The Atherosclerosis Risk in the Communities study (ARIC) is a prospective population-based cohort study of 15,792 individuals recruited from four US communities (Washington County, Maryland; Forsyth County, North Carolina; Jackson, Mississippi; and suburbs of Minneapolis, Minnesota). The Institutional Review Board of the participating institutions approved the ARIC study protocol and study participants provided written informed consent. Participants aged 45 to 64 years were recruited to the cohort in 1987 to 1989. This cohort was established to study the natural history of atherosclerosis, and the study consisted of one baseline visit (visit 1) between 1987 and 1989 and three follow-up visits (visits 2, 3, and 4) administered three years apart. Details of the study design have been previously published 18.

This analysis was limited to participants who were Caucasian or African American; few participants reported other races (n = 48). We excluded participants who did not report their gout status at visit 4 (n = 4,269) and those with prevalent gout at cohort entry, defined as the self-report of gout onset prior to the baseline visit (n = 419). Participants with missing baseline information on the main covariates of interest were not included (n = 265; sex, race, estimated glomerular filtration rate (eGFR), Body Mass Index (BMI), alcohol intake and hypertension). Participants who were older, had diabetes, lower education and higher serum urate were more likely not to be included in this analysis. However, baseline hypertension, BMI, eGFR, use of an antihypertensive medication (including diuretic use) or sex were associated with being included in this analysis. Therefore, it is unlikely that our results are not generalizable to the ARIC cohort.

There were 10,791 participants included in this study of the association of anemia and gout. This study hypothesis was developed a priori and the sole focus of this analysis.

At baseline, each participant had their blood drawn and the procedures have been described elsewhere 19. Hemoglobin was measured using automated hematology analyzers (Coulter S + IV (calibration S-Cal), Beckman Coulter, Inc, Fullerton, CA, USA) at two sites, Coulter S + III and Coulter S + IV (calibration S-Cal) at one site and Technicon H-6000 (calibration Fisher, Technicon Corporation, Tarrytown, NY, USA) at one site. Anemia was defined as baseline hemoglobin less than or equal to 12 g/dL for female participants and less than or equal to 13.5 g/dL for male participants. The cutpoint for women is the same as the World Health Organization (WHO) definition of anemia 20. However, for men, we chose 13.5 g/L for anemia rather than the WHO cutpoint for men (13.0 g/L) 20 because it approximated the lowest deciles for men (12 g/dL was the lowest decile for women) and was consistent with previous ARIC studies 11.

At ARIC visit 4, participants were asked, "Has a doctor ever told you that you had gout?" Participants who answered, "Yes," to the gout query then reported the age at gout diagnosis. The outcome of interest was incident gout based on self-reported onset of gout at an age greater than their age at ARIC visit 1. Our previous research suggests that self-report of a physician diagnosis of gout is a reliable (three-year reliability kappa = 0.73) and a sensitive (sensitivity = 84%) measure of gout 21.

Other covariates of interest at baseline (1989) included self-reported age (in years), sex (male or female), race (white or African American), education (<12 years, 12 to 16 years and 17 to 21 years) and alcohol intake (grams/week). Additionally, we included clinical factors that were thought to be related to anemia or gout, including diabetes (fasting glucose level ≥126 mg/dL, nonfasting glucose level ≥200 mg/dL, reported physician-diagnosis of diabetes, or history of medication use for diabetes), measured BMI (kg/m²), coronary heart disease (CHD) (myocardial infarction from adjudicated visit 1 ECG data, history of myocardial infarction by self-report or history of heart or arterial surgery, report of coronary bypass, balloon angioplasty, angioplasty of coronary arteries), hypertension (self-report of medication to treat hypertension, or a measured blood pressure ≥140/90 mm Hg) or diuretic use in the two weeks prior to the baseline visit. Self-report of a history of heart failure or coronary heart failure (CHF) at visit 4 was also considered. Serum creatinine was estimated using a modified kinetic Jaffé reaction. Glomerular filtration rate (eGFR) was estimated using the CKD-Epi equation 22 and categorized as ≥90, 60 to 90, or <60 ml/min/1.73 m².

Serum urate concentrations were measured with the uricase method at visit 1 in mg/dL. The reliability coefficient of serum urate was 0.91, and the coefficient of variation was 7.2% in a sample of 40 individuals with repeated measures taken at least one week apart 23.

The mean and standard deviation (SD), as well as the prevalence of the covariates, were calculated and compared by baseline anemia status using a t-test and chi-squared test, respectively.

Using a Cox Proportional Hazards model, the hazard ratio (HR) and 95% confidence interval (CI) of incident gout by baseline anemia was estimated with age as the time scale. The unadjusted survival functions were plotted using Kaplan-Meier plots. Models were adjusted for confounders of the association of anemia and gout, including sex and race, as well as baseline measures of BMI, categorical eGFR and alcohol intake. In a separate model, we additionally adjusted for baseline diabetes, hypertension, CHD, history of CHF and diuretic use. All of these measures were considered confounders of the gout and anemia relationship, based on exploratory data analysis and previous research 52425. We hypothesized that the presence of uterine fibroids may explain the observed association of anemia and gout in the female population. Thus, in an analysis limited to women, we adjusted for surgical menopause as a proxy because fibroids are the leading cause of hysterectomy. Additionally, we adjusted for baseline serum urate level to test whether the association of anemia and gout was independent of serum urate levels.

We tested whether the association of anemia and incident gout differed by sex, race and eGFR (<60 ml/min/1.73m² or < 60 ml/min/1.73m²). The stratified models were adjusted for the previously listed confounders. Using a Wald test, we assessed whether these factors were effect measure modifiers (statistical interaction) of the association of anemia and incident gout.

As a sensitivity analysis, we restricted the population to those with an eGFR> 30 ml/min/1.73m² and tested the association of anemia and gout. Additionally, we examined whether the association of anemia and incident gout differed when the model was adjusted for time-varying eGFR, including estimates of GFR from visit 1 and 2. We tested whether socio-economic status factors and other risk factors explained the association between anemia and gout: education, smoking status, fibrinogen (an acute phase reactant), use of non-steroidal anti-inflammatory drugs (NSAIDs), mean corpuscular volume and for women menopausal status and history of surgical menopause.

All analyses were performed in SAS, version 9.1 (SAS Institute, Cary, NC, USA). The cumulative incidence plots were constructed in Stata, version 11 (Stata Corp, College Station, TX, USA).

A total of 10,791 ARIC participants met the study criteria. The study population was 43% male, 21% African American and the mean age at cohort entry was 54 years (SD = 5.70). The mean hemoglobin level was 13.9 g/dL (SD = 1.37); for men 14.9 g/dL (SD = 1.05) and for women 13.1 g/dL (SD = 1.05) (Figure 1). At baseline, 1,084 (10%) participants were classified as having anemia; 66% of the participants with anemia were women.

There were 271 cases of incident gout. Of the 271 gout cases, 39% were female and 38% were African American. The mean age at baseline for the gout cases was 54 (SD = 5.9) and the mean age at gout diagnosis was 59 (SD = 6.4). At baseline, 64% of the women with gout were post-menopausal. The mean serum urate level at baseline was 8.0 mg/dL (SD = 2.0) for those who developed gout during follow-up.

Table 1 lists the study population characteristics and gout risk factors by baseline anemia status. Participants with anemia were more likely to be female and African American race than those without anemia. Those with anemia had a lower alcohol intake, serum urate level and lower educational attainment. Participants with eGFR <60 ml/min/1.73m² were more likely to have anemia (2.6% vs. 1.9%, P-value <0.001 for trend).

Incident gout occurred more frequently in participants with anemia at baseline compared with those without anemia (nine-year cumulative incidence: 4.3% vs. 2.3%, P-value <0.001), Table 1. Figure 2 presents the Kaplan-Meier cumulative incidence function of gout by baseline anemia status, suggesting that by age 70, the cumulative incidence of gout was 4.4% in those without anemia and 10.2% in those with anemia. Table 2 lists the unadjusted and adjusted HR of incident gout by baseline anemia status. Participants with anemia were twice as likely to develop gout than participants without anemia (HR = 2.01, 95% CI: 1.46, 2.76). Baseline anemia was associated with a 1.7-fold (95% CI: 1.24, 2.41) increase in the hazard of incident gout after adjusting for clinical risk factors for gout and confounders of the anemia and gout association. Even after adjustment for serum urate level, anemia was associated with incident gout (HR = 1.83, 95% CI: 1.30, 2.57).

There was no evidence of effect measure modification by race (P-value = 0.83) or kidney function (P-value = 0.29). However, there was limited support for an interaction by sex (P-value = 0.06). Among female participants, the HR of incident gout by baseline anemia status was 2.38 (95% CI: 1.47, 3.84) compared to a HR of 1.43 (95% CI: 0.88, 2.30) for male participants.

In an analysis restricted to those who had an eGFR greater than 30 ml/min/1.73 m², the association of anemia and incident gout was similar to the main analysis (Adjusted HR: 1.80; 95% CI: 1.29, 2.52). Adding visit 2 eGFR to the model as a time-varying confounder did not change the results (results not shown). Although we did not have a direct measure of socioeconomic status, the results did not differ when education and baseline smoking status were added to the multivariate model (Adjusted HR: 1.73, 95% CI: 1.23, 2.42). Adjusting for baseline NSAIDs use did not alter the association between anemia and gout (Adjusted HR: 1.74, 95% CI: 1.25, 2.43). Additionally, in the models restricted to women, the results did not change when we additionally adjusted for menopause (Adjusted HR: 2.02, 95% CI: 1.24, 3.28). When we adjusted for baseline history of surgical menopause, as a proxy for fibroids 26, the association of anemia and gout did not change (Adjusted HR: 2.43, 95% CI: 1.51, 3.92). Finally, adjusting for mean corpuscular volume, as a proxy for hemoglobinopathy, did not alter the association of anemia and gout (Adjusted HR: 1.83, 95% CI: 1.30, 2.58).

To our knowledge, this is the first study to identify anemia as a risk factor for incident gout. This is one of the largest biracial studies of gout, which included both men and women with gout. Anemia was established prior to the onset of gout, thus supporting temporality. Additionally, we were able to control for kidney function at baseline and other important risk factors for both gout and anemia. For example, we adjusted for race, hypertension and kidney function, which are known common risk factors for gout and anemia 54649. Finally, our measure of socioeconomic status and education did not alter our results. Additionally, ARIC is one of the few studies that has measures of serum urate and collected data on incident gout during follow-up. This allowed us to test whether the association of anemia and gout remained after adjustment for serum urate levels.

As with any epidemiology study there are limitations to our study. The main limitation of our study was that gout was self-reported by participants at visit 4 and incidence of gout was based on reported age at onset. However, our previous work has suggested that self-reported gout and age of onset is both sensitive and reliable 21. Participants had to survive until visit 4 and be healthy enough to attend the clinical visit to be included in this study. This may induce bias if those who attended visit 4 were different from the study population with respect to the association between anemia and gout. However, if anemia is non-differentially related to mortality, then this would bias our results towards the null, suggesting that our results are an underestimate of the true association. The etiology of participants' anemia could not be determined in this study and we were unable to establish whether participants had a hemoglobinopathy. Additionally, it is not known whether participants were taking iron supplements. Finally, the hemoglobin cutpoints for anemia were relatively high as severe anemia, hemoglobin <11 g/dL, was rare in this population as it would be in the general population. However, previous studies of anemia have used hemoglobin levels and not established the etiology 1011. We cannot be assured that we have eliminated residual confounding, although we have adjusted for the main confounders as was previously noted.