Abstract
Objective
Controversy exists surrounding the health effects of added sugar (AS) and sugar-sweetened beverage (SSB) intakes, primarily due to a reliance on self-reported dietary intake. The purpose of the current investigation was to determine if a 6-month intervention targeting reduced SSB intake would impact δ13C AS intake biomarker values.
Design
A randomized controlled intervention trial. At baseline and at 6 months, participants underwent assessments of anthropometrics and dietary intake. Fasting fingerstick blood samples were obtained and analysed for δ13C value using natural abundance stable isotope MS. Statistical analysis included descriptive statistics, correlational analyses and multilevel mixed-effects linear regression analysis using an intention-to-treat approach.
Setting
Rural Southwest Virginia, USA.
Subjects
Adults aged ≥18 years who consumed ≥200 kcal SSB/d (≥837 kJ/d) were randomly assigned to either the intervention (n 155) or a matched-contact group (n 146). Participants (mean age 42·1 (sd 13·4) years) were primarily female and overweight (21·5 %) or obese (57·0 %).
Results
A significant group by time difference in δ13C value was detected (P<0·001), with mean (sd) δ13C value decreasing in the intervention group (pre: −18·92 (0·65) ‰, post: −18·97 (0·65) ‰) and no change in the comparison group (pre: −18·94 (0·72) ‰, post: −18·92 (0·73) ‰). Significant group differences in weight and BMI change were also detected. Changes in biomarker δ13C values were consistent with changes in self-reported AS and SSB intakes.
Conclusions
The δ13C sugar intake biomarker assessed using fingerstick blood samples shows promise as an objective indicator of AS and SSB intakes which could be feasibly included in community-based research trials.
Keywords: Added sugar, Sugar-sweetened beverage, Dietary biomarker, Isotope, Dietary assessment
In spite of a large body of evidence linking sugar-sweetened beverage (SSB) intake with adverse health outcomes, including increased risk of type 2 diabetes( 1 , 2 ), weight gain( 1 , 3 ) and obesity( 1 , 2 ), significant controversy exists surrounding the health effects of added sugar (AS) and SSB( 1 , 4 , 5 ) due primarily to a reliance on self-reported dietary intake methodologies( 5 – 7 ). This controversy has recently extended into the health and economic policy arena (e.g. references 8–10). Public health guidelines have recommended that consumers limit AS intake and replace SSB with water( 1 ), yet critics of these guidelines cite the use of research utilizing memory-based dietary recall methods as a fatal flaw, with significant public health consequences( 8 ). The availability of an objective indicator of dietary intake, such as an AS intake biomarker, could overcome this research limitation( 10 ).
Because corn and cane plants employ the rare C4 photosynthetic pathway, their sugars naturally contain a high concentration of 13C (the heavy stable isotope of carbon) relative to 12C( 11 ). After digested food is absorbed from the intestine into the bloodstream, the carbon within food becomes carbon within tissues. Thus, a high 13C:12C (measured as ‘δ13C’) in human blood may reflect a high 13C:12C in the diet, with corn and cane sugars as important contributors( 11 ). Because approximately half of AS is consumed in the form of SSB( 12 ) and the majority of the AS consumed is ultimately derived from corn and cane plants( 11 ), the δ13C value of human blood may be affected by SSB intake( 11 , 13 , 14 ).
Talking Health is a randomized controlled trial which aimed to reduce SSB intake among residents of rural, health-disparate communities( 15 ). Participants were randomly assigned to either a 6-month intervention targeting a reduction in SSB intake (SIPsmartER) or a matched-contact comparison group (MoveMore). This trial represents the first randomized controlled trial directly targeting SSB intake reduction to include assessments of the δ13C biomarker, using fingerstick blood samples( 13 ). Our objective was to determine if group differences in biomarker δ13C values were evident over the 6-month intervention period.
Materials and methods
Detailed design and methods for the Talking Health trial have been reported previously( 15 ). Briefly, eligible individuals (English-speaking adults ≥18 years of age, who consumed ≥200 kcal SSB/d (≥837 kJ/d) and self-reported no contraindications for physical activity) were recruited and enrolled from eight counties/cohorts in Southwest Virginia, USA, each spaced approximately 3 months apart. Within each cohort, participants were randomly assigned to SIPsmartER (n 155) or MoveMore (n 146). For randomization, equal numbers of envelopes were prepared containing the name of each study condition; participants selected a sealed envelope to determine their assigned condition. SIPsmartER targeted decreasing SSB consumption, with the primary goal of ≤8 fluid ounces (≤237 ml) per day. MoveMore targeted physical activity promotion and did not contain content related to SSB or other dietary factors. Conditions were matched in terms of contact (i.e. three small-group classes, one live teach-back call, eleven interactive voice response calls) and structure( 15 ). Participants were compensated with a gift card for completing assessments (baseline, $US 25; month 6, $US 50).
At baseline and at 6 months, participants underwent assessments of height and weight, measured in light clothing without shoes (scale model 310GS; Tanita, Tokyo, Japan). AS sugar and SSB intakes were assessed using three 24 h recalls, obtained using the five-step multiple pass method( 16 ). Recalls were collected within a two-week baseline testing period. Recalls were analysed using nutritional analysis software (Nutrition Data System for Research (NDS-R 2011), University of Minnesota, Minneapolis, MN, USA). The NDS-R database primarily utilizes the US Department of Agriculture’s Nutrient Data Laboratory for its food composition data, and is supplemented by information from food manufacturers and data available in the scientific literature. Imputation procedures are applied to minimize missing values. The database is 100 % complete for AS( 17 ).
Fasting blood samples were obtained via routine fingersticks (One Touch Fine Point Lancet; Johnson & Johnson Company, New Brunswick, NJ, USA). Blood samples were blotted on to sterilized binder-free glass microfibre filters (Whatman, type GF/D, 2·5 cm diameter; Whatman, Inc., Piscataway, NJ, USA), air-dried and then analysed for δ13C value using natural abundance stable isotope MS, as per our previous work( 13 , 14 ). Samples were analysed in triplicate; the analytical error associated with each measurement in the current investigation was 0·039 ‰. Stable isotope values are reported using standard δ-notation in units of ‘per mil’ (‰) relative to international standards (Vienna Pee Dee Belemnite (VPDB)). The unit ‘per mil’ is standard within stable isotope reporting and refers to the number of units out of 1000; similar to how ‘per cent’ refers to the number of units out of 100. Human blood samples have lower δ13C values than the VPDB standard, therefore the δ13C values presented here are less than 0. A δ13C value of human blood that is closer to 0, representing a higher 13C:12C in the diet, corresponds to a higher AS or SSB intake( 11 , 14 , 18 ). Alanine was used as an internal laboratory standard for carbon. A more detailed description of this biomarker technique, which includes the δ13C values of several dietary sources of AS and SSB, as well as foods with ‘low’ δ13C values which may contribute to total sugar intake (e.g. fruit), has been previously published( 18 ).
Five women who were pregnant were excluded from the present analysis (n 4, SIPsmartER; n 1, MoveMore). Descriptive statistics were used to summarize baseline demographic characteristics. The t test was used to compare group means; the χ 2 test was used to compare proportions across groups. Multilevel mixed-effects linear regression analyses were performed using the statistical software package Stata version 13 (2013) to account for clustering of individuals within cohorts. Results of intention-to-treat are presented. The models included controls for the following baseline covariates: age, gender, race/ethnicity, income, education level, health literacy level, employment status, number of children, smoking status and BMI. Correlational analyses were also performed.
Results
Demographic characteristics of participants (n 296) are presented in Table 1. Participants (mean age 42·1 (sd 13·4) years) were primarily female and Caucasian, and almost half of the sample (43 %) reported an annual household income of ≤$US 14 999. Most participants were overweight (21·5 %) or obese (57·0 %). There were no significant demographic differences between groups. At month 6, the overall retention rate was 74 %, which was not statistically different between conditions. Group differences over time (both P<0·05) were noted in BMI and weight, which decreased in SIPsmartER (−0·21 (95 % CI −0·35, −0·06) kg/m2, −0·5 (95 % CI −0·9, 0·0) kg) but did not change in MoveMore (0·10 (95 % CI −0·09, 0·30) kg/m2, 1·0 (95 % CI −0·2, 0·4) kg).
Table 1.
Baseline participant characteristics in the full sample, and in the SIPsmartER intervention and MoveMore matched-contact comparison groups, rural Southwest Virginia, USA
| Full sample (n 296) | SIPsmartER (n 151) | MoveMore (n 145) | Test statistic, | ||||
|---|---|---|---|---|---|---|---|
| Characteristic | Mean or n | sd or % | Mean or n | sd or % | Mean or n | sd or % | P value† |
| Age (years), mean and sd | 42·1 | 13·4 | 41·7 | 13·4 | 42·4 | 13·3 | t=−0·41 |
| P=0·68 | |||||||
| Gender, n and % | |||||||
| Male | 56 | 19 | 30 | 20 | 26 | 18 | χ 2=0·18 |
| Female | 240 | 81 | 121 | 80 | 119 | 82 | P=0·67 |
| Race, n and % | |||||||
| Caucasian | 275 | 93 | 137 | 91 | 138 | 95 | χ 2=2·22 |
| African American | 13 | 4 | 10 | 6 | 3 | 2 | P=0·14 |
| More than one race | 7 | 2·5 | 3 | 2 | 4 | 3 | |
| Other | 1 | 0·5 | 1 | 1 | 0 | 0 | |
| Ethnicity, n and % | |||||||
| Hispanic/Latina | 3 | 1 | 2 | 1 | 1 | 0·5 | χ 2=0·48 |
| P=0·79 | |||||||
| Education level, n and % | |||||||
| ≤High-school graduate | 93 | 31 | 49 | 32·5 | 44 | 30 | χ 2=0·15 |
| Some college or greater | 203 | 69 | 102 | 67·5 | 101 | 70 | P=0·70 |
| Annual household income, n and % | |||||||
| ≤$US 14 999 | 126 | 43 | 69 | 46 | 57 | 39 | χ 2=6·89 |
| $US 15 000–34 999 | 94 | 32 | 52 | 35 | 42 | 29 | P=0·08 |
| $US 35 000–54 999 | 39 | 13 | 18 | 12 | 21 | 15 | |
| ≥$US 55 000 | 37 | 12 | 12 | 8 | 25 | 17 | |
| Weight status | |||||||
| Weight (kg), mean and sd | 90·6 | 25·4 | 90·7 | 26·4 | 90·4 | 24·3 | t=0·09 |
| P=0·93 | |||||||
| BMI (kg/m2), mean and sd | 33·0 | 9·1 | 33·3 | 9·3 | 32·7 | 9·0 | t=0·49 |
| P=0·62 | |||||||
The t test was used to compare group means; the χ 2 test was used to compare proportions across groups.
Changes in biomarker δ13C values and self-reported dietary intake over the 6-month intervention period are presented in Table 2. A significant group by time difference in δ13C value was detected, with mean δ13C value decreasing (i.e. reflecting a reduction in AS and SSB intakes) in the intervention group (Table 2). Group changes in biomarker δ13C values over time were largely consistent with changes in self-reported AS and SSB intakes. Total sugar intake declined in the intervention group and there was a significant group difference over time in total sugar intake. Significant correlations were noted between self-reported SSB intake and δ13C values at baseline (r=0·259, P<0·001) and at month 6 (r=0·280, P<0·001). Changes in self-reported SSB intake and δ13C values were not significantly associated (r=0·066, P=0·261), nor were changes in self-reported AS intake and δ13C values (r=0·041, P=0·487). The correlation between changes in percentage of total energy from AS and δ13C value was not statistically significant (r=0·101, P=0·084). As expected, changes in total sugar intake and δ13C values were not significantly associated (total sugar, percentage of total energy: r=0·099; total sugar, grams: r=0·040; both P≥0·05).
Table 2.
δ13C sugar intake biomarker values and self-reported dietary intake (n 296) at baseline and after the 6-month intervention in the SIPsmartER intervention and MoveMore matched-contact comparison groups, rural Southwest Virginia, USA
| Baseline† | Month 6† | Adjusted change, baseline to month 6‡ | P value, | |||||
|---|---|---|---|---|---|---|---|---|
| Variable | Group | Mean | sd | Mean | sd | Mean | 95 % CI | group by time§ |
| δ13C (‰) | SIPsmartER | −18·92 | 0·65 | −18·97 | 0·65 | −0·05 | −0·10, 0·01 | <0·001 |
| MoveMore | −18·94 | 0·72 | −18·92 | 0·73 | 0·02 | −0·04, 0·08 | ||
| SSB (kcal/d)|| | SIPsmartER | 496 | 374 | 268 | 297 | −227*** | −326, −127 | <0·001 |
| MoveMore | 377 | 287 | 325 | 319 | −53** | −88, −17 | ||
| SSB (fluid ounces/d)¶ | SIPsmartER | 43 | 31 | 24 | 24 | −19*** | −28, −10 | <0·01 |
| MoveMore | 33 | 24 | 28 | 27 | −5*** | −7, −2 | ||
| AS (% of total energy) | SIPsmartER | 22 | 12 | 17 | 12 | −5*** | −7, −3 | <0·001 |
| MoveMore | 21 | 11 | 20 | 12 | −1 | −2, 0 | ||
| AS (g/d) | SIPsmartER | 108 | 92 | 74 | 88 | −34*** | −46, −22 | <0·001 |
| MoveMore | 95 | 66 | 89 | 70 | −6 | −14, 2 | ||
| TS (% of total energy) | SIPsmartER | 26 | 12 | 22 | 13 | −5*** | −7, −2 | <0·01 |
| MoveMore | 26 | 11 | 25 | 11 | −1 | −3, 1 | ||
| TS (g/d) | SIPsmartER | 130 | 98 | 94 | 97 | −36*** | −49, −23 | <0·001 |
| MoveMore | 117 | 67 | 110 | 75 | −8 | −17, 1 | ||
| Energy (kcal/d) | SIPsmartER | 1975 | 1100 | 1690 | 1099 | −285*** | −434, −136 | <0·05 |
| MoveMore | 1766 | 640 | 1723 | 682 | −44 | −136, 49 | ||
| Protein (% of energy) | SIPsmartER | 14·6 | 4·2 | 16·4 | 5·0 | 1·8*** | 0·9, 2·8 | NS |
| MoveMore | 15·2 | 4·3 | 15·9 | 4·4 | 0·7 | −0·02, 1·4 | ||
| Fat (% of energy) | SIPsmartER | 33·6 | 7·9 | 35·2 | 9·4 | 1·6* | 0·3, 2·9 | NS |
| MoveMore | 33·6 | 6·7 | 34·6 | 7·9 | 0·9 | −0·1, 1·9 | ||
| Carbohydrate (% of energy) | SIPsmartER | 51·0 | 10·4 | 47·6 | 11·3 | −3*** | −5, −0·04 | NS |
| MoveMore | 50·4 | 9·0 | 48·9 | 9·9 | −1·5*** | −2·5, −0·4 | ||
SSB, sugar-sweetened beverages; AS, added sugar; TS, total sugar.
Means and standard deviations are not adjusted for covariates.
Within-group statistical significance indicated by asterisks: *P<0·05, ** P<0·01, *** P<0·001.
Models control for baseline covariates including age, gender, race/ethnicity, income, education level, health literacy level, employment status, number of children, smoking status and BMI.
1 kcal=4·184 kJ.
1 fluid ounce=29·57 ml.
To determine if weight loss impacted our findings related to changes in biomarker values, we developed a separate model to predict change in δ13C value for each group and included changes in weight, AS and SSB intakes as predictors. The model for the MoveMore comparison group was not significant (P=0·4695). The model for the SIPsmartER group was significant (P=0·008) and the only significant predictor of change in δ13C value was change in SSB kcal (P=0·019). This suggests that weight loss does not impact change in δ13C values when holding SSB and AS intake changes at the same level.
Discussion
The Institute of Medicine and others have highlighted the need for novel methods to objectively assess dietary intake, including biomarkers of food and nutrient intakes( 10 , 19 ). Common limitations of existing biomarker techniques include cost, participant burden and degree of invasiveness( 20 ). The current investigation describes results from the first randomized controlled trial evaluating an SSB intake reduction intervention to include assessment of the δ13C sugar intake biomarker with particular application to SSB. Group differences in biomarker δ13C values and in self-reported AS and SSB intakes were detected over the intervention period, and in weight and BMI. Importantly, this biomarker was evaluated using minimally invasive fingerstick blood samples obtained in a field research setting. These findings suggest that the δ13C value of fingerstick blood shows promise as a biomarker of AS – and by extension – SSB intake, which could be feasibly included in large-scale, community-based research trials.
Mean δ13C value in this sample was higher than that reported in a university community( 14 ), which is expected given the high SSB consumption reported by participants (~400–500 kcal/d; 1674–2092 kJ/d). Although the significant correlations between biomarker values and self-reported SSB intake at baseline and at month 6 were considered modest (i.e. r~0·3)( 21 ), these correlations may underestimate biomarker validity due to dietary intake under-reporting( 22 ), which is a particular problem when reporting intake of socially undesirable foods such as SSB( 10 , 23 ).
Strengths of the current investigation include a large sample size, the low degree of invasiveness for sample collection and the investigation of changes in biomarker values in response to an intervention targeting a reduction in SSB consumption. Limitations include the use of self-reported AS and SSB intakes as the method for biomarker comparison and the inability of the δ13C biomarker to detect some forms of AS such as beet sugar, honey and maple syrups. These represent minor contributors of AS to the US diet (i.e. 22 %) compared with corn- and cane-derived sweeteners (i.e. 78 %)( 11 ); however, we did not assess these forms of AS in this sample.
The availability of an objective indicator of AS and SSB intake could overcome a commonly cited methodological limitation of research investigating the health effects of AS and SSB consumption( 5 – 7 , 10 ). Additional research is warranted to investigate the validity of the δ13C sugar intake biomarker in children and adolescents, who consume high amounts of AS, and in controlled feeding trials.
Acknowledgements
Financial support: This work was supported by the National Institutes of Heath (Principal Investigator J.M.Z., grant number 1R01CA154364). The National Institutes of Heath had no role in the design, analysis or writing of this article. Conflict of interest: The authors declare no conflicts of interest. Authorship: All authors formulated the research question and contributed to the study design. V.E.H. and J.M.Z. collected the data. A.H.J. performed the laboratory analyses. B.M.D., W.Y. and J.M.Z. conducted the data analysis. All authors contributed to manuscript drafting/revisions. Ethics of human subject participation: This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Virginia Tech Institutional Review Board. Written informed consent was obtained from all subjects.
References
- 1. US Department of Health and Human Services & US Department of Agriculture (2015) Scientific Report of the 2015 Dietary Guidelines Advisory Committee. Washington, DC: USDHHS and USDA. [Google Scholar]
- 2. Basu S, McKee M, Galea G et al. (2013) Relationship of soft drink consumption to global overweight, obesity, and diabetes: a cross-national analysis of 75 countries. Am J Public Health 103, 2071–2077. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Malik VS, Pan A, Willett WC et al. (2013) Sugar-sweetened beverages and weight gain in children and adults: a systematic review and meta-analysis. Am J Clin Nutr 98, 1084–1102. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Bray GA & Popkin BM (2014) Dietary sugar and body weight: have we reached a crisis in the epidemic of obesity and diabetes?: health be damned! Pour on the sugar. Diabetes Care 37, 950–956. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Kahn R & Sievenpiper JL (2014) Dietary sugar and body weight: have we reached a crisis in the epidemic of obesity and diabetes?: we have, but the pox on sugar is overwrought and overworked. Diabetes Care 37, 957–962. [DOI] [PubMed] [Google Scholar]
- 6. Pereira MA (2014) Sugar-sweetened and artificially-sweetened beverages in relation to obesity risk. Adv Nutr 5, 797–808. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Bingham S, Luben R, Welch A et al. (2007) Epidemiologic assessment of sugars consumption using biomarkers: comparisons of obese and nonobese individuals in the European Prospective Investigation of Cancer Norfolk. Cancer Epidemiol Biomarkers Prev 16, 1651–1654. [DOI] [PubMed] [Google Scholar]
- 8. Archer E, Pavela G & Lavie C (2015) The inadmissability of What We Eat In America and NHANES dietary data in nutrition and obesity research and the scientific formulation of national dietary guidelines. Mayo Clin Proc 90, 911–926. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. BloombergView (2015) Soda Tax Bubbles Up. http://www.bloombergview.com/articles/2015-02-24/soda-tax-bubbles-up (accessed October 2015).
- 10. Subar A, Freedman L, Tooze J et al. (2015) Addressing current criticism regarding the value of self-report dietary data. J Nutr 145, 2639–2645. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Jahren AH, Bostic JN & Davy BM (2014) The potential for a carbon stable isotope biomarker of dietary sugar intake. J Anal Atom Spectrom 29, 795–816. [Google Scholar]
- 12. Huth PJ, Fulgoni VL, Keast DR et al. (2013) Major food sources of calories, added sugars, and saturated fat and their contribution to essential nutrient intakes in the US diet: data from the National Health and Nutrition Examination Survey (2003–2006). Nutr J 12, 116. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hedrick VE, Davy BM, Wilburn GA et al. (2016) Evaluation of a novel biomarker of added sugar intake (δ13C) compared with self-reported added sugar intake and the Healthy Eating Index-2010 in a community-based, rural US sample. Public Health Nutr 19, 429–436. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Davy BM, Jahren AH, Hedrick VE et al. (2011) Association of δ13C in fingerstick blood with added-sugar and sugar-sweetened beverage intake. J Am Diet Assoc 111, 874–878. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Zoellner J, Chen Y, Davy B et al. (2014) Talking Health, a pragmatic randomized-controlled health literacy trial targeting sugar-sweetened beverage consumption among adults: rationale, design & methods. Contemp Clin Trials 37, 43–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Moshfegh AJ, Rhodes DG, Baer DJ et al. (2008) The US Department of Agriculture Automated Multiple-Pass Method reduces bias in the collection of energy intakes. Am J Clin Nutr 88, 324–332. [DOI] [PubMed] [Google Scholar]
- 17. University of Minnesota, Nutrition Coordinating Center (2016) Nutrition Data Systems for Research. Nutrient Completeness. http://www.ncc.umn.edu/products/nutrient-completeness/ (accessed March 2016).
- 18. Jahren AH, Saudek C, Yeung EH et al. (2006) An isotopic method for quantifying sweeteners derived from corn and sugar cane. Am J Clin Nutr 84, 1380–1384. [DOI] [PubMed] [Google Scholar]
- 19. Food and Nutrition Board, Institute of Medicine (2007) Dietary Reference Intakes: Research Synthesis Workshop Summary. Washington, DC: National Academies Press. [Google Scholar]
- 20. Thompson FE, Subar AF, Loria CM et al. (2010) Need for technological innovation in dietary assessment. J Am Diet Assoc 110, 48–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Dancey CP & Reidy J (2004) Statistics Without Maths for Psychology: Using SPSS for Windows. Harlow: Prentice-Hall. [Google Scholar]
- 22. Willett WC & Lenart E (1998) Reproducibility and validity of food-frequency questionnaires. In Nutritional Epidemiology, 2nd ed., pp. 101–147 [WC Willett, editor]. New York: Oxford University Press. [Google Scholar]
- 23. Johnson RK, Yon BA & Hankin JH (2008) Dietary assessment and validation. In Research: Successful Approaches, 3rd ed., pp. 187–204 [ER Monsen & L Van Horn, editors]. Chicago, IL: American Dietetic Association. [Google Scholar]