Study design and population
NHANES is an ongoing program conducted by the National Center for Health Statistics (NCHS) to evaluate the health and nutritional status of non-institutionalized adults and children in the US.The program began in the early 1960s and examines a nationally representative sample of about 5,000 persons each year from 15 counties across the US. The NHANES program consists of two main components: face-to-face interviews and physical examinations conducted in mobile examination center (MEC). The interviews include demographic, socioeconomic, dietary, and health-related questions, and the examinations consist of medical, dental, physiological measurements, and laboratory tests. Written informed consent was obtained from all participants. The survey protocol of NHANES program was approved by the institutional review board of the NCHS. In the present study, data of 10 cycles from 1999 to 2018 were extracted for analysis, and a total of 14,208 adult participants with normal range UACR values and without DM were enrolled in the final analysis. The details of the inclusion and exclusion process are shown in Figure 1.
Measurement of UACR
A random urine was collected for each participant in the MEC, and urine specimens were processed, stored and shipped to University of Minnesota, Minneapolis, MN for analysis. Urinary albumin was measured using a solid-phase fluorescent immunoassay described by Chavers et al. The assay has a standard curve of 0.5–20 μg/mL of albumin and is reproducible, accurate, and sensitive for detecting low levels of urinary albumin. Urinary creatinine was assessed with a CX3 analyzer using a Jaffé rate reaction method before 2007 and the enzymatic method using the Roche/Hitachi Modular P Chemistry Analyzer after 2007. The equation recommended in the NHANES website was used to adjust urine creatinine levels from the 2005-2006 cycles. Detailed information on the laboratory methodology is illustrated on NHANES website [16]. UACR was calculated using the formula: UACR (mg/g) = urine albumin (ug/mL) / urine creatinine (mg/dL) × 100.
Assessment of outcomes
The outcome variables included indicators of DM and the prevalence of IR and prediabetes. Indicators of DM included FPG (mmol/L), HbA1c (%), FSI (μU/mL), and HOMA-IR. HOMA-IR was conducted in accordance with FPG and FSI levels as follows: [FPG (mmol/L) × FSI (μU/mL)] / 22.5. The Homeostatic Model Assessment of insulin sensitivity (HOMA-IS) were calculated with the formula: 22.5 / [FPG (mmol/L) × FSI (μU/mL)]. HOMA-IR > 2.6 was used as a determination criterion of IR in this study [17]. The diagnosis of prediabetes was defined as FPG ≥ 5.7 mmol/L and < 7.0 mmol/L, oral glucose tolerance test 2 h plasma glucose ≥ 7.8 mmol/L and < 11.0 mmol/L, HbA1c ≥ 5.7% and < 6.5%, and self-reported of prediabetes [18].
Assessment of covariates
Data of age, sex, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), race, marital status, education level, poverty income ratio, smoking status, alcohol consumption, physical activity (PA), hypercholesterolemia, hypertension, cardiovascular disease (CVD), chronic kidney disease (CKD), estimated glomerular filtration rate (eGFR), anti-hypertensive agents, and anti-hyperlipidemia agents were extracted. BP was measured by trained medical professionals using a mercury sphygmomanometer with an appropriately sized cuff. Three consecutive measurements after participants rested quietly for 5 minutes were obtained, and the average of the three measurements was calculated and used for analysis. BMI was calculated as weight in kilograms divided by height in meters squared. Smoking status was classified into never smoker (smoked < 100 cigarettes in life), former smoker (smoked ≥ 100 cigarettes in life and smoke not at all now), or current smoker (smoked ≥ 100 cigarettes in life and smoke some days or every day). Alcohol consumption was grouped into nondrinker, mild-to-moderate drinker (< 3 drinks per day for females, < 4 drinks per day for males, or binge drinking on less than 5 days per month), or heavy drinker (≥ 3 drinks per day for females, ≥ 4 drinks per day for males, or binge drinking on ≥ 5 days per month). PA was quantified by metabolic equivalent (MET) minutes of moderate to vigorous PA per week and calculated using the following formula: PA (MET-min/week) = MET × weekly frequency × duration of each PA. The eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration Equation [19]. Hypertension was defined as self-reported of hypertension, an average SBP ≥ 140 mmHg or/and an average DBP ≥ 90 mmHg, or use of anti-hypertensive agents. Hyperlipidemia was defined as total cholesterol ≥ 200 mg/dL, high-density lipoprotein cholesterol ≤ 40 mg/dL in males and ≤ 50 mg/dL in females, or use of anti-hyperlipidemia agents. CVD was defined as a composite of self-reported diagnosis of coronary heart disease, congestive heart failure, heart attack, angina, and stroke. CKD was defined as an eGFR < 60 mL/min/1.73 m2 or the presence of albuminuria.
Statistical analysis
As NHANES is designed as a complex, multi-stage, probability sampling study to recruit participants representative of the civilian, non-institutionalized US population, sample weights, clustering, and stratification must be considered in statistical analysis. According to the NHANES analytic guidelines, a ten-year weight (1999-2018) was generated to obtain weighted percentages adjusted to the US adult population. All statistical analysis was conducted using R software (version 3.3.3, S. Urbanek & H.-J. Bibiko, © R Foundation for Statistical Computing), with a P < 0.05 (2-tailed) was considered significant.
UACR levels were divided into four groups based on quartiles: Q1, 0.12-4.05 mg/g; Q2, 4.05-5.87 mg/g; Q3, 5.87-9.30 mg/g; Q4, 9.30-30.00 mg/g. Characteristics of the study participants were expressed as mean ± standard error for continuous variables, and counts (percentages) for categorical variables. Comparisons of characteristics by the UACR quartiles were conducted using one-way analysis of variance for continuous variables and the Rao-Scott χ2 test for categorical variables adjusted for sampling weights. Weighted multivariate linear regression models were constructed to analyze the associations of UACR with indicators of DM. Weighted multivariate logistic regression models were constructed to assess the associations of UACR with IR and prediabetes. Three models were constructed for the analysis. Model 1 adjusted for age, sex, and race. Model 2 further adjusted for BMI, SBP, DBP, education level, marital status, poverty income ratio, MET, smoking status, alcohol consumption. Model 3 further adjusted for eGFR, hypertension, hypercholesterolemia, CKD, CVD, anti-hypertensive agents, and anti-hyperlipidemia agents. In addition, UACR was natural log-transformed and included in above analyses as a continuous variable. Results of the regression analyses were reported as weighted β with their respective 95% confidence intervals (CIs) for multivariate linear regression analysis and odds ratios (ORs) with their respective 95% CIs for multivariate logistic regression analysis.
Restricted cubic spline (RCS) analysis was performed to visually estimate the dose-response relationships of UACR with indicators of DM. Non-linearity was estimated using likelihood ratio tests comparing models with and without the cubic spline term, with a P < 0.05 indicating dose-response relationship in a non-linear manner. Subgroup analyses for the association between UACR and prediabetes were performed by age (< 50 years, ≥ 50 years), sex (male, female), race (non-Hispanic white, other races), BMI (< 30 kg/m2, ≥ 30 kg/m2), smoking status (non-smoker, former/current smoker), alcohol consumption (non-drinker, drinker), hypertension (yes, no) and hyperlipidemia (yes, no). Natural log-transformed UACR was included in regression models as a continuous variable to assess the significance of interactions. Sensitivity analyses were performed to verify the reliability of the results of this study. Firstly, to eliminate the potential confounding effects of CKD and CVD on the associations between UACR with IR and prediabetes, participants with CKD and CVD were excluded and the analysis was reconducted. Secondly, based on previous studies indicating a significant association between blood lipids with DM, blood lipids including triglyceride, total cholesterol, and low-density lipoprotein cholesterol were further adjusted in Model 3 to eliminate their potential confounding effects on the association. Last, the inflammatory markers including dietary inflammatory index (DII) and systemic immune-inflammation index (SII) were also included in the models to eliminate the potential impact of inflammatory on the association. The details of results are shown in Supplementary Files.