A Dual-Carbon-and-Nitrogen Stable Isotope Ratio Model Is Not Superior to a Single-Carbon Stable Isotope Ratio Model for Predicting Added Sugar Intake in Southwest Virginian Adults^,

Abstract

Background: An objective measure of added sugar (AS) and sugar-sweetened beverage (SSB) intake is needed. The δ¹³C value of finger-stick blood is a novel validated biomarker of AS/SSB intake; however, nonsweetener corn products and animal protein also carry a δ¹³C value similar to AS sources, which may affect blood δ¹³C values. The δ¹⁵N value of blood has been proposed as a “correction factor” for animal protein intake.

Objectives: The objectives were to 1) identify foods associated with δ¹³C and δ¹⁵N blood values, 2) determine the contribution of nonsweetener corn to the diet relative to AS intake, and 3) determine if the dual-isotope model (δ¹³C and δ¹⁵N) is a better predictor of AS/SSB intake than δ¹³C alone.

Methods: A cross-sectional sample of southwest Virginian adults (n = 257; aged 42 ± 15 y; 74% overweight/obese) underwent dietary intake assessments and provided finger-stick blood samples, which were analyzed for δ¹³C and δ¹⁵N values by using natural abundance stable isotope mass spectrometry. Statistical analyses included ANOVAs, paired-samples t tests, and multiple linear regressions.

Results: The mean ± SD daily AS intake was 88 ± 59 g and nonsweetener corn intake was 13 ± 13 g. The mean δ¹³C value was −19.1 ± 0.9‰, which was significantly correlated with AS and SSB intakes (r = 0.32 and 0.39, respectively; P ≤ 0.01). The δ¹³C value and nonsweetener corn intake and the δ¹⁵N value and animal protein intake were not correlated. AS intake was significantly greater than nonsweetener corn intake (mean difference = 76.2 ± 57.2 g; P ≤ 0.001). The δ¹³C value was predictive of AS/SSB intake (β range: 0.28–0.35; P ≤ 0.01); however, δ¹⁵N was not predictive and minimal increases in R² values were observed when the δ¹⁵N value was added to the model.

Conclusions: The data do not provide evidence that the dual-isotope method is superior for predicting AS/SSB intakes within a southwest Virginian population. Our results support the potential of the δ¹³C value of finger-stick blood to serve as an objective measure of AS/SSB intake. This trial was registered at clinicaltrials.gov as NCT02193009.

Introduction

The consumption of added sugars (ASs)⁵ and sugar-sweetened beverages (SSBs) provides 16% and 7%, respectively, of total energy intake in US adults (1, 2), and excessive intake of ASs and SSBs has been identified as a contributor to the increased prevalence of obesity and related comorbidities (3–5). ASs are sugars that are added to foods before or after processing and preparation as well as sugars added at the table (6). SSBs include regular sodas, fruit drinks, energy drinks, and tea or coffee sweetened with a caloric sweetener (5). Despite the implementation of many public policies (taxation, AS/SSB recommendations) (7–9), existing research on AS and health outcomes is often limited by its reliance on self-reported dietary intake, which makes it increasingly difficult to evaluate the effectiveness of AS/SSB reduction interventions. Although dietary records and recalls are considered the “gold standards” of dietary assessment (10–12), because of the self-reported subjective nature of these methods, under- or overreporting of certain dietary items may occur (e.g., underreporting of socially undesirable high fat and/or AS foods such as SSBs) (11, 13). The ability to objectively measure the effectiveness of AS/SSB policies on dietary consumption patterns is limited, as evidenced by a recent review acknowledging the need for objective measures of SSB intake (14). Thus, an objective measure, such as a dietary biomarker (15–19), could facilitate a more accurate determination of the impact of ASs and SSBs on the health status of the US population (20).

The δ¹³C value of blood as a novel biomarker of AS and SSB intake (20) has shown preliminary validity in several clinical laboratory–based investigations (i.e., data were collected in a clinical research laboratory) (21–24) and within community-based settings (i.e., data were collected in a field setting) (25–27). Corn [e.g., high-fructose corn syrup (HFCS)] and cane sugars, which are derived from C4 plants, exhibit high δ¹³C values relative to sugars derived from more common C3 plants because of enzymatic differences in the biosynthetic pathways of sugar fixation between the 2 types of plants (22, 28, 29). Because the δ¹³C value of living tissues is considered first and foremost to reflect the ultimate carbon source for metabolism (30), it may be possible to recognize the isotopic signature of AS consumption via tissue isotopic analysis (29). Although this approach has shown promise as an intake biomarker of AS/SSB intake (21, 31), several limitations need to be addressed before its validation for use in a clinical research setting.

First, although the previously mentioned C4 sugars comprise a majority of AS intake among Americans (20, 32), the consumption of C3 sugars (e.g., beet sugar, honey, and maple syrup) would not contribute to a high δ¹³C value, the mechanism upon which the biomarker is premised (21). However, C3 plants represent only 22% of ASs consumed in America compared with C4 sources of ASs (see Table 1). Second, because nonsweetener corn products also exhibit high δ¹³C values, the consumption of these items may affect δ¹³C values (20). Furthermore, the consumption of nonsweetener corn products (per capita availability = 26 kg) is much lower than sweetener corn and cane sugar consumption (per capita availability = 91 kg) based on USDA data (33). Third, animal protein consumption may affect δ¹³C values because corn is a primary food source of milk and meat animals in the United States (28); thus, corn-fed meat possesses a δ¹³C signature intermediate between C3 and C4 plants (20, 28). It has been suggested that the δ¹⁵N value may be used as a “correction factor” to account for animal protein consumption because of elevated δ¹⁵N values of animal products relative to plant products (20, 26, 35). This may be accomplished by using a dual-isotope model to explain AS intake by using δ¹³C as a predictor and δ¹⁵N as a covariate, which may increase the biomarker’s sensitivity for AS intake (20, 26). Nash et al. (26, 27, 35) showed the utility of this dual-isotope model in the RBC fraction from a Yup’ik population with a high marine animal intake. To date, no investigations have examined the ability of a dual-isotope method to predict AS intake or the impact of nonsweetener corn and terrestrial animal protein consumption on δ¹³C blood values within a US population.

TABLE 1.

Total energy contribution of added sugar to US diet, by type¹

Sweetener	Total energy consumed, %	Relative contributions of added sugar sources, %
C4 plant sources		78
High-fructose corn syrup	4.6
Other corn syrups	1.5
Cane sugar	4.0
C3 plant and other sources		22
Beet sugar	2.7
Other sweeteners (e.g., honey, maple syrup)	0.2
Total added sugars	13.0

¹Data are from the USDA Economic Research Service Sugar Yearbook tables 2012–2013 (33) and reference 20.

The objective of this investigation was to assess the ability of a dual-isotope model (δ¹³C and δ¹⁵N) to predict AS/SSB intake compared with a single-isotope model (δ¹³C) within a southwest Virginian adult population while concurrently exploring potential confounds to the biomarker. The specific aims of this investigation were as follows: 1) to identify which food sources (ASs, SSBs, nonsweetener corn products, and animal protein) are associated with δ¹³C values; 2) to determine if δ¹⁵N blood values are associated with animal protein intake within this population; 3) to assess the relative amounts of nonsweetener corn products (e.g., corn oil, corn starch), sweetener corn products, and AS intake to determine the magnitude of nonsweetener corn’s potential to confound interpretation of the AS biomarker; and 4) to determine if the dual-isotope model is a better predictor of AS/SSB intake than a model with only the δ¹³C value. Age is controlled for in the models, because previous investigations identified an effect of age on the model when predicting the δ¹³C value (25, 26, 36). We hypothesized that the δ¹³C value is correlated with AS/SSB and with nonsweetener corn and animal protein intakes and that the δ¹⁵N value is correlated with animal protein intake. We also hypothesized that the intake of nonsweetener corn is minimal relative to the consumption of sweetener corn and that using the dual-isotope method of δ¹³C and δ¹⁵N values to predict AS /SSB intake, while controlling for age, will increase the model’s explained variance.

Methods

Subjects and design.

Two hundred fifty-seven participants were included in this investigation, which used baseline data from 2 investigations. The first investigation (n = 60) included data from a previous cross-sectional trial that aimed to validate a 15-item beverage intake questionnaire (BEVQ-15) (21, 37). Participants (adults aged >21 y) were recruited from a local university community from June 2008 to June 2009. The second investigation (n = 197) is an ongoing trial [“Talking Health” (38)], which is a 6-mo, community-based, randomized controlled trial that targets SSB consumption behaviors among low-socioeconomic adult (>18 y) residents in rural southwest Virginia. Participants were recruited from April 2012 to March 2014. All participants from the validation study (n = 60) had δ¹⁵N values analyzed; however, only a subset of the Talking Health participants had δ¹⁵N values analyzed because the decision to measure both δ¹⁵N and δ¹³C was made after the start of data collection. Thus, a δ¹⁵N value sample size of 115 participants was used for data analyses. These studies were conducted according to the guidelines laid down in the Declaration of Helsinki. The Virginia Tech Institutional Review Board approved the study protocols, and participants provided written informed consent before enrollment.

Methods.

Participants underwent assessments of height, which was measured in meters without shoes by using a portable stadiometer, and weight, which was measured wearing light clothing without shoes to the nearest 0.1 kg by using a digital scale (model 310GS; Tanita). BMI was calculated, and dietary intake assessed by using three 24-h dietary recalls [Talking Health (38)] or one 4-d food intake record [BEVQ-15 validation study (37)]. Records and recalls were collected by trained research technicians who were supervised by a registered dietitian. The dietary intake records and recalls were analyzed by using Nutrition Data System for Research (NDSR) nutritional analysis software (39). Participants also provided demographic information (age, gender, race/ethnicity) and completed the BEVQ-15 (37, 40–42), which is a validated measure of habitual SSB consumption. All participants completed both a three-day 24-h recall or a 4-d food record as well as the BEVQ-15 (n = 257).

Fasting whole-blood samples were provided via a routine finger-stick and blotted onto sterilized Whatman spun-glass filters (type GF/D, 2.5 cm; GE Healthcare). Punches, 3.1 mm in diameter, were collected from air-dried samples, loaded into high-purity tin capsules, and quantitatively combusted to carbon dioxide and nitrogen in a Costech ECS 4010 Elemental Analyzer (Costech Analytical) coupled to a Delta V Advantage Isotope Ratio Mass Spectrometer (Thermo Fisher). Stable isotope values are reported in standard δ-notation relative to international standards (Vienna Pee Dee Belemnite for carbon and atmospheric AIR for nitrogen] by using the following equation:

where R is the ratio of heavy:light isotope (¹³C:¹²C or ¹⁵N:¹⁴N). Reported values represent the mean of 3 analyses (SD of the 3 measurements never exceeded 0.05‰).

Data analysis.

Average grams of ASs were calculated by NDSR, and average grams of nonsweetener corn (whole corn, corn chips and snacks, corn tortillas, corn-based cereal grains, corn-based prepared cereals, processed meat and dairy products, and condiments) in food products were extracted from the component/ingredient level of dietary intake by using NDSR (39). Because grams of ASs are extracted from items containing ASs (e.g., gelatin), we used a similar approach to extract grams of nonsweetener corn. For whole-corn food items (corn kernels, popcorn), the weight of the product (g; excluding water weight) was considered the total nonsweetener corn product amount. Similarly, for processed food items containing corn (corn tortillas, chips, cereal), the contribution of nonsweetener corn intake to the product was estimated from the total weight of the product (excluding water weight and grams of ASs). Animal protein intake (g) was extracted from the food group amount of dietary intake by using NDSR (39).

Statistical analyses were performed by using SPSS statistical analysis software (version 21.0 for Windows, 2012; IBM). Descriptive statistics (means ± SDs and frequencies) are reported for demographic characteristics. One-factor ANOVA evaluated differences in BMI and dietary variables between varying age groups (18–24, 25–44, 45–64, and ≥65 y old); differences between groups were assessed by using Tukey’s post hoc tests. Pearson correlations were used to determine associations between δ¹³C and δ¹⁵N values and intakes of ASs, nonsweetener corn products, and animal protein. Paired-samples t tests were used to examine differences in mean AS and nonsweetener corn product intakes. Multiple linear regression (MLR) models using δ¹³C and δ¹⁵N values of finger-stick blood to predict AS and SSB intakes were used. Skewness, kurtosis, and Kolmogorov-Smirnov tests were used to confirm normal data distribution on the following variables: δ¹³C and δ¹⁵N values, age, and intakes of ASs and SSBs. AS and SSB variables were log-transformed due to nonnormal data distribution. Variables were entered into an MLR model by using the “enter” method with either 1 or 2 independent variables (δ¹³C value or δ¹³C and δ¹⁵N values, with age as a covariate). Tolerance and variance inflation factor statistics were used to determine if multicollinearity issues were present, and both were found to be acceptable (i.e., tolerance >0.10 and variance inflation factor <10). By using the MLR model, prediction equations for AS and SSB intakes, which accounted for animal protein intake (via the δ¹⁵N value) and age, were developed. Missing data were addressed by using list-wise deletion methods for MLR and case-by-case deletion for paired t tests and ANOVA. The multiple regression analyses recommended approach (n ≥ 50 + 8m, where m equals the number of predictor variables) to detect a moderate effect size with 80% power and an α of 0.05 was applied (43). The a priori hypothesis included a maximum of 3 predictor variables per model; therefore, ≥74 participants provide sufficient power.

Results

Demographic and dietary characteristics.

Participants (n = 257) were primarily female and white, with a mean age of 42 ± 15 y (range: 18–89 y). Although BMI was widely distributed (in kg/m²; range: 17.3–71.7), 74% of the sample was considered overweight or obese (BMI ≥25) (Table 2). Total intakes of energy, ASs, SSBs, total nonsweetener corn, and animal protein are presented in Table 3. Participants aged 25–44 y consumed significantly more ASs than those who were older than 45 y (mean difference = 24.6 g/d) and more SSBs than those who were older than 65 y (mean difference = 237 kcal/d). No significant differences between age groups were found for total energy, nonsweetener corn, or animal protein intakes.

TABLE 2.

Participant characteristics¹

	n (%)
Gender
Male	60 (23)
Female	197 (77)
Race
White	235 (91)
African American	12 (5)
Asian	4 (1.5)
Other	2 (1)
More than 1 race	4 (1.5)
Hispanic/Latino	2 (1)
BMI, kg/m2
Underweight, ≤18.4	3 (1)
Normal weight, 18.5–24.9	66 (25.5)
Overweight, 25–29.9	64 (25)
Obese, ≥30	124 (48.5)
Class 1, 30–34.9	46 (18)
Class 2, 35–39.9	32 (12.5)
Class 3, ≥40	46 (18)

¹n = 257.

TABLE 3.

BMI and selected dietary intakes of southwest Virginian adults stratified by age¹

		Age
Characteristics	Total	18–24 y	25–44 y	45–64 y	≥65 y	P (ANOVA)
n	257	33	117	91	16
BMI, kg/m2	31.8 ± 9	28.3 ± 9a	33.8 ± 10b	31.3 ± 8a,b	27.2 ± 5a	≤0.01
Total energy,2 kcal/d	1867 ± 675	1869 ± 619	1961 ± 727	1756 ± 641	1804 ± 516	0.19
Added sugars,2 g/d	88.8 ± 58.8	84.0 ± 43.0a,b	102 ± 67.2a	77.8 ± 51.7b	61.3 ± 32.8b	≤0.01
Sugar-sweetened beverages,3 kcal/d	359 ± 347	392 ± 380a	420 ± 370a	299 ± 300a,b	183 ± 264b	≤0.05
Total nonsweetener corn,2 g/d	12.6 ± 13.3	8.7 ± 8.6	14.9 ± 14.6	11.4 ± 12.4	11.5 ± 14.1	0.06
Animal protein,2 g/d	45.9 ± 21.8	49.9 ± 20.2	46.2 ± 22.5	44.4 ± 22.2	44.4 ± 18.4	0.50
δ13C, ‰	−19.1 ± 0.9	−19.0 ± 0.8a,b	−19.0 ± 0.9a	−19.3 ± 0.7b,c	−19.7 ± 0.5c	≤0.001
δ15N,4 ‰	7.4 ± 0.5	7.2 ± 0.5a	7.5 ± 0.5a	7.4 ± 0.4a	7.2 ± 0.3a	≤0.05

¹Values are means ± SDs unless otherwise indicated. Labeled means in a row without a common letter differ, P ≤ 0.05.

²Assessed by dietary records and recalls.

³Assessed by the BEVQ-15 (37, 40, 41).

⁴n = 115, 17, 52, 36, and 10 for total adults and ages 18–24, 25–44, 45–64, and ≥65 y, respectively.

Finger-stick δ¹³C values ranged from −22.3‰ to −17.0‰ and δ¹⁵N values ranged from 6.1‰ to 8.6‰ (Table 3). Individuals aged 25–44 y demonstrated significantly higher δ¹³C values than did those aged 45–64 y (mean difference ± SE = 0.3 ± 0.1, P = 0.02) and those older than 65 y (mean difference ± SE = 0.8 ± 0.2, P ≤ 0.01). The overall association of the δ¹⁵N value and age was significant; however, post hoc tests were unable to detect with significant confidence which pairs of means differed between age groups.

Aims 1 and 2: determining associations of dietary variables with δ¹³C and δ¹⁵N.

The δ¹³C value was significantly correlated with intakes of ASs (r = 0.32, P ≤ 0.01) and SSBs (r = 0.39, P ≤ 0.01) but was not significantly correlated with intake of total nonsweetener corn (r = 0.06, P = 0.31); in addition, no significant correlations of δ¹³C or δ¹⁵N values were found with animal protein intake (r = 0.11, P = 0.08, and r = 0.12, P = 0.22, respectively) in this sample. However, when different sources of animal protein intake were examined, it was found that both intake of fish/shellfish (r = 0.12, P ≤ 0.05) and red meat (r = 0.16, P ≤ 0.05) were significantly correlated with δ¹³C values. No significant correlations were found for poultry intake and δ¹³C or for any animal protein sources and δ¹⁵N values. Fish/shellfish, poultry, and red meat intake comprised 18%, 53%, and 29%, respectively, of total animal protein consumption.

Aim 3: determining relative dietary contributions of nonsweetener corn and AS.

Mean AS intake ranged from 5.4 to 330 g/d, and mean total nonsweetener corn intake ranged from 0 to 86.0 g/d. Intake of ASs was significantly greater than intake of total nonsweetener corn (mean difference = 76.2 ± 57.2 g/d, P ≤ 0.001) (Figure 1).

FIGURE 1.

Contributions of added sugar and nonsweetener corn to the diet of southwest Virginian adults. Values are means ± SDs, n = 257. Means without a common letter differ, P ≤ 0.05.

Aim 4: Predicting AS and SSB intake from δ¹³C and δ^\5N values.

To assess the utility of the dual-isotope method, 4 MLR models (n = 115) were generated (Table 4). In model 1a, only δ¹³C was a significant predictor of AS intake (P ≤ 0.01); however; in model 2a, both the δ¹³C value and age were significant predictors of SSB intake (P ≤ 0.001). With the addition of the δ¹⁵N value, models 1b and 2b remained significant; however, minimal changes in R² were observed (i.e., R² change: 0.21 to 0.22). In model 1b predicting AS intake, the β-weights were no longer significant; however, for model 2b predicting SSB intake, the β-weights for the δ¹³C value and age remained relatively consistent and significant (P ≤ 0.01). In both models, the δ¹⁵N value was not predictive of AS or SSB intake. Figure 2 shows the relation between reported and predicted AS (Figure 2A) and SSB (Figure 2B) intakes using the following equations generated from the MLR models with the use of δ¹³C and δ¹⁵N values and age as predictors (AS R² = 0.11, SSB R² = 0.22; both P ≤ 0.001):

TABLE 4.

Predicting added sugar and sugar-sweetened beverage intake in southwest Virginian adults from δ¹³C and δ¹⁵N¹

Model and predictors	β	P	95% CI	R2	P
Added sugars, g/d
Model 1a
δ13C	0.28	≤0.01	0.07, 0.31	0.09	≤0.01
Age	−0.11	0.23	−0.01, 0.01
Model 1b
δ13C	0.19	0.10	−0.20, 0.28	0.11	≤0.01
δ15N	0.15	0.18	−0.09, 0.45
Age	−0.11	0.23	−0.09, 0.45
Sugar-sweetened beverages, kcal/d
Model 2a
δ13C	0.35	≤0.001	0.40, 1.12	0.21	≤0.001
Age	−0.28	≤0.001	−0.06, −0.01
Model 2b
δ13C	0.29	≤0.01	0.19, 1.09	0.22	≤0.001
δ15N	0.10	0.90	−0.44, 1.18
Age	−0.28	≤0.01	−0.06, −0.01

¹n = 115. The dependent variables (added sugar, sugar-sweetened beverages) were log-transformed. β, standardized β-weight.

FIGURE 2.

Associations of reported and predicted added sugar and sugar-sweetened beverage intake of southwest Virginian adults (n = 115). Values are log-transformed.

Discussion

This investigation contributes new information to a growing body of literature addressing δ¹³C as a potential biomarker of AS and SSB intakes (20–23, 25–29), by reporting findings from a large sample of community-dwelling US adults using the minimally invasive finger-stick sampling method. We explored potential confounds (age, nonsweetener corn intake, and animal protein intake) on the basis of findings from previous investigations (20, 23, 25–27) and found that age was a significant predictor of SSB but not AS intake, nonsweetener corn intake relative to ASs was significantly lower, nonsweetener corn and total animal protein intakes were not significantly correlated with the δ¹³C value, and total animal protein intake was not correlated with the δ¹⁵N value. We evaluated the ability of a dual-isotope method (δ¹³C and δ¹⁵N values) to predict AS and SSB intakes in a US population using an approach similar to that of Nash et al. (26, 27) in a Yup’ik population. Our findings were not consistent with our hypothesis or with the findings of Nash et al. (26) in that the addition of the δ¹⁵N value (dual-isotope method) did not significantly increase the explained variance of the model when predicting AS and SSB intakes. We note that key attributes of diet differed between the Yup’ik population studied by Nash et al. (26) and the population of our study: within our study, AS intake was slightly higher (89 vs. 74 g), SSB intake was substantially higher [2.5 vs. 1.4 servings (240 mL/8 fluid ounce serving size)], and marine and terrestrial animal protein intake was lower (1% vs. 18% and 8% vs. 11% of total energy intake, respectively). In addition, the mean δ¹³C value for this sample (−19.1‰) was greater than that reported in previous investigations [−19.3‰ (23) and −19.8‰ (26)].

Age.

δ¹³C values decreased significantly with age, but differences in δ¹⁵N values across age groups were not significant. In contrast, Nash et al. (26, 36) reported differences by age in δ¹³C values as well as in δ¹⁵N values, which may be attributed to the relative higher intake of marine animal protein as compared with that of the population in our study. Because age was significantly correlated with the δ¹³C value and intakes of ASs and SSBs in this investigation, and it was shown to be a significant predictor of the δ¹³C value in a previous study (25), the present investigation used age as a covariate when predicting AS and SSB intakes. Differences in AS and SSB intakes across age groups were similar to a comparable study sample (26), and others reported that different age groups consume varying amounts of ASs and SSBs (i.e., younger adults consume more ASs and SSBs than do older adults) (2, 44), which is consistent with findings from the present investigation.

Nonsweetener corn.

Nonsweetener corn consumption was comparable to intakes from Nash et al. (26) (11 vs. 12.6 g), and both studies showed no significant correlations of nonsweetener corn intake with δ¹³C blood values. This may be due in part to the significant differences in consumption of nonsweetener corn compared with other C4 sources of ASs (cane sugar, HFCS). In 2010, via the USDA’s Economic Research Service, the per capita availability of cane/beet sugar was 30 kg, total corn sweeteners was 29 kg, and nonsweetener corn products was 15 kg (relative contributions of cane vs. beet sugar are 45% and 55%, respectively) (32). When taking the relative contribution of beet sugar into consideration, the proportion of per capita availability of nonsweetener corn was less than one-third of the availability of ASs (cane and corn sweeteners). In comparison to this sample, the proportion of nonsweetener corn consumption to ASs was 16%, which was slightly less than US per capita availability. Two additional investigations also explored the impact of nonsweetener corn intake on the δ¹³C value within US populations. Fakhouri et al. (23) found no significant correlation between the δ¹³C value and nonsweetener corn intake (56% of participants were African American), which was assessed as a percentage of consumers (e.g., the percentage of people who reported consuming nonsweetener corn on a 24-h dietary recall) as opposed to actual total gram intake. Conversely, Yeung et al. (31) found a significant, but weak, correlation between the δ¹³C value and whole-corn intake (no race categories provided). Although the present investigation did not show a significant impact of nonsweetener corn consumption on δ¹³C blood values, due to the majority of participants being white, there are potential implications to these findings because other populations may consume greater amounts of nonsweetener corn products [e.g., Hispanic populations (45)].

Animal protein.

In this sample, no significant correlations between animal protein intake (marine or terrestrial) and δ¹⁵N values were found; however, δ¹³C values were found to be positively associated with both marine and terrestrial animal protein. The contrast between our results and those of previous studies (with other sample substrates not including finger-stick samples), which established significant associations between animal protein intake and both δ¹³C and δ¹⁵N values, may be attributable to multiple geocultural dietary factors (34). Relative to Nash et al. (26), the intake of marine animals by our sample population, which possess higher δ¹⁵N values relative to their terrestrial counterparts (34), was significantly lower. Furthermore, terrestrial protein intake, which comprised a majority of animal protein in our sample population, was not associated with δ¹⁵N in their sample population (26). Petzke et al. (46) found an association between δ¹³C and δ¹⁵N values and meat intake in a German population consuming primarily terrestrial animal proteins [and higher animal protein intake compared with this investigation (71 vs. 46 g/d)]. An additional investigation, based in the United Kingdom, found positive associations between δ¹³C and δ¹⁵N values and fish protein; however, only δ¹⁵N values were associated with terrestrial animal protein intake (47). The differences between our results and those in European populations may be explained by the higher overall meat intake in their population, resulting in larger intake variations between omnivorous and vegetarians in their sample set (47), or by the different feeding practices in German vs. American livestock; grass-fed cattle, which predominate in Europe, may possess higher overall δ¹⁵N values than those fed corn-based diets, possibly due to the fact that conventionally grown crops fed to American cattle are depleted in δ¹⁵N relative to grass/hay due to low δ¹⁵N values of synthetic fertilizers used in their production (48). Further studies within an American population are warranted to determine the association between animal protein intake and tissue isotope values.

Predicting AS and SSB intakes from δ¹³C and δ^\5N values.

By using only the δ¹³C value and age to predict AS and SSB intakes, this sample produced higher R² values than a previous study (26) (ASs = 0.09 vs. 0.03; SSBs = 0.21 vs. 0.05) and higher β-weights for the δ¹³C value (ASs = 0.28 vs. 0.20; SSBs = 0.35 vs. 0.21). Age was not a significant predictor of AS intake, although it was for SSB intake. This provides additional evidence that age should be included in AS/SSB prediction models, because the δ¹³C value and age were both significant predictors and the δ¹³C value produced a β-weight comparable to that reported by Nash et al. (26). When the δ¹⁵N value was added as a predictor, minimal changes were shown in R² values and β-weights for the δ¹³C value and age in the SSB model. In addition, the δ¹⁵N value was not considered a significant predictor in either model, which contrasts with previous findings (26).

This investigation extends findings by Nash et al. (26, 27) who investigated a unique population who consumes greater quantities of fish and marine animal protein than the US adult population. The present study inferentially evaluated the effects of animal protein intake in grams [i.e., absolute intake vs. percentage of energy (26)] and nonsweetener corn intake in grams [vs. percentage of consumers (23)]. We nevertheless acknowledge several limitations of this investigation. The consumption of C3 plants (beet sugar and maple syrup) is not reflected by this AS biomarker and this sample’s analysis of AS does not distinguish between C3 and C4 sources. However, because C3 plants provide only ∼22% of ASs in the US diet (which translates to ∼3% of total energy intake) (Table 1) and SSBs are typically sweetened by C4 sweeteners (HFCS), this may account for the lower R² value when predicting AS intake than when predicting SSB intake. In addition, AS intake was assessed via food intake records and recalls, whereas SSB intake was determined by the BEVQ-15 (37, 40, 41), which possibly explains the higher correlation of the δ¹³C value to habitual SSB intake rather than recent AS consumption, because the whole-blood δ¹³C value reflects intake over several months (20). However, it should be noted as a limitation that there is a potential for different levels of misreporting/error between the food records/recalls and the BEVQ-15. Typical correlations of nutritional biomarkers to their respective dietary variables average ∼0.39; however, a wide range of correlations have been reported when evaluating dietary biomarkers (0.03–0.73) (11). Nonetheless, acceptable correlations for this area of research range from 0.5 to 0.7 (11). Although the reported values are below this threshold (∼0.2), as with any study using self-reported dietary intake over-/underreporting errors are possible, thus demonstrating the need for future controlled feeding studies to further refine this AS/SSB biomarker approach.

In conclusion, the analysis of these cross-sectional data does not provide additional evidence that the dual-isotope method (i.e., measurement of both the δ¹³C and δ¹⁵N value) is superior for predicting AS and SSB intakes within a US population. Data do suggest that the δ¹³C value of finger-stick blood is an objective measure of AS and SSB intake and that age should be considered as a significant predictor of AS and SSB consumption. Future directions should include further assessing the impact of various types of animal proteins (specifically fish) on the dual-isotope method and exploring the potential of this biomarker within racially diverse populations (23), as well as in children and adolescents. The δ¹³C value of finger-stick blood is a promising biomarker of AS and SSB intake, which could be used in community and field-type settings and in large-scale epidemiologic trials because of the ease of sample collection.