Dapagliflozin reduces uric acid concentration, an independent predictor of adverse outcomes in DAPA-HF.

AIMS: Blood uric acid (UA) levels are frequently elevated in patients with heart failure and reduced ejection fraction (HFrEF), may lead to gout and are associated with worse outcomes. Reduction in UA is desirable in HFrEF and sodium-glucose cotransporter 2 inhibitors may have this effect. We aimed to examine the association between UA and outcomes, the effect of dapagliflozin according to baseline UA level, and the effect of dapagliflozin on UA in patients with HFrEF in the DAPA-HF trial. METHODS AND
RESULTS: The association between UA and the primary composite outcome of cardiovascular death or worsening heart failure, its components, and all-cause mortality was examined using Cox regression analyses among 3119 patients using tertiles of UA, after adjustment for other prognostic variables. Change in UA from baseline over 12 months was also evaluated. Patients in tertile 3 (UA ≥6.8 mg/dl) versus tertile 1 (<5.4 mg/dl) were younger (66.3 ± 10.8 vs. 68 ± 10.2 years), more often male (83.1% vs. 71.5%), had lower estimated glomerular filtration rate (58.2 ± 17.4 vs. 70.6 ± 18.7 ml/min/1.73 m2 ), and more often treated with diuretics. Higher UA was associated with a greater risk of the primary outcome (adjusted hazard ratio tertile 3 vs. tertile 1: 1.32, 95% confidence interval [CI] 1.06-1.66; p = 0.01). The risk of heart failure hospitalization and cardiovascular death increased by 7% and 6%, respectively per 1 mg/dl unit increase of UA (p = 0.04 and p = 0.07). Spline analysis revealed a linear increase in risk above a cut-off UA value of 7.09 mg/dl. Compared with placebo, dapagliflozin reduced UA by 0.84 mg/dl (95% CI -0.93 to -0.74) over 12 months (p < 0.001). Dapagliflozin improved outcomes, irrespective of baseline UA concentration.
CONCLUSION: Uric acid remains an independent predictor of worse outcomes in a well-treated contemporary HFrEF population. Compared with placebo, dapagliflozin reduced UA and improved outcomes irrespective of UA concentration.

Introduction

Uric acid (UA) is the final product of purine metabolism and blood levels reflect dietary intake of purines, synthesis of UA by xanthine oxidase and excretion of UA, mainly by the kidneys. Consequently, UA may be elevated in heart failure because elimination is reduced due to impaired kidney function and because diuretics impair uric acid excretion. , , , , UA may also be increased because of increased production due to greater xanthine oxidase activity in patients with heart failure. , As a result, hyperuricaemia is common in heart failure and higher UA is associated with worse clinical outcomes. , , , The association between higher UA and worse outcomes persists after adjustment for renal function, diuretic use and dose, and natriuretic peptide levels. , , Whether this is due to unmeasured confounding, or a directly injurious effect of UA is unknown. , , , , , , , , , , However, UA increases cytokine and chemokine production, promotes inflammation, impairs endothelial function and activates the renin–angiotensin system. , , , , , , , , , , In addition, UA may be a marker of oxidative stress as xanthine oxidase generates superoxide along with UA. , , , , , , , , , , Hyperuricaemia can also lead to gout which is common in patients with heart failure, is difficult to manage and may lead to and prolong hospitalization. , Therefore, drugs are frequently used, prophylactically, to reduce UA in patients with heart failure, with approximately 15%–20% of patients treated in this way.

For these reasons, the effect of therapies for heart failure on UA is of interest and agents that lower UA have even been investigated as a potential treatment for heart failure. , The angiotensin receptor blocker losartan inhibition has been shown to reduce UA in patients without heart failure and neprilysin inhibition also reduces UA in patients with both heart failure with reduced (HFrEF) and preserved ejection fraction. , , Recently, sodium–glucose cotransporter 2 (SGLT2) inhibitors have been shown to reduce hospitalization and death in patients with HFrEF. , , These drugs also reduce UA in patients with diabetes, although the exact mechanism of this effect is not understood and whether SGLT2 inhibitors also reduce UA in patients without diabetes is unknown. , , ,

Therefore, we assessed the effect of dapagliflozin on UA in patients with HFrEF, with and without type 2 diabetes, enrolled in the Dapagliflozin and Prevention of Adverse‐outcomes in Heart Failure trial (DAPA‐HF). We also examined whether UA remains an independent predictor of adverse outcomes in patients receiving optimum contemporary treatment for HFrEF.

Methods

DAPA‐HF was a randomized double‐blind, placebo‐controlled, event‐driven trial in patients with HFrEF, with or without type 2 diabetes. The design, baseline characteristics, and primary results are published. , , Ethics Committees for the 410 participating institutions in 20 countries approved the protocol and all patients gave written informed consent.

Study patients and treatment

Patients in New York Heart Association (NYHA) functional class II–IV, with a left ventricular ejection fraction (LVEF) ≤40%, and an elevated N‐terminal pro‐B‐type natriuretic peptide (NT‐proBNP) concentration, were eligible if receiving standard pharmacological and device therapy. The key exclusion criteria were: type 1 diabetes mellitus, symptomatic hypotension/systolic blood pressure <95 mmHg, and an estimated glomerular filtration rate (eGFR) <30 ml/min/1.73 m2. Dapagliflozin10 mg was compared to a matching placebo, taken once daily in addition to standard treatment.

Measurement of uric acid

Blood samples were taken at randomization and 52 weeks. UA was measured using stored EDTA plasma in a central laboratory using an automated platform and the manufacturer's calibration and quality control materials (c311, Roche Diagnostics, Burgess Hill, UK). The coefficient of variation was 2.0% for a low control and 2.8% for a high control.

Outcomes

The primary trial outcome was the composite of worsening heart failure event (heart failure hospitalization or urgent visit for heart failure requiring intravenous therapy) or cardiovascular death, whichever occurred first. In this study, we investigated the association between baseline UA and the risk of the primary outcome, its composites, and all‐cause mortality. We also examined the effect of dapagliflozin according to baseline UA analysed as a categorical and continuous variable (see below).

In addition, we examined the effect of dapagliflozin on UA level after randomization (difference between baseline and 12‐month measurement) and initiation of new UA‐lowering therapy.

Statistical analysis

Serum UA levels at baseline were categorized into tertiles. Baseline characteristics according to serum UA tertile are presented as frequencies and percentages for categorical variables and means with standard deviation or medians with interquartile range (IQR) for continuous variables. A non‐parametric test for trend across groups, an extension of the Wilcoxon rank sum test, was used to examine for variation in baseline characteristics across UA tertiles. Use of oral loop diuretics at baseline was grouped in categories of furosemide equivalents: 40 mg furosemide = 20 mg torsemide = 1 mg bumetanide. Non‐loop diuretics were categorized as thiazides or as ‘other’.

Incidence rates for each outcome of interest are presented per 100 person‐years of follow‐up. Event rates in each UA tertile were estimated by the Kaplan–Meier method and compared using the log‐rank test. Cox proportional hazards regression models stratified by diabetes status and adjusted for heart failure hospitalization (except for all‐cause mortality) and randomized treatment group were used to compare hazard ratios (HR) with 95% confidence intervals (CI) for outcomes according to UA tertiles. In multivariable models, the HR was further adjusted for the following baseline characteristics: age, sex, pulse, systolic blood pressure, body mass index, atrial fibrillation, diabetes status, aetiology of heart failure, LVEF, NYHA functional classification, NT‐proBNP (log), eGFR, non‐loop diuretic use, loop diuretic use dose and use of an angiotensin‐converting enzyme inhibitor, angiotensin receptor blocker or angiotensin receptor–neprilysin inhibitor.

The association between UA and each outcome was also assessed using a restricted cubic spline with five knots, using UA of 7.0 mg/dl as a reference in the same multivariable‐adjusted model. The proportional hazards assumption was evaluated using plots of Schoenfeld residuals versus log time and found valid, as was the assumption of linearity of continuous variables.

The effect of dapagliflozin compared to placebo on each outcome across UA tertile was examined using Cox regression stratified by diabetes status and adjusted for previous heart failure hospitalization (except for all‐cause death). Likelihood ratio tests are reported to examine for any interaction between UA category and treatment effect.

The treatment effect of dapagliflozin on UA was assessed using a linear regression model adjusted for baseline value and diabetes status. This was repeated for subgroups of interest. The efficacy of dapagliflozin compared with placebo on the primary endpoint over serum UA as a continuous variable was modelled as a fractional polynomial.

All analyses were conducted using Stata version 16.1 (College Station, TX, USA). A p‐value <0.05 was considered statistically significant.

Results

Of the 4744 randomized patients, 3119 (65.7%) had UA measured at baseline (not all countries participated in the biomarker sub‐study). The mean UA concentration was 6.1 ± 1.7 (median 5.9, IQR 4.9–7.1) mg/dl. Mean UA was 6.2 ± 1.7 mg/dl in men and 5.7 ± 1.6 mg/dl in women (p < 0.001). Overall, 686 participants (14.5%) were prescribed a UA‐lowering agent at baseline and, of these, 531 had UA measured; the mean UA level in these patients was 5.8 ± 1.6 (median 5.7, IQR 4.7–6.7) versus 6.2 ± 1.7 (median 6.0, IQR 5.0–7.1) mg/dl in those not receiving UA‐lowering therapy (p < 0.001). The prevalence of hyperuricaemia (UA >7.0 mg/dl for men and >6.0 mg/dl for women) was 31.6% (29.4% men and 39.2% women). The prevalence of hyperuricaemia was lower in those taking UA‐lowering therapies: 23.9% (32.4% in women and 22.7% in men) vs 33.1% (40.0% in women and 31.0% in men), respectively. Overall, 188 patients (6.0%) had a UA ≥9 mg/dl and 488 participants (10.3%) had a history of gout.

Baseline characteristics according to uric acid

Patient characteristics according to tertile of UA are shown in Table  . Patients with higher UA were younger (66.3 ± 10.8 years vs. 68.0 ± 10.2 years) and more often male. There was a higher proportion of Asian and Black participants among those with higher, compared to lower, UA. In general, patients with higher UA had an overall profile suggesting more advanced HFrEF: Kansas City Cardiomyopathy Questionnaire total symptom score (KCCQ‐TSS) was lower (worse); a history of prior heart failure hospitalization was more common; NT‐proBNP and heart rate were higher; and LVEF, eGFR and systolic blood pressure were lower. Patients with higher UA more commonly suffered from atrial fibrillation and type 2 diabetes mellitus and had a higher body mass index. Loop diuretic treatment was more frequent in patients with higher UA (91.2% vs. 71.5%) (Table  ). Patients with a higher UA were also treated with a mineralocorticoid receptor antagonist (MRA) and digoxin more often than those with a lower UA.

Table 1

Baseline characteristics according to uric acid tertile and overall

	Uric acid			p‐value	All patients (n = 3119)
	Tertile 1 (n = 1086)	Tertile 2 (n = 1052)	Tertile 3 (n = 981)
Uric acid, mg/dl	4.4 ± 0.7	6.0 ± 0.4	8.1 ± 1.2	<0.001	6.1 ± 1.7
Age, years	68.0 ± 10.2	67.5 ± 10.3	66.3 ± 10.8	<0.001	67.3 ± 10.4
Sex, n (%)				<0.001
Female	310 (28.5)	217 (20.6)	166 (16.9)		693 (22.2)
Male	776 (71.5)	835 (79.4)	815 (83.1)		2426 (77.8)
Race, n (%)				0.033
Asian	187 (17.2)	205 (19.5)	184 (18.8)		576 (18.5)
Black	20 (1.8)	26 (2.5)	41 (4.2)		87 (2.8)
Other	5 (0.5)	3 (0.3)	4 (0.4)		12 (0.4)
White	874 (80.5)	818 (77.8)	752 (76.7)		2444 (78.4)
Region, n (%)				0.34
Asia/Pacific	180 (16.6)	198 (18.8)	183 (18.7)		561 (18.0)
Europe	619 (57.0)	595 (56.6)	558 (56.9)		1772 (56.8)
North America	181 (16.7)	172 (16.3)	171 (17.4)		524 (16.8)
South America	106 (9.8)	87 (8.3)	69 (7.0)		262 (8.4)
NYHA functional class, n (%)				0.16
II	775 (71.4)	731 (69.5)	652 (66.5)		2158 (69.2)
III	309 (28.5)	317 (30.1)	325 (33.1)		951 (30.5)
IV	2 (0.2)	4 (0.4)	4 (0.4)		10 (0.3)
Heart rate, bpm	69.9 ± 10.8	71.1 ± 11.3	71.6 ± 11.5	<0.001	70.8 ± 11.2
Systolic blood pressure, mmHg	123.6 ± 15.4	122.4 ± 16.0	121.2 ± 15.9	<0.001	122.4 ± 15.8
Left ventricular ejection fraction, %	31.9 ± 6.2	31.2 ± 6.7	30.4 ± 7.3	<0.001	31.2 ± 6.8
NT‐proBNP, pg/ml, median (IQR)	1283.9 (781.2–2252.2)	1369.6 (839.6–2479.8)	1747.5 (987.0–3073.9)	<0.001	1421.3 (852.0–2564.5)
KCCQ‐TSS, median (IQR)	79.2 (60.4–93.8)	79.2 (61.5–93.8)	76.0 (58.3–89.6)	<0.001	78.1 (60.4–91.7)
Body mass index, kg/m2	27.6 ± 5.6	28.7 ± 5.9	29.4 ± 6.1	<0.001	28.5 ± 5.9
Ischaemic aetiology of HF, n (%)	652 (60.0)	627 (59.6)	571 (58.2)	0.008	1850 (59.3)
Prior HF hospitalization, n (%)	468 (43.1)	470 (44.7)	483 (49.2)	0.015	1421 (45.6)
Atrial fibrillation, n (%)	388 (35.7)	452 (43.0)	438 (44.6)	<0.001	1278 (41.0)
Type 2 diabetes mellitus, n (%)	411 (37.8)	436 (41.4)	455 (46.4)	<0.001	1302 (41.7)
eGFR, ml/min/1.73 m2	70.6 ± 18.7	65.6 ± 18.0	58.2 ± 17.4	<0.001	65.0 ± 18.8
eGFR, mL/min/1.73 m2, n (%)				<0.001
<60	301 (27.8)	412 (39.2)	558 (56.9)		1271 (40.8)
≥60	783 (72.2)	640 (60.8)	423 (43.1)		1846 (59.2)
Implantable cardioverter defibrillator, n (%)	342 (31.5)	340 (32.3)	299 (30.5)	0.13	981 (31.5)
Cardiac resynchronization therapy, n (%)	103 (9.5)	78 (7.4)	81 (8.3)	0.0002	262 (8.4)
Medical therapy, n (%)
Loop diuretic	776 (71.5)	873(83.0)	895(91.2)	<0.001	2544 (81.6)
Thiazide diuretic	90(8.3)	88(8.4)	105(10.5)	<0.001	283(9.1)
Other diuretic (non‐MRA)	12(1.1)	16(1.5)	9(0.9)	0.06	37(1.2)
ACE inhibitor	608 (56.0)	613 (58.3)	550 (56.1)	0.03	1771 (56.8)
ARB	295 (27.2)	261 (24.8)	239 (24.4)	<0.001	795 (25.5)
Sacubitril/valsartan	132 (12.2)	134 (12.7)	117 (11.9)	0.02	383 (12.3)
Beta‐blocker	1038 (95.6)	1008 (95.8)	942 (96.0)	0.36	2988 (95.8)
MRA	746 (68.7)	744 (70.7)	731 (74.5)	<0.001	2221 (71.2)
Digoxin	127 (11.7)	149 (14.2)	204 (20.8)	<0.001	480 (15.4)

Data are means ± standard deviation, unless otherwise indicated. Percentages may not total 100 due to rounding. To convert NT‐proBNP from pg/ml to ng/L multiply by 1.

ACE, angiotensin‐converting enzyme; ARB, angiotensin receptor blocker; eGFR, estimated glomerular filtration rate; HF heart failure; IQR interquartile range; KCCQ‐TSS, Kansas City Cardiomyopathy Questionnaire total symptom score; MRA, mineralocorticoid receptor antagonist; NT‐proBNP, N‐terminal pro‐B‐type natriuretic peptide; NYHA, New York Heart Association.

Clinical outcomes according to uric acid

The rates of the pre‐specified clinical outcomes according to baseline UA tertile are shown in Table  and online supplementary Figure  and according to UA displayed as a continuous variable in Figure  . The primary composite outcome occurred more frequently in patients with higher UA although, after adjustment for other prognostic variables (including NT‐proBNP, diuretic dose and eGFR), the greater risk was only apparent in those in the highest tertile (6.8–13.7 mg/dl), using tertile 1 as reference. Spline analysis suggested a linear increase in risk above a serum concentration of around 7.09 mg/dl. The unadjusted HR per unit increase in UA above 7 mg/dl was 1.35 to 1.39 for the endpoints of interest and the adjusted HRs ranged from 1.13 to 1.18 (Figure  and online supplementary Table  ). eGFR slopes according to baseline UA tertile are shown in online supplementary Figure  ; eGFR slope did not vary by UA tertile.

Table 2

Risk of various endpoints according to uric acid levels at randomization

	No. events	Crude rate per 100 py	Unadjusted HR (95% CI)	p‐value	Adjusted HR (95% CI) a	p‐value
Cardiovascular death or worsening HF event b
Uric acid tertile
T1: <5.4 mg/dl	150	9.7 (8.2–11.4)	1.00 (reference)		1.00 (reference)
T2: 5.4–6.7 mg/dl	172	11.4 (9.7–13.2)	1.15 (0.93–1.44)	0.2	1.03 (0.82–1.29)	0.83
T3: 6.8–13.7 mg/dl	250	18.9 (16.7–21.4)	1.87 (1.52–2.29)	<0.0001	1.32 (1.06–1.66)	0.01
UA per 1 mg/dl unit increase			1.19 (1.14–1.25)	<0.001	1.08 (1.03–1.14)	0.0017
Cardiovascular death
Uric acid tertile
T1: <5.4 mg/dl	95	5.9 (4.8–7.2)	1.00 (reference)		1.00 (reference)
T2: 5.4–6.7 mg/dl	83	5.2 (4.2–6.4)	0.86 (0.64–1.16)	0.33	0.77 (0.57–1.05)	0.10
T3: 6.8–13.7 mg/dl	146	10.3 (8.7–12.1)	1.69 (1.30–2.19)	<0.001	1.18 (0.89–1.58)	0.25
UA per 1 mg/dl unit increase			1.18 (1.12–1.26)	<0.001	1.06 (0.99–1.14)	0.07
HF hospitalization
Uric acid tertile
T1: <5.4 mg/dl	89	5.8 (4.7–7.1)	1.00 (reference)		1.00 (reference)
T2: 5.4–6.7 mg/dl	119	7.9 (6.6–9.4)	1.34 (1.02–1.77)	0.04	1.17 (0.88–1.55)	0.29
T3: 6.8–13.7 mg/dl	158	12.0 (10.2–14.0)	1.96 (1.51–2.54)	<0.001	1.29 (0.97–1.72)	0.08
UA per 1 mg/dl unit increase			1.20 (1.14–1.27)	<0.001	1.07 (1.00–1.14)	0.04
All‐cause mortality
Uric acid tertile
T1: <5.4 mg/dl	122	7.6 (6.3–9.1)	1.00 (reference)		1.00 (reference)
T2: 5.4–6.7 mg/dl	102	6.4 (5.3–7.7)	0.83 (0.64–1.08)	0.26	0.76 (0.58–0.99)	0.05
T3: 6.8–13.7 mg/dl	166	11.7 (10.0–13.6)	1.5(1.19–1.89)	0.0007	1.09 (0.84–1.42)	0.51
UA per 1 mg/dl unit increase			1.14 (1.08–1.21)	<0.001	1.03 (0.97–1.1)	0.28

CI, confidence interval; HF, heart failure; HR, hazard ratio; py, person‐years; T, tertile.

Stratified by diabetes status and adjusted for the following baseline variables: history of HF hospitalization, treatment group assignment, age, sex, pulse, systolic blood pressure, body mass index, atrial fibrillation, diabetes, aetiology of HF, left ventricular ejection fraction, New York Heart Association functional classification, N‐terminal pro‐B‐type natriuretic peptide (log), estimated glomerular filtration rate, non‐loop diuretic use, loop diuretic use dose, and use of angiotensin‐converting enzyme inhibitor, angiotensin receptor blocker or angiotensin receptor–neprilysin inhibitor.

Worsening HF event includes unplanned HF hospitalization or urgent visit for worsening HF requiring intravenous diuretic therapy.

Figure 1

Association between baseline uric acid and the risk of the primary composite outcome (worsening heart failure event or cardiovascular death), cardiovascular death, heart failure hospitalization and all‐cause mortality (restricted cubic spline analysis). Model adjusted for randomized treatment (dapagliflozin), age, sex, pulse, systolic blood pressure, body mass index, history of heart failure hospitalization, atrial fibrillation, diabetes status, aetiology of heart failure, left ventricular ejection fraction, New York Heart Association functional classification, N‐terminal pro‐B‐type natriuretic peptide (log), estimated glomerular filtration rate, non‐loop diuretic use, loop diuretic use dose, and use of an angiotensin‐converting enzyme inhibitor, angiotensin receptor blocker or angiotensin receptor–neprilysin inhibitor.

Effect of dapagliflozin on outcomes according to uric acid

The benefit of dapagliflozin compared with placebo was consistent for all pre‐specified outcomes across the range of UA, whether UA was examined as a categorical (tertile) or continuous variable (Table  and Figure  ).

Table 3

Effect of randomized treatment on outcomes according to uric acid tertile

	Tertile 1 (n = 1086)		Tertile 2 (n = 1052)		Tertile 3 (n = 981)		p for interaction
	Dapagliflozin (n = 546)	Placebo (n = 540)	Dapagliflozin (n = 545)	Placebo (n = 507)	Dapagliflozin (n = 491)	Placebo (n = 490)
Cardiovascular death or worsening HF event
n (%)	65 (11.9)	85 (15.7)	90 (16.5)	82 (16.1)	105 (21.4)	145 (29.6)
Rate (95% CI)	8.2 (6.4–10.5)	11.3 (9.1–13.9)	11.3 (9.2–13.9)	11.5 (9.2–14.2)	15.7 (13.0–19.0)	22.3 (18.9–26.2)
Hazard ratio a (95% CI)	0.71(0.52–0.99)		1.00 (0.74–1.35)		0.70 (0.55–0.90)		0.16
Cardiovascular death
No (%)	39 (7.14)	56 (10.37)	41 (7.52)	42 (8.28)	71 (14.46)	75 (15.31)
Rate (95% CI)	4.8 (3.5–6.5)	7.1 (5.5–9.2)	4.9 (3.6–6.7)	5.5 (4.1–7.5)	10.1 (8.0–12.8)	10.4 (8.3,13.1)
Hazard ratio a (95% CI)	0.65 (0.43–0.98)		0.90 (0.58–1.38)		0.96 (0.69–1.33)		0.32
HF hospitalization
No (%)	42 (7.7)	47 (8.7)	60 (11.0)	59 (11.6)	60 (12.2)	98 (20.0)
Rate (95% CI)	5.3 (3.9–7.2)	6.2 (4.7–8.3)	7.5 (5.8–9.7)	8.2 (6.4–10.6)	9.0 (6.9–11.6)	15.0 (12.3–18.3)
Hazard ratio a (95% CI)	0.84 (0.56–1.28)		0.92 (0.64–1.32)		0.59 (0.43–0.82)		0.16
All‐cause mortality
No (%)	49 (9.0)	73 (13.5)	51 (9.4)	51 (10.1)	82 (16.7)	84 (17.1)
Rate (95% CI)	6.0 (4.5–7.9)	9.3 (7.4–11.6)	6.1 (4.6–8.0)	6.7 (5.1–8.8)	11.7 (9.4–14.5)	11.7 (9.4–14.5)
Hazard ratio a (95% CI)	0.63 (0.44–0.90)		0.91 (0.62–1.35)		1.00 (0.73–1.35)		0.14

CI, confidence interval; HF, heart failure.

Hazard ratio for treatment adjusted for history of HF hospitalization (apart from all‐cause death) and stratified by diabetes status.

Figure 2

Effect of dapagliflozin compared with placebo on the primary composite outcome, cardiovascular death, heart failure hospitalization and all‐cause mortality according to baseline uric acid level (fractional polynomial analysis), displayed as a hazard ratio in the solid black line with the shaded area representing 95% confidence intervals. The interrupted red line represents the line of null effect (i.e. hazard ratio of 1.00), below which the estimates favour dapagliflozin. Treatment effect for each outcome (except all‐cause mortality) adjusted for history of heart failure hospitalization and stratified by diabetes status.

Effect of dapagliflozin on uric acid level

At 52 weeks after randomization, the placebo‐corrected reduction in UA from baseline was 0.84 mg/dl (95% CI −0.93 to −0.74; p < 0.001) (Figure  ). The reduction according to baseline UA tertile was: T1 (UA <5.4 mg/dl) −0.75 (−0.89 to −0.61) mg/dl; T2 (UA 5.4–6.7 mg/dl); −0.83 (−0.98 to −0.67) mg/dl; T3 (UA >5.8 mg/dl) −0.94 (−1.14 to −0.74) mg/dl (all changes p < 0.001; no interaction between effect of dapagliflozin and UA tertile). A ‘waterfall plot’ of change in UA is shown in online supplementary Figure  .

Figure 3

(A) Mean (and standard error of the mean, SE) uric acid (UA) by randomized treatment at 0 and 52 weeks. (B) Effect of treatment (dapagliflozin vs. placebo) on UA across subgroups of interest (figure panel shows change in UA from baseline). ARNI, angiotensin receptor–neprilysin inhibitor; CKD, chronic kidney disease; Hba1c, glycated haemoglobin; MRA, mineralocorticoid receptor antagonist. High/low Hba1c defined as levels above/below median value of 6.1%. Dispersion lines represent 95% confidence intervals.

The reduction in UA was consistent in most subgroups of interest, including patients treated with an angiotensin receptor blocker or sacubitril/valsartan at baseline and patients treated with other uricosuric drugs and drugs inhibiting UA production (Figure  ). However, there was evidence of a greater reduction in UA in patients without diabetes and lower glycated haemoglobin (HbA1c).

A total of 2500 patients had UA levels checked at 12‐month follow‐up. The proportion of those achieving a level of <6.0 mg/dl at 12 months was 72.7% (n = 930/1278) in the dapagliflozin group and 51.2% in the placebo group (n = 626/1222) (p < 0.001).

Use of uric acid‐lowering agents before and after randomization

At baseline, 664 patients (14.0%) were taking a treatment inhibiting UA production (allopurinol, febuxostat, or topiroxostat) and 24 participants (0.5%) were treated with a drug increasing UA excretion (benzbromarone, probenecid, or sulfinpyrazone). A UA‐lowering agent was initiated in 104 (4.4%) patients after randomization in the placebo group, as compared to 51 (2.1%) among those assigned to dapagliflozin (between‐group p <0.001). The number of patients who had a serious adverse event related to gout during follow‐up was 4/2368 (0.17%) in the placebo group versus 2/2368 (0.08%) among those receiving dapagliflozin (non‐serious adverse events related to gout were not collected).

Use of other drugs after randomization

No patient in either group was started on colchicine after randomization. During follow‐up, patients in the dapagliflozin group were less likely to have an increase in diuretic dose (odds ratio 0.74, 95% CI 0.57–0.96) and more likely to have a reduction in diuretic dose (odds ratio 1.6, 95% CI 1.21–2.11).

Discussion

Many patients in DAPA‐HF had an elevated UA and higher UA was associated with a greater risk of the primary outcome of worsening heart failure or cardiovascular death in this contemporary HFrEF cohort receiving excellent conventional therapy. Spline analysis indicated a linear risk above a UA concentration of approximately 7.09 mg/dl and the risk of the primary endpoint increased by 9% for each 1 mg/dl increase in UA (cardiovascular death increased by 6% for each 1 mg/dl). The elevation of risk related to UA remained even after adjustment for other prognostic variables including natriuretic peptides. The benefits of dapagliflozin on the trial primary and secondary outcomes were consistent across the range of UA concentrations at baseline. Dapagliflozin lowered UA and conventional UA‐lowering agents were initiated significantly less frequently in patients assigned to dapagliflozin compared with placebo (Graphical Abstract).

The finding that dapagliflozin lowered UA in patients with chronic HFrEF treated with contemporary medications is important because hyperuricaemia is common in this population. Using recommended sex‐specific cut‐offs of 6.0 mg/dl (∼360 µmol/L) in females and 7.0 mg/dl (∼420 µmol/L) in males, we found that 29.4% of men and 39.2% of women had hyperuricaemia, considerably higher than the 5%–20% prevalence reported in the general population but consistent with other studies in heart failure. , , , ,

The independent prognostic importance of UA continues to be debated. For example, while the GISSI‐HF investigators reported an association between death from cardiovascular causes, death from any cause and hospitalization for heart failure, their multivariate analysis did not include natriuretic peptides. More recently, UA was found to remain predictive of outcomes in PARADIGM‐HF, even after adjustment for NT‐proBNP level and our data support this observation.

Although higher UA is related to worse outcomes, the explanation for this association is not clear and a cause‐and‐effect relationship has not been established. Specifically, several randomized controlled trials have failed to demonstrate a benefit of the non‐selective xanthine oxidase inhibitor, allopurinol and its metabolite oxypurinol, in patients with heart failure, although none of these was a large mortality/morbidity trial. , , , , , There are no completed trials with the novel xanthine oxidase inhibitors, febuxostat and topiroxostat, in patients with heart failure. ,

While not shown to improve heart failure outcomes, lowering UA is still needed in some patients, primarily to reduce the risk of gout or recurrence of gout. Overall, 10.3% of our patients had a history of gout, 14.5% were prescribed a UA‐lowering treatment at baseline, and 6.0% had a UA ≥9 mg/dl, the suggested threshold for initiating prophylactic UA‐lowering therapy. Intolerance of conventional UA‐lowering treatments is common, particularly in some Asian ethnic groups, and serious adverse effects may occur, including hypersensitivity reactions (e.g. Stevens–Johnson syndrome with allopurinol). Drug interactions are also common (including with furosemide and angiotensin‐converting enzyme inhibitors). , Consequently, avoidance of the use of these drugs is preferable. Therefore the finding that dapagliflozin reduced UA concentration is potentially clinically relevant. Of interest, the reduction in UA at 52 weeks in DAPA‐HF was 0.84 mg/dl (95% CI −0.93 to −0.74), which was more than twice the mean placebo‐corrected reduction in serum UA of 0.37 (95% CI 0.42−0.31) mg/dl with empagliflozin at 52 weeks in EMPA‐REG OUTCOME. Although it is difficult to compare across trials and in different medical conditions, a recent systematic review and network meta‐analysis has suggested there may be differences in the size of UA reduction with different SGLT2 inhibitors, at least in people with type 2 diabetes. Although the average reduction in UA was still modest in DAPA‐HF, it did result in more patients achieving an ideal UA level at 12 months (<6.0 mg/dl) compared with placebo (72.7% vs. 51.2%) and the rate of initiation of a new UA‐lowering agent was halved over the median follow‐up of 18.2 months (2.1% vs. 4.4%; p <0.001). We did not have data on gouty flares, although a reduction in these was demonstrated in the CANVAS trials (from 2.6 patients with an event per 1000 person‐years in the placebo group to 2.0 per 1000 person‐years in the canagliflozin group; HR 0.64, 95% CI 0.41–0.99; p = 0.046).

Although the UA‐lowering action of dapagliflozin was both statistically and clinically significant, the size of the reduction was modest and around a half to a third of that observed with xanthine oxidase inhibitors, albeit in patients with higher baseline UA concentrations. Importantly, however, the UA‐lowering action of dapagliflozin was similar in patients treated and not treated with a conventional UA‐lowering agent that is, appears to be mechanistically distinct.

Sodium–glucose cotransporter inhibitors are thought to reduce UA by increasing the rate of urinary UA excretion. Non‐reabsorbed glucose is thought to compete for the facilitated glucose transporter member 9 isoform 2 in the proximal renal tubule, a major regulator of urate homeostasis. This may explain why there was an interaction between baseline diabetes status (and HbA1c level) and the UA‐lowering efficacy of dapagliflozin, whereby the reduction in UA was greater in patients without type 2 diabetes (and in those with a lower HbA1c).

This ancillary UA‐lowering property of SGLT2 inhibitors has also been shown for some other treatments for heart failure, notably losartan and sacubitril/valsartan. , Recently, vericiguat, a stimulator of soluble guanylate cyclase, was also demonstrated to reduce UA with a placebo corrected reduction of 10.0% (95% CI 3%–16%) with a dose of 10 mg compared to the 13.6% (95% CI 12.1%–15.2%) reduction demonstrated in DAPA‐HF.

The benefits of dapagliflozin on clinical outcomes were not modified by baseline UA, with consistent reductions in the primary endpoint, its components, heart failure hospitalization and all‐cause death across the range of UA measured at baseline. The results were consistent whether the outcomes were analysed by UA tertile or using UA as a continuous variable.

The mechanism of the UA‐lowering effect of SGLT2 inhibitors in heart failure is unknown. Broadly, this action could reflect increased excretion or reduced production of UA. We did not measure urinary UA levels and, clearly, that would be of interest. We know of no evidence that SGLT2 inhibitors reduce xanthine oxidase activity and we found that dapagliflozin reduced UA levels as much in patients taking conventional UA‐lowering therapy as in those not. However, it would be of interest to examine the effect of SGLT2 inhibitors on plasma xanthine oxidase activity. There has also been speculation that SGLT2 inhibitors might act directly or indirectly on urate transporters in the kidney tubules, and this could be probed by studies combining an SGLT2 inhibitor with other drugs such as verinurad, which is a specific inhibitor of urate transporter 1 (URAT1).

Our study has several limitations. This analysis was not pre‐specified and retrospective analyses of this type may be subject to residual/unmeasured confounding. Patients were enrolled in a clinical trial with specific entry criteria, including the exclusion of individuals with an eGFR <30 ml/min/1.73 m2. We had only a single repeat measurement of UA at 12 months.

Conclusions

Uric acid was an independent predictor of worse outcomes in DAPA‐HF even after multivariable adjustment including natriuretic peptides. Compared with placebo, dapagliflozin reduced UA and improved outcomes irrespective of UA concentration in patients with HFrEF.

Figure S1. Cumulative incidence of the primary composite endpoint, cardiovascular death, heart failure hospitalization and all‐cause mortality according to baseline uric acid tertile.

Click here for additional data file.

Figure S2. Change in estimated glomerular filtration rate from baseline with slope analysis over days 14–720, by treatment group.

Click here for additional data file.

Figure S3. ‘Waterfall plot’ of change in uric acid between baseline and 52 weeks in each treatment group.

Click here for additional data file.

Table S1. Risk of various endpoints according to uric acid levels at randomization, per 1 mg/dl unit increase above 7 mg/dl.

Click here for additional data file.