The δ¹³C Value of Fingerstick Blood Is a Valid, Reliable, and Sensitive Biomarker of Sugar-Sweetened Beverage Intake in Children and Adolescents

Abstract

Background

Reliance on self-reported dietary intake methods is a commonly cited research limitation, and dietary misreporting is a particular problem in children and adolescents. Objective indicators of dietary intake, such as dietary biomarkers, are needed to overcome this research limitation. The added sugar (AS) biomarker δ¹³C, which measures the relative abundance of ¹³C to ¹²C, has demonstrated preliminary validity in adults.

Objective

The purpose of this investigation was to determine the comparative validity, test-retest reliability, and sensitivity of the δ¹³C biomarker to detect AS and sugar-sweetened beverage (SSB) intake using fingerstick blood samples in children and adolescents.

Methods

Children (aged 6–11 y, n = 126, 56% male, mean ± SD age: 9 ± 2 y) and adolescents (aged 12–18 y, n = 200, 44% male, mean ± SD age: 15 ± 2 y) completed 4 testing sessions within a 3-wk period. Participants’ height, weight, demographic characteristics, and health history were determined at the first session; 24-h recalls were obtained at each visit and fingerstick blood samples were collected at visits 1 and 3. Samples were analyzed for δ¹³C value using natural abundance stable isotope mass spectrometry. δ¹³C value was compared with dietary outcomes in the full sample, and in child and adolescent subgroups. t Tests and correlational analyses were used to assess biomarker validity and reliability, whereas logistic regression and area under the receiver-operator characteristic curve (AUC) were used to evaluate sensitivity.

Results

Reported mean ± SD AS consumption was 82.2 ± 35.8 g/d and 329 ± 143 kcal/d, and SSB consumption was 222 ± 243 mL/d and 98 ± 103 kcal/d. Mean δ¹³C value was –19.65 ± 0.69‰, and was lower in children than in adolescents (–19.80 ± 0.67‰ compared with –19.56 ± 0.67‰, P = 0.002). δ¹³C values were similar across sessions (visit 1: –19.66 ± 0.68‰; visit 3: –19.64 ± 0.68‰; r = 0.99, P < 0.001) and were associated (P < 0.001) with intake of total AS (grams, kilocalories: r = 0.29) and SSB (milliliters, kilocalories: r = 0.35). The biomarker was able to better discriminate between high and low SSB consumers than high and low AS consumers, as demonstrated by the AUC (0.75 and 0.62, respectively).

Conclusions

The δ¹³C biomarker is a promising, minimally invasive, objective biomarker of SSB intake in children and adolescents. Further evaluation using controlled feeding designs is warranted. Registered at clinicaltrials.gov as NCT02455388.

Introduction

Childhood obesity has become a major public health concern, with 17% of children and adolescents (in the US) now considered obese (1). Excessive consumption of added sugar (AS) has been suggested as a contributor to this public health problem (2–5). Despite recommendations from major health organizations (6–8), children and adolescents consume AS in excess of the 10% intake level currently recommended (9), with children consuming ∼16% of their total calories from AS (7). Sugar-sweetened beverages (SSBs) represent the primary source of AS, accounting for 8% of total caloric intake (10). Excessive AS intake, particularly in the form of SSBs, has demonstrated an association with obesity (11, 12). High AS consumption is also associated with an increased risk for dental caries (13), chronic diseases such as diabetes, hypertension, ischemic heart disease and stroke (6, 14, 15), stunted growth and development (9), and an inadequate micronutrient intake (8, 9, 16, 17).

Although research findings have suggested a relation between AS and SSB intake and several chronic diseases (6, 14, 15, 18–22), a causal relation is difficult to determine due to the limitations of self-reported dietary assessment methodologies (23). Misreporting is more pronounced in younger populations who may have difficulty estimating portion sizes, conceptualizing foods, understanding food preparation, and recalling all foods and beverages consumed (23–25). Furthermore, it is common for people of all ages to underreport foods deemed socially undesirable (e.g., sugar-rich foods) and those that are consumed as snacks (25), which makes the relation between dietary intake and disease prevention or progression challenging to assess. The Institute of Medicine and others have highlighted the need to develop and validate dietary biomarkers in order to objectively assess dietary intake (26, 27). Currently, there are no validated predictive dietary biomarkers of AS intake in children and adolescents—a population that consumes even greater amounts of AS than adults (28).

Naturally occurring variations in stable isotope ratios, including ¹³C:¹²C (reported as δ¹³C), have been used as objective measures of diet in modern humans (29–34). The δ¹³C value of a given plant sugar reflects the biochemical pathways used by the plant during photosynthesis. “C4” plants, such as corn and sugar cane, give rise to sugar with high δ¹³C value, while “C3” plants, which include the majority of fruits and vegetables, give rise to sugars with much lower δ¹³C values (33, 38). Therefore, the consumption of corn and cane sugars and their derivatives is reflected in the δ¹³C value of adult human tissues using a variety of biological sample types, including a minimally invasive fingerstick blood sample (29, 35). The δ¹³C AS biomarker has demonstrated promising preliminary evidence of validity and reliability in adults (26, 29, 31–33, 35). However, tissue turnover times may vary according to growth stage, which could impact biomarker values (35, 36). Thus, the purpose of this investigation was to determine the comparative validity and test-retest reliability of the δ¹³C biomarker of AS intake, particularly in the form of SSBs, using fingerstick blood samples in children and adolescents aged 6–18 y.

Methods

Subjects

Three-hundred sixty-four children (aged 6–11 y) and adolescents (aged 12–18 y) were screened online from a local university community in southwestern Virginia, between January 2014 and September 2015. Three-hundred twenty-six children and adolescents completed all study sessions (90% completion rate). Interested participants were enrolled in the study following parental consent and if they met eligibility criteria, as follows: aged 6–18 y; can read, write, and speak English; and willing to comply with all study procedures. Study participation required the completion of 4 laboratory sessions within a 3-wk period (Figure 1). Study sessions were primarily scheduled in the early afternoon to early evening period (i.e., 1500–1900), on days that would allow the dietary recalls to reflect 3 weekdays and 1 weekend day, on nonconsecutive days. The Virginia Tech Institutional Review Board approved the study protocol. This investigation is registered at clinical trials.gov as NCT02455388.

FIGURE 1.

Recruitment and procedures for a cross-sectional study assessing the reliability, validity, and sensitivity of the δ ¹³C biomarker of added sugar intake in children and adolescents.

Procedures

Following parental consent and participant assent, a health history questionnaire was administered to gather information regarding participants’ age, race, ethnicity, medical conditions, and medication use. Height was measured in meters without shoes using a wall-mounted stadiometer, and weight was measured in light street clothing without shoes to the nearest 0.1 kg (Scale Tronix 5002; Carol Stream, IL); values were used to calculate BMI and BMI percentile (37); BMI percentile was used to categorize the weight status of participants. Resting energy expenditure (REE) (38) was calculated to identify potential underreporters. Participants were classified as underreporters if their reported total caloric intake was <80% of their estimated REE.

At each session, a record-assisted 24-h dietary recall (24-HR) was administered by a trained research assistant using the multiple-pass method; visual aids, including food diagrams and food models, were used. The four 24-HRs were obtained on nonconsecutive days, including 1 weekend day, in a manner consistent with NHANES methodology (24, 25, 39). However, all participants were encouraged to have an adult present to aid in completing the 24-HR. The 24-HRs were analyzed with the use of NDS-R (Nutrition Data System for Research 2013; University of Minnesota, Minneapolis, MN) to quantify dietary intake variables using a 4-d average, including usual AS and SSB intake.

Non-fasting fingerstick blood samples were obtained at 2 of the 4 visits (visits 1 and 3; Figure 1) via routine fingersticks (One Touch Fine Point Lancet; Johnson & Johnson Company). Blood samples were blotted onto sterilized, binder-free glass microfiber filters (Whatman type GF/D, 2.5-cm diameter; Whatman, Inc.) and air-dried. Blood samples were analyzed in triplicate for δ¹³C value via natural abundance stable isotope mass spectrometry (Isoprime; Micromass UK Ltd), as described previously (29). δ¹⁵N values, measuring the relative abundance of ¹⁵N to ¹⁴N, were also determined and evaluated as possible covariates to control for the potential confounders of meat (i.e., livestock consuming corn feed) consumption on δ¹³C values (32, 40). Total variability across the 3 measurements was within 0.04‰.

Statistical methods

Statistical analyses were performed using statistical analysis software (Medcalc 17.8, STATA 14, SPSS 22). Descriptive statistics (mean ± SD) are reported for participant demographic characteristics and dietary intake variables [total kilocalories, total dietary AS (grams, kilocalories), and total SSBs (milliliters, kilocalories)]. Paired-sample t tests, independent-sample t tests, chi-square tests and one-factor ANOVA were used to evaluate differences within and between assessment methods and group differences. Test-retest reliability (i.e., δ¹³C values obtained at the 2 laboratory visits) was assessed using the intraclass correlation coefficient (ICC), and a Pearson correlation coefficient between δ¹³C values and reported dietary AS intake evaluated comparative validity. Lastly, binary logistic regression analyses and non-parametric receiver operator curves (ROCs) were used to evaluate the diagnostic value of δ¹³C biomarker using the fingerstick blood. Participant AS and SSB intakes based on 24-HR recalls were dichotomized into high and low subgroups; participants were characterized as high AS consumers if their AS consumption (kilocalories) constituted >20% of total calories (i.e., twice the recommended 10% of total kilocalories) and as high SSB consumers if their SSB intake (kilocalories) was >128 kcal/d (twice the recommended intake suggested by the American Heart Association) (9, 41). The area under the ROC (AUC) was used as a measure for the diagnostic accuracy of the δ¹³C biomarker, with values closer to 1.0 indicating greater ability to distinguish between low and high AS and SSB consumers. Logistic regression analyses were performed separately for gender and for the age groups of children (aged 6–11 y) and adolescents (aged 12–18 y). The variability in ROC AUC was assessed to ensure that the difference in the AUC was not significantly different. ROC analyses were evaluated with and without potential covariates (BMI, δ¹⁵N, age, and gender).

Results

Participant demographic characteristics, dietary intake, and biomarker values are presented in Table 1. The sample (n = 326) was balanced with regard to sex (49% males, 51% females), while 93% of participants considered themselves white (2% were Asian, 1% were African American, and 1% were other) and non-Hispanic (93%). Age ranged from 6 to 18 y (mean age: 12 ± 3 y). The sample comprised 126 participants who were children (aged 6–11 y; 44% girls) and 200 participants who were adolescents (aged 12–18 y; 56% female). Mean BMI was in the normal range with most participants categorized between the 5th and 85th percentile. Age group differences were noted between the percentage of children and adolescents classified as underweight and obese (P < 0.05). As expected, adolescents had greater REE and reported energy intake than did children. Approximately 12% of the sample were classified as underreporters; however, reliability (r = 0.992, P < 0.001) and validity (AS g/d, kcal/d: r = 0.23, P < 0.001; SSB mL/d: r = 0.35, P < 0.001 and SSB kcal/d: r = 0.35, P < 0.001) results were not different after excluding potential underreporters.

TABLE 1.

Participant characteristics of the full sample and among children and adolescent subgroups in a study evaluating the validity and reliability of the δ¹³C value of fingerstick blood as a biomarker of added sugar intake¹

Participant	Total	Children	Adolescents
characteristics	sample	(6–11 y)	(12–18 y)
Total number of participants, n	326	126	200
Male, n (%)	159 (49)	71 (56)	88 (44)*
Female, n (%)	167 (51)	55 (44)	112 (56)*
Age, y	12 ± 3	9 ± 2	15 ± 2
BMI status, n (%)
Underweight (<5th percentile)	12 (4)	9 (7)	3 (2)*
Normal weight (5th to <85th percentile)	254 (78)	101 (80)	153 (77)
Overweight (85th to <95th percentile)	35 (11)	11 (9)	24 (12)
Obese (≥95th percentile)	25 (8)	5 (4)	20 (10)*
BMI, kg/m2	20.0 ± 4.5	17.0 ± 2.8	21.8 ± 4.4*
BMI-for-age, percentiles	56.1 ± 27.9	49.6 ± 28.4	60.1 ± 26.8*
Resting energy expenditure, kcal/d	1891 ± 472	1736 ± 412	1992 ± 485*
Reported energy intake, kcal/d	2063 ± 552	1862 ± 418	2190 ± 588*
δ13C, ‰	–19.65 ± 0.69	–19.80 ± 0.67	–19.56 ± 0.67*
Added sugar
g/d	82.2 ± 35.8	73.2 ± 28.1	87.8 ± 38.9*
kcal/d	329 ± 143	293 ± 112	351 ± 155*
Sugar-sweetened beverage
mL/d	222 ± 243	177 ± 210	251 ± 254*
kcal/d	98 ± 103	79 ± 88	110 ± 110*

¹Values are means + SDs unless otherwise indicated. Chi-square tests were used to compare proportions across groups; independent-sample t tests were used to compare age group means; BMI-for-age percentiles were calculated according to CDC standards (37); resting energy expenditure was calculated based upon data from the Institute of Medicine (38); reported energy intake was the average energy intake from the four 24-h dietary recalls. *Different from children, P < 0.05.

Among children and adolescents in this sample, reported total daily AS intake was similar to the mean intake of AS by persons >2 y (83.9 g, 336 kcal/d) in the United States (8, 42). AS and SSB intakes were significantly different between children and adolescents (P < 0.001), with children reporting a lower consumption than adolescents.

Mean δ¹³C values at visits 1 and 3, respectively, were –19.66 ± 0.68‰ (range –22.26‰ to –17.79‰) and –19.64 ± 0.68‰ (range –22.39‰ to –17.84‰). Biomarker measurements were strongly correlated across the 2 visits (ICC = 0.99, P < 0.001) and for subsequent analyses the δ¹³C value at visit 1 was used. (Note: the following results do not differ when using the visit 2 δ¹³C value, or the average of the 2 measures.) The δ¹³C values differed between children and adolescents with lower values noted in children (i.e., farther from zero, reflecting a lower AS and SSB intake), which is consistent with self-reported AS and SSB intake. In the full sample, significant differences in δ¹³C value were noted between normal weight and obese participants (mean difference = –0.551‰, P < 0.001), but not between normal weight and overweight participants (mean difference = –0.196‰, P = 0.11) or by gender (mean difference = 0.112‰, P = 0.14).

Significant correlations were noted between δ¹³C values and self-reported dietary variables including: total AS grams per day and kilocalories per day (r = 0.23, P < 0.001), SSB milliliters per day (r = 0.35, P < 0.001) and kilocalories per day (r = 0.35, P < 0.001). Scatterplots (Figure 2) show that δ¹³C was significantly higher in the high AS and high SSB consumers. δ¹³C value was correlated (P < 0.01) with BMI (r = 0.28), δ¹⁵N (r = 0.63), and age (r = 0.17), and only marginally with gender (P = 0.07). After controlling for BMI, δ¹⁵N, and age, all differences in δ¹³C values between high and low AS and SSB consumers were maintained.

FIGURE 2.

δ¹³C values of fingerstick blood in children and adolescents with low and high SSB (A) and AS (B) intakes. Values are significantly higher in the high-SSB (high, n = 98; low, n = 228) and high-AS (high, n = 82; low, n = 244) consumers. **Means differ, P < 0.01. After controlling for age, BMI, and δ¹⁵N (‰), these differences were maintained, P < 0.01. AS, added sugar; SSB, sugar-sweetened beverage.

δ¹³C discriminated low SSB consumers from high consumers with a sensitivity of 52% and a specificity of 77% (at a cut-off of >–19.31‰; AUC = 0.70, SE = 0.03). Binary logistic regression with block entry using BMI, δ¹⁵N and gender provided a model [Model 1: 26.91 + (1.20 × δ¹³C) + (0.09 × BMI) – (0.65 × δ¹⁵N) – (0.74 × gender)] that significantly improved the ability to distinguish low from high SSB consumers with a specificity of 84% and a sensitivity of 60% (at a cut-off of >–19.51‰; AUC = 0.75, SE = 0.03; ROC AUC comparison: P < 0.01; see Table 2). Adding age to the above model did not improve the discrimination between high and low SSB consumers (AUC = 0.75). δ¹³C discriminated low AS consumers from high AS consumers with a sensitivity of 83% and a specificity of 38% (at a cut-off of >–19.85‰; AUC = 0.62, SE = 0.03). Adding BMI, δ¹⁵N, gender, and age did not improve the discrimination between high and low AS consumers (AUC = 0.61). The discrimination capability of δ¹³C was not significantly different between children and adolescents, or between males and females.

TABLE 2.

AUC for the associations between AS and SSB intake with δ¹³C in children and adolescents¹

	AS (AUC)	SSB (AUC)
Full sample	0.61 ± 0.03	0.75 ± 0.03
Gender
Male (n = 159)	0.59 ± 0.05	0.76 ± 0.04
Female (n = 167)	0.64 ± 0.05	0.68 ± 0.05
Age group
Children (n = 126)	0.70 ± 0.06	0.78 ± 0.05
Adolescents (n = 200)	0.60 ± 0.05	0.74 ± 0.04

¹All values are means ± SEs. Analyses for the full sample and age groups adjusted for δ¹⁵N, BMI and gender; analyses by gender adjusted for δ ¹⁵N, BMI and age. AS, added sugar; SSB, sugar-sweetened beverage.

Discussion

Our findings represent the first investigation to evaluate the validity, reliability, and sensitivity of the δ¹³C AS biomarker using fingerstick blood samples in children and adolescents. Only one prior investigation has evaluated the δ¹³C biomarker in youth, using hair samples from 51 children in the Alaskan native Yup'ik population (43). That investigation did not include a direct assessment of dietary intake, but used δ¹³C values from hair samples to demonstrate associations between biomarker-based estimates of AS intake with carious tooth surfaces. In the present investigation, the biomarker demonstrated strong test-retest reliability, similar to findings from previous investigations (29, 32). The δ¹³C biomarker demonstrated comparative validity as it was moderately correlated with AS intake, expressed in g and kcal, and SSB intake, expressed in mL and kcal. The association between δ¹³C and SSB intake was stronger than when compared with AS intake, consistent with the literature (29, 32–44). This is likely explained by SSBs being comprised of 99% corn and cane sweeteners (36), while AS could be derived from a wider variety of sugar sources (i.e., C4 or C3 plants) (35). This may also explain the ROC results, which indicated that the biomarker was able to better discriminate between high and low SSB consumers than high and low AS consumers. While the δ¹³C biomarker can be useful for identifying high and low AS and SSB consumers, the AUC for these variables was within the acceptable range (0.7–0.9) (45) for SSB but not for AS. In part, this could be attributed to the use of self-reported dietary data, which may underestimate biomarker validity due to both dietary underreporting and the variability in self-reported intake (46). Therefore, this technique merits further investigation using controlled feeding study designs in children and adolescents, given its potential to contribute to the investigations of the health consequences of AS and SSB intake (6, 8, 9, 13–16, 43, 47, 48).

The current investigation reported mean intakes of SSB intake for children (79 kcal/d) and adolescents (110 kcal/d) that were below the current estimates of 129 and 350 kcal/d from SSBs alone for children and adolescents, respectively, determined using a single 24-HR, in the United States (2). Johnson et al. (8) found that the mean AS intake, using two 24-HRs, for children aged 4–8 y was ∼336 kcal/d while adolescent males and females reported much higher intakes (467 and 371 kcal/d, respectively); our results are consistent with the current literature, indicating that our sample is generalizable with respect to AS and SSB intake in children. These estimates of AS and SSB intake far exceed recommendations made by the Dietary Guidelines for Americans (9), the American Heart Association (8, 41), and the WHO (6).

Strengths of this investigation include a large sample size (n = 326) with a wide age range (6–18 y). The approach used also allowed us to evaluate differences in dietary intake and biomarker values between children and adolescents. Utilizing fingerstick blood samples minimized the degree of invasiveness, which is a commonly cited limitation of biochemical data collection, particularly among children, and no adverse effects were reported.

We acknowledge several limitations in our study. Despite a large sample size, there was a lack of racial and ethnic diversity in our sample; minority populations are more likely to consume SSBs than their white counterparts (2). Pubertal status was not assessed; however, this was addressed by dichotomizing the sample into child and adolescent subgroups, which did not present any significant differences in AUC values. The δ¹³C biomarker is limited to assessing intake of sugar from C4 plants, which demonstrate a conspicuously high δ¹³C value (30, 34, 35). Although ∼75% of high AS foods carry the C4 signature in their carbon isotope composition (35, 36), there are a few other sweeteners with a C3 signature, including beet sugar, honey, and maple syrup, that can be used in products. However, these comprise the minority of AS used in the United States (35). Furthermore, the δ¹³C value in whole blood cannot distinguish between corn consumption and corn-derivative consumption (33, 35, 49). However, we and others have reported no significant differences in the relation between δ¹³C and SSB consumption when using corn consumption as a covariate (40, 44). An additional complication may be livestock fed corn products, which are then ingested as meat. To address this, the δ¹⁵N value was tested as a covariate to control for meat intake using methods proposed elsewhere (31, 32, 50). However for this study, the addition of δ¹⁵N did not enhance the ability of δ¹³C values to predict AS intake, as was previously seen in populations of mostly white US residents (40). The limitations of self-reported dietary assessment methodologies, including misreporting, are also acknowledged (23). Lastly, these results only apply to the timeframe studied (i.e., 3 wk). Longer-term studies are needed to determine whether the δ¹³C value assessed in fingerstick blood is a valid indicator of AS and SSB intake in children and adolescents over the course of ≥1 mo.

In conclusion, the δ¹³C biomarker holds promise as a minimally invasive, objective indicator of AS and SSB intake in children and adolescents. This biomarker has the potential to overcome limitations posed by self-reported dietary data that may aid in discovering consistent and causal relations between AS and SSB intake and health (26, 29, 35, 51). Further research is warranted using a controlled feeding study design, where AS intake and/or SSB intake is manipulated. This approach could also determine the dose-response relations between δ¹³C values and AS and SSB intake (52), which could potentially be translated into clinically relevant thresholds to objectively characterize the AS and SSB consumption of children and adolescents. As MS technology advances to include portable, lower-cost mass spectrometers (53), a valid stable isotope biomarker of AS and SSB intake could have substantial public health impact both for research and clinical practice.

Funding for this project was provided by NIH R21 HD078636 (to BMD); Virginia Tech Fralin Translational Obesity Research Center; Virginia Tech Institute for Society, Culture and Environment.

Author disclosures: CRM, CEH, AHR, JS, SKR, VEH, HAR, JCD, MIF, and BMD, no conflicts of interest.

Abbreviations used

AS

added sugar

ICC

intraclass correlation coefficient

REE

resting energy expenditure

ROC

receiver operating curve

SSB

sugar-sweetened beverage

24-HR

24-h dietary recall