Abstract
Background: The World Health Organisation recommends a free sugars intake of <10% due to the connection between sugar intake and non-communicable diseases including type 2 diabetes and obesity (1). An accurate tool is needed to measure free sugars intake on an individual and population level, as current tools, dependent on self-reported intake, are prone to recall error and bias (2, 3).
Objective: This study aims to measure the association between self-reported dietary intake from a sugar-specific FFQ (including free sugars), and 𝛿13C in hair samples as a predictive biomarker of sugar intake.
Method: This 10-week study compared predicted sugar intake using δ13C in a 3cm hair sample, collected at the baseline and final visit, to self-reported free sugars, total sugar, added sugars, and sucrose intake using a sugar-specific FFQ (week 4 and 7), in a subsample of 50 healthy volunteers from the Dunedin public. δ13C was analysed using isotopic ratio mass spectrometry. Associations were assessed using multiple regression analysis. Correlations were determined using Spearman’s correlations. Cross-classification was used to determine whether the participants were classed in the same intake tertile for both methods. Spearman’s correlation analysis was also used to determine consistency between the two δ13C biomarker measures, and the week 4 and 7 sugar-specific FFQ. Secondary analysis was conducted for total sugar intake from sugar-sweetened beverages (SSBs).
Results: Multiple regression analysis adjusting for δ15N, age, sex, and body fat percentage showed an increase of 1‰ of δ13C was related to an expected 25% higher free sugars intake as reported by the FFQ, however this was not statistically significant (95% CI=0.62, 2.52). Associations between δ 13C predicted sugar intake and self-reported total sugar, added sugars, and sucrose were not statistically significant (p>0.05). Correlation coefficients between δ13C predicted sugar intake, with and without adjusting for δ15N, and self-reported intake were very low for free sugars (r=-0.03 and 0.2) and added sugars. Correlations were higher for total sugar and sucrose, however, overall still low (≦35) (4). Cross-classification between predicted and self-reported free sugars intake by tertiles showed only 26.81% of participants were accurately classified. Results were the same for added sugars, with moderately stronger results for total sugar and sucrose. Week 4 and 7 of the sugar-specific FFQ were moderately correlated for free sugars (r=0.54). Baseline and final results for the biomarker were very strongly correlated (r=0.92). Secondary analysis revealed a significant association between self-reported total sugar intake from SSBs and predicted sugar intake using δ13C. Multiple regression analysis showed a significant association between the two estimates (p<0.5), with moderate correlations, using δ13C (r=0.42), and δ13C adjusted for δ15N (r=0.45), and 36.56% of participants were classified into accurate tertiles. The correlation between the week 4 and 7 FFQ was also stronger (r=0.79).
Conclusion: Based on this subsample, there no evidence of a significant association between predicted sugar intake using δ13C in hair, and self-reported intake using a sugar-specific FFQ except for total sugar intake from SSBs. Future validation studies are required to determine any possible uses for δ13C in hair as a prediction of sugar intake.