Reassessing the Cardiovascular Safety of Febuxostat: Implications of the Febuxostat Versus Allopurinol Streamlined Trial

The US Food and Drug Administration (FDA)–mandated Cardiovascular Safety of Febuxostat and Allopurinol in Patients with Gout and Cardiovascular Morbidities (CARES) trial, published in 2018, demonstrated increased all-cause mortality and death from cardiovascular causes in participants randomized to receive febuxostat compared with allopurinol (1). The outcome of this trial and the subsequent FDA Drug Safety Communication and Boxed Warning (2) resulted in substantial reductions in febuxostat use in the US (3). The European Medicines Agency (EMA)–mandated Febuxostat versus Allopurinol Streamlined Trial (FAST), published in 2020, demonstrated no increased risk of composite cardiovascular events, cardiovascular disease (CVD) mortality, or all-cause mortality with febuxostat as compared with allopurinol (4). We will discuss implications of these new findings for gout management.

CARES trial findings

As described in our previous commentary (5), CARES was a multicenter, double-blind, noninferiority cardiovascular outcomes trial in 6,190 patients with gout and established CVD, with a median follow-up period of 32 months (Table 1) (1). Though CARES was a large clinical trial, there were important limitations including very high rates of study medication discontinuation (>50% of participants), large amounts of missing data (45% of participants did not complete all trial visits), and concerns about incomplete capture of cardiovascular and mortality events. Moreover, the majority of mortality events in the CARES trial (~85%) occurred when participants were not taking urate-lowering therapy (ULT). Hence, results of the EMA-mandated FAST study were widely anticipated.

Table 1.

Comparison of the CARES trial and the FAST study^*

	CARES study (ref. 1)	FAST study (ref. 4)
Trial design	Prospective, randomized, multicenter noninferiority trial	Prospective, randomized, multicenter noninferiority trial
Blinding	Double-blind, blinded–end point adjudication	Open-label, blinded–end point adjudication
Setting	US, Mexico, and Canada	Scotland, England, Denmark, and Sweden
Start date	April 2010	December 2011
No. of participants randomized	6,190	6,128
Study population characteristics	Gout, history of major CVD, serum urate level ≥7.0 mg/dl (0.42 mmoles/liter) or ≥6.0 mg/dl (0.36 mmoles/liter), inadequately controlled gout	Gout, age ≥60 years, already receiving allopurinol, at least 1 additional cardiovascular risk factor
Relevant exclusions	MI or stroke within 60 days prior to screening, severe renal impairment	MI or stroke in the previous 6 months, congestive heart failure (NYHA class III or IV), severe renal impairment
% with established CVD	100	33.4
% with tophaceous gout	21.3	10.2
Run-in period	No	Yes, to optimize allopurinol dose to achieve serum urate levels <6 mg/dl (0.36 mmoles/liter) prior to randomization
Final daily febuxostat dose	Median dose 40 mg; 61.0% receiving 40 mg, 39.0% receiving 80 mg	Median dose 80 mg; 97.5% receiving 80 mg, 2.5% receiving 120 mg
Final daily allopurinol dose	Median dose 300 mg; 21.8% receiving 200 mg, 44.6% receiving 300 mg, 33.6% receiving ≥400 mg	Median dose 300 mg; 10.0% receiving 100 mg, 23.3% receiving 200 mg, 50.9% receiving 300 mg, 15.8% receiving ≥400 mg
Primary end point	Composite of death from cardiovascular cause, nonfatal MI, nonfatal stroke, or unstable angina with urgent revascularization	Composite of hospitalization for nonfatal MI or biomarker-positive acute coronary syndrome, nonfatal stroke, or death due to a cardiovascular event
Follow-up strategy for primary analysis	Study visits and telephone follow-up	Record linkage to centralized databases for hospitalizations, deaths, and cancer diagnoses, in addition to study visits and telephone follow-up
Primary analysis	Modified ITT noninferiority analysis with a noninferior limit of HR of 1.3	On-treatment noninferiority analysis with a noninferior limit of HR of 1.3
Median follow-up duration, days	Febuxostat 968; allopurinol 942	All 1,467
% lost to follow-up	45.0	5.8
% assigned drug discontinuation	Febuxostat 57.3; allopurinol 55.9	Febuxostat 32.4; allopurinol 16.5
% prescribed colchicine for gout flare prophylaxis (first 6 months)	Febuxostat 84.1; allopurinol 83.8	Febuxostat 71.4; allopurinol 52.6
Primary end point result	Primary ITT analysis: Febuxostat (10.8%) was noninferior to allopurinol (10.4%) (HR 1.03 [97% CI 0.87–1.23])	Primary on-treatment analysis: Febuxostat (5.6%) was noninferior to allopurinol (7.9%) (HR 0.85 [95% CI 0.70–1.03]) Secondary ITT analysis: Febuxostat (8.4%) was noninferior to allopurinol (9.3%) (HR 0.89 [95% CI 0.75–1.06])
All-cause deaths	Primary ITT analysis: There were more deaths with febuxostat (7.8%) than with allopurinol (6.4%) (HR 1.22 [95% CI 1.01–1.47])	Primary on-treatment analysis: Febuxostat (3.8%) was noninferior to allopurinol (5.7%) (HR 0.75 [95% CI 0.59–0.95]) Secondary ITT analysis: Febuxostat (7.2%) was noninferior to allopurinol (8.6%) (HR 0.84 [95% CI 0.71–1.01])
Death from cardiovascular causes	Primary ITT analysis: There were more deaths with febuxostat (4.3%) than with allopurinol (3.2%) (HR 1.34 [95% CI 1.03–1.73])	Primary on-treatment analysis: Febuxostat (2.0%) was noninferior to allopurinol (2.7%) (HR 0.91 [95% CI 0.66–1.27]) Secondary ITT analysis: Febuxostat (3.8%) was noninferior to allopurinol (4.0%) (HR 0.96 [95% CI 0.74–1.23])
Serum urate outcome	Similar proportion of participants had serum urate levels <6 mg/dl; more in the febuxostat group had serum urate levels <5 mg/dl	Mean follow-up serum urate level 3.6 mg/dl in the febuxostat group and 5.0 mg/dl in the allopurinol group
Gout flare outcome	Febuxostat 0.68 flares per patient-year; allopurinol 0.63 flares per patient-year	Febuxostat at least 1 flare (18 flares per 100 patient-years); allopurinol 20 flares per 100 patient-years

^*CARES = Cardiovascular Safety of Febuxostat and Allopurinol in Patients with Gout and Cardiovascular Morbidities; FAST = Febuxostat versus Allopurinol Streamlined Trial; CVD = cardiovascular disease; MI = myocardial infarction; NYHA = New York Heart Association; ITT = intent-to-treat; HR = hazard ratio; 97% CI = 97% confidence interval.

The FAST study

The FAST study was a prospective, randomized, open-label, blinded–end point, noninferiority cardiovascular outcomes trial in 6,128 participants with gout and at least 1 additional cardiovascular risk factor, who were already receiving allopurinol (Table 1) (4). Following a run-in period in which allopurinol doses were optimized to achieve a serum urate level of <6 mg/dl, participants were randomized to either resume allopurinol at the optimized dose or begin treatment with febuxostat at 80 mg/day (increasing to 120 mg daily if required to achieve a serum urate level <6 mg/dl) after a washout period of 7–21 days. During the median follow-up period of ~43 months, the primary (on-treatment) analysis of major adverse cardiovascular events (MACE) (Table 1) revealed that febuxostat was noninferior to allopurinol, with an adjusted hazard ratio (HR) of 0.85 (95% confidence interval [95% CI] 0.70–1.03). In the febuxostat group, 3.8% of participants died (of any cause), compared with 5.7% of participants in the allopurinol group (HR 0.75 [95% CI 0.59–0.95]). Febuxostat (median dose 80 mg/day) led to lower serum urate levels than allopurinol (median dose 300 mg/day) throughout the study follow-up (mean 3.6 mg/dl versus 5.0 mg/dl). End points were assessed in a blinded manner by an independent clinical events classification committee.

FAST participant retention was excellent (94%). However, there was differential withdrawal of ULT during the first year (32.4% in the febuxostat group versus 16.5% in the allopurinol group), and more participants in the febuxostat group (71.4%) received colchicine as gout flare prophylaxis compared with participants in the allopurinol group (52.6%). The fact that participants had previously been receiving allopurinol (for a median duration of 6 years) and had achieved the target urate level of <6 mg/dl by the time of randomization likely contributed to higher intolerance or discontinuation of febuxostat in this open-label study. Additional intent-to-treat (ITT) analyses, which kept the original trial assignment until the end of the follow-up period without differential loss, demonstrated results consistent with the primary findings (HR for MACE 0.89 [95% CI 0.75–1.06] and HR for all-cause death 0.84 [95% CI 0.71–1.01]), which supports the trial’s internal validity. Similarly, unblinded switching to febuxostat, which was considered more potent than allopurinol, could have contributed to more prescriptions for colchicine gout flare prophylaxis in the febuxostat arm after the post-randomization washout period. Given the cardiovascular-protective effect of low-dose colchicine (6,7), these participants could have experienced the cardiovascular benefits, at least while exposed to colchicine. Nevertheless, subgroup analyses of participants not exposed to colchicine for gout flare prophylaxis showed consistent findings (HR 0.84 in on-treatment analysis and 0.82 in ITT analysis), suggesting that colchicine was probably not a significant factor.

Participants in the FAST study were recruited mainly in primary care settings, reflecting the clinical setting in which most gout is managed. As such, the study population (33% with CVD) is broadly applicable to the gout population at large. The FAST study excluded people who experienced myocardial infarction (MI) or stroke in the preceding 6 months or who had severe heart failure or chronic kidney disease, whereas in the CARES trial, patients who had experienced MI or stroke were excluded only if the event had occurred within 60 days prior to screening and a history of major CVD was required for enrollment in the CARES trial. The CARES trial also enrolled a higher proportion of participants with tophaceous gout (21% versus 10%), indicative of more severe disease. Considering the trial design, FAST findings should be most generalizable to febuxostat use after allopurinol use. However, if cardiovascular risk is not affected by (pre-trial) allopurinol use among people with gout (thus, not causing selection bias such as depletion of participants susceptible to cardiovascular conditions), generalizability of cardiovascular outcomes in the FAST study would also be analogous to that of a trial without universal pre-trial exposure to allopurinol (e.g., CARES).

To date, there is no high-level evidence for the mortality or cardiovascular impact of allopurinol, leaving unclear the inference of the generalizability of the FAST study in relation to trials without a lead-in exposure to allopurinol. Overall rates of serious adverse events (SAEs) were similar between groups, and there were fewer neoplasms, including malignant neoplasms, in the febuxostat group in the FAST study. Nevertheless, patients who had experienced AEs or other tolerability issues associated with the initiation of allopurinol would have been selected out before the FAST study, while febuxostat was an incident exposure, thus not benefiting from such a selection process.

Remaining uncertainties after the FAST trial

Below we discuss some remaining uncertainties after the FAST study, some of which are more readily addressable than others.

1. The absence of a placebo arm in the FAST study or CARES trial makes it unclear whether allopurinol or febuxostat has any impact on cardiovascular events compared with no use of ULT in people with gout. However, use of a placebo would be ethically challenging given the indications for ULT in these trial participants with gout (8).

2. While the results from the subgroup analysis among those not exposed to colchicine prophylaxis were consistent with the main findings, formal mediation analysis of post-randomization exposure to colchicine would be valuable in this at-risk gout population.

3. A formal time-varying analysis would have been helpful to assess potential reasons for ULT discontinuation beyond postulating simple reluctance and resistance to switching to a new drug (febuxostat) from an effective drug.

4. There was a lower risk of the primary cardiovascular end point for febuxostat (adjusted HR 0.66 [95% CI 0.51–0.86]) in the subgroup with baseline serum urate levels <5 mg/dl, but no such difference among those with baseline serum urate levels ≥5 mg/dl. Clarifying associations between serum urate levels and cardiovascular risk in the FAST study would be valuable.

5. Additional data on requirements for ULT, particularly after discontinuation of febuxostat, would be helpful, as findings of the ITT analysis could potentially be impacted if participants began treatment with the other urate-lowering agent (including reinitiating allopurinol after stopping febuxostat) during the trial follow-up.

6. The potential role of gout flares in cardiovascular risk remains unclear. Neither the CARES nor the FAST trial provided characterization of flare burden (severity, duration, frequency) that might mediate cardiovascular events.

Implications of the FAST findings together with CARES trial data

Notwithstanding some uncertainties associated with the FAST study, and the differences from the CARES trial (Table 1), the FAST findings suggest that the CARES trial could have been critically hampered by its high rate of loss to follow-up (45%), as shown by its post hoc ascertainment nullifying the mortality risk associated with febuxostat use (1). Despite the higher febuxostat doses in the FAST study (median 80 mg daily, versus 40 mg daily in CARES), febuxostat tended to have a lower mortality risk than allopurinol in the FAST study, significantly lower in the on-treatment analysis (primary approach), by ~25%, as discussed above. It remains unclear whether further lowering serum urate levels with a higher febuxostat dose mediated this finding. While the FAST findings may not clarify all the concerns raised by the CARES trial for patients with gout and established CVD, it is also important to remember that >55% of CARES participants discontinued study medication and 45% were lost to follow-up. With discontinuation and rates of loss to follow-up this extreme, it is very difficult to ensure the internal validity of the findings, as noted in the conflicting findings from the fuller post hoc ascertainment (1). Furthermore, the CARES trial showed internal inconsistency between primary MACE and CVD mortality end points, whereas FAST findings were internally consistent. Finally, results of several recent large-scale pharmacoepidemiologic studies also support the findings of the FAST study but not the CARES trial (9–11).

If results from the FAST study had demonstrated any evidence of increased risk of mortality, it would have furthered the concerns raised by the CARES trial findings. However, FAST findings did not demonstrate any such signal, and there was actually a suggestion of survival benefit associated with febuxostat use in the FAST study, even with a mean doubling of the febuxostat exposure dose. While there are a number of differences between the CARES study and the FAST study, the most important one threatening internal validity is the rate of loss to follow-up in the CARES trial (45%, versus 6% in FAST). To this end, the FAST study is considered to have superior internal validity compared with the CARES study, regardless of generalizability (external validity). Naturally, generalizability matters as well; however, without internal validity, i.e., valid effect estimates, generalizability is meaningless. Thus, based on the current evidence, it is our view that the collective verdict on the cardiovascular safety of febuxostat should rely more on results from the FAST study than CARES results. To this end, we support the FAST authors’ suggestion that regulatory agencies update their guidance on the cardiovascular risk associated with the use of febuxostat (4).