Association between urinary biomarkers of total sugars and sucrose intake and BMI in a cross-sectional study

Obesity is an important modifiable risk factors for chronic diseases. While there is increasing focus on the role of dietary sugars, there remains a paucity of data establishing the association between sugar intake and obesity in the general public. The objective of this study was to investigate associations of estimated sugar intake with odds for obesity in a representative samples of English adults. We used data from 434 participants of the 2005 Health Survey of England. Biomarkers for total sugar intake were measured in 24h urine samples and used to estimate intake. Linear and logistic regression analyses were used to investigate associations between estimated intake and measures of obesity (BMI, waist circumference and waist-to-hip ratio) and obesity risk., respectively. Estimated sugars intake was significantly associated with BMI, waist circumference and waist-to-hip ratio, and these associations remained significant after adjustment for estimated protein intake. Estimated sugars intake was also associated with increased odds for obesity based on BMI (OR 1.02; 95% CI 1.00; 1.04 per 10 g), waist-circumference (OR 1.03; 95% CI 1.01; 1.05) and waist-to-hip ratio (OR 1.04; 95% CI 1.02; 1.06); all OR estimates remained significant after adjusting for estimated protein intake. Our results show a significant association between biomarker-estimated total sugars intake and both measures of obesity and obesity risk, confirming positive associations between total sugar intake, measures of obesity and obesity risk. This biomarker could be used to monitor the efficacy of public health interventions.

and disaccharides added to foods and beverages by the manufacturer, cook or consumer, 23 and sugars naturally present in honey, syrups, fruit juices and fruit juice concentrates" (1)) 24 have received increasing attention from the WHO 1 as well as the UK government 2 and the 25 UK's Scientific Advisory Committee on Nutrition (SACN) 3 . While sugar intake is often 26 associated with an increased risk of obesity 4 , the evidence available from observational 27 studies is more ambiguous and shows significant associations for sugar-sweetened 28 beverages (SSB) 5,6 only, but fails to show consistent associations for intake of sugars as 29 nutrients [6][7][8][9] . However, in most observational studies, sugar intake was assessed using self -30 report. It is likely that this has introduced bias, especially as underreporting of diet has been 31 found to be more prevalent among obese people [10][11][12] and it is sugar-rich foods that are most 32 commonly underreported 13 . It is possible that reporting bias contributes to the observed 33 inverse associations between sugar intake and BMI. 34 Urinary sugars have been investigated 14,15 and validated 16,17 as dietary biomarkers of 35 total sugars (i.e., the sum of intrinsic, milk and free sugars) and sucrose 18 and can help to 36 resolve the discrepancy between self-reported and true intake. This biomarker relies on the 37 precision was 1.77 and 3.80 %, respectively for an internal quality control sample containing 97 1 % N. The limit of quantification for the test was 0.018 % N. 98 2.5 Biomarker-based estimates of total sugars and protein intake 99 Estimated total sugars intake was calculated based on a calibration equation for the sugars 100 biomarker developed from a feeding study conducted in the UK (16), which describes the 101 association between the biomarker and true intake 17

Statistical analyses
113 All data were processed using R 24 version 3.3.2. Urinary fructose, urinary sucrose and the 114 sum of 24-hour urinary fructose and sucrose were skewed to the right and log 2 -transformed, 115 whereas biomarker estimates of total sugars and protein intakes were used without 116 transformation. The ratio of urinary sugars and estimated sugar intake, to urinary nitrogen 117 and estimated protein intake, respectively, were log 2 -transformed. We used the ratios of 118 estimated total sugars to protein intake or urinary sugars to urinary nitrogen to investigate the 119 effect of sugars while controlling for dietary composition. Unadjusted models were used 120 when investigating associations between estimated total sugars intake and BMI and obesity 121 risk (based on WHO definition either as BMI ≥ 30 kg/m 2 or waist-to-hip ratio > 0.85 for 122 women and > 0.90 for men), given the calibration equation for the sugars biomarker which 123 we used to estimate total sugars intake included age and sex. Models with uncalibrated 124 urinary fructose, uncalibrated urinary sucrose or estimated protein intake and BMI and 125 obesity risk were adjusted for age and sex. Associations with BMI were investigated using 126 linear regression models; OR for obesity (as estimate of risk) was estimated using logistic 127 regression. Urinary nitrogen or estimated protein intake was included in the models to control 128 for protein intake as a contributor to energy intake. P<0.05 was used as threshold for 129 statistical significance.  Table  133 1. Complete data on age, sex, BMI, waist-to-hip-ratio and 24h urine volume were available 134 for 298 women and 200 men (n=498). Due to missing samples or insufficient volume, not all 135 samples could be analysed for urinary biomarkers; data on urinary sugars and nitrogen are 136 available for 261 women and 173 men only (n=434). 137 p<0.001) than those in the remaining sample. There were however no statistically significant 147 differences in BMI, waist circumference, waist-to-hip ratio or protein intake. circumference and waist-to-hip ratio), adjusted for age and sex, are shown in Table 2. We 153 found a significant positive association for 24h urinary sucrose with all measures of obesity; 154 these associations were strengthened when including 24h urinary fructose and 24h urinary 155 nitrogen in the model. Total urinary sugars were significantly associated only with waist 156 circumference and waist-to-hip ratio, although the former association became non-significant 157 after adjusting for urinary nitrogen. There were no associations between any marker and total 158 urinary fructose, whereas total urinary nitrogen was significantly associated with BMI and 159 waist circumference, but not waist-to-hip ratio. 160 Table 2: Associations between 24h excretion of sucrose, fructose and nitrogen and BMI, waist-circumference and waist-to-hip-ratio (β and 95% CI). Data were log2-transformed and models are adjusted for age and sex. Estimates in each column represent a separate model. Total urinary sucrose was positively associated with obesity risk when using waist 1 circumference as the obesity marker, and this association became stronger when adjusting 2 for 24h urinary fructose, and for both 24h urinary fructose and 24h urinary nitrogen (Table 3). 3 Total urinary sucrose was also positively associated with obesity risk when using waist-to-4 hip-ratio as obesity marker, but only after adjustment for urinary fructose and urinary fructose 5 and nitrogen. We found no statistically significant increase in obesity risk when using BMI as 6 the obesity marker. 7 Table 3: Associations between 24h uriny excretion of sucrose, fructose and nitrogen and and odds for obesity (OR and 95% CI). Data were log2-transformed and models are adjusted for age and sex. Estimates in each column represent a separate model. 3.3 Associations between estimated intake, measures of obesity and obesity risk 1 Estimated total sugars and protein intake were positively associated with BMI, waist 2 circumference and waist-to-hip ratio, both independently and when combined in the same 3 model (Table 4; Fig 1). They were also positively associated with obesity risk when using 4 waist-to-hip-ratio as the obesity marker. However, associations were weaker for BMI and 5 waist circumference. Significant associations were observed only for estimated protein intake 6 (both independently and in the combined model) and estimated sugar intake when using BMI 7 as the obesity marker, and only for estimated sugar intake in the combined model when 8 using waist circumference. 9

markers. 11
Associations between estimated sugars and protein intake and BMI, waist circumference and 12 waist-to-hip ratio in men (brown circles) and women (blue triangles) 13 Table 4: Associations between estimated total sugars and protein intake and BMI (β and 95% CI per 10 g) and obesity risk (OR and 95% CI per 10 g  19 We investigated the ratios of estimated total sugars to protein intake or urinary sugars to 20 urinary nitrogen to investigate the effect of sugar intake while controlling for dietary 21 composition (Table 5). These data showed a positive association between the urinary 22 sucrose-to-nitrogen ratio and measures of obesity, especially after adjustment for urinary 23 fructose-to-nitrogen ratio. The increase in urinary sucrose to nitrogen ratio was associated 24 with statistically significant increased risk of obesity (waist circumference and waist-to-hip 25 ratio) after adjusting for urinary fructose to nitrogen ratio in the model. We found no 26 statistically significant increase in obesity risk with estimated total sugars to nitrogen intake, 27 urinary sugars to nitrogen ratio or urinary fructose to nitrogen ratio. 28 Table 5:

Associations between ratio of sugar-to-protein intake, BMI and obesity risk
Associations between ratio of sugars and protein intake, and ratio of urinary sugars and nitrogen and BMI (β and 95% CI) and obesity risk (OR and 95% CI).

29
Estimates in each column represent a separate model. In this study, we have used exclusively biomarker and biomarker-estimated data and not selfreported data to investigate associations between sugar intake and obesity risk. In our study population, using biomarker-based intake estiamtes, sugars were significantly associated with BMI, waist circumference and waist-to-hip ratio, and these associations remained significant after adjustment for biomarker-based protein intake. Estimated sugars intake was also associated with increased odds for obesity as measured by BMI, waist-circumference and waistto-hip ratio. The association between sugar intake and obesity risk in the general public is difficult to investigate because of the known limitation of self-reported dietary assessment, in particular the tendency to underreport the intake of perceived unhealthy foods and foods with high sugar content, especially among overweight individuals 11 . Indeed, observational studies relying on selfreported intake have long produced inconsistent results and generated controversy. Consistent data are available only for an association between obesity and sweetened beverages but not total sugar intake 9 . The objective assessment of sugar intake using a dietary biomarker 16,17 relies on total daily sucrose and fructose excretion and therefore the availability of 24h urine samples, which are often not available. Previously, this biomarker has been adapted for use with spot urine samples, showing a significant association between sucrose intake and BMI in two subsets of a cohort study, EPIC Norfolk 4,19 . However, while the biomarker measured in 24-h urine samples has been thoroughly validated, this is not the case for the biomarker measured in spot urine samples. Controlled feeding studies are needed to investigate and characterize the use of sucrose and fructose from spot urine as a biomarker for sugars 18 .
Total sugars and protein intake in our women (117 g/d, 80 g/d) and men (162 g/d; 102 g/d) estimated using biomarkers was higher than in the 2008/9 UK National Diet and Nutrition Survey (NDNS) (women: 78 g/d men, 66 g/d: men: 107 g/d, 89 g/d) 25 . An explanation for this discrepancy, in particular for total sugars intake, is that the NDNS relies on self-reported data and underreporting, in particular of sugar intake, is likely.
Our data showed a significant association between biomarker-estimated total sugar intake and both measures of obesity and obesity risk, confirming positive associations between total sugar intake, measures of obesity and obesity risk. The main strengths of this study are that the samples are a representative selection of the English population and that 24h urine samples were available. Moreover, the calibration equation that was used to calibrate the biomarker and generate estimate of sugars intake was developed in a UK feeding study under a UK diet. Even though no data on energy intake were available in the study, we were able to partially control for non-sugars energy using an objective measure of protein intake. Limitations include the small sample size; many associations were of borderline statistical significance and a larger study would allow further exploration. Furthermore, the application of biomarkers assumes an equilibrium, i.e., that participants do not change their body composition 26 , which information was not available, and could have introduced bias.
There was no information about stomach ulcers -which increase gastrointestinal permeability for (unhydrolized) sucrose -or impaired kidney function, for example as a result of type 2 diabetes, which could affect urinary fructose and sucrose excretion. Previous research has shown that neither obesity nor stomach ulcers have a significant impact on the biomarker used 19,27 , but there is a paucity of data investigating the effect of impaired renal function. As sucrose is excreted rapidly and almost completely in urine 28 , it is unlikely that diabetic kidney disease affects urinary sucrose concentrations. The physiological processes are more complex for fructose as it involves active reabsorption in the kidney 29 and higher urinary fructose concentrations have been observed in patients with diabetes 30 , although it is not clear whether this is due to impaired kidney function. This would result in an overestimation of sugar intake in those participants.
While BMI is commonly used to diagnose obesity, there are some limitations due to its inability to discriminate between fat and lean mass 31 . We have therefore also included waist circumference and waist-to-hip ratio in our analyses and the results are comparable. Indeed, associations between estimated sugar intake and obesity risk are stronger when using waistcircumference and waist-to-hip ratio as measures of obesity.
A possible explanation for the association between sugar intake and measures of obesity could be that sugar intake is an important contributor of energy intake. The paucity of validated recovery biomarkers for fat and total carbohydrate intake makes it difficult to assess total energy intake without double labelled water 32 or retrospectively. Protein is the only macronutrient for which intake can be estimated reliably with a dietary biomarker, total urinary nitrogen excretion 12,23 . In the UK, protein contributed approximately 15% to 20% of total daily energy intake 25 and we have therefore used biomarker-estimated protein intake to partially adjust for non-sugar energy intake. Urinary nitrogen excretion was positively associated with BMI and waist circumference, but not waist-to-hip ratio. Independently, estimated protein intake was also associated with many measures of obesity and obesity risk based on BMI and waist-to-hip rato.
These associations remained significant when combining sugar and protein in the same model, although both became slightly attenuated (Fig 2). We have explored these relationships further by using uncalibrated biomarker data. Our data show a strong association between urinary sucrose and measures of obesity, as well as using les) nd r s obesity risk based on waist circumference and waist-to-hip ratio. These associations were generally strengthened when including sucrose and fructose in the same model. Conversely, there were no significant associations for urinary fructose and only few associations were significant for total urinary sugars.
These results suggest that the associations between sugar intake and measures of obesity are mainly driven by sucrose. In contrast to fructose, which is derived from dietary fructose and hydrolysed sucrose and extensively metabolised, the only source of urinary sucrose is dietary sucrose [14][15][16]33 , making it more sensitive to changes in sucrose intake, the main contributor to intake of free sugars in the UK. Furthermore, high-fructose corn syrup (HFCS) or isoglucose was not commonly used in England at the time of the study as import and production was tighly controlled as part of the European Union sugar regime (Commission Regulation (EC) No 314/2002). Therefore the main source of dietary fructose were fruit and fruit products, such that fructose was most likely a surrogate marker of their intake.
Our results show that urinary sugars can be used to estimate sugar intake in the general population when 24h urine samples are available. In the context of current discussions regarding sugar intake and the recently updated WHO recommendations on sugars intake (1), the biomarker could be used to monitor the efficacy of public health interventions. Furthermore, we showed significant associations between sugar intake and BMI, confirming results of previous observations in EPIC Norfolk 4,19 . It is the first time that such an association has been shown in a nationally-representative sample of the general population using a validated biomarker. Our data also show significant associations between protein intake and measures of obesity and risk, however, in contrast to protein, sucrose is not an essential part of the human diet and intake can be reduced without adverse effects.