The Breath Carbon Isotope Ratio Reflects Short-term Added-Sugar Intake in a Dose-Response, Crossover Feeding Study of 12 Healthy Adults

ABSTRACT

Background

Objective dietary biomarkers are urgently needed for a wider range of foods and nutrients. The breath carbon isotope ratio (CIR; measured as δ¹³C values) has potential as a noninvasive measure of short-term added sugar (AS) intake but has not been evaluated in a controlled-feeding study.

Objective

The aim was to evaluate the effect of short-term AS intake on breath CIR in a dose-response, randomized, crossover feeding study.

Methods

Six men and 6 women, aged 25 to 60 y, were randomly assigned to a balanced sequence of 5 dietary treatments. Three treatments delivered low (0 g/d), medium (75 g/d), or high (150 g/d) amounts of AS over the course of a single day's breakfast and lunch and 2 switched high and low intake amounts between breakfast and lunch. Experimental meals delivered 60% of daily energy and added-sugar targets. There was a washout period of 1–2 wk between treatments. Breath was collected at 2-h intervals from 08:00 (fasting) to 16:00 h. Breath CIR was measured using cavity ring-down spectroscopy, and the effects of dietary treatments and baseline were evaluated using multivariate linear regression.

Results

Breath CIR showed a significant response to increasing AS intake at all sampling time points (all P < 0.0001), with a dose-response of 0.030 (95% CI: 0.024, 0.037) ‰/g. Fasting breath CIR (baseline) influenced postfeeding breath CIR at all sampling time points (P < 0.0001); however, effect sizes were largest in the morning. For afternoon-collected samples (14:00 and 16:00), the effect of recent AS intake (lunch) was 4-fold greater than the effect of previous added-sugar intake (breakfast).

Conclusions

These findings support the potential of the breath CIR as a biomarker of short-term AS intake in healthy US adults. More work is needed to evaluate other potential dietary effects and whether multiple breath collections could capture daily AS intake.

See corresponding commentary on page 457.

Introduction

There have been recent calls to identify and evaluate unbiased dietary biomarkers for a broader range of nutrients and foods than are currently available (1). Tissue and molecular carbon isotope ratios (CIRs) are being explored as biomarkers of added sugars (AS) and sugar-sweetened beverages (SSBs) due to the high CIR of corn and sugar cane (2, 3), which are the source of the majority of AS (>70%) and nearly all SSBs (>99%) in the United States (4). A limitation is that associations of whole-tissue CIR with AS and/or SSB intake are often relatively weak and nonspecific (5, 6). However, Cook and colleagues (7) demonstrated a strong relation between recent sugar intake and the CIR of blood glucose (R² = 0.90), suggesting such a measure might have potential as a biomarker of short-term AS intake. Tools for assessing diet over short time frames, like the 24-h recall or metabolomic biomarkers, can provide improved information on habitual diet when feasible to collect repeatedly over time (8–10).

The strong relation of recent sugar intake with the CIR of blood glucose suggests that there may be a similar association with the CIR of breath carbon dioxide due to the use of glucose as a metabolic fuel. Increased breath CIR in response to dietary sugars was observed when ¹³C was first explored as a tool for metabolic labeling and diagnostic studies in the late 1970s, when researchers noted that unlabeled sucrose used to increase the palatability of isotopically labeled aminopyrine significantly increased the baseline breath CIR (11). Subsequent work confirmed that corn and cane sugars have naturally elevated CIRs relative to most other foods (12, 13) due to their use of the C4 photosynthetic pathway for fixing carbon dioxide (14, 15). A recent study observed that the change in breath CIR in response to a dose of corn-derived glucose at natural abundance was large enough to be used to monitor rates of carbohydrate metabolism (16). However, little attention has been paid to the potential of the breath CIR as a tool to assess dietary intake of AS or SSBs. Breath CIR would be both noninvasive and feasible to collect repeatedly, as participants can easily self-collect breath samples. Cavity ring-down laser spectroscopy permits rapid, inexpensive measurement of the CIR of breath carbon dioxide in breath samples (17, 18); however, it is not known how sensitive breath CIR is to changes in AS intake.

Here we evaluate the effect of short-term AS intake on the CIR of breath carbon dioxide in a crossover feeding study, with 2 aims. The first was to evaluate the dose-response of breath CIR to 3 amounts of AS intake (low, medium, and high). The second was to evaluate the effect of a change in AS intake on subsequent measures of breath CIR. The results of this study will inform whether the breath CIR has potential as a biomarker of AS intake.

Methods

Recruitment

This study was conducted at The Center for Alaska Native Health Research clinic on the University of Alaska Fairbanks (UAF) campus during the summer of 2015. Participants were UAF faculty, staff, or students, recruited by fliers and e-mail announcements, who met the following criteria: ≥18 y age, weight stable, nonpregnant, and free from metabolic disease and food allergies. This study was reviewed and approved by the UAF Institutional Review Board (project #680491-10). We recruited 12 participants, 6 women and 6 men. All participants met with study staff prior to the beginning of the study to determine eligibility, provide informed consent, review experimental diets to ensure acceptability, and collect baseline demographic and anthropometric measures. These included age, race/ethnicity, height, weight, percentage of body fat, BMI, and self-reported activity level. Participants also reported smoking behaviors and use of dietary supplements and medications. Weight and body composition were measured with a Tanita body-composition analyzer (model TBF-300A). For inclusion in the study, participants were screened for weight stability; specifically, they were asked if they were following a weight-loss or athletic training regimen that would likely alter their body mass or composition during the period of the study.

Experimental design

The Breath Biomarker of Added Sugar Study (BBASS) was a short-term controlled-feeding study with a randomized crossover design. On each study day, fasted participants were provided breakfast at 08:05 and lunch at 12:05 h and breath samples were taken at 2-h intervals from 08:00 (fasted) to 16:00 h. There were 5 experimental treatments in total, 3 in which AS intake amount was constant across breakfast and lunch [“constant” treatments: low AS (LL; 0 g/d), medium AS (MM; 75 g/d), and high AS (HH; 150 g/d)], and 2 in which AS intake amounts were switched between breakfast and lunch [“switch” treatments: low to high AS (LH) and high to low AS (HL)]. The MM dose of 75 g/d was selected to approximately match US adult mean intake (NHANES 2011–2012) (19). Participants completed the “constant” AS treatments first, with 1 female and 1 male participant randomly assigned to each of the 6 possible treatment orders. All 12 participants completed all of the constant AS treatments. Once participants completed the 3 constant AS treatments, they were randomly assigned to 1 of the 2 switch treatments. From each of the 6 existing diet combinations, 1 participant was randomly assigned to HL and the other to LH. Two participants were unable to complete their final switch treatment, one in HL and one in LH. Thus, the sample size for the switch treatments was n = 11. Between each dietary treatment was a washout period of 1–2 wk; thus, the project took each participant 7–10 wk to complete.

Experimental procedures

Fasted participants arrived at the Center for Alaska Native Health Research clinic at 08:00, provided a fasting breath sample, and had weight and body composition measured for that participation day. Body temperature was measured using an Exergen TemporalScanner thermometer (Exergen Corporation) to verify that participants did not have a fever, as acute infection can alter breath stable isotope ratios (20, 21). Fasting status was queried and participants had the opportunity to report noncompliance without penalty. Participants ate breakfast in the clinic suite and then completed a computerized Automated Self-Administered 24-hour Recall (ASA24^® 2014; National Cancer Institute, Bethesda, MD). Participants completed the recalls on laptops at the clinic suite under supervision, which allowed study staff to assist with questions or problems. Participants returned to the clinic to provide breath samples at 10:00, 12:00, 14:00, and 16:00 h. Participants ate lunch in the clinic suite after they had provided the 12:00 breath sample. After providing the 16:00 sample, participants estimated minutes of physical activity that day on a 7-point scale: 0–20 min, 20–40 min, 40–60 min, 60–80 min, 80–100 min, 100–120 min, and >120 min. Smoking behavior was assessed at the conclusion of the study as current smoking status (yes/no) and number of cigarettes typically smoked 1) daily, 2) before 08:00, and 3) within study hours (08:00–16:00).

In addition to the 12 BBASS study participants, 3 female volunteers completed an extension study in which they followed the HH dietary treatment, but breath samples were collected every 10 min from 08:00 until 16:00 instead of every 2 h. These extension study participants were new to the study and were not recruited from the previously enrolled study participants.

Diets

All experimental diets were formulated in consultation with a registered dietician. Diets were 50% carbohydrate, 30% fat, and 20% protein. There were 6 treatment meals: low-, medium-, and high-AS breakfast, and low-, medium-, and high-AS lunch. The menus were created for a 2000-kcal daily requirement and then adjusted to create 4 calorie levels (2000 kcal, 2400 kcal, 3000 kcal, and 3400 kcal). Participants’ daily energy requirements were predicted from their height, weight, age, and self-reported activity level using the Harris-Benedict equation (22) and they were assigned to the nearest calorie level. Participants were required to consume all provided foods and beverages and were asked at the end of their first study day whether they felt like the diets were providing too much or too little food. Based on this feedback, 3 participants were switched to a different calorie level for their remaining study days. Participants’ AS and energy-intake amounts were allocated across meals as follows: 25% breakfast, 35% lunch, and 40% dinner. Thus, the experimental diets provided 60% of daily energy and daily AS targets (Table 1).

TABLE 1.

AS, macronutrient, and energy content of the 2000-kcal/d experimental diets¹

	Carbohydrates, g
	AS	Other CHO	Protein, g	Fat, g	Energy, kcal
Breakfast
Low AS2	0 (0)	62.50 (67.5)	25 (26.6)	16.7 (16.6)	500 (497.1)
Medium AS	18.75 (18.8)	43.75 (47.7)	25 (25.9)	16.7 (16.6)	500 (499.9)
High AS	37.5 (37.5)	25 (31.3)	25 (23.6)	16.7 (16.2)	500 (505.5)
Lunch
Low AS	0 (0.16)	87.5 (91.7)	35 (35.9)	23.3 (23.5)	700 (705.7)
Medium AS	26.25 (26.5)	61.25 (61.6)	35 (35.6)	23.3 (24.4)	700 (703.8)
High AS	52.5 (52.7)	35 (36.9)	35 (35.3)	23.3 (24.5)	700 (706.1)

¹Data are shown as “target (actual),” as they varied slightly in macronutrient content due to variation in food composition and menu requirements. AS, added sugar; CHO, carbohydrate.

²Low, Medium, and High AS diets were designed to reflect daily intakes of 0, 75, and 150 g/d AS, respectively. Diets provided 60% of total daily energy and AS targets (25% in breakfast and 35% in lunch), as dinner was not included. The masses for meals providing larger daily energy intake were larger, to maintain a constant proportion of energy intake.

Study foods are given in Supplemental Table 1. Low-AS meals were composed entirely of foods with no AS, whereas medium- and high-AS meals progressively replaced carbohydrate-rich meal components with alternatives containing AS. For example, the low-AS breakfast served shredded wheat cereal with raisins, almonds, milk, apple juice, and hard-boiled egg, whereas the high-AS breakfast served Honey Smacks cereal (Kellog's Company) with milk, almonds, powdered fruit drink (Tang; Mondelez International), and hard-boiled egg. Processed-food brands served in the no-AS treatment were carefully selected based on their ingredient list to ensure no AS [e.g., sugar-free bread, crackers (Triscuit; Nabsico), luncheon meat].

Participants were instructed not to consume any foods or beverages outside of the provided meals (other than water, coffee, or tea without additives) for the study period, and were given the opportunity to report lapses without penalty. These were minor and infrequent (8% of participant days) and were noted for potential exclusion from statistical analysis.

Breath collection and analysis

Breath was collected in 500-mL contoured media bags with an open-flow male connector (Isomark, LLC) using a 2-cm length of Tygon tubing over the hose barb as a disposable mouthpiece. Participants were instructed to partially exhale (approximately one-third of the exhalation) before collecting the remaining exhalation into the bag.

The CIR of breath samples was measured within 24 h of collection using a Picarro G2131-i cavity ring-down spectroscopy isotopic carbon dioxide analyzer (Picarro, Inc.), calibrated periodically using 2 certified gas mixtures of 1% CO₂ (δ¹³C values = −29.5‰ and −9.7‰) in balance nitrogen. Reference breath samples were run prior to each batch of samples to confirm analyzer stability. Samples were withdrawn from collection bags using a glass 50-mL gas-tight syringe (TOMAC Interchangeable; American Hospital Supply) fitted with a 1-way valve, pumping 3 times to ensure complete mixing. They were injected into a 200-mL/min stream of carbon dioxide–free air using a Genie^TM Plus syringe pump at a rate of 7 mL/min (Kent Scientific) to achieve a 30:1 dilution. The stream of diluted sample was subsampled via an open split and drawn through the analyzer at a flow rate of 30 mL/min. Diluted sample (CO₂) was monitored during sample analysis and the syringe pump flow rate was adjusted if the dilution proved to be too high or low. Once the analyzer readings stabilized, the sample CIR was recorded as a mean over ∼2 min of sample gas analysis. Because the analyzer records data every 7 s, these means were based on ∼17 measurements. By convention, CIRs are presented as δ¹³C values referenced to an international isotope standard, Vienna Pee Dee Belemnite (V-PDB), as follows:

(1)

where the ¹³C/¹²C of V-PDB is 0.0112372. Analytical precision was assessed as the SD of repeated measurements of reference breath samples; these were typically within 0.05‰.

Food analysis

The CIRs of all food components of the experimental diets were analyzed at the Alaska Stable Isotope Facility using elemental analysis–isotope ratio MS. Food samples were either oven dried at 60°C or freeze dried and homogenized and 0.2- to 0.6-mg portions were weighed into tin capsules using a microbalance (Sartorius CP2P; Sartorius AG). The capsules were crushed for introduction into the autosampler of an Elemental Analyzer (ECS 4010; Costech Analytical), interfaced with a DeltaV Isotope Ratio Mass Spectrometer via the Conflo IV Interface (Thermo Scientific). As above, food CIRs were reported as δ¹³C values (‰). A laboratory reference material (peptone) was run every 10 samples for quality control, and the SD of those values was within 0.3‰. Elemental analysis–isotope ratio MS provides concurrent measurement of nitrogen isotope ratios, which are reported as δ¹⁵N values relative to an international isotope standard, atmospheric nitrogen (AN), as follows:

(2)

where the ¹⁵N/¹⁴N of AN is 0.0036765.

Statistical analyses

Statistical analyses were conducted in JMP Pro version 13.1.0. Differences in the δ¹³C values of types of study foods were evaluated using ANOVA, using Tukey's honest significant difference test for pairwise comparisons. The effects of LL, MM, and HH intake were assessed at each experimental time point (08:00, 10:00, 12:00, 14:00, and 16:00 h) using linear models (n = 36), with sex, age, BMI, current smoking, physical activity, and the fasting breath δ¹³C value as covariates. Age, BMI, current smoking, and physical activity were not significant at any time point and were removed from all final models. Sex was a significant covariate at the 10:00 sampling time point and was retained in that model only. The effect of AS consumed at breakfast versus that consumed at lunch on afternoon breath δ¹³C values was evaluated in the LL, LH, HL, and HH dietary treatments, using linear models with the fasting breath δ¹³C as a covariate (n = 46). To evaluate dose-response we used the slope of the relationship between breath δ¹³C measured at the afternoon time points (14:00 and 16:00) and total grams of AS consumed (breakfast + lunch). Total grams of AS consumed varied across 5 dietary treatments (LL, LH, MM, HL, HH) and 4 dietary intake amounts (2000, 2400, 3000, and 3400 kcal/d); thus, it was treated as a continuous variable. Dietary intake of participants was evaluated from the 24-h recalls collected at the beginning of each study day; for each main study participant there were either 4 (n = 2) or 5 (n = 10) 24-h recalls, and for extension study participants there was 1 (n = 3) 24-h recall. Custom dietary variables were created from the Individual Foods and Nutrients output file of the ASA24. Foods were coded 0/1 for AS, SSBs, or “meat” (red meat, poultry, or egg), multiplied by grams of sugar (for AS and SSBs), grams of protein (for meat protein), or grams of fat (for meat fat), and summed for each 24-h recall. The associations of dietary intakes of AS (grams/day), SSB (grams of sugar/day), meat protein (grams/day), and meat fat (grams/day) with the mean fasting breath δ¹³C value were evaluated using Pearson correlation. Normality of residuals was confirmed for all linear models using the Shapiro-Wilks test.

Results

Participant characteristics are presented in Table 2. Participants were 50% female and their median age was 30 y (range = 25–60 y). One participant was Alaska Native (Iñupiaq) and the other 11 were White. Forty-two percent of participants were either overweight (3/12) or obese (2/12) based on BMI (Table 2). Three participants were current smokers. Participants reported a mean of 20–40 min of daily exercise (prior to 16:00) across the 5 study days. Participant body mass varied by <1% across all study weeks when measured as differences from week 1 (mean = −0.1% ± 1.0%, maximum = −3.1%; n = 58 participant-weeks). The 3 extension study participants were female and White, ranging in age from 37 to 45 y.

TABLE 2.

Demographic characteristics of the study participants

Characteristics	n (%)
Total participants	12 (100)
Female	6 (50)
Age
25–34y	7 (58)
35–44y	2 (17)
45–54y	1 (8)
55–64y	2 (17)
Race
White	11 (92)
Alaska Native (Inupiaq)	1 (8)
BMI (kg/m2)
<25	7 (58)
25–29.9	3 (25)
≥30	2 (17)
Current smoker
No	9 (75)
Yes	3 (25)
Supplement use1
No	9 (75)
Yes	3 (25)
Physical activity2
0–20 min/d	3 (25)
20–40 min/d	6 (50)
40–60 min/d	3 (25)

¹Three participants reported taking a multivitamin.

²Mean of same-day physical activity scores reported at the 16:00 sample collection over 5 study days.

The δ¹³C values of BBASS study foods varied widely depending on food type (Figure 1, Supplemental Table 1). Sweetened foods consisting almost entirely of AS (SSBs and powdered sugar candy) had the highest δ¹³C values (mean δ¹³C value = −11.3‰ ± 0.2‰), followed by unsweetened, animal-based foods (mean δ¹³C value = −19.8‰ ± 2.5‰). The lowest δ¹³C values were in unsweetened, plant-based foods (mean δ¹³C value = −27.6‰ ± 3.1‰). Sweetened, plant-based foods (“mixed”) had intermediate δ¹³C values that overlapped both plant- and animal-based foods (mean δ¹³C value = −24.3‰ ± 0.6‰).

FIGURE 1.

Carbon isotope ratios (presented as δ¹³C values) of food items used in the BBASS. Horizontal lines indicate category means (± SD). Means without a common letter differ (ANOVA with Tukey's HSD, P < 0.05). Food categories are as follows: “Plant” indicates plant-based foods without added sugar (n = 15); “Mixed” indicates plant-based foods with added sugars (n = 4); “Animal” indicates animal-based foods without added sugar (meat and dairy, n = 4); and “Sugar” indicates processed foods for which nearly all calories derive from added sugar (SSBs, powdered sugar candy, n = 3). BBASS, Breath Biomarker of Added Sugar Study; HSD, honest significant difference; SSB, sugar-sweetened beverage.

Breath δ¹³C values increased with increasing AS intake at all measurement time points (Figure 2A, Table 3). At 2 h postbreakfast (10:00), breath δ¹³C values were 1.8‰ and 0.9‰ higher in the HH and MM treatments, respectively, relative to the LL treatment. Sex was a significant covariate at the 10:00 time point, with the breath δ¹³C values of female participants 0.6‰ higher than those of male participants; however, this effect disappeared in the subsequent sampling time points. The effect of AS intake was similar but slightly lower in samples collected at 12:00 (Table 3). The largest effect of AS intake was observed 2 h following the start of lunch (14:00), in which breath δ¹³C values were 3.7‰ and 0.9‰ higher in the HH and MM treatments, respectively, relative to the LL treatment (Figure 2A, Table 3). Effect sizes were similar but slightly lower in the breath samples collected later in the afternoon (16:00). The dose-response of breath CIR, measured as the slope of afternoon breath CIR regressed on total (breakfast + lunch) AS intake (95% CI), was 0.031‰/g (0.024, 0.039‰/g) AS for samples taken at 14:00 and 0.029‰/g (0.022, 0.037‰/g) AS for samples taken at 16:00. When afternoon breath CIR measurements were averaged, the dose-response was 0.030‰/g (0.024, 0.037‰/g ) AS.

FIGURE 2.

Breath δ¹³C values at each sampling time point in the Breath Biomarker of Added Sugar Study (mean ± SE). (A) Participants who were maintained at the same amount of AS intake across breakfast and lunch. (B) Participants who were switched between high- and low-AS intake between breakfast and lunch in reference to those who were maintained at LL and HH. (C) Three extension study participants who were provided the HH dietary treatment and provided breath samples every 10 min throughout the study period. AS treatments were formulated to represent AS intakes of 0 g/d (low), 75 g/d (medium) and 150 g/d (high), with breakfast and lunch delivering 25% and 35% of daily calories and AS, respectively. AS, added sugar(s); HH, high, high; HL, high, low; LH, low, high; LL, low, low; MM, medium, medium.

TABLE 3.

Effect of AS treatment on breath CIR in 12 adults at 3 amounts of AS intake, controlling for baseline (fasting breath CIR)¹

Time point and effect2	Estimate (SE)	t	P	R 2
10:00
Intercept	−8.6 (1.2)	−7.4	<0.0001	0.89
Treatment
HH	1.8 (0.2)	9.1	<0.0001
MM	0.9 (0.2)	4.6	<0.0001
Sex, female	0.6 (0.2)	3.3	0.0028
Fasting breath CIR	0.7 (0.1)	13.0	<0.0001
12:00
Intercept	−13.4 (1.4)	−9.6	<0.0001	0.73
Treatment
HH	1.3 (0.3)	5.2	<0.0001
MM	0.7 (0.3)	2.8	0.009
Fasting breath CIR	0.5 (0.1)	7.6	<0.0001
14:00
Intercept	−16.5 (1.3)	−12.6	<0.0001	0.91
Treatment
HH	3.7 (0.2)	15.8	<0.0001
MM	0.9 (0.2)	3.8	0.0006
Fasting breath CIR	0.3 (0.1)	6.1	<0.0001
16:00
Intercept	−17.4 (1.3)	−13.4	<0.0001	0.90
Treatment
HH	3.4 (0.2)	14.9	<0.0001
MM	0.9 (0.2)	3.9	0.0005
Fasting breath CIR	0.3 (0.1)	5.7	<0.0001

¹ n = 36. AS, added sugar; CIR, carbon isotope ratio, expressed as a δ value [δ¹³C (‰) = (¹³C/¹²C_sample/¹³C/¹²C_standard − 1) × 1000]; HH, high AS breakfast and lunch; LL, low AS breakfast and lunch; MM, medium AS breakfast and lunch.

²Dietary treatment effect estimates are expressed relative to LL treatment.

Baseline contributed significantly to breath δ¹³C values at all time points; however, the effect was greatest in the morning-collected samples (10:00, 12:00) (Table 3). Analyses of afternoon breath δ¹³C values using the “switched” (LH, HL) and corresponding nonswitched (LL, HH) dietary treatments showed that, while both breakfast and lunch AS intake significantly affected afternoon breath δ¹³C values, the effect of lunch AS intake was greater than the carry-over effect of breakfast AS intake (Table 4, Figure 2B).

TABLE 4.

Effects of breakfast and lunch AS treatment on breath CIR in 12 adults at 4 amounts of AS intake (LL, LH, HL, HH), controlling for baseline (fasting breath CIR)¹

Time point and effect2	Estimate (SE)	t	P	R 2
14:00
Intercept	−16.1 (1.4)	−11.4	<0.0001	0.89
Breakfast, H	0.7 (0.2)	4.1	0.0002
Lunch, H	2.9 (0.2)	16.3	<0.0001
Fasting breath CIR	0.4 (0.1)	6.0	<0.0001
16:00
Intercept	−16.4 (1.3)	−12.3	<0.0001	0.89
Breakfast, H	0.6 (0.2)	3.5	0.0013
Lunch, H	2.8 (0.2)	16.6	<0.0001
Fasting breath CIR	0.4 (0.1)	6.4	<0.0001

¹ n = 46. AS, added sugar; CIR, carbon isotope ratio, expressed as a δ value [δ¹³C (‰) = (¹³C/¹²C_sample/¹³C/¹²C_standard − 1) × 1000]; H, high AS; HH, high AS breakfast and lunch; HL, high AS breakfast, low AS lunch; L, low AS; LH, low AS breakfast, high AS lunch; LL, low AS breakfast and lunch.

²Dietary effects are expressed relative to the L treatment.

In the HH treatment, breath δ¹³C peaked at 2 h postmeal and had decreased by 4 h postmeal, although not back to prefeeding levels (Figure 2A, C). Breath samples collected every 10 min in 3 extension study participants consuming the HH dietary treatment similarly showed peaks in breath δ¹³C at ∼1.5 to 2.5 h postmeal, although the timing of the peak varied slightly among individuals (Figure 2C).

Fasting breath δ¹³C values were variable both within and among participants. Within participants, fasting breath δ¹³C values ranged by up to 4.3‰, whereas among participants mean fasting breath δ¹³C values ranged 5.1‰ (mean = −23.5‰ ± 1.3‰). Mean fasting breath δ¹³C was associated with self-reported intakes of AS (r = 0.57, P = 0.0258), SSBs (r = 0.79, P = 0.0004), protein from meat (r = 0.66, P = 0.008), and fat from meat (r = 0.79, P = 0.0005); however, these relations were influenced by a single high-intake individual (Figure 3). When this individual was excluded, associations with SSBs (r = 0.54, P = 0.048) and fat from meat (r = 0.54, P = 0.045) remained significant.

FIGURE 3.

Associations of mean fasting breath δ¹³C values with mean dietary intakes of AS (ln g/d) (A), sugar from SSBs (ln g/d) (B), meat protein (ln g/d) (C), and meat fat (ln g/d) (D), where “meat” includes red meat, poultry, and eggs. Each data point is a mean of 4–5 breath δ¹³C values and 4–5 intake measurements from 24-h recalls, except for extension study participants (gray-filled symbols), which had only 1 measurement of fasting breath and one 24-h recall. A high outlier value for breath δ¹³C is indicated with a triangle symbol; when excluded, only SSBs and meat fat remain significantly associated with fasting breath δ¹³C values (P < 0.05). AS, added sugar(s); SSB, sugar-sweetened beverage.

Discussion

The CIR of breath carbon dioxide increased with AS intake in this randomized, crossover, dose-response feeding study of 12 male and female, mostly White adults. Breath CIR peaked ∼2 h following a meal containing AS and remained elevated for at least 4 h following the meal. The dose-response was 0.030‰/g (0.024, 0.037‰/g) cumulative AS intake; thus, the effect of a 10-g increase in AS intake (0.30‰) would be large relative to analytical precision (0.05‰). More recent AS intake (<4 h prior) had a larger effect on the breath CIR than less recent AS intake (>4 h prior). There was a significant effect of fasting breath CIR on the postfeeding breath CIR, although its effect waned throughout the day and by the afternoon was small relative to the effect of AS intake. The fasting breath CIR was significantly associated with habitual intakes of SSBs and fat from meat, although our power to detect such associations was low. These findings support the potential of the breath CIR as a measure of recent AS intake.

In the morning sampling time points, the effect of the high-AS breakfast on the breath CIR was double that of the medium-AS breakfast, matching the increase in AS dose. In contrast, in the afternoon sampling time points, the effect of the HH treatment on the breath CIR was 4-fold higher than that of the MM treatment. This difference cannot be explained by the AS dose alone, which doubled from the medium-AS lunch to the high-AS lunch. It may reflect the cumulative effect of high AS intake during both breakfast and lunch; alternatively, it may involve how much of the AS was consumed as SSBs. In the medium- and high-AS breakfasts, SSBs provided 46% and 57% of AS, respectively. In contrast, in the medium- and high-AS lunches, SSBs provided 0% and 58% of AS, respectively. Our findings suggest that AS consumed as SSBs may have a larger effect on breath CIR than those consumed in foods. Further research is needed to clarify the time course and effect of SSB intake on the breath CIR.

Our data support the hypothesis that the postfeeding breath CIR is increased by short-term AS intake, peaking at 2 h postexposure and remaining elevated for >4 h. These kinetics are similar to those of some urinary metabolomic biomarkers, including alkyl-resorcinols for whole-grain consumption (23), proline betaine for citrus fruits (24), and 2-furoylglycine for coffee (25). For biomarkers with relatively rapid turnover rates, urine collection over a 24-h period may be required to capture daily intake (26), whereas for those with slower turnover rates, the collection of a first morning void may be sufficient (27). It is possible that repeated breath sampling across a day could enable estimation of daily AS intake, similar to a 24-h urine collection, as breath collection is noninvasive and easy to self-administer. Such approaches to sampling over a day's typical intake patterns will likely be needed in the further evaluation of this candidate biomarker.

Consistent with a number of published studies on the isotope ratios of US foods, BBASS study foods consisting entirely of AS had the highest CIR, whereas unsweetened, plant-based foods had the lowest CIR (12, 13, 28–30). Unsweetened, animal-based foods also had elevated CIRs relative to plant-based, unsweetened foods (30, 31), overlapping in CIR with plant-based foods containing AS (“mixed”). Intake of animal-based foods, particularly meat, may influence the CIR of breath, as they do the CIRs of serum, plasma, RBCs, hair, feces, and urine (5, 6, 32, 33). Due to the dose-response nature of our study design for AS intake, we were unable to assess the effect of varying meat intake here. We expect AS to have a greater effect on breath CIR due to its higher CIR relative to animal-derived foods; however, this assumption and the effect of varying meat intakes on breath CIR remain to be investigated. It will be important to assess the response of breath CIR to AS intake in the context of varying meat intake in future evaluations of this biomarker. It is also important to bear in mind that the mechanism underlying this proposed measure is the elevated CIR of sugars deriving from corn and sugar cane, which are C4 plants. This measure would not be expected to work in contexts where sugars derive predominantly from sugar beet, a C3 plant—for example, much of Europe. Conversely, such a measure could potentially work better in contexts where sugars derive exclusively from sugar cane or corn. In 2018, worldwide sugar production from sugar cane was almost 7 times greater than sugar production from sugar beets (34).

We observed suggestive associations of the fasting breath CIR with self-reported intake of both SSB sugars and fat from meat (red meat, poultry, and eggs). During fasting, we expect the CIR of breath to reflect the CIR of energy stores, particularly stored fat, which we expect should integrate habitual dietary sources of fat and carbohydrate. Evidence from this study suggests that SSBs and animal fat have the largest effects, as would be predicted based on the CIRs of those foods (3, 13, 31); however, we note that our study had low power to detect associations. Future studies involving more participants will be needed to understand whether and how dietary intake affects fasting breath CIR.

The natural abundance CIR of breath carbon dioxide has also been explored as a proxy for metabolic substrate use in both human and animal studies (35). Lipids have naturally lower CIRs than nonlipid molecules from the same carbon source, due to an enzymatic preference for ¹²C in lipid synthesis (36). It follows that a shift from fat to carbohydrate metabolism should be accompanied by an increase in the natural abundance CIR of breath. Studies of exercising humans have found increases in breath δ¹³C values of up to 2‰ with increasing exercise intensity, as participants shifted toward greater carbohydrate use (37, 38). Some evidence also suggests that the breath CIR can also decline during fasting, with greater catabolism of fat reserves (21, 39), and during severe, acute immune response that provokes major, complex changes in metabolic flux (20, 40). Because of these effects, it is important that breath samples collected for the purpose of dietary evaluation be collected at rest and in individuals who are not showing signs of active infection.

This study had several strengths and limitations. The primary strength was the randomized crossover design of this dose-response controlled-feeding study. Participants were highly compliant, with 100% completing the first, dose-response phase of the study and 83% completing the second, diet-switch phase, and relatively few reporting consumption of nonstudy food or drinks. A strength of the breath CIR measurement by cavity ring-down spectroscopy is its simple, noninvasive collection and rapidity of measurement. A limitation of the study was its small sample size. While the study was well powered to detect changes in breath CIR in response to 3 amounts of recent AS intake, the small sample size limited our ability to evaluate associations between longer-term AS intake and the fasting breath CIR. The study was designed to allow participation during a workday and thus did not include foods consumed in the evening. We expect similar results for evening breath CIR to those observed in the afternoon, with more recent AS intake having a greater effect on breath CIR than less recent AS intake. Finally, as is typical for a dose-response study, 1 nutrient is manipulated in the context of a common background diet. This design does not permit evaluation of how other dietary variables may affect the biomarker measurement. Studies in the context of habitual dietary variation would be a logical next step in the evaluation of this biomarker.

In summary, the results of this study support the potential of the natural abundance breath CIR as a measure of short-term AS intake. Breath CIR responded strongly to increasing AS intake, with the peak response occurring 2 h postexposure and effects remaining for at least 4 h postexposure. The advantages of the breath CIR as a candidate biomarker include its simple, noninvasive collection and its ease of measurement. However, more study is needed to clarify the effect of other foods with elevated CIRs, such as animal products or specific forms of AS such as SSBs, and whether repeated breath collection and measurement could adequately capture daily intake.

Notes

This project was supported by the National Institute of General Medical Sciences (NIGMS) of the NIH through grant number P30GM103325.

Author disclosures: The authors report no conflicts of interest.

The contents of this article are solely the responsibility of the authors and do not necessarily represent the official views of the NIGMS or the NIH.

Supplemental Table 1 is available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: AN, atmospheric nitrogen; AS, added sugar(s); ASA24, Automated Self-Administered 24-h Recall; BBASS, Breath Biomarker of Added Sugar Study; CIR, carbon isotope ratio; HH, high-high AS treatment; HL, high-low AS treatment; LH, low-high AS treatment; LL, low-low AS treatment; MM, medium-medium AS treatment; SSB, sugar-sweetened beverage; UAF, University of Alaska Fairbanks; V-PDB, Vienna Pee Dee Belemnite.