The positive association between serum uric acid, impaired fasting glucose, impaired glucose tolerance, and diabetes mellitus in the ELSA-Brasil study

There is a conflict in the literature regarding the association between serum uric acid (SUA) levels and glycemic status. Therefore, we evaluated the association between SUA level and glycemic status – impaired fasting glucose (IFG), impaired glucose tolerance (IGT), and diabetes mellitus – and insulin resistance, in a large Brazilian study. This is a cross-sectional, observational study with 13,207 participants aged 35-74 years, at baseline (2008-2010) of the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). A multinomial regression analysis was performed to test the association between SUA and glycemic status (IFG, IGT, and newly diagnosed type 2 diabetes at the cohort baseline) after adjustments by age, sex, skin color, body mass index, physical activity, smoking, alcohol consumption, comorbidities, and medicines use. Logistic regression model was used to evaluate the association between SUA and insulin resistance by HOMA-IR. Stratified analyses by sex were performed. The mean age (standard deviation) was 51.4 (8.9) years, 55.2% of participants were women. There were 1,439 newly diagnosed diabetes. After all adjustments, higher SUA was associated with IFG, IGT, and diabetes, with odds ratio (OR) = 1.15 (95%CI: 1.06; 1.25), 1.23 (95%CI: 1.14; 1.33), and 1.37 (95%CI: 1.24; 1.51), respectively. There was association between SUA levels and insulin resistance with OR = 1.24 (95%CI: 1.13; 1.36). In analysis stratified by sex, higher SUA persisted independently associated with impaired glycemic status. Our results suggest that a higher SUA levels were significantly associated with glycemic status in a large Latin American population, mainly among women. blood pressure, prevalent coronary heart disease, and glomerular filtration rate; Model 3 – Model 2 plus physical activity, smoking, alcohol and dairy intake, fructose and animal protein consumption, triglycerides, and HDL cholesterol; Model 4 – Model 3 plus use of medicines that may interfere with uricemia (diuretics, losartan, fenofibrate, acetylsalicylic acid, gout medica-tions). Tests were conducted on whether uric acid levels and fructose intake were effect modifiers on the associations of glycemic status by multiplicative interaction terms between uric acid/fructose confounding factors, such as skin color, glomerular filtration rate, and use of interfering drugs. Another significant adjustment was dietary intake of dairy products and animal protein recognized as factors that can be protective or increase the risk of hyperuricemia and diabetes, respectively 12,32 . We also tested whether SUA and skin color were effect modifiers on glycemic status associations by multiplicative interaction terms between uric acid/skin color, with no change in OR. A positive, statistically significant, association persisted after we tested for effect modification by fructose intake, a sugar associated with the rapid uric acid hepatic synthesis, insulin resistance, and hyperglycemia 12 .


Introduction
Diabetes mellitus is a public health problem. Being a highly prevalent condition that increases cardiovascular risk and associated morbimortality, diabetes is associated with high health expenditures and large economic impact on health systems 1 . About 463 million adults were diagnosed with diabetes worldwide by 2019 2 . If effective interventions are not performed, diabetes will be present in over 700 million adults, by 2025 1 .
Several risk factors that contribute to the development of diabetes are well known, such as obesity and sedentary lifestyle 1 . Uric acid is the final product of purine metabolism under the action of the xanthine oxidase enzyme 3 . Higher serum uric acid (SUA) levels are associated with the development of gout, cardiovascular diseases, renal disease, and inflammatory process 3 . The association between elevated SUA levels and increased risk of hyperglycemia has been reported by some authors 4,5,6,7,8 . A meta-analysis showed that elevation of 1mg/dL in the SUA level was associated with a 17% risk increase of developing diabetes. Moreover, it was suggested that the increase of 1mg/dL on uricemia would be comparable to the increase of 1kg/m 2 in body mass index (BMI) for the development of type 2 diabetes 4 . In contrast, other studies, including some with genetic evaluation, have not shown positive association between higher uricemia and hyperglycemia 9,10,11 . Studies have found that urate is involved in oxidative stress, systemic inflammation, insulin sensibility, and intrahepatic fructose metabolism 12 .
SUA test is an accessible and simple laboratorial test, and is a potentially modifiable factor. Investigating whether SUA levels are associated with diabetes, pre-diabetes status, and insulin resistance may be relevant for clinical and public health purposes. To date, these studies are lacking in a large mixed-race and multi-ethnic populations. Brazil is a developing country, with more than 210 million inhabitants distributed in the territory of continental dimensions, and with some epidemiological profile distinct from other countries. We did not find Latin America studies evaluating the association between uricemia and diabetes, and epidemiological studies about uricemia are scarce and small. Meta-analysis evaluated do not include studies with specifically Latin American populations.
Thus, we aim to investigate a possible association between SUA levels and impaired glycemic status; and between SUA and insulin resistance, at the baseline of a large Brazilian prospective study, the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil).

Methods
This is a cross-sectional study using ELSA-Brasil baseline data, collected from August 2008 to December 2010. The study population consisted of 15,105 participants aged 35-74 years. This multicenter Brazilian cohort study involves five public universities and one research institution, from three of the five major regions of the country. It aims to investigate epidemiological, clinical, and molecular aspects of chronic non-communicable diseases -mainly cardiovascular disease and diabetes 13,14 .
Data collection started in 2008, according to study manuals and following international regulation for storage and data analysis. Interviews were performed after all participants signed the free consent form. Additional interviews, examinations, and laboratory tests were performed subsequently 13,14 . The ELSA-Brasil study was supported by the Brazilian Ministry of Health and the Brazilian Ministry of Science and Technology. More information is available at http://www.elsa.org.br/.
Our hypothesis is that there is a positive association between SUA levels and diabetes, impaired fasting glucose (IFG), and impaired glucose tolerance (IGT). For this analysis, the exclusion criteria were: participants with previous self-reported medical diagnosis of diabetes or in use of oral antidiabetic medications or insulin (n = 1,407); with reported bariatric surgery (n = 120) or allopurinol use (n = 153); without previous diabetes information status (n = 4); with missing values for fasting glucose (FG) (n = 3), glycated hemoglobin A1c (n = 14), oral glucose tolerance test (OGTT) (n = 192) and uric acid (n = 5). The final sample was composed of 13,207 participants (87.4%) (Figure 1).

Laboratory tests
After overnight fasting, biological material samples were collected from all participants and were initially frozen and stored, and then transported to the central laboratory (University of São Paulo). All laboratory analysis were carried out in a single research center for better standardization 15 .
The standard OGTT was performed in all participants of this study. Samples were collected for FG, after two hours of oral overload with 75g of anhydrous glucose 15 .
Uric acid was measured by uricase method (enzymatic colorimetric) in the Siemens 1200 (Siemens Healthcare Diagnostics, Deerfield, Unites States). Hyperuricemia was defined on the basis of genderspecific cut-off points (> 7.2mg/dL for men, and > 6.0mg/dL for women) 15 . FG was measured by an enzymatic hexokinase (enzymatic colorimetric) method in Siemens 1200, glycated hemoglobin (A1c) by a high performance liquid chromatography (HPLC) in Bio-Rad Laboratories (Hercules, United States), and insulin by an immunoenzymatic assay, in Centaur Siemens equipment.
Creatinine measurement was based on the Jaffe's method in ADVIA 1200 Siemens equipment. The glomerular filtration rate (GFR) was estimated by serum creatinine, age, and sex, using the abbreviated form of Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) 16 .
Insulin resistance was assessed by homeostasis model assessment of insulin resistance (HOMA-IR) by the [FG (mg/dL) x 0.0555 X plasma fasting insulin (mUI/L)] / 22.5. Insulin resistance was defined by HOMA-IR ≥ 90% percentile (P90) of the variable distribution 18 .
Sex was categorized as male or female; the age was analyzed in years; the race /skin color was selfreported and categorized as black, brown, white, and others (including Asians and Indians); and the schooling was categorized in elementary, high school, and higher education. Diabetes in first-degree relatives was self-reported (family history of diabetes) 13 .
Prevalent coronary heart disease was defined by the report of myocardial infarction and/or myocardial revascularization 13 .
Anthropometry was performed by trained examiners using standardized techniques and equipment. BMI was calculated using weight (kg) divided by square of height (m 2 ). Health-related behaviors were assessed by the participants' self-report. Physical activity in leisure time was evaluated by the International Physical Activity Questionnaire (IPAQ) in the modified long version and participants were classified into three groups: low, moderately, or very active, according to the sum of the weekly metabolic equivalents -a combination of type, frequency, and duration of the physical activities performed 13 . Participants were classified as never smoked, ex-smokers, or smokers. Ex-smokers were those who smoked at least 100 cigarettes throughout a lifetime and current smokers, who still smoked at the time of the interview 13 .
The intake of dairy, animal protein, fructose, and alcoholic beverages were assessed by a validated semi quantitative Food Frequency Questionnaire (FFQ) 19,20,21 . Frequency and the amount of dairy consumed were assessed, resulting in the evaluation of the intake categories in servings/day. Alcohol consumption was analyzed as grams/week and fructose as grams/24h 19 .
Arterial hypertension was defined by a self-reported medical diagnosis of hypertension, use of anti-hypertensive drugs, or blood pressure ≥ 140x90mmHg at the moment of evaluation (sum of two measures, after five minutes of rest in the sitting position) 13 .

Statistical analysis
Continuous variables were described by means and standard deviation (SD) or median value and interquartile range (IQR; 25th to 75th percentile), according to the distribution of the data. Categorical variables were described by frequencies and proportions. ANOVA test was used to compare means, and Kruskal-Wallis for the medians across glycemic status categories.
The association of variable glycemic status with SUA levels (continuous) and other explanatory variables was investigated by univariate analysis. All the variables with p-value < 0.20 and those clinically relevant, regardless of whether p < 0.20 in the univariate analysis, were pre-selected to enter multinomial logistic regression models with "glycemic status" as the response variable, and SUA levels as the variable of interest. A low number of participants had chronic kidney disease (n = 521). Sensitivity analysis showed that the exclusion of them did not affect the results. Therefore, we are presenting analysis with adjustment by GFR rather than with the exclusion of these participants.
Models were consecutively adjusted as follows: Model 1 -glycemic status, SUA (continuous), age (continuous), sex, race/skin color, schooling, family history of diabetes; Model 2 -Model 1 plus BMI, systolic and diastolic blood pressure, prevalent coronary heart disease, and glomerular filtration rate; Model 3 -Model 2 plus physical activity, smoking, alcohol and dairy intake, fructose and animal protein consumption, triglycerides, and HDL cholesterol; Model 4 -Model 3 plus use of medicines that may interfere with uricemia (diuretics, losartan, fenofibrate, acetylsalicylic acid, gout medications). Tests were conducted on whether uric acid levels and fructose intake were effect modifiers on the associations of glycemic status by multiplicative interaction terms between uric acid/fructose Cad. Saúde Pública 2021; 37(9):e00255920 intake. Additionally, tests were conducted on whether SUA and skin color were effect modifiers on the associations of glycemic status by multiplicative interaction terms between uric acid/skin color. We performed sensitivity analysis with categorical BMI, stratified into four categories according to nutritional status: underweight, eutrophic (reference), overweight and obesity.
The strength of the associations between the variables in the regression analysis was verified by odds ratio (OR) and confidence intervals (95%CI).
Logistic regression models were used to test the association between SUA levels and insulin resistance (yes/no), as the response variable, with the same adjustments and including multiplicative interaction terms between uric acid/fructose intake.
Sex-specific analysis were performed, and for the female sex we also included estrogens therapy used for postmenopausal, in Model 4.
All analyses were performed using the statistical software Stata 14 (https://www.stata.com) with a 5% significance level and the p-values were two sided.
The study was approved by the Research Ethics Committees of the six participating centers and the Brazilian National Commission for Ethics in Research (CONEP/MS 976/2006).

Results
Data from 13,207 participants were analyzed. The mean (standard deviation) age was 51.4 (±8.9) years, with 55.2% of females. Most participants had university degree (54%), and self-reported race/ skin color as white (53.2%). A family history of diabetes was present in 37%. The mean SUA was 5.5mg/dL (±1.5), with 6.4mg/dL (±1.4) for men and 4.8mg/dL (±1.2) for women. The prevalence of hyperuricemia in the total sample was 18.4% (n = 2,430), 25.1% among men and 13.0% among women.
Insulin resistance was defined by HOMA-IR value of 4.18, which corresponds to P90. Insulin resistance was evidenced in 1,327 participants, 10% of the final sample, most of them male.
In Table 1, data of the sample study is depicted according to the glycemic status. After all adjustments, higher SUA levels were significantly associated with impaired glycemic status. For each 1mg/ dL increase in SUA levels, the odds of IFG, IGT and diabetes, was, respectively, 15%, 23% and 37% higher ( Table 2). In the overall sample, we adjusted for the waist-to-hip ratio and there was no relevant change in OR (data not shown). Table 2 shows the sensitivity analysis, stratified by sex. The positive association between SUA levels and altered glycemic status remained significant after stratification. After sensitivity analysis with categorical BMI, the OR of IFG, IGT and diabetes was 1.17 (95%CI: 1.08; 1.27), 1.25 (95%CI: 1.16; 1.35) and 1.40 (95%CI: 1.26; 1.54), respectively, maintaining the positive association. The association between higher SUA levels and altered glycemic status was stronger among women. SUA levels were positive and significantly associated with insulin resistance in the general population OR = 1,24 (95%CI: 1.13; 1.36) and after stratification by sex (Table 3). Higher SUA had a strong association with insulin resistance, mainly among women.

Discussion
This study investigated the association between glycemic status and SUA levels in men and women at the baseline of a large cohort of Brazilian adults. After adjusting for demographic, anthropometric, health conditions, laboratory variables, and for the intake of some food groups, alcoholic beverages, and use of medicines, it was observed that higher SUA levels were independently associated with abnormal glycemic status and insulin resistance. These results were demonstrated in both genders, with higher association among female participants, and without influence of estrogenic drugs used for hormonal postmenopausal therapy. Some authors have demonstrated an association between higher SUA levels and hyperglycemia. Prospective data from the original Framingham Heart Study, showed there was a high independent risk of future diabetes development in those with higher uricemia, in a positive linear progression; including in young adults, reaching a 20% higher risk for each 1mg/dL increase in SUA 22 .
Our findings were consistent with the literature. Kramer et al. 6 investigated the incidence of diabetes in individuals with normal glycemia, IFG, and IGT, with mean age of 63.3 years. The risk of diabetes increases by 60% for each 1mg/dL higher baseline SUA, after several adjustments. The SUA cut-off of 5.35mg/dL had a 100% negative predictive value for incident diabetes, after almost two decades of follow-up.
However, in contrast to this study, other researchers did not find an association between SUA and altered glycemic status. A study with American adults from the third National Health and Nutrition Examination Survey 10 found an inverse association between higher SUA levels and diabetes, but there is no mention about adjustments for drugs such as allopurinol. A Chinese cohort 9 with more than 4,400 participants also evaluated the association between SUA, xanthine oxidase activity, and diabetes development. There was an association between SUA and diabetes, but this association was lost after adjustment for xanthine oxidase activity. The authors concluded that elevated xanthine oxidase activity, but not SUA, has been associated with an increased risk of developing diabetes.
Another study 11 evaluated the association between SUA levels and diabetes among more than 24,000 participants of eight European countries. Although a positive association between the increase Cad. Saúde Pública 2021; 37(9):e00255920 in SUA level and risk of diabetes was found, there was no causal relationship between the variables after analysis by Mendelian randomization, using a genetic score with 24 uric acid associated loci. Other studies have also shown an association between SUA and incident diabetes. Mendelian randomization, however, did not provide evidence for a causal link between them 23,24 .
A meta-analysis emphasized that urate plays an important role in insulin resistance; and higher SUA was associated with an increased risk of developing diabetes 4 . Hyperuricemia reduces the bioavailability of nitric oxide -fundamental for the cellular uptake of glucose -which could be a mechanism that would, partially, explain insulin resistance 3,25 . On the other hand, hyperinsulinemia leads to an increase in SUA due to a reduction in uric acid secretion; since insulin stimulates the expression of the urate transporter 1 (URAT1), which increases renal urate reabsorption, generating accumulation of plasma uric acid 3,25 . The rapid hepatic synthesis of uric acid from fructose(a sugar commonly added to beverages, soft drinks, and processed foods) causes mitochondrial oxidative stress, via activation of NADPH oxidase, which stimulates insulin resistance and fat accumulation regardless of excessive caloric intake 12 . The likely mechanism in which fructose induces hyperuricemia is due to the increased degradation of adenosine triphosphate to adenosine monophosphate, a precursor of uric acid 12 .
A review article evaluated uricemia as a marker of metabolic syndrome. It was observed that the association between hyperuricemia, cardiovascular disease, and elevation of proinflammatory cytokines implicated in the development of atherosclerosis 26 . Increased levels of markers and mediators of inflammation are related to changes in blood glucose and in insulin sensitivity. Uric acid may induce oxidative stress on adipocytes, by inflammatory cytokines synthesis via NADPH oxidase, generating insulin resistance 12 . Interleukin-6 and tumor necrosis factor α (TNF-α) appear to interfere with insulin, signaling pathways and beta cell function. TNF-α facilitates serine phosphorylation of the insulin receptor substrate 1, altering the phosphatidylinositol 3-kinase pathway and reducing insulin sensitivity 12,27 .
Inhibition of β-cell function by SUA may occur through several insulin signaling pathways, such as the adenosine monophosphate-activated protein kinase pathway. This is due to the overproduction of reactive oxygen species and oxidative damage, causing inhibition of proliferation and growth of β-cells. The kinase pathway is regulated by extracellular signaling and the nuclear transcription factor kappa B 28,29 .
The pathogenesis of diabetes is complex and multifactorial, in which overweight, sedentary lifestyle, dietary habits, genetic factors, and age are highly relevant factors 1,17 . Despite all the knowledge about these factors, the prevalence of diabetes has increased in a worrisome way, with Brazil being the fifth country with the largest number of individuals with diabetes 1,2 .
The population of this study, according to the data presented, shows a high risk of insulin resistance and altered glycemic status development, since most participants were middle-aged, with the mean BMI of 26.7kg/m 2 (±4.6), and reported a sedentary behavior. We adjusted for BMI to avoid confusion bias due to the causal relationship between obesity and diabetes. Even after conducting a sensitivity analysis with categorical BMI, stratified into four categories according to nutritional status, the association between SUA and impaired glycemic status remained positive.
The most relevant positive aspects of our study are the rigor in the collection and processing of data and the large sample size, which allowed the analysis of men and women, separately. It has been shown that there are strong sex-specific effects in the genetic basis of urate production and excretion 8,30 . We adjusted it in females using hormonal postmenopausal therapy, as there is evidence that estrogen exerts an uricosuric effect, interfering with uricemia 31 . Other positives aspects are the fact that it is a multicenter study carried out in different Brazilian regions, with diverse genetic profile. We adjusted for important confounding factors, such as skin color, glomerular filtration rate, and use of interfering drugs. Another significant adjustment was dietary intake of dairy products and animal protein recognized as factors that can be protective or increase the risk of hyperuricemia and diabetes, respectively 12,32 . We also tested whether SUA and skin color were effect modifiers on glycemic status associations by multiplicative interaction terms between uric acid/skin color, with no change in OR. A positive, statistically significant, association persisted after we tested for effect modification by fructose intake, a sugar associated with the rapid uric acid hepatic synthesis, insulin resistance, and hyperglycemia 12 .
Cad. Saúde Pública 2021; 37(9):e00255920 Finally, this is a cross-sectional study and it is not possible to show a causal relation between explanatory variables and the outcomes assessed. No residual confounding factors were considered, such as the xanthine oxidase activity and genetic profile.
Our work is innovative because we approach the association of SUA and altered glycemic status in the Latin American population, in a large sample, with almost 14,000 participants. With an estimated population of 50% white and 43% black and brown, this study proves to be remarkably diverse from the studies that addressed the topic, essentially with a Caucasian or Asian population.
We also made adjustments and an interaction term for the fructose intake and the consumption of dairy products, which may be related to SUA levels and glycemia.
Despite being a cross-sectional analysis, this study is related to a large Latin American sample, and Brazil has the fifth largest diabetes population in the world 2 , many still without diagnosis and treatment.
In sum, despite all current scientific knowledge of pathogenesis, diagnosis, and therapy, the high incidence of type 2 diabetes -and all the consequences of its diagnosis -remain a major focus on public health, with worldwide interest. In our study, we sought to clarify the association between SUA, a low cost, simple and accessible biomarker, and hyperglycemia, trying to better understand a residual factor that influences the development of diabetes. There was a positive, independent, and significant association between SUA levels and altered status of glycemia, diabetes, and insulin resistance, mainly in women.
SUA is an accessible and simple laboratorial test, and a potentially modifiable factor. Investigating the SUA levels, especially in patients diagnosed with pre-diabetes and insulin resistance, may be relevant for clinical and public health purposes. Further studies can be developed to answer challenging questions about this association in populations with diverse genetic background.