Chronic inflammatory conditions, such as rheumatoid arthritis and gout, are associated with significant increase in cardiovascular morbidity and mortality. Several studies have reported a higher risk of cardiovascular disease in patients with gout [1–3]. The evidence supporting the association of hyperuricemia with increased cardiovascular risk [1, 4–7] indicates that this may be at least one potential mechanism for this increased cardiovascular risk in patients with gout.

Allopurinol, a purine analog, acts as a competitive substrate for xanthine oxidase, an enzyme that converts hypoxanthine to xanthine and xanthine into uric acid. Thus, it reduces the level of serum urate and acts as a urate-lowering therapy (ULT). Xanthine oxidase generates superoxide anions and other oxidative free radicals that are associated with oxidative stress [8]. Allopurinol reduces uric acid production and oxidative stress, processes that have been hypothesized to be associated with endothelial dysfunction, inflammation and progression of atherosclerosis [9]. Thus, allopurinol might have beneficial effects on cardiac physiology [10, 11].

Recent studies have debated whether allopurinol has a true cardio-protective effect, i.e., does it reduce the risk of outcomes such as myocardial infarction (MI)? A case-control study with a prevalent user design using the French pharmacoepidemiology database found that allopurinol use did not significantly reduce the risk of non-fatal MI, odds ratio (OR) was 0.80 (95% CI, 0.59, 1.09) [12]. In contrast, in a nested case-control study using Spanish primary care database, current allopurinol use and not past use, was associated with reduction in non-fatal MI with odds ratios of 0.52 (95% CI, 0.33 to 0.83) and 0.99 (95% CI, 0.61, 1.61), respectively [13]. Yet another study of gout patients using Scottish database reported that compared to non-use, allopurinol use was associated with higher (not lower) adjusted hazards of cardiovascular hospitalization (including coronary artery disease (CAD), hypertensive heart disease, heart failure, stroke, other cardiovascular diseases) of 1.25 (95% CI, 1.10, 1.41) [14]. These three studies differed in the main outcome assessed (MI vs. non-fatal MI vs. cardiovascular hospitalization including heart failure) and the country setting (France vs. Spain vs. Scotland), which could have explain the differences in  the results between these three studies.

Key limitations of the previous studies included the use of a prevalent- user design [12, 14], the potential lack of adequate statistical power [12, 13], a short follow-up [13], potential survival bias since non-fatal MI cases were included [12–14], and the non-exclusion of patients with previous MI [14]. A prevalent-user design (used in 2 of the 3 studies) is likely more biased than a new-user design, due to the potential inclusion of factors on the causal pathway and the lack of an ability to capture early events (especially harmful effects) after starting a medication. Incident user design, in which medication exposure time begins with a new start of a medication, allows for more valid treatment comparisons compared to mixing ongoing users with new users in the same analysis [15]. Thus, based on the current knowledge, it is unclear whether allopurinol treatment is independently associated with better cardiac outcomes.

Gout is commonly associated with several comorbidities, such as hypertension, diabetes etc. One-third of gout patients have diabetes [16]. Diabetes, as a major cardiovascular disease risk factor, affects 26 million Americans, is among the top 10 causes of death in the U.S., and cost $245 billion in direct and indirect costs in 2012 [17]. Despite the common occurence of both conditions and associated morbidity, studies of cardiovascular outcomes in patients with gout and diabetes are lacking.

We were interested in investigating cardiovascular (CV) events in patients with gout and concomitant diabetes, since this group of patients is at high risk of CV events, given that diabetes is an established CV risk factor and hyperuricemia (the key underlying abnormality in gout) has been implicated as a CV risk factor [1, 4–7]. Therefore our study objectives were to assess whether in patients with gout with diabetes, allopurinol therapy was associated with cardiovascular benefit as demonstrated by a lower incidence of acute cardiovascular events (myocardial infarction [MI] or stroke).

Study methods and results are reported as recommended by the Strengthening of Reporting in Observational studies in Epidemiology (STROBE) statement [18].

In this study, we found that current use of allopurinol was associated with a statistically significant and clinically meaningful reduction of the risk of acute cardiovascular events (incident stroke or MI) in patients with gout and diabetes. We also provided crude incident rates for acute cardiovascular events in patients with gout and diabetes, by current versus prior allopurinol use. Several findings in this study deserve further discussion.

Our finding of 33% hazard reduction of incident stroke or MI with current allopurinol use compared to prior allopurinol use in patients with gout and diabetes is novel. We are unaware of other large studies that have examined this question. We chose to study this patient population (gout and diabetes) due to a higher risk of cardiovascular events. The beneficial cardiovascular effect of allopurinol may be related to several potential mechanisms including: (1) allopurinol associated attenuation of intercellular adhesion molecule-1 levels noted in patients after a recent stroke [27]; (2) improvement in endothelial function and reduction of oxidative stress with allopurinol [28–33]; and (3) reduction in glycosylated hemoglobin (HbA1C) with high-doses of allopurinol [33]. The potential mechanisms of cardiovascular risk reduction with reduction of serum urate, including oxidative stress and inflammation, were recently reviewed [34]. The hazards were decreased by 29% in the prevalent user analyses. This is consistent with a greater risk reduction seen in the incident user vs. prevalent user design, and might explain the lack of finding an association if prevalent user design is used and the sample size is smaller or events fewer.

Our findings must be interpreted in the context of current knowledge in this area. Most previous studies examined stroke as part of a composite cardiovascular outcome that included multiple cardiovascular outcomes, not just stroke and MI. In a 7-year follow-up of a 2-year RCT in patients with chronic kidney disease with glomerular filtration rate of < 60 ml/min, adjusted for race, sex and renal function, allopurinol use was associated with lower risk of 0.43 (95% CI, 0.21–0.88) of cardiovascular events (including CAD, cerebrovascular disease, heart failure and peripheral vascular disease) [35]. In an observational study of 187 patients with hypertensive nephropathy, allopurinol use was associated with a reduced hazard of 0.34 (p = 0.04) for cardiovascular disease (CAD, heart failure and stroke) [36]. In a study using Scottish database, allopurinol use was associated with no difference in adjusted hazards of cardiovascular events (non-fatal myocardial infarction, non-fatal stroke and cardiovascular mortality) compared to non-users 1.10 (95% CI, 0.95–1.26) [14]. Differences in the conditions included in the composite cardiovascular outcome, study populations and study designs, likely explain the differences in findings.

The three key studies examining the risk of MI reported that compared to non-use, allopurinol use was associated with a 48% decrease [13] vs. a 25% increase [14] vs. no significant change [12] in the risk of MI. The studies were similar in that they all used clinical databases, assessed non-fatal MI and controlled for several cardiovascular risk factors. However, several differences were evident. Two studies used prevalent-user design [12, 14], which puts them at the risk of adjusting for the mediating factors and missing early events. Thus, it was not surprising that the two studies that used prevalent design showed increased risk [14] or no difference in risk [12], while the only new user design study showed risk reduction for MI [13]. Our study, using a new user design, also shows statistically significant benefits of allopurinol for risk reduction of acute cardiovascular events, and these findings were confirmed in several sensitivity analyses, including an analysis of prevalent allopurinol users (to allow comparability to the two previous studies). Our study is among the first to show allopurinol’s protective effect on acute cardiovascular event, i.e., incident stroke or MI, in patients with gout and diabetes. However, our study differs in one important aspect from the previous cited studies, i.e., we focused on patients with both gout and diabetes. These patients likely have a higher risk of acute cardiovascular events than patients with only one of the two risk factors. This likely enriched our population for more acute cardiovascular events and likely made our study adequately powered, as intended.

This finding has a practical implication for patients and providers. For patients with both gout and diabetes, allopurinol has an added benefit of reduction of the risk of acute MI or stroke. The risk reduction is not only statistically significant, but also clinically meaningful (33% reduction). Patients should be informed of this potentially beneficial effect, when allopurinol is first prescribed. This risk reduction may potentially be even greater in patients with other cardiovascular disease risk factors.

Our study findings must be interpreted considering several limitations. The diagnosis of acute cardiovascular events, our study outcome, was based on the presence of ICD-9-CM codes, which may lead to misclassification bias. However, previous studies have shown the these codes for cardiovascular outcomes were accurate [24]. There is also evidence of a reasonable accuracy for codes used for covariates i.e., cardiovascular risk factors [23–25]. Serum urate levels were available for a small proportion of patients and therefore, we could not examine whether these associations were or were not mediated via the reduction of serum urate levels. The database also did not allow us to include other variables such as body mass index, smoking, diet and exercise, over the counter aspirin use, and family history of cardiovascular disease, important covariates/risk factors for MI and stroke. We had considered excluding patients with cancer, but decided to include them in our study to improve the generalizability of study findings. A very small proportion of patients on allopurinol (<0.5%) are usually receiving short-term allopurinol for prophylaxis against tumor lysis syndrome, since they are receiving chemotherapy [37]; we doubt that this study limitation influenced our results significantly. Examination of effect of allopurinol on other outcomes, including renal failure, dialysis or death, or rare side effects such as allergic reactions, was beyond the scope of current study. Future studies should examine these outcomes.

The contribution of healthy adherer effect to the observed allopurinol’s beneficial effect for these outcomes can not be ruled out, since patients adherent to/using allopurinol may have other healthy behaviors, such as high adherence to other medications, healthier lifestyle, preventive care etc., which may all reduce the risk of stroke and MI. Whether these findings can be generalized to populations other than diabetes and gout is not known, and further study is needed in this area. Reasons for allopurinol discontinuation can be other than non-adherence, such as allopurinol cessation due to severe renal disease (which is not a contraindication for allopurinol use, since allopurinol can be used even in patients on severe kidney dysfunction and also dialysis), suspected or confirmed allergic reactions (which are extremely rare), the concomitant/new use of azathioprine for the rheumatic disease or organ transplantation (which is uncommon/rare), or switching to febuxostat, another xanthine oxidase inhibitor launched in 2009 (which would have likely decreased the observed difference, but likely did not have a major impact given the use of febuxostat in <7% patients with gout) [38]. These confounding biases would have likely reduced the differences, and likely made our current estimates conservative, since most of these conditions would likely increase the risk of MI and/or stroke in current allopurinol non-users.

Study strengths include a large sample size, the use of a sample that allows findings to be generalizable to cohorts with gout or diabetes in the U.S., the use of a new user design that reduces bias in observational studies [39–41], the inclusion of multiple risk factors for MI/stroke, and the robustness of estimates in sensitivity analyses.

In conclusion, in robust analyses of data from a representative sample of Americans, we found that current allopurinol use was independently associated with a lower risk of incident stroke or MI in patients with gout and diabetes. Our findings provide evidence for a potential cardiovascular protective effect of allopurinol in patients with gout and diabetes, a group of patients at high risk for cardiovascular events. Future studies need to explore the factors that mediate this protective effect, and examine what proportion of variability in this protective effect is explained by serum urate reduction and/or anti-oxidant effect of allopurinol.