th Grade 282 10.0 (0.6) 9-11th Grade 716 11.6 (0.5) High School Grad/GED or Equivalent 1195 10.9 (0.3) Some College or Associate degree 1882 9.7 (0.3) College Graduate or above 1643 8.5 (0.2) <.001a Last dental visit Within last year 3104 9.7 (0.2) More than 1 year but less than 5 years 1760 9.5 (0.3) More than 5 years, never or does not know 856 10.5 (0.6) .516a Could not get dental care (last year) Yes 1384 11.6 (0.3) No 4234 9.3 (0.2) <.001b Body Mass Index Underweight 99 9.4 (1.0) Normal 1670 9.2 (0.4) Overweight 1769 9.5 (0.2) Obesity 2151 10.4 (0.2) <.001a Smoking status Never smoker 3454 8.5 (0.2) Ex-smoker 979 10.7 (0.3) Current smoker 1284 12.1 (0.5) <.001a a Kruskal-Wallis test. b Mann-Whitney-Wilcoxon test.

Bivariable analysis of untreated dental caries is shown in Table 2. Age and gender were not associated with this outcome, while remaining socioeconomic variables, access to dental care and ability to get it, smoking status and BMI were associated with the prevalence of untreated dental caries.

TABLE 2. Prevalence of untreated caries by sociodemographic and health characteristics in the NHANES cycles 2011-2014 No. participants % Untreated Caries P-valuea Total 5597 26.0% Age group 20-29 1431 29.0 30-39 1414 28.6 40-49 1408 23.2 50-59 1344 23.0 .334 Sex Male 2682 26.9 Female 2915 25.1 .298 Race/ethnicity Non-Hispanic white 2170 21.9 Mexican American 706 35.7 Other Hispanic 511 28.6 Non-Hispanic black 1278 40.7 Non-Hispanic Asian 722 16.1 Other race—including multi-racial 210 32.1 <.001 Family Income-to-Poverty Threshold Ratio >1 3952 21.6 ≤1 1257 45.9 <.001 Education Less Than 9th Grade 274 44.0 9-11th Grade 678 48.2 High School Grad/GED or Equivalent 1160 38.7 Some College or Associate degree 1850 26.1 College Graduate or above 1633 9.9 <.001 Last dental visit Within last year 3069 16.5 More than 1 year but less than 5 years 1722 35.0 More than 5 years, never or does not know 806 50.9 <0.001 Could not get dental care (last year) Yes 1362 52.2 No 4136 18.5 <.001 Body Mass Index Underweight 91 26.2 Normal 1635 22.8 Overweight 1737 22.6 Obesity 2105 31.7 <.001 Smoking status Never smoker 3429 20.2 Ex-smoker 962 21.3 Current smoker 1203 46.5 <.001

Table 3 shows associations between DMFT and dietary measures. The only NOVA group associated with DMFT was processed culinary ingredients, with a higher DMFT at the top tertile. Total energy intake was inversely associated with mean DMFT. No crude estimates were statistically significant. After adjustment, processed food dietary contribution in the second (RR 0.92; 95% CI 0.87-0.98) and the fourth (RR 0.89; 95% CI 0.84-0.94) quartiles of G3%E were associated with lower DMFT. Consumption of unprocessed or minimally processed foods was not associated with any of the outcomes in our study. A higher intake of processed culinary ingredients also was associated with fewer decayed surfaces (RR 0.66; 95% CI 0.44-0.99). Consumption of ultra-processed foods was associated with caries experience at its top quartile (RR 1.10; 95% CI 1.04-1.16) after adjustment for confounders.

TABLE 3. Association between DMFT index and daily energy intake contribution quantiles for each NOVA group and total energy intake, NHANES, 2011-2014 Quantile range N DMFT (± SE) DMFT – Rate Ratio (95% Confidence interval) Crude Adjusteda Unprocessed or minimally processed foods Quartile 1 (reference) 0.0-17.5 1275 9.9 (0.3) Quartile 2 17.6-26.6 1302 10.0 (0.2) 1.02 (0.96-1.07) 1.00 (0.94-1.06) Quartile 3 26.7-37.7 1364 9.5 (0.3) 0.97 (0.89-1.05) 0.98 (0.93-1.05) Quartile 4 37.8-98.3 1779 9.6 (0.3) 0.97 (0.90-1.06) 0.98 (0.93-1.02) Processed culinary ingredients Tertile 1 (reference) 0.0-1.9 1774 9.6 (0.3)b Tertile 2 2.0-5.1 1913 9.5 (0.4)b 1.01 (0.92-1.12) 0.99 (0.93-1.06) Tertile 3 5.1-51.5 2033 10.1 (0.2)b 1.06 (1.00-1.14) 0.99 (0.94-1.05) Processed foods Quartile 1 (reference) 0.0-3.0 1637 10.4 (0.5) Quartile 2 3.1-7.6 1529 9.4 (0.4) 0.89 (0.81-0.98) 0.92 (0.87-0.98) Quartile 3 7.7-14.6 1348 9.9 (0.2) 0.94 (0.85-1.03) 0.94 (0.88-1.01) Quartile 4 14.7-76.6 1206 9.3 (0.3) 0.89 (0.81-0.96) 0.89 (0.84-0.94) Ultra-processed foods Quartile 1 (reference) 0-44.5 1604 9.5 (0.3)b Quartile 2