Assessment of Cardiovascular Risk in Older Patients with Gout Initiating Febuxostat versus Allopurinol: a Population-based Cohort Study

Abstract

Background

Hyperuricemia and gout are associated with increased risk of cardiovascular disease (CVD). Xanthine oxidase inhibitors (XOI), allopurinol and febuxostat, are the mainstay of urate lowering treatment for gout and may have different effects on cardiovascular risk in patients with gout.

Methods

Using U.S. Medicare claims data (2008-2013), we conducted a cohort study for comparative cardiovascular safety of initiating febuxostat versus allopurinol among gout patients aged ≥65 years. The primary outcome was a composite endpoint of hospitalization for myocardial infarction (MI) or stroke. Secondary outcomes were individual endpoints of hospitalization for MI, stroke, coronary revascularization, new and recurrent heart failure (HF), and all-cause mortality. We used propensity score (PS) matching with a ratio of 1:3 to control for confounding. We estimated incidence rates (IR) and hazard ratios (HR) for primary and secondary outcomes in the PS-matched cohorts of febuxostat and allopurinol initiators.

Results

We included 24,936 febuxostat initiators PS-matched to 74,808 allopurinol initiators. The median age was 76 years, 52% were male, and 12% had cardiovascular disease at baseline. The incidence rate (IR) per 100 person-years for the primary outcome was 3.43 in febuxostat and 3.36 in allopurinol initiators. HR for the primary outcome was 1.01 (95%CI 0.94-1.08) in the febuxostat compared to allopurinol groups. Risk of secondary outcomes including all-cause mortality was similar in both groups, except for a modestly decreased risk of HF exacerbation (HR 0.94, 95%CI 0.91-0.99) in febuxostat initiators. The HR for all-cause mortality associated with long-term use of febuxostat (>3 years) was 1.25 (95%CI 0.56-2.80) versus allopurinol. Subgroup and sensitivity analyses consistently showed similar cardiovascular risk in both groups.

Conclusion

Among a cohort of 99,744 older Medicare patients with gout, overall there was no difference in the risk of MI, stroke, new onset HF, coronary revascularization, or all-cause mortality between patients initiating febuxostat compared with allopurinol. However, there seemed to be a trend toward an increased, albeit not statistically significant, risk for all-cause mortality in patients who used febuxostat for over 3 years versus allopurinol for over 3 years. The risk of HF exacerbation was slightly lower in febuxostat initiators.

INTRODUCTION

Gout, a disorder characterized by monosodium urate crystal deposition in the joints, is one of the most common inflammatory arthropathies, affecting nearly 8.3 million individuals in the United States.^{1, 2} Although the association between gout and cardiovascular disease (CVD) has been well-documented, the evidence for a causal relationship between xanthine oxidase inhibitors (XOI) and CVD remains equivocal.^3-6 While some studies have demonstrated a protective effect of allopurinol against myocardial infarction (MI), cardiovascular (CV) outcomes, and all-cause mortality, other studies have shown no benefit on heart failure (HF).^7-11 A recent cohort study comparing gout patients on XOI with non-users who have hyperuricemia showed that XOI initiation had no effect on cardiovascular risk.¹² The current literature for XOI-mediated CVD risk reduction remains inconsistent.^13-16

Indeed, the concern for XOI-related increased CVD risk has even been raised. In the original two phase III randomized controlled trials (RCT) of febuxostat [APEX (Allopurinol and Placebo-Controlled Efficacy Study of Febuxostat) and FACT (Febuxostat versus Allopurinol Control Trial)], febuxostat (at 80 and 120mg/day) was more effective at lowering serum uric acid levels compared with allopurinol (at 300mg/day) over 1 year.^{17, 18} Despite the more potent urate-lowering effect of febuxostat, the incidence of major adverse cardiovascular events including nonfatal MI, nonfatal stroke, and cardiovascular death in both trials was numerically, though not statistically significantly, higher with febuxostat than allopurinol, and rates of cardiovascular events did not correlate with febuxostat dose.^17-19 These initial findings prompted the U.S. Food and Drug Administration (FDA) to mandate additional safety evaluations. Subsequently, the phase IIIB RCT CARES trial (Cardiovascular Safety of Febuxostat and Allopurinol in Patients with Gout and Cardiovascular Morbidities) was conducted to further investigate the cardiovascular safety of febuxostat compared with allopurinol in gout patients with known cardiovascular comorbidities.²⁰ The CARES trial ultimately showed no difference in the combined risk of cardiovascular death, nonfatal MI, nonfatal stroke, and unstable angina with urgent coronary revascularization for febuxostat compared with allopurinol. However, the individual risks of cardiovascular mortality and all-cause mortality were 1.2 to 1.3 times higher in febuxostat initiators. A similar RCT titled FAST is also underway in Europe.²¹ Given the relatively limited evidence and overall high prevalence of CVD in gout patients, we sought to examine the risk of cardiovascular events in gout patients with and without baseline CVD who started febuxostat or allopurinol (both at typical and equipotent dosing) using longitudinal comprehensive medical and pharmacy dispensing claims data from the U.S. Medicare.

METHODS

The data and study materials cannot be made available to other researchers due to the data use agreement with the Centers for Medicare and Medicaid Services.

Data Source

We conducted a cohort study among gout patients initiating febuxostat versus allopurinol using longitudinal claims data from Medicare Parts A/B/D from January 1, 2008 to December 31, 2013. Medicare, a federal health insurance program in the U.S., provides coverage for legal residents aged 65 years and older, patients younger than 65 with certain disabilities, and those with end stage renal disease (ESRD) requiring dialysis or transplant. Medicare Part A covers inpatient care. Part B coverage encompasses physician services including outpatient visits, laboratory testing, and imaging. Finally, part D provides outpatient prescription drug coverage.²² As a result, the study database includes patient information on demographics, diagnosis and procedure codes from outpatient visits, emergency room visits and acute care hospitalizations, and dispensing records of prescription drugs. The study protocol and waiver for patient informed consent were approved by the Institutional Review Board of the Brigham and Women’s Hospital.

Study Cohort

Patients aged 65 years and older with a diagnosis of gout based on the International Classification of Diseases, 9^th Revision, Clinical Modification (ICD-9-CM) codes 274.00, 274.01, 274.02, 274.03, 274.81, 274.82, 274.89 or 274.9 were eligible for the study. Patients with uric acid nephrolithiasis were not included. We identified initiators of febuxostat or allopurinol using the national drug codes in Part D claims. The index date was the drug initiation date. Drug initiation was defined as having at least 365 days free of a given drug. In other words, allopurinol initiators were allowed to use febuxostat in the 365 days prior to the first prescription of allopurinol and vice versa. This was to reflect typical prescription practices for febuxostat initiation. Naivety to both drugs prior to index date was examined in subsequent sensitivity analyses (see Statistical Analysis).

Exclusion criteria included: age younger than 65 years on the index date, less than 365 days of insurance eligibility in Parts A/B/D prior to the index date, no active claim in the 365 days prior to the index date, use of pegloticase or rasburicase in the 365 days prior to the index date, and ESRD/dialysis in the 365 days prior to the index date.

Outcome Definition

The primary outcome was defined as a composite endpoint of hospitalization for MI or stroke (excluding transient ischemic attacks (TIA)). Secondary outcomes included: hospitalization for MI, stroke, coronary revascularization, HF subdivided into new onset HF or HF exacerbation, and all-cause mortality. Cause-specific mortality was not available in the Medicare database. New onset HF was defined as hospitalization for HF among patients with no baseline history of HF based on the primary inpatient diagnosis. HF exacerbation was defined as hospitalization for HF in patients with a baseline history of HF based on the primary inpatient diagnosis. These outcomes were identified with previously validated claims-based algorithms with the positive predictive value over 80%. ^23-25

Covariates Assessment

To adjust for potential confounders between the two XOI groups, we assessed 81 prespecified baseline variables in the 365 days prior to the index date or on the index date (see Table 1 for partial list of covariates). Covariates included demographic data (age, sex, place of residence), index year, cardiovascular comorbidities (i.e., MI, stroke, coronary revascularization), other medical comorbidities (i.e., CKD-any stage, diabetes, mellitus, hyperlipidemia), gout-related medications (i.e., probenecid, colchicine, NSAIDs, steroids), other medications (i.e., ACE inhibitors, angiotensin receptor blockers, diuretics), and healthcare utilization patterns (i.e., number of emergency room visits, outpatient visits, various tests ordered). To better account for potential confounding by various comorbidities, we also used a comorbidity score that incorporated 20 different medical conditions including HF, renal failure, respiratory disease, cirrhosis, and malignancy.²⁶

Table 1. Baseline characteristics of the 1:3 PS matched cohort.

	Febuxostat	Allopurinol	Standardized Difference (%)
N	24,936	74,808	--
Demographics
Age- years (median, IQR)	76 (70-82)	76 (71-82)	0
Male (%)	52.3	52.3	0.1
Race -White (%)	76.4	76.2	0.4
U.S. Region
- Midwest (%)	19.3	19.1	0.6
- Northeast (%)	18.7	19.0	−0.6
- South (%)	42.5	42.4	0.3
- West (%)	19.4	19.5	−0.4
Cardiovascular Comorbidities
Myocardial infarction	3.5	3.6	−0.2
Stroke	7.2	7.3	−0.2
Coronary revascularization	2.3	2.3	0.1
Heart failure (%)	35.7	35.8	−0.2
Recent heart failure (60 days) (%)	5.1	5.1	−0.1
Venous thromboembolism (%)	7.7	7.6	0.4
Hypertension (%)	95.4	95.4	−0.1
Peripheral vascular disease (%)	19.7	19.7	−0.1
Cardiovascular disease (%)	12.2	12.2	−0.2
Other Comorbidities
Hyperlipidemia (%)	82.8	82.9	−0.2
Chronic kidney disease (%)	57.1	58.0	−1.7
Chronic obstructive pulmonary disease (%)	32.5	32.5	−0.1
Diabetes mellitus (%)	55.0	55.2	−0.6
Malignancy (%)	21.7	21.5	0.4
Renal stone (%)	5.6	5.7	−0.3
Liver disease (%)	7.1	7.1	0.2
Obesity (%)	17.5	17.3	0.6
Sleep apnea (%)	9.5	9.3	0.5
Smoking (%)	6.0	5.9	0.3
Comorbidity score (median, IQR)	3 (1-6)	3 (1-6)	0
Gout-related Medications
Colchicine (%)	42.8	43.1	−0.7
NSAIDS/COXIB (%)	41.0	41.0	0.1
Opioids (%)	49.4	49.0	0.6
365d-cumulative prednisone equivalent dose, mg (median, IQR)	0 (0-240)	0 (0-210)	0
Any steroid use (365 days) (%)	41.4	41.2	0.3
Recent steroid use (90 days) (%)	29.4	29.2	0.4
Other Medications
ACE inhibitors/angiotensin receptor blockers (%)	68.2	68.3	−0.2
Beta blockers (%)	47.1	47.1	−0.1
Calcium channel blockers (%)	42.8	43.1	−0.5
Diuretics (%)	76.5	76.6	−0.4
Nitrates (%)	17.3	17.4	−0.2
Noninsulin anti-diabetic drugs (%)	30.1	30.3	−0.4
Insulin (%)	14.9	15.1	−0.5
Anticoagulants (%)	21.2	21.1	0.4
Antiplatelets (%)	17.8	18.0	−0.5
Statins (%)	61.4	61.5	−0.1
Other lipid lowering drugs (%)	17.2	17.3	−0.2
Health Care Utilization Pattern
No. of emergency room visits (median, IQR)	0 (0-1)	0 (0-1)	0
No. of all outpatient visits (median, IQR)	14 (8-21)	13 (8-21)	0
No. of prescription drugs (median, IQR)	15 (11-20)	15 (11-20)	0
Hospitalization (%)	33.7	33.7	0
No. of cardiology visits (median, IQR)	0 (0-2)	0 (0-2)	0
No. of rheumatology visits (median, IQR)	0 (0-0)	0 (0-0)	0.1
CRP test ordered (%)	22.1	22.0	0.2
ECG ordered (%)	63.7	64.0	−0.5
Echocardiogram ordered (%)	2.1	2.2	−0.3
Cardiac stress test ordered (%)	16.3	16.6	−0.8
Hemoglobin A1c ordered (%)	54.4	54.7	−0.6
Lipid/cholesterol test ordered (%)	80.8	80.9	−0.3
Uric acid test ordered (%)	86.5	87.1	−1.7
Serum creatinine test ordered (%)	97.9	98.1	−1.3

Recent history of MI (60 day), stroke (60 day), alcoholism, and phosphate binders were present in <1% of patients. For all covariates, standardized differences were less than 10%.

IQR = interquartile range, NSAID = non-steroidal anti-inflammatory drug, COXIB = Cox-2 inhibitor, ACE = angiotensin converting enzyme, ED = emergency department, CRP = C-reactive protein, ECG = electrocardiogram

Statistical Analysis

Main Analysis (Propensity Score Matching, Primary & Secondary Outcomes)

For confounding adjustment, we used propensity score (PS) matching in which all the baseline covariates were simultaneously adjusted for. The PS was defined as the probability of receiving febuxostat versus allopurinol given patients’ baseline characteristics and calculated based on multivariable logistic regression models that incorporated baseline variables such as demographic information, medical comorbidities, medications, and healthcare utilization patterns listed in Table 1. Using nearest-neighbor matching within a “caliper” of 0.05 on the PS scale, febuxostat initiators were matched to allopurinol initiators with a fixed ratio of 1:3 to optimize the size of the febuxostat group and overall study cohort. The fixed ratio of 1:3 was maintained throughout all subsequent main, subgroup and sensitivity analyses. We compared baseline characteristics of febuxostat and allopurinol initiators before and after PS matching. Variables with standardized differences <10% between the two groups were considered well-balanced after PS matching. ^{27, 28}

For the primary ‘as-treated’ analysis, follow-up started on the day after the index date and ended on the earliest date of the following censoring events: drug discontinuation, last day of study database, insurance disenrollment (part A/B/D), occurrence of outcome, death, or nursing home admission. The last drug available date was defined as the last dispensing date plus days of supply with a 30-day grace period. Treatment adherence was calculated using a proportion of days covered (PDC), where PDC (%) was equal to the number of days covered by prescriptions multiple by 100, divided by the total number of days of follow-up.

For the secondary ‘intention-to-treat 365-day (ITT-365d)’ analysis, follow-up time was truncated on the 366^th day after the index date unless patients were censored based on the previously mentioned criteria except drug discontinuation. This analysis was conducted to address the potential for lower adherence over long term follow-up.

Incidence rates (IR) and 95% confidence intervals (CI) were calculated for the previously mentioned primary and secondary outcomes amongst the PS-matched groups separately. Cox proportional hazards regression compared the risk of primary and secondary outcomes in the PS-matched cohorts of febuxostat and allopurinol initiators. Cumulative incidence plots between treatment groups were compared. We assessed the proportional hazards assumption by testing the significance of the interaction term between exposure and follow-up time and found that the assumption was violated for the all-cause mortality analysis (p=0.02 for the interaction term). Therefore, we conducted Cox regression stratified by follow-up time for all-cause mortality. In addition, we ran follow-up time-stratified Cox regression for the primary outcome.

Subgroup Analyses

We conducted two subgroup analyses. The first analysis was an a priori defined PS-matched subgroup analysis by the presence or absence of baseline CVD (defined as history of MI, hospitalized unstable angina, coronary or cerebral revascularization, stroke, or hospitalized TIA). IR and HR were calculated for both primary and secondary outcomes. In the second subgroup analysis, we identified patients with high CV risk, similar to the CARES inclusion criteria (i.e., peripheral vascular disease or diabetes mellitus in addition to the previously defined CVD).

Sensitivity Analyses

We also performed two sensitivity analyses. The first analysis was limited to patients who initiated equipotent dosing of febuxostat (≥40mg daily) versus allopurinol (≥300mg daily) on the index date. The second analysis was limited to patients who initiated equipotent dosing of febuxostat (≥40mg daily) versus allopurinol (≥300mg daily) on the index date and were naïve to both drugs prior to index date.

All analyses were done using SAS 9.4 statistical software (SAS Institute, Cary, N.C.).

RESULTS

Cohort Selection

Application of inclusion and exclusion criteria resulted in a total of 331,134 patients with gout aged 65 years or older continuously enrolled in Medicare Parts A, B and D for at least 365 days prior to initiation of febuxostat (n=26,233) and allopurinol (n=304,901) (Supplemental Figure 1). Following 1:3 PS matching, 95% of febuxostat initiators (n=24,936) and 25% of allopurinol initiators (n=74,808) were included in the study.

Patient Characteristics

Baseline demographics and clinical characteristics of each group after 1:3 PS matching are summarized in Table 1. (See Supplementary Material Appendix Table 1 for baseline characteristics before PS matching). Among febuxostat users, the median (interquartile range (IQR)) age was 76 (70-82) years and 52% were male. Among allopurinol users, the median (IQR) age was 76 (71-82) years and 52% were male. In both groups, 12% had CVD at baseline. Hypertension (95%), CKD (58%), diabetes (55%), and heart failure (36%) were common comorbidities in both groups. Use of gout-related medications including colchicine (43%), NSAIDs (41%), and steroids (41%) was also common amongst all users. All the baseline covariates were well-balanced between the PS-matched groups with a standardized difference <10%.²⁷

The mean [standard deviation (SD)] follow-up time was 1.1 (1.1) years among febuxostat initiators and 1.2 (1.2) years among allopurinol initiators. There were 5,013 (20.1%) febuxostat and 18,235 (24.4%) allopurinol initiators, who had over 730 days of follow-up time. Of the febuxostat initiators, 30.4% had been on allopurinol at some point during the 365 days prior to febuxostat initiation. Among allopurinol initiators, 0.4% had been on febuxostat during the 365 days prior to allopurinol initiation.

Patterns of Febuxostat and Allopurinol Treatment

In the febuxostat group, the median (IQR) PDC was 93.85% (51.96-100) up to 180 days and 89.34% (48.15-100) up to 365 days. In the allopurinol group, the median (IQR) PDC was 85.08% (50.28-100) up to 180 days and 79.78% (33.33-98.92) up to 365 days. Among febuxostat initiators, 98.89% were on a dosage of 40mg or higher per day. For allopurinol initiators, 30.67% were on a daily dosage of 300mg of higher. Of febuxostat initiators, 13.2% had a dose increase during follow-up compared to 22.8% of allopurinol initiators.

Risk of Cardiovascular Events

In the primary as-treated analysis, the IR per 100 person-years for the primary outcome (i.e., hospitalization for MI or stroke) was 3.43 (95% CI 3.22-3.66) in febuxostat and 3.36 (95% CI 3.25-3.49) in allopurinol initiators (Table 2). In the ITT-365d analysis, the IR for the primary outcome was also similar in the two groups: IR per 100 person-years of 3.72 (95% CI 3.45-4.00) for febuxostat and 3.83 (95% CI 3.67-4.00) for allopurinol initiators. The hazard ratio (HR) for the primary outcome was 1.01 (95% CI 0.94-1.08) in the febuxostat compared with allopurinol initiators (Table 2). Cumulative incidence plots also showed null results for the primary outcome with the log-rank test p-value of 0.83 (Figure 1a). In the Cox regression analysis stratified by follow-up time (i.e., treatment duration) for the primary outcome, the HR (95% CI) associated with febuxostat versus allopurinol was 0.84 (0.73-0.98) for 0-1 year of follow-up, 0.88 (0.61-1.25) for 1-2 years, 0.76 (0.42-1.39) for 2-3 years, and 1.17 (0.45-3.05) for >3 years.

Table 2. Risk of cardiovascular events in febuxostat versus allopurinol initiators: 1:3 PS-matched analysis.

	Febuxostat (n=24,936)				Allopurinol (n=74,808)
Outcome	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)
As-treated analysis
Primary outcome
MI or stroke	935	27,251	3.43 (3.22-3.66)	1.01 (0.94-1.08)	3,105	92,264	3.36 (3.25-3.49)	Ref
Secondary outcomes
MI	596	27,440	2.17 (2.00-2.35)	1.03 (0.94-1.13)	1,935	92,962	2.08 (1.99-2.18)	Ref
Stroke	372	27,609	1.35 (1.22-1.49)	0.98 (0.87-1.10)	1,272	93,487	1.36 (1.29-1.44)	Ref
Coronary revascularization	719	27,209	2.64 (2.46-2.84)	0.95 (0.87-1.03)	2,525	91,815	2.75 (2.65-2.86)	Ref
All-cause mortality	1,144	27,809	4.11 (3.88-4.36)	0.95 (0.89-1.02)	4,022	94,219	4.27 (4.14-4.40)	Ref
ITT365-d analysis
Primary outcome
MI or stroke	711	19,132	3.72 (3.45-4.00)	0.97 (0.89-1.06)	2,146	55,986	3.83 (3.67-4.00)	Ref
Secondary outcomes
MI	442	19,192	2.30 (2.10-2.53)	0.97 (0.87-1.08)	1,334	56,190	2.37 (2.25-2.51)	Ref
Stroke	285	19,254	1.48 (1.32-1.66)	0.95 (0.83-1.09)	876	56,357	1.55 (1.45-1.66)	Ref
Coronary revascularization	570	19,133	2.98 (2.74-3.23)	0.93 (0.84-1.02)	1,804	55,946	3.23 (3.08-3.38)	Ref
All-cause mortality	995	19,317	5.15 (4.84-5.48)	0.94 (0.88-1.01)	3,092	56,571	5.47 (5.28-5.66)	Ref

^*IR is per 100 person-years; IR= incidence rate, HR=hazard ratio, CI=confidence interval, MI= myocardial infarction, ITT _365-d=intention-to-treat analysis up to the 365^th day of follow-up

Figure 1. Cumulative incidences of the composite endpoint of MI or stroke and all-cause mortality.

Among the 1:3 propensity score matched cohort of febuxostat and allopurinol initiators, the cumulative incidences of the composite endpoint of MI or stroke (a) and all-cause mortality (b) were compared with the log-rank test (p=0.8 for MI or stroke and p=0.15 for all-cause mortality).

The risk of developing secondary outcomes was also similar between the two groups. In the as-treated analysis, the HR in febuxostat initiators was 1.03 (95% CI 0.94-1.13) for MI, 0.98 (95% CI 0.87-1.10) for stroke, 0.95 (95% CI 0.87-1.03) for coronary revascularization, and 0.95 (95% CI 0.89-1.02) for all-cause mortality (Table 2). Cumulative incidence plots for all-cause mortality was also consistent with the log-rank test p-value of 0.15 (Figure 1b). In the Cox regression analysis stratified by follow-up time for all-cause mortality, the HR (95% CI) associated with febuxostat was 0.75 (0.66-0.86) for 0-1 year of follow-up, 0.85 (0.63-1.15) for 1-2 years, 0.72 (0.53-1.54) for 2-3 years, and 1.25 (0.56-2.80) for >3 years.

The ITT-365d analyses showed consistent results as well for the secondary outcomes.

Risk of Heart Failure

For new onset HF hospitalizations, the IR per 100 person-years was 5.71 (95% CI 5.37-6.06) for febuxostat initiators compared to 5.41 (95% CI 5.23-5.60) for allopurinol initiators in the as-treated analysis (Table 3). The HR for new-onset HF in the as-treated analysis was 1.05 (95% CI 0.98-1.12). For HF exacerbations, the IR per 100 person-years was 42.70 (95% CI 41.16-44.29) for febuxostat initiators compared to 44.06 (95% CI 43.18-44.96) for allopurinol initiators in the as-treated analysis. The HR for HF exacerbation in the as-treated analysis was 0.94 (95% CI 0.91-0.99). Although this latter HR was statistically significantly lower than 1.0, the degree of risk reduction appeared modest. ITT-365d analysis yielded similar results with no difference between the two groups for new onset HF and borderline risk reduction for HF exacerbation.

Table 3. Risk of heart failure (HF) in febuxostat initiators versus allopurinol initiators: 1:3 PS-matched analysis.

	Febuxostat					Allopurinol
Outcome	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)
As-treated analysis
New-onset HF †	15,929	1,056	18,500	5.71 (5.37-6.06)	1.05 (0.98-1.12)	47,787	3,437	63,507	5.41 (5.23-5.60)	Ref
HF exacerbation ‡	8,977	2,855	6,687	42.70 (41.16-44.29)	0.94 (0.91-0.99)	26,931	9,426	21,394	44.06 (43.18- 44.96)	Ref
ITT365-d analysis
New-onset HF †	15,929	760	12,892	5.90 (5.49-6.33)	0.99 (0.91-1.07)	47,787	2,250	37,723	5.97 (5.72-6.22)	Ref
HF exacerbation ‡	8,977	2,646	5,300	49.92 (48.06-51.86)	0.93 (0.89-0.98)	26,931	8,127	15,137	53.69 (52.53- 54.87)	Ref

^*IR is per 100 person-years

^† among the subgroup of patients with no baseline history of HF

^‡among the subgroup of patients with baseline history of HF, only counting the 1^st exacerbation after the index date

IR= incidence rate, HR=hazard ratio, CI=confidence interval, HF=heart failure, ITT _365-d=intention-to-treat analysis up to the 365^th day of follow-up

Subgroup analyses

In the first subgroup analysis by baseline CVD, we noted no significant difference in the primary outcome between febuxostat and allopurinol initiators (Table 4). For the secondary outcome of all-cause mortality, the HR for febuxostat versus allopurinol was 0.97 (95% CI 0.90-1.04) in those without baseline CVD and 0.85 (95% CI 0.72-0.99) among those with baseline CVD. In the second subgroup analysis for high CV risk defined similar to the CARES’ inclusion criteria, we also found no difference in both primary and secondary outcomes, including all-cause mortality (Table 5).

Table 4. Subgroup analysis by baseline CVD: 1:3 PS-matched as-treated analysis.

	Febuxostat					Allopurinol
Outcome	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)
Without baseline CVD
MI or stroke	21,821	726	24,526	2.96 (2.75-3.18)	0.99 (0.91-1.08)	65,463	2,471	83,379	2.96 (2.85-3.08)	Ref
MI	21,821	457	24,683	1.85 (1.69-2.03)	0.98 (0.88-1.08)	65,463	1,579	83,911	1.88 (1.79-1.98)	Ref
Stroke	21,821	291	24,813	1.17 (1.05-1.32)	1.00 (0.88-1.15)	65,463	977	84,420	1.16 (1.09-1.23)	Ref
Coronary revascularization	21,821	577	24,501	2.36 (2.17-2.56)	0.92 (0.84-1.01)	65,463	2,098	82,859	2.53 (2.43-2.64)	Ref
All-cause mortality	21,821	946	24,972	3.79 (3.55-4.04)	0.97 (0.90-1.04)	65,463	3,309	84,984	3.89 (3.76-4.03)	Ref
With baseline CVD
MI or stroke	3,067	207	2,655	7.80 (6.80-8.93)	0.97 (0.83-1.13)	9,201	693	8,803	7.87 (7.31-8.48)	Ref
MI	3,067	139	2,686	5.18 (4.38-6.11)	0.98 (0.81-1.19)	9,201	461	8,938	5.16 (4.71-5.65)	Ref
Stroke	3,067	79	2,726	2.90 (2.32-3.61)	1.03 (0.80-1.32)	9,201	252	9,093	2.77 (2.45-3.14)	Ref
Coronary revascularization	3,067	146	2,634	5.55 (4.72-6.52)	1.10 (0.91-1.33)	9,201	435	8,820	4.93 (4.49-5.42)	Ref
All-cause mortality	3,067	195	2,765	7.05 (6.13-8.12)	0.85 (0.72-0.99)	9,201	756	9,235	8.19 (7.62-8.79)	Ref

^*IR is per 100 person-years; IR= incidence rate, HR=hazard ratio, CI=confidence interval, MI= myocardial infarction, CVD=cardiovascular disease

Table 5. Subgroup analysis by high CV risk: 1:3 PS-matched as-treated analysis.

	Febuxostat					Allopurinol
Outcome	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)
Without high CV risk
MI or stroke	16,174	674	16,943	3.98 (3.69-4.29)	0.97 (0.89-1.06)	48,522	2,280	56,347	4.05 (3.88-4.22)	Ref
MI	16,174	448	17,062	2.63 (2.39-2.88)	1.00 (0.90-1.11)	48,522	1,479	56,814	2.60 (2.47-2.74)	Ref
Stroke	16,174	252	17,208	1.46 (1.29-1.66)	0.96 (0.84-1.11)	48,522	866	57,335	1.51 (1.41-1.61)	Ref
Coronary revascularization	16,174	509	16,920	3.01 (2.76-3.28)	0.91 (0.82-1.00)	48,522	1,840	56,089	3.28 (3.13-3.43)	Ref
All-cause mortality	16,174	801	17,337	4.62 (4.31-4.95)	0.93 (0.86-1.00)	48,522	2,847	57,821	4.92 (4.75-5.11)	Ref
With high CV risk
MI or stroke	8,645	253	10,212	2.48 (2.19-2.80)	1.03 (0.90-1.19)	25,935	832	34,790	2.39 (2.23-2.56)	Ref
MI	8,645	143	10,279	1.39 (1.18-1.64)	1.01 (0.84-1.22)	25,935	479	35,029	1.37 (1.25-1.50)	Ref
Stroke	8,645	116	10,303	1.13 (0.94-1.35)	1.04 (0.85-1.28)	25,935	378	35,123	1.08 (0.97-1.19)	Ref
Coronary revascularization	8,645	202	10,198	1.98 (1.73-2.27)	1.02 (0.87-1.19)	25,935	668	34,725	1.92 (1.78-2.08)	Ref
All-cause mortality	8,645	332	10,371	3.20 (2.88-3.57)	0.97 (0.86-1.10)	25,935	1,162	35,366	3.29 (3.10-3.48)	Ref

^*IR is per 100 person-years; IR= incidence rate, HR=hazard ratio, CI=confidence interval, MI= myocardial infarction, CV=cardiovascular

Sensitivity analyses

For the sensitivity analysis limited to patients who initiated febuxostat ≥40mg daily versus allopurinol ≥300mg daily on the index date, the risk for the primary outcome (hospitalization for MI or stroke) was similar between the PS-matched febuxostat and allopurinol groups with a HR of 1.05 (95% CI 0.94-1.18) (Table 6). Results for the secondary outcomes including MI, stroke, coronary revascularization, new onset HF, HF recurrence, and all-cause mortality were all consistent with the main analyses.

Table 6. Sensitivity analysis- Risk of cardiovascular events in febuxostat (≥40mg/d) versus allopurinol (≥300mg/d) initiators: 1:3 PS-matched as-treated analysis.

	Febuxostat					Allopurinol
Outcome	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)
Primary outcome
MI or stroke	12,252	404	13,848	2.92 (2.65-3.22)	1.05 (0.94-1.18)	36,756	1,359	49,755	2.73 (2.59-2.88)	Ref
Secondary outcomes
MI	12,252	242	13,938	1.74 (1.53-1.97)	1.08 (0.93-1.24)	36,756	796	50,144	1.59 (1.48-1.70)	Ref
Stroke	12,252	178	13,993	1.27 (1.10-1.47)	1.05 (0.89-1.24)	36,756	605	50,314	1.20 (1.11-1.30)	Ref
Coronary revascularization	12,252	358	13,779	2.60 (2.34-2.88)	1.01 (0.90-1.14)	36,756	1,247	49,379	2.53 (2.39-2.67)	Ref
New onset HF	9,021	471	10,461	4.50 (4.11-4.93)	1.00 (0.91-1.11)	27,063	1,687	37,948	4.45 (4.24-4.66)	Ref
HF exacerbation	3,261	948	2,556	37.09 (34.80- 39.53)	0.97 (0.90-1.04)	9,783	3,274	8,866	36.93 (35.68- 38.22)	Ref
All-cause mortality	12,252	426	14,091	3.02 (2.75-3.32)	0.93 (0.84-1.04)	36,756	1,620	50,720	3.19 (3.04-3.35)	Ref

^*IR is per 100 person-years; IR= incidence rate, HR=hazard ratio, CI=confidence interval, MI= myocardial infarction, HF=heart failure

When we further restricted the cohort to those who initiated equipotent dosing (i.e., at least 40mg/d febuxostat or 300mg/d allopurinol) and had no prior use of either febuxostat or allopurinol before the index date, there was no difference in risk between the two groups for the primary outcome (HR 0.96, 95% CI 0.85-1.08) (Table 7) as well as all secondary outcomes.

Table 7. Sensitivity analysis- Risk of cardiovascular events in febuxostat (≥40mg/d) versus allopurinol (≥300mg/d) initiators naïve to both drugs: 1:3 PS-matched as-treated analysis.

	Febuxostat					Allopurinol
Outcome	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)	N	Event (n)	Person- years	IR * (95% CI)	HR (95% CI)
Primary outcome
MI or stroke	11,220	368	12,080	3.05 (2.75-3.37)	0.96 (0.85-1.08)	33,660	1,346	42,943	3.13 (2.97-3.31)	Ref
Secondary outcomes
MI	11,220	224	12,151	1.84 (1.62-2.10)	0.97 (0.84-1.13)	33,660	811	43,284	1.87 (1.75-2.01)	Ref
Stroke	11,220	158	12,219	1.29 (1.11-1.51)	0.97 (0.81-1.15)	33,660	575	43,489	1.32 (1.22-1.44)	Ref
Coronary revascularization	11,220	308	12,047	2.56 (2.29-2.86)	0.97 (0.86-1.10)	33,660	1,100	42,691	2.58 (2.43-2.73)	Ref
New onset HF	7,785	454	8,667	5.24 (4.78-5.74)	1.05 (0.95-1.17)	23,355	1,541	31,460	4.90 (4.66-5.15)	Ref
HF exacerbation	3,412	1,022	2,561	39.91 (37.54- 42.44)	1.00 (0.93-1.07)	10,236	3,393	8,891	38.16 (36.90- 39.47)	Ref
All-cause mortality	11,220	420	12,295	3.42 (3.10-3.76)	0.93 (0.83-1.03)	33,660	1,597	43,848	3.64 (3.47-3.83)	Ref

^*IR is per 100 person-years; IR= incidence rate, HR=hazard ratio, CI=confidence interval, MI= myocardial infarction, HF=heart failure

DISCUSSION

In our study of U.S. Medicare patients with gout, initiation of febuxostat compared with allopurinol was not associated with a change in risk of cardiovascular events for MI, stroke, new onset HF, coronary revascularization, or all-cause mortality. This was observed in the main as-treated as well as intention-to-treat analyses truncated at one-year of follow-up. However, in a follow-up time-stratified analysis, we observed a trend toward a greater risk of all-cause mortality in the febuxostat group with over 3 years of follow-up versus the allopurinol group with over 3 years of follow-up. We also found a modestly decreased risk for HF exacerbation associated with febuxostat versus allopurinol in the primary as-treated (HR 0.94, 95% CI 0.91-0.99) and intention-to treat analysis (HR 0.93, 95% CI 0.89-0.98).

In the subgroup analysis, there was no difference in the risk of cardiovascular events between the two groups with and without baseline CVD, except for a modestly decreased risk of all-cause mortality in febuxostat users with baseline CVD (HR 0.85, 95% CI 0.72-0.99). Among patients with high CV risk at baseline, we noted no difference in the risk of cardiovascular events or all-cause mortality between the two drugs. Finally, in the two sensitivity analyses where we compared 1) equipotent index dosages (febuxostat ≥40mg versus allopurinol ≥300mg daily) and 2) equipotent index dosages (febuxostat ≥40mg versus allopurinol ≥300mg daily) with naivety to both drugs prior to index date, we found no difference in cardiovascular risk between febuxostat and allopurinol users.

Although the original phase III randomized trials for febuxostat (APEX and FACT) revealed a numerically higher but statistically non-significant risk for adverse cardiovascular events in febuxostat initiators compared to allopurinol initiators, the rates of CV events did not correlate with febuxostat dosage and the number of cardiovascular events did not increase over the duration of the study.^{17, 18} In APEX, the number of adverse cardiovascular events ranged from 1 to 5 (<1-2%) in each of the febuxostat groups (80, 120, and 240mg) versus 1 (<1%) in the allopurinol group.¹⁷ In FACT, the number of adverse cardiovascular events was 1 (<1%) in each of the febuxostat groups (80mg and 120mg) versus 0 (0%) in the allopurinol group.¹⁸ In both trials, the adverse cardiovascular events were considered to be unlikely related to the study drug. Nonetheless, the FDA required the sponsor to collect additional safety data.

The recent CARES trial on CV risk of febuxostat versus allopurinol showed that the individual risk of cardiovascular mortality and all-cause mortality were both higher in febuxostat initiators.²⁰ There was no difference in risk of the composite endpoint of cardiovascular death, nonfatal MI, nonfatal stroke, and unstable angina with urgent coronary revascularization between the two drug groups. Similar to CARES, we found no difference in individual risk for MI, stroke, or coronary revascularization between the two groups. However, we also found no difference in risk of all-cause mortality in the primary analysis, as well as subgroup analyses limited to those with baseline CVD or baseline “high cardiovascular risk” (i.e., CVD, including PVD or DM as seen in CARES).

The discrepancy in results for mortality between CARES and our study may be related to differences in the underlying populations. CARES was restricted to patients with high cardiovascular risk defined as those with a history of major cardiovascular or cerebrovascular disease including MI, hospitalized unstable angina, coronary or cerebral vascularization procedure, stroke, hospitalized TIA, peripheral vascular disease or diabetes mellitus with micro/macrovascular complications. Our main analysis included both patients with and without CVD. Our Medicare study population was also older with a more equal sex distribution compared to CARES (52% male with a median age of 76 years in our trial versus 84% male with a median age of 64 years in CARES). However, in our follow-up time-stratified analysis, the risk for all-cause mortality appeared to be increased, albeit not statistically significant, among patients who used febuxostat over 3 years compared with those who used allopurinol over 3 years. We were unable to assess cause-specific deaths including cardiovascular mortality due to limitations with the Medicare database.

There are several strengths to our study. By using an active comparator design (febuxostat versus allopurinol rather than febuxostat versus placebo), we increased the overlapping characteristics of the two groups and minimized unmeasured confounding.²⁹ We increased generalizability with a large sample size, performed a comprehensive covariate adjustment via propensity score matching, and used validated algorithms for outcome measurement. In addition, unlike the CARES study, we included patients representative of Medicare enrollees regardless of baseline cardiovascular comorbidities.

Limitations included misclassification bias as participant eligibility was largely dependent on diagnosis code, although this was likely minimized by the additional inclusion criteria of being prescribed at least one urate lowering drug. Second, Medicare claims data did not provide information on cause-specific mortality, family history of CVD, severity of gout, and use of over-the-counter medications such as NSAIDs or aspirin, which could have led to residual confounding. Third, with regards to generalizability, our gout cohort was roughly 52% male, which may appear low; however, this prevalence is comparable to other U.S. Medicare gout studies and may reflect the older age of the cohort.^{11, 30} Fourth, mean follow-up time was around 1.2 years, which led to less precise estimates for the long-term effects of febuxostat on cardiovascular risk as well as all-cause mortality although our study still included many patients (n=23,317) with more than 2 years of follow-up. Fifth, as the aim of our study was to determine comparative cardiovascular safety of febuxostat and allopurinol, we did not examine the risk of cardiovascular events associated with XOIs compared with untreated gout patients. Finally, our study did not include participant serum urate levels so it is possible that participants were under-dosed and inadequately treated. Any instances of suboptimal gout treatment, however, likely reflect real-life patterns of urate-lowering treatment allocation in the U.S., as prior studies on medication use and serum uric acid monitoring in gout patients on urate-lowering therapy demonstrate widespread suboptimal dosing.³

CONCLUSIONS

In this retrospective cohort study of 99,744 patients with gout over age 65 years enrolled in Medicare, we noted no overall difference in the risk for myocardial infarction, stroke, coronary revascularization, new heart failure, or all-cause mortality between febuxostat and allopurinol initiators. However, we noted a trend toward an increased risk, not statistically significant, for all-cause mortality in long-term users of febuxostat (over 3 years) versus long-term users of allopurinol. The risk of HF exacerbation was slightly lower among febuxostat initiators versus allopurinol. Subgroup and sensitivity analyses showed consistent results.

CLINICAL PERSPECTIVE

What is new?

• In the CARES trial, febuxostat was non-inferior to allopurinol with respect to rates of adverse cardiovascular events, but febuxostat users had a greater risk of cardiovascular mortality and all-cause mortality.

• In this cohort study of 99,744 older patients with gout enrolled in Medicare, overall we found no difference in cardiovascular risk including myocardial infarction, stroke, coronary revascularization, new heart failure, or all-cause mortality between febuxostat and allopurinol initiators.

What are the clinical implications?

• Among older gout patients with and without cardiovascular comorbidities, the risk of cardiovascular events and all-cause mortality was similar between febuxostat and allopurinol initiators.

• However, there is a suggestion for an increased risk of all-cause mortality associated with long-term use of febuxostat versus allopurinol.

ABBREVIATIONS

CI

confidence interval

CKD

chronic kidney disease

COXIB

Cox-2 inhibitor

CRP

C-reactive protein

CV

cardiovascular

CVD

cardiovascular disease

ECG

electrocardiogram

ED

emergency department

HF

heart failure

HR

hazard ratio

ICD-9

International Classification of Diseases, Ninth Revision

IQR

interquartile range

IR

incidence rate

ITT

intention-to-treat

MI

myocardial infarction

NSAID

nonsteroidal anti-inflammatory drug

PDC

proportion of days covered

PS

propensity score

RCT

randomized controlled trial

XOI

xanthine oxidase inhibitor