Association between cumulative serum urate and development of diabetes type II: the Kailuan Study

The Kailuan Study [12] is a prospective cohort study conducted in the Kailuan community in Tangshan City, China. From June 2006 to October 2007, a total of 101,510 participants (81,110 men and 20,400 women) aged 18–98 years were recruited to participate in the Kailuan Study. They were subsequently followed-up in three subsequent visits in 2008–2009, 2010–2011, and 2012–2013. The current analysis is based on a subgroup of 36,277 individuals (27,077 men and 9200 women) with complete follow-up data available and without diabetes type II mellitus prior to their last visit (Fig. 1). Subjects diagnosed with myocardial infarction, stroke, cancer, or pregnancy during the visits were excluded.

Information on age, sex, disease history, medication (including insulin, oral hypoglycemic agents, diuretics, and antihypertensive drug use), physical activity, smoking, alcohol intake, and education was collected via questionnaires at baseline and each of the three follow-up visits. Height, weight, and blood pressure were examined by trained physician staff. All measurements were performed using standardized protocols described previously [12]. Blood samples were collected from the antecubital vein after overnight fasting. Fasting blood glucose was measured using the hexokinase/glucose-6-phosphate-dehydrogenase method. Total cholesterol and triglycerides were assessed enzymatically. High-density lipoprotein cholesterol and low-density lipoprotein cholesterol levels were determined using a direct test method (inter-assay coefficient of variation < 10%; Mind Bioengineering Co. Ltd., Shanghai, China). SU concentrations were determined using oxidase method. All biochemical variables were assessed at the central laboratory of Kailuan General Hospital with using a Hitachi autoanalyzer (Hitachi 747; Hitachi, Tokyo, Japan).

Diabetes type II is diagnosed as fasting glucose ≥ 7 mmol/L, or fasting glucose ≤ 7 mmol/L with the use of insulin or oral hypoglycemic agents. Systolic and diastolic blood pressures were averaged across 2 measurements. Hypertension was defined as a mean systolic blood pressure ≥ 140 mmHg and/or a mean diastolic blood pressure ≥ 90 mmHg, or a mean systolic blood pressure < 140 mmHg and/or a mean diastolic blood pressure < 90 mmHg after treatment. Estimated glomerular filtration rate (mL/min/1.73 m²) was calculated by CKD-EPI. Body mass index was calculated as body weight (kg) divided by the square of body height (m²). Smoking was defined as consumption of 1 cigarette every day for at least 1 year. Drinking was defined as consuming on average 100-mL wine (or alcohol content over 50%) per day for more than 1 year. Physical activity was defined as performing at least 30-min exercise more than 3 times a week. Education was defined as high school education level or above. Positive family history of diabetes type II means first-degree relative has been diagnosed with diabetes type II.

CumSU was defined as the summation of average SU for each pair of consecutive examinations multiplied by the time between these two consecutive visits in years:

[((SU₁ + SU₂)/2) × time_1–2] + [((SU₂ + SU₃)/2) × time_2–3] + [((SU₃ + SU₄)/2) × time_3–4]_, where SU₁, SU₂, SU_3, and SU₄ indicate SU at examinations 1 (baseline), 2, 3, and 4, respectively [13‐16]. Time_1–2, time_2–3, and time_3–4 indicate the participant-specific time interval between consecutive examinations 1–4 in years. For men, CumSU was categorized in μmol/L × year as < 1499 (Q₁ group), 1499–1739 (Q₂ group), 1739–1971 (Q₃ group), 1971–2261 (Q₄ group), and ≥ 2261 (Q₅ group). For women, cumSU was categorized as < 1259 (Q₁ group), 1259–1435 (Q₂ group), 1435–1602 (Q₃ group), 1602–1841 (Q₄ group), and ≥ 1841 (Q₅ group).

Statistical analyses were performed with SPSS 13.0. Continuous variables were described as mean (standard deviation, SD). Comparison among different groups was analyzed by one-way analysis of variance. Categorical variables were described as percentages and were compared using chi-square tests (χ² tests). Logistic regression model was used to estimate the risk of diabetes type II with cumSU metrics. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. We fitted three multivariate models. Model 1 is a univariate logistic regression analysis without adjustment. Model 2 is adjusted by baseline age and sex from model 1. Model 3 is adjusted by education level, physical activity, smoking, alcohol consumption, family history of diabetes type II, baseline hypertension, diuretics usage, body mass index, fasting blood glucose, SU, total cholesterol, triglycerides, and estimated glomerular filtration rate. Logistic regression analysis was used to evaluate the effect of one SD increase of cumSU on the new onset of diabetes type II. The significance level was set at P < 0.05.

Out of the 101,510 individuals participating in the baseline examination, 47,828 subjects underwent all 4 visits. The number of cases excluded due to diabetes type II diagnosis prior to the last visit was 6762. Moreover, 903, 1412, and 247 cases were excluded due to diagnosis of myocardial infarction, stroke, and cancer, respectively, during any visit. Two hundred eighty-six individuals were excluded due to pregnancy. We also excluded 1941 participants due to missing data of blood glucose and SU. The remaining 36,277 participants were included in the present study (Fig. 1). Among them, 9200 participants were women, and 27,077 cases were men. The mean age of the participants was 47.33 (11.48) years. For women, the average age was 46.88 (10.57) years. For men, the average age was 47.49 (11.77) years.

Q₁ group occupies 19.99% of all the subjects, whereas Q₂–Q₅ occupies 19.97%, 20.05%, 19.97%, and 20.02%, respectively. As compared with the baseline data obtained in 2006, there were significant differences in sex, age, hypertension, diuretics use, body mass index, fasting blood glucose, SU, total cholesterol, triglycerides, estimated glomerular filtration rate, family history of diabetes type II, smoking, alcohol intake, physical activity, and education among the 5 groups (Table 1, P < 0.05).

The incidence of new-onset diabetes type II in the last visit was 3.34%, with 1212 cases diagnosed during the last visit. In Q₁, Q₂, Q₃, Q₄, and Q₅, 166, 206, 239, 258, and 343 cases were diagnosed, respectively. The incidence of new-onset diabetes type II was 2.29%, 2.84%, 3.29%, 3.56%, and 4.72%, in Q₁–Q₅ groups, respectively. In the incidence of new-onset diabetes type II, men scored higher than women in the Q₁, Q₂, and Q₃ groups, while no differences between men and women in Q₄ and Q₅ groups were observed (Fig. 2).

As compared with Q₁ group, the odds ratio of new-onset diabetes type II incidence was 1.25, 1.45, 1.58, and 2.12 in Q₂, Q₃, Q₄, and Q₅ groups, respectively. After complete adjustment (model 3), OR in Q₂–Q₄ was slightly increased while OR in Q₅ group decreased from 2.12 to 1.80. The trend was more obvious for women than men. Multivariate adjusted ORs for the incidence of diabetes type II corresponding to a 1-S.D. increase in cumSU were 1.20 in men and 1.52 in women. We found a similar trend across age groups (Table 2).

To exclude the influence of hypertension and antihypertensive therapy on the association between cumSU and incidence of diabetes type II, a sensitivity analysis was performed after excluding 15,945 patients with hypertension. We found the association between cumSU and incidence of diabetes type II was unaffected (Table 3).