Professional Documents
Culture Documents
Circulation
Circulation
Circulation
BACKGROUND: Outcome trials in patients with type 2 diabetes mellitus have Michael E. Nassif, MD
demonstrated reduced hospitalizations for heart failure (HF) with sodium-glucose et al
co-transporter-2 inhibitors. However, few of these patients had HF, and those that
did were not well-characterized. Thus, the effects of sodium-glucose co-transporter-2
inhibitors in patients with established HF with reduced ejection fraction, including
those with and without type 2 diabetes mellitus, remain unknown.
pro b-type natriuretic peptide) and (2) proportion of patients with ≥5-point increase
in HF disease-specific health status on the Kansas City Cardiomyopathy Questionnaire
overall summary score, or a ≥20% decrease in NT-proBNP.
CONCLUSIONS: In patients with heart failure and reduced ejection fraction, use
of dapagliflozin over 12 weeks did not affect mean NT-proBNP but increased the
proportion of patients experiencing clinically meaningful improvements in HF-related
health status or natriuretic peptides. Benefits of dapagliflozin on clinically meaningful
Key Words: biomarkers ◼ health
HF measures appear to extend to patients without type 2 diabetes mellitus. status ◼ heart failure ◼ outcomes
◼ SGLT2 inhibitors
CLINICAL TRIAL REGISTRATION: URL: https://www.clinicaltrials.gov. Unique
identifier: NCT 02653482. Sources of Funding, see page XXX
https://www.ahajournals.org/journal/circ
H
eart failure (HF) remains a leading cause of hos- patients, both with and without T2D.
pitalization in the United States and Western Eu-
rope.1,2 Patients who have HF with reduced ejec- METHODS
tion fraction (HFrEF) have a high burden of debilitating The authors declare that all supporting data are available
symptoms, and are at especially high risk of disease pro- within the article (and its online-only Data Supplement).
gression, including frequent hospitalizations, require- DEFINE-HF was a randomized, double-blind, placebo con-
ment for advanced therapies, and death.3 Despite ther- trolled, multi-center trial that enrolled patients with HFrEF,
apeutic advances developed over the past 2 decades,4 defined as an established diagnosis of HF for at least 16
outcomes for patients with HFrEF remain poor, and weeks, left ventricular ejection fraction ≤40%, NYHA class
II-III symptoms, and an elevated NT-proBNP (N-terminal pro
many interventions demonstrating benefit on hard out-
b-type natriuretic peptide) or BNP (B-type natriuretic peptide).
comes (such as mortality) may not improve health sta- Patients were randomized to treatment with dapagliflozin
tus (symptoms, function and quality of life).5 Improving 10 mg daily or matching placebo in addition to guideline
symptom burden in HFrEF is a key goal of management, directed standard of care therapy for 12 weeks (Appendix SI
highly predictive of lower risk of future HF hospitaliza- in the online-only Data Supplement). The trial was conducted
tion, and increasingly recognized by practice guidelines across 26 sites in the United States (US). The list of participat-
and regulators as an important therapeutic target.6,7 ing sites and investigators can be found in Appendix SII in
Notably, the recent Food and Drug Administration draft the online-only Data Supplement. The primary objectives of
guidance to industry on drug development in HF has the trial were to evaluate the effects of dapagliflozin on HF
disease-specific biomarkers (NT-proBNP) and health status, as
endorsed measures of patients’ symptoms and physical
assessed by the Kansas City Cardiomyopathy Questionnaire
function as acceptable (and important) end points.8 (KCCQ).18
Sodium-glucose co-transporter-2 inhibitors (SGLT- DEFINE-HF was an investigator-initiated trial, with the
2i) are a class of medications developed for treatment concept developed, and trial sponsored and executed by
of type 2 diabetes mellitus (T2D), which lower plasma the national coordinating center at Saint Luke’s Mid America
glucose concentrations via increased urinary excretion Heart Institute in collaboration with the academic Executive
Committee, independent of the funding source (Astra responsiveness and interpretability have been independently
ORIGINAL RESEARCH
Zeneca). The study was monitored by an independent data established. Scores are transformed to a range of 0 to 100, in
safety and monitoring committee. Institutional review boards which higher scores reflect better health status.18
approved the study for all sites, and all patients provided Six-minute walk tests were conducted by site study per-
ARTICLE
informed consent for research participation. The trial was sonnel, based on protocol instructions.
conducted in accordance with the ICH E6(R1) Guidelines of
Good Clinical Practice and the Declaration of Helsinki.
Concomitant Medications
The trial was designed to enroll patients on optimal doses of
Patient Selection guideline-directed medical therapy for HFrEF, though adjust-
Full inclusion and exclusion criteria can be found in Appendix ment of standard therapies, including attempted up-titration
SIII in the online-only Data Supplement. Adult ambulatory of neurohormonal antagonists was permitted if clinically
patients with or without T2D, established HF for at least 16 indicated. In patients with T2D, plans for risk reduction of
weeks, LVEF ≤40%, and NYHA class II-III HF were recruited hypoglycemia included suggested 20% reduction in the total
from 26 sites in the US, and screened for participation. Key daily dose of insulin or insulin secretagogues (ie, sulfonyl-
exclusion criteria were recent hospitalization (within 30 days) urea, metiglinides) for patients with baseline hemoglobin A1c
for decompensated HF, eGFR <30 mL/min/1.73m2 at the (HbA1c) of ≤7%. All participants were counseled on the signs
screening visit (using modified MDRD equation), and history and symptoms of hypoglycemia and the appropriate treat-
of type 1 diabetes mellitus.19 ment and were monitored for recent hypoglycemic events at
each study visit.
Trial Design
Patients considered potentially eligible who agreed to par- Trial End Points
ticipate and provided informed consent entered a 2-week Dual primary end points were (1) the average of 6- and
screening phase, during which their eligibility was confirmed 12-week mean NT-proBNP and (2) a composite of the propor-
based on central core laboratory evaluation (NT-proBNP ≥400 tion of patients that achieved a meaningful improvement in
pg/mL, or BNP ≥100 pg/ml), eGFR ≥30 mL/min/1.73m2, and health status (≥5-point increase in average of 6- and 12-week
clinical stability was ensured. Patients confirmed eligible were KCCQ-OS) or NT-proBNP (≥20% decrease in average of 6- and
randomized 1:1 to oral dapagliflozin 10 mg or matching pla- 12-week NT-proBNP).20,21 Key secondary end points included
cebo once daily (Figure I in the online-only Data Supplement). proportion of patients with meaningful change in KCCQ,
Before administration of the first dose of dapagliflozin or pla- and NT-proBNP at each time point, mean BNP and propor-
cebo, patients underwent a physical exam, trial-related labo- tion of patients with meaningful change in BNP, functional
Downloaded from http://ahajournals.org by on October 7, 2019
ratory assessments, completion of the KCCQ questionnaire status based on 6-minute walk test, change in weight, systolic
and a 6-minute walk test. Patients then entered a 12-week blood pressure and HbA1c. Exploratory end points included
treatment period, during which they were followed via 4 a composite of hospitalization for HF or urgent HF visits (see
phone call visits, as well as 2 in-person study visits (at 6 and Appendix SI in the online-only Data Supplement).
12 weeks), at which time repeat physical exams, laboratory All serious adverse events were reported by investigators.
assessments, KCCQ questionnaires, and 6-minute walk tests A blinded Clinical Events Committee adjudicated all deaths,
were completed. At week 12, study medication was discon- hospitalizations for heart failure, urgent heart failure visits,
tinued, and patients were followed for 1 additional week, at myocardial infraction and transient ischemic attack/stroke
the end of which another in-person study visit was conducted events. In addition, investigator reported adverse events of
to assess for any intercurrent safety events, and to collect special interest included acute kidney injury (defined as dou-
additional laboratory samples. bling of serum creatinine), diabetic ketoacidosis, volume
depletion, severe hypoglycemia (defined as blood glucose
<54 mg/dL (<3.0 mmol/L) and requiring assistance), and
Study Measurements lower limb amputations.
Levels of NT-proBNP and BNP, and all other study laboratory
assessments were analyzed at a Quest Diagnostics central
laboratory, blinded to treatment assignment (NT-ProBNP assay Statistical Analysis
Roche electrochemiluminescent method on Elecsys© plat- Patient disposition is reported including all patients who
form, Pro-BNP II reagent by Roche/Cobas; BNP assay chemi- signed the informed consent. All primary and secondary effi-
luminescent method on Siemens ADVIA Centaur® platform). cacy end points were evaluated using the modified intention
The KCCQ questionnaire was completed at each time to treat data set, which included all randomized patients who
point by patients without assistance by site study staff (as received at least 1 dose of study medication and had at least
validated). The KCCQ is a 23-item, self-administered instru- 1 evaluable end point. Patients with no evaluable follow-up
ment that quantifies physical function, symptoms (frequency, data for a particular outcome (eg, NT-proBNP) were excluded
severity and recent change), quality of life, and social func- from these analyses. The on-treatment data set includes all
tion. In the KCCQ, an overall summary score (KCCQ-OS) can patients in the modified intention to treat data set; however,
be derived from the physical function, symptoms, quality efficacy end point measurements in the on-treatment data
of life and social function domains. KCCQ clinical summary set were excluded for follow-up time point(s) during which
score (KCCQ-CS) includes the physical function and symp- patients were off the study medication at the time the cor-
toms domains. For each domain, the validity, reproducibility, responding measurements were obtained. The on-treatment
data set was used for sensitivity analyses. The safety analysis Prespecified supportive analyses were also performed that
ORIGINAL RESEARCH
set included all patients who received at least 1 dose of study included KCCQ-CS, which specifically evaluates patients’
medication and was used for all safety analyses. symptom burden and physical function, instead of KCCQ-OS,
Baseline characteristics were evaluated overall and by within a similar composite end point of clinically meaningful
ARTICLE
treatment group. Continuous measures were summarized change in NT-proBNP and KCCQ over 12 weeks.
by mean±standard deviation and compared using Students For secondary end points, binary outcomes were ana-
t tests. Categorical variables were summarized by frequency lyzed in a manner analogous to that of the second dual
and percent and compared using Chi square or Fisher’s exact primary outcome, and continuous outcomes were analyzed
tests, as appropriate. in a manner analogous to that of the first dual primary out-
For the first dual primary end point, a generalized linear come, although appropriate distributions and link functions
mixed model was used to estimate the effect of treatment were chosen for the given outcomes. Models also provided
on the average of 6- and 12-week mean NT-proBNP values, separate effect estimates for the 6 and 12-week time points.
adjusting for log baseline NT-proBNP, history of T2D, eGFR, and Additional prespecified supportive analyses assessed the sub-
age. A gamma distribution and log link function were used to domains of KCCQ-OS separately.
account for the skewed nature of NT-proBNP. For the second For the exploratory outcome of time to first adjudicated
dual primary end point that includes both health status mea- HF event (HF hospitalization or urgent HF visit) a Cox propor-
sures and biomarker measurement, the proportion of patients tional hazard model was used to assess the effect of treatment
achieving meaningful change in KCCQ (a ≥5-point increase in versus placebo on the time to first occurrence of adjudicated
average of 6- and 12-week KCCQ-OS) or NT-proBNP (≥20% HF event, adjusting for history of T2D, eGFR, and age. Safety
decrease in average of 6- and 12-week NT-proBNP), were outcomes were assessed using descriptive statistics only, and
calculated by treatment group. A logistic regression model no P values were calculated.
was used to assess the effect of dapagliflozin versus placebo.
Models were adjusted for log baseline NT-proBNP, baseline Sample Size Calculations
KCCQ, history of T2D, eGFR, and age. For the dual primary biomarker end point, a sample size of
For the primary end points, several subgroup analyses 110 for each group was projected to achieve 80% power with
were prespecified, including T2D status, age (<65, ≥65 years), α=0.05 to detect a difference in mean adjusted NT-proBNP
sex (male, female), race, baseline NT-proBNP (< median, ≥ between the 2 groups of at least 302 pg/mL from baseline to
median), baseline LVEF (≤30%, >30%), atrial fibrillation, 12 weeks. The assumptions for this calculation were derived
baseline KCCQ-OS (<70, ≥70), baseline eGFR (<60, ≥60 mL/ from a previous trial where the estimated standard deviation
min/1.73 m2), cause of cardiomyopathy (ischemic, non-isch- for NT-proBNP was 1250 pg/mL.25 For the dual primary health
emic), baseline renin angiotensin aldosterone system inhibi- status and biomarker end point, a sample size of 110 for each
Downloaded from http://ahajournals.org by on October 7, 2019
tor type (angiotensin receptor neprilysin inhibitor (ARNI), group was projected to achieve 80% power with α=0.05 to
angiotensin-converting enzyme inhibitors, angiotensin recep- detect an absolute difference in the proportions between the
tor blockers, neither), baseline loop diuretic dose (furosemide 2 groups of 18% from baseline to 12 weeks. There was an
equivalent mean daily dose: ≤ 40 mg, >40 mg). anticipated non-completion rate of approximately 13%, so
Both primary hypotheses were tested at the 2-sided the final sample size per group was estimated at 125 patients
5% significance level. The statistical analysis plan prespeci- per treatment group.
fied that in the case of non-significant findings for the first
dual primary end point, the second dual primary end point
would be tested at the 2-sided 5% significance level but will
be regarded as an exploratory end point, and a nominal P
RESULTS
value will be produced. No adjustments for multiplicity were Patients and Treatment
made for secondary or exploratory end points. Prespecified
responder analyses were conducted examining proportions
A total of 510 patients were screened, of which 263
of patients with significant worsening, no change, and clini- qualified and were subsequently randomized: 131 to
cally important improvement in NT-proBNP and KCCQ at the dapagliflozin and 132 to placebo (Figure 1). Baseline
end of the treatment period. These analyses used clinically characteristics were not different between the 2 groups
meaningful threshold for NT-proBNP (as well as BNP) of 20% (Table 1). Across the entire cohort, the mean age was
or greater change, based on previous studies demonstrating 61.3 years, 73% were male, and 40% African American.
that change in the excess of 20% demonstrates more than The mean duration of HF was 7.1 years and more than
random or physiologic variation in NT-proBNP, correlates with 85% had been hospitalized for HF at least once before
echocardiographic markers of adverse left ventricular remod- study enrolment. Overall, 62% (166 patients) had T2D,
eling, and is associated with clinical events in patients with and 40% had atrial fibrillation; median body mass index
established HF.22–24 We also used previously established mean-
was 30.6 (interquartile range (IQR), 27.5–36.3). NYHA
ingful thresholds for change in KCCQ (5 to <10 point (small–
moderate), 10 to <20 point (moderate-large), and ≥20 point
class II symptoms were present in 66%, with class III
(very large) change).20,21 symptoms in 34%. The frequency of guideline-directed
Sensitivity analysis was performed, including per protocol optimal medical therapy for HFrEF was high: 97% of
evaluation using the on-treatment data set, in which only the patients were on beta blockers, 61% on mineralocorti-
end points ascertained while patients were receiving dapa- coid antagonists, 59% of patients were on angiotensin-
gliflozin or placebo were included. converting enzyme inhibitors or angiotensin receptor
ORIGINAL RESEARCH
ARTICLE
Figure 1. Trial flow chart.
blockers, 33% ARNI, 62% had implanted cardiac defi- 1.03–3.06, nominal P value=0.039; Figure 2B). The re-
brillators, including 35% with cardiac resynchronization sults were consistent within subgroups of patients with
therapy. The majority (86%) were on a loop diuretic. and without T2D, as well as across all other prespeci-
Mean eGFR was 69 mL/min/1.73m2, and median urine fied subgroups (Figure 2B; all P values for interaction
albumin creatinine ratio was 37 mg/gm. The median were non-significant). The results were also consistent
Downloaded from http://ahajournals.org by on October 7, 2019
NT-proBNP level at randomization was 1136pg/mL (IQR, when KCCQ-CS was analyzed instead of KCCQ-OS
615–2267 pg/mL) and mean LVEF was 26%. (Figure 2B), and in the on-treatment sensitivity analyses
Every randomized patient received at least 1 dose (Table II in the online-only Data Supplement).
of dapagliflozin or placebo. One patient in each of the
dapagliflozin and placebo groups died during the study.
Premature permanent treatment discontinuation oc-
Secondary End Points
curred in 11 patients in the dapagliflozin group, and 12 A numerically greater proportion of patients treated with
in the placebo group (see Table I in the online-only Data dapagliflozin versus placebo had a clinically meaningful
Supplement which outlines detailed reasons for dis- ≥5 point improvement in KCCQ-OS at 6 weeks (47% vs
continuation), resulting in 119 patients in each group 36%), and 12 weeks (43% vs 33%); with adjusted odds
completing the trial on treatment. Safety follow-up was ratios favoring dapagliflozin both at 6 and 12 weeks
competed in all patients; no patients withdrew consent (OR 1.8, 95% CI: 1.04–3.12, P=0.03; OR 1.7, 95% CI
or were lost to follow-up. 0.98–3.1, P=0.06, respectively). Results were similar
when using KCCQ-CS, with improvements in physical
limitations and symptoms diverging over time (KCCQ-
Primary End Points CS improvement of ≥5 points with dapagliflozin versus
In the intent-to-treat analysis, there was no between- placebo: 6 weeks, 44% vs 38%; 12 weeks: 47% vs
group difference in the biomarker dual primary out- 32%), with adjusted odds ratios favoring dapagliflozin
come, the average 6 and 12 week adjusted mean NT- both at 6 and 12 weeks (OR 1.5, 95% CI 0.86–2.70,
proBNP between patients treated with dapagliflozin P=0.14; OR 2.4, 95% CI 1.31–4.2, P<0.01, respective-
versus placebo (1133 pg/dL (95% CI, 1036–1238) vs ly). In the supportive analyses, mean KCCQ-OS was 3.7
1191 pg/dL (95% CI 1089–1304), adjusted ratio 0.95; points higher (P=0.037), and mean KCCQ-CS was 4.6
95% CI 0.84–1.08, P=0.43; Figure 2A). For the second points higher (P=0.007) at 12 weeks in patients treated
dual primary outcome, a greater proportion of patients with dapagliflozin versus placebo. Three of 4 domains of
treated with dapagliflozin had a clinically meaningful KCCQ improved significantly with dapagliflozin versus
improvement of ≥5 points in KCCQ-OS or at least a placebo at 12 weeks (total symptoms, physical limita-
20% reduction in NT-proBNP, as compared with pla- tions, quality of life), with the benefits increasing over
cebo (61.5% vs 50.4%, adjusted OR 1.8, 95% CI time (Figure 3). The only domain that did not show a sig-
Values are shown as absolute numbers (percentages) and mean±SD. ACE-I indicates angiotensin converting enzyme-inhibitor; ARB, angiotensin II receptor blocker; ARNI,
angiotensin receptor-neprilysin inhibitor; BNP, B-type natriuretic peptide; CRT, creatinine; CV, cardiovascular; DPP, Dipetidyl Pepdidase; eGFR, estimated glomerular filtration
rate; GLP-1RA, glucagon-like peptide-1 receptor agonist; HF, heart failure; ICD, internal cardiac defibrillator; KCCQ-OS, Kansas City Cardiomyopathy Questionnaire–Overall
Summary Score; MRA, mineralocorticoid receptor antagonist; NYHA, New York Heart Association; and NT-proBNP, N-terminal pro B-type natriuretic peptide.
*Blood pressure and heart rate measured from noninvasive cuff measurements for patients in sinus rhythm and manual pulse and blood pressure for patients in atrial fibrillation.
ORIGINAL RESEARCH
ARTICLE
Figure 2. Effects of dapagliflozin versus
placebo on the primary end points.
A, Effects of dapagliflozin vs placebo on
NT-proBNP (N-terminal pro b-type natriuretic
peptide) levels over 12 weeks. Adjusted mean
NT-proBNP values with whiskers equal to 1.5×
interquartile range. B, Proportion of patients
achieving clinically meaningful improvement in
heart failure disease-specific health status or
natriuretic peptides (overall, and within prespec-
ified subgroups).
Downloaded from http://ahajournals.org by on October 7, 2019
nificant difference between dapagliflozin and placebo Although there was no difference between dapagliflozin
was the social limitations score. In the responder analy- and placebo in the proportion of patients with 20% reduc-
sis, fewer patients treated with dapagliflozin versus tion in NT-proBNP at 6 weeks (34.4 vs 33.3%; adjusted OR
placebo had deterioration, or experienced no change, 1.1, 95% CI: 0.6–1.9, P=0.74), a greater proportion of pa-
whereas greater proportion of patients treated with tients treated with dapagliflozin experienced ≥20% reduc-
dapagliflozin experienced small-moderate, moderate- tion in NT-proBNP at 12 weeks (44.0 vs 29.4%; adjusted
large, and very large improvements in KCCQ-OS and OR 1.9, 95% CI: 1.1–3.3, P=0.02; Table 2). Patients treated
KCCQ-CS (Figure 4A and 4B). with dapagliflozin had greater reduction in average of 6
Figure 3. Effects of dapagliflozin versus placebo on KCCQ (overall, and within specific domains).
A, Kansas City Cardiomyopathy Questionnaire (KCCQ) overall summary score (KCCQ-OS) over 12 weeks. B, KCCQ clinical summary score (KCCQ-CS) over 12
weeks. C, KCCQ total symptom score over 12 weeks. D, KCCQ quality of life score over 12 weeks, E, KCCQ physical limitation score over 12 weeks, F, KCCQ
social limitation score over 12 weeks.
and 12 week adjusted mean BNP versus placebo (232 pg/ walk distance at week 12: 304 meters vs 301 meters;
dL (95% CI: 210–256) vs 268 pg/dL (95% CI: 243–296), P=0.79). There were no significant between-group dif-
P=0.04; Figure II in the online-only Data Supplement). As ferences in weight, hemoglobin A1c, or systolic blood
with NT-proBNP, while there was no difference between pressure, although most values were numerically lower
dapagliflozin and placebo in the proportion of patients with dapagliflozin versus placebo (Table 3). Virtually all
with 20% decline in BNP at 6 weeks (43.5% vs 35.4%; results were consistent within subgroups of patients
adjusted OR 1.5, 95% CI: 0.9–2.7, P=0.15), a greater pro- with and without T2D, and across all other subgroups
portion of patients treated with dapagliflozin experienced (data not shown) for all secondary end points, with all
≥20% reduction in BNP at 12 weeks (50.0% vs 33.9%; P values for interaction being non-significant, except
adjusted OR 2.0, 95% CI: 1.1–3.4, P=0.01; Table 2). In for weight. Patients without T2D experienced a mean
the responder analysis, fewer patients treated with dapa- weight reduction of -1.33 kg (95% CI -2.41 to -0.23
gliflozin versus placebo had moderate-large increase or no kg, P=0.018) with dapagliflozin versus placebo; while
change, whereas a greater proportion of patients treated patients with T2D had a mean change in weight of 0.24
with dapagliflozin experienced moderate-large reduction kg (95% CI -0.76 to 1.25 kg, P=0.64) with dapagliflozin
in NT-proBNP and BNP (Figure 4C and 4D). versus placebo (P for interaction =0.08). The per-proto-
There was no difference in 6-minute walk distance col sensitivity analyses demonstrated consistent results
between dapagliflozin and placebo (adjusted mean (Table III in the online-only Data Supplement).
ORIGINAL RESEARCH
ARTICLE
Figure 4. Responder analyses of clinically meaningful changes in KCCQ and natriuretic peptides.
A, Clinically meaningful changes in Kansas City Cardiomyopathy Questionnaire overall summary score (KCCQ-OS) with dapagliflozin vs placebo at 12 weeks. B,
Clinically meaningful changes in KCCQ clinical summary score (KCCQ-CS) with dapagliflozin vs placebo at 12 weeks. C, Clinically meaningful changes in NT-proB-
NP (N-terminal pro b-type natriuretic peptide) levels with dapagliflozin vs placebo at 12 weeks. D, Clinically meaningful changes in BNP levels with dapagliflozin vs
placebo at 12 weeks.
Downloaded from http://ahajournals.org by on October 7, 2019
Exploratory Clinical End Points injury in the dapagliflozin group and 1 (<1%) in the
placebo group. The number of severe hypoglycemic
In total, there were 25 patients with adjudicated HF
events was 1 (<1%) in the dapagliflozin group versus 1
events (18 hospitalizations for HF, and 7 urgent HF vis-
(<1%) with placebo; both events occurred in patients
its) during the study period.
with T2D. There were no episodes of diabetic ketoaci-
In the time to first event analysis, there was no sig-
dosis or amputations.
nificant difference in adjudicated HF events with dapa-
gliflozin versus placebo, including hospitalization for
HF or urgent HF visits (HR 0.84, 95% CI: 0.35–1.97,
P=0.62; Table IV in the online-only Data Supplement).
DISCUSSION
In this investigator-initiated, multi-center, double blind,
placebo-controlled trial in patients with HFrEF, with or
Safety Outcomes without T2D, treatment with dapagliflozin did not sig-
Serious adverse events and adverse events of special nificantly reduce the adjusted mean NT-proBNP levels
interest observed with dapagliflozin and placebo are over 12 weeks as compared with placebo. However,
reported in Table 4. In total, there were 85 reported a greater proportion of patients treated with dapa-
adverse events with dapagliflozin versus 82 with pla- gliflozin versus placebo experienced a clinically mean-
cebo. Adverse events leading to treatment discontinu- ingful improvement in health status (KCCQ-OS) or a
ation occurred in 11 patients with dapagliflozin and clinically meaningful reduction in NT-proBNP over the
12 patients with placebo. There was 1 death in the treatment period, which was the second dual primary
dapagliflozin group (<1%) caused by worsening HF, outcome. Moreover, in secondary analyses directly as-
and 1 death in the placebo group (<1%) caused by sessing the impact of treatment on patient reported
sudden cardiac death. There were 2 transient ischemic outcomes, dapagliflozin had favorable effects on health
attacks, no myocardial infarctions, and no strokes in status at the end of the study, especially on the KCCQ
the dapagliflozin group and no transient ischemic at- clinical summary score (symptoms and physical func-
tacks, 1 stroke, and 4 myocardial infarctions in the pla- tion). The proportions of patients with ≥20% decrease
cebo group. There was 1 (<1%) event of acute kidney in NT-proBNP and BNP were also higher, and mean BNP
Table 2. Binary Secondary End Points at 6 and 12 Weeks for Patients Treated With Dapagliflozin Versus Placebo
ORIGINAL RESEARCH
Adjusted Odds
Binary Secondary End Points Dapagliflozin Placebo Ratio* 95% CI P Value
ARTICLE
KCCQ OS increase ≥5 points (6w) 59 (46.8%) 46 (35.9%) 1.83 1.04, 3.21 0.03
KCCQ OS increase ≥5 points (12w) 54 (42.9%) 41 (32.5%) 1.73 0.98, 3.05 0.06
KCCQ CS increase ≥5 points (6w) 56 (44.4%) 49 (38.3%) 1.53 0.86, 2.72 0.15
KCCQ CS increase ≥5 points (12w) 59 (46.8%) 40 (31.7%) 2.35 1.31, 4.22 <0.01
NT-proBNP decrease ≥20% (6w) 43 (34.4%) 42 (33.3%) 1.10 0.63, 1.92 0.74
NT-proBNP decrease ≥20% (12w) 55 (44.0%) 37 (29.4%) 1.90 1.09, 3.31 0.02
BNP decrease ≥20% (6w) 54 (43.5%) 45 (34.5%) 1.52 0.86, 2.67 0.15
BNP decrease ≥20% (12w) 62 (50.0%) 42 (33.9%) 1.99 1.15, 3.45 0.01
Values are shown as absolute numbers (percentages) for the binary outcomes, with odds ratios and 95 percent confidence intervals; *adjusted for corresponding
baseline value, history of type 2 diabetes mellitus, age and baseline estimated glomerular filtration rate. BNP indicates B-type natriuretic peptide; KCCQ CS, Kansas
City Cardiomyopathy Questionnaire clinical summary score; KCCQ OS, Kansas City Cardiomyopathy Questionnaire overall summary score; and NT-proBNP, N-terminal
pro B-type natriuretic peptide.
levels were lower, in patients treated with dapagliflozin treated patients experiencing substantial increase or
by the end of the treatment period. no change, and greater proportions of dapagliflozin-
Responder analyses, especially when using KCCQ- treated patients experiencing substantial reductions
CS, which quantifies both symptoms and physical limi- in natriuretic peptides versus placebo. These analyses
tations, indicated that a higher proportion of patients used a threshold of 20% or greater change; based on
treated with dapagliflozin experienced small-moder- previous studies which demonstrated that change in
ate, moderate-large, and very large improvements, and the excess of 20% demonstrates more than random
fewer patients treated with dapagliflozin experienced or physiologic variation in NT-proBNP, correlates with
deterioration or no change in health status. Health sta- echocardiographic markers of adverse left ventricular
tus improvements with dapagliflozin versus placebo remodeling, and is associated with clinical events in
were seen across all KCCQ domains, except for the patients with established HF.22–24 Notably, these results
social limitations score. Similar effects of dapagliflozin were achieved in the context of high rates of back-
Downloaded from http://ahajournals.org by on October 7, 2019
versus placebo were seen in the responder analyses ex- ground guideline-directed optimal medical therapy
amining NT-proBNP and BNP with fewer dapagliflozin- for HFrEF, including >95% patients treated with beta
Table 3. Continuous Secondary End Points at 6 and 12 Weeks for Patients Treated With Dapagliflozin Versus Placebo
Values are shown as adjusted means (95 percent confidence intervals) for continuous variables. BNP indicates B-Type Natriuretic Peptide; KCCQ CS, Kansas City
Cardiomyopathy Questionnaire clinical summary score; KCCQ OS, Kansas City Cardiomyopathy Questionnaire overall summary score; and NT-proBNP, N-Terminal
Pro B-Type Natriuretic Peptide.
*Adjusted for corresponding baseline value, history of type 2 diabetes mellitus, age, and baseline estimated glomerular filtration rate.
ORIGINAL RESEARCH
Dapagliflozin patients with T2D, with these cardiovascular benefits
(n=131) Placebo (n=132) having little to do with their effect on plasma glucose
ARTICLE
All-cause death 1 (0.8%) 1 (0.8%) or HbA1c. Whether this class of medications can also
Cardiovascular death 1 (0.8%) 1 (0.8%) effectively improve prognosis and symptom burden in
Nonfatal MI 0 (0.0%) 4 (3.0%)
patients with established HF (specifically HFrEF) was not
previously explored. Moreover, whether they would be
Stroke 0 (0.0%) 1 (0.8%)
helpful in patients without T2D was also not established.
Acute kidney injury 1 (0.8%) 1 (0.8%)
While large HF outcomes trials are currently underway
Diabetic ketoacidosis 0 (0.0%) 0 (0.0%) with several agents in the class, DEFINE-HF is the first
Volume depletion event 12 (9.2%) 7 (5.3%) randomized controlled trial to explore the early effects
Severe hypoglycemic event 1 (0.8%) 1 (0.8%) of an SGLT-2i on natriuretic peptides and health status.
Lower limb amputation 0 (0.0%) 0 (0.0%) The 3.7 point mean improvement in KCCQ-OS and
4.6 point mean improvement in KCCQ-CS with dapa-
Drug adverse event 3 (2.3%) 3 (2.3%)
gliflozin versus placebo after 12 weeks of treatment
AE leading to study drug 11 (8.4%) 12 (9.1%)
discontinuation
compare favorably with the 2 most recently approved
therapies for HFrEF. Specifically, ivabradine demon-
Serious adverse event 30 (22.9%) 24 (18.2%)
strated a 2.6 point mean improvement in KCCQ-OS,
Values are shown as absolute numbers (percentages) for patients with and 1.9 point mean improvement in KCCQ-CS over 22
events. AE indicates adverse event; and MI, myocardial infarction.
months of treatment, and ARNI demonstrated 1.3 and
0.9 point improvements in KCCQ-OS and KCCQ-CS,
blockers, 33% on ARNI, 59% on angiotensin-convert- respectively, after 8 months of treatment.27,28
ing enzyme inhibitors/angiotensin receptor blockers, There are several reasons that may explain why
and over 60% on mineralocorticoid antagonists, as dapagliflozin did not significantly reduce the first dual
well as frequent use of devices, including cardiac resyn- primary outcome (mean adjusted NT-proBNP) while
chronization therapy. There were modest numeric im- it did improve the second primary outcome (clinically
provements in weight, HbA1c, systolic blood pressure, meaningful changes in NT-proBNP and health status).
and 6-minute walk distance in patients treated with SGLT-2i exert modest osmotic diuretic effect by prevent-
ing glucose reabsorption in the proximal tubule, and
Downloaded from http://ahajournals.org by on October 7, 2019
as the benefits of fewer symptoms, less physical limita- which allowed an appropriate interpretation of the ef-
ORIGINAL RESEARCH
tions, and better quality of life become more apparent fects SGLT-2i on HF-related health status and natriuretic
to patients in their social activities. peptides on top of state-of-the art guideline-directed
ARTICLE
Despite the large improvement in patient reported medical and device therapy for HFrEF. Furthermore, we
physical limitations because of HF with dapagliflozin ensured high adherence to study procedures, with a
versus placebo (based on KCCQ physical limitations do- small proportion of patients prematurely discontinuing
main), we did not observe a corresponding significant assigned treatments, and 100% completing the study,
improvement in 6-minute walking distance. There are with no withdrawals of consent or loss to follow-up.
several potential explanations for this seeming discrep- Finally, and perhaps most importantly, this is the first
ancy. First (and most likely) reason is that the 6-minute time that early clinical cardiovascular benefits from this
walk test is more heavily dependent on the effort of the emerging class of medications have been observed in
operator and the patient at 1 point in time (the day of HF patients without T2D.
the clinic visit), which may potentially create significant Despite the strengths in study design and execution,
variability, and make it more difficult to standardize. our findings should be interpreted in the context of sev-
Second, comorbidities unrelated to heart failure (such eral potential limitations. DEFINE-HF was not powered
as orthopedic limitations) can affect 6-minute walking to evaluate effects on HF hospitalizations or mortality.
distance even if HF-related physical limitations have im- Second, its short duration of follow up precludes an
proved. In contrast, KCCQ is only dependent on the assessment on the durability of the observed benefit
patient (self-standardized), has a 2-week recall period, on HF disease-specific health status. Third, all patients
and it is specific to HF-related physical limitations. Third, were enrolled at sites in the US, and while this makes
while there is a correlation between change in 6-minute the study applicable to the US population with HFrEF,
walking distance and change in KCCQ, it is a relatively its generalizability outside the US is uncertain. Large
modest correlation (r=0.28).30 Finally, it may take time outcome trials, including DAPA-HF (Study to Evaluate
for the patient perception of less physical limitations the Effect of Dapagliflozin on the Incidence of Wors-
(KCCQ) to translate to a better objective measurement ening Heart Failure or Cardiovascular Death in Patients
of physical function (6-minute walking distance). with Chronic Heart Failure; NCT03036124), EMPER-
Dapagliflozin demonstrated a reassuring safety pro- OR-Reduced (Empagliflozin Outcome Trial in Patients
file, consistent with previous studies. Dapagliflozin was With Chronic Heart Failure With Reduced Ejection
well tolerated with only 11 of 131 patients stopping Fraction; NCT 03057977) and SOLOIST-WHF (Effect of
Downloaded from http://ahajournals.org by on October 7, 2019
drug because of adverse events, which was similar to Sotagliflozin on Cardiovascular Events in Patients With
the number of patients that stopped placebo. These Type 2 Diabetes Post Worsening Heart Failure; NCT
safety data are important, as DEFINE-HF is one of the 03521934) will ultimately address these limitations.
first rigorous studies to report on the tolerability of SGLT-
2i for patients without T2D. Indeed, there have been Conclusion
potential concerns with volume depletion in patients
with HFrEF on high doses of loop diuretics; however, Among patients with HFrEF with and without T2D re-
we found only a small numerical increase in volume ceiving optimal guideline directed medical therapy, the
depletion events with dapagliflozin (12 total events vs 7 addition of dapagliflozin for 12 weeks did not affect
with placebo). Likewise, we did not observe any imbal- the mean NT-proBNP but significantly increased the
ance in severe hypoglycemic events in patients with or proportion of patients experiencing clinically meaning-
without T2D, although investigators were encouraged ful improvements in HF disease-specific health status
to consider a modest reduction in the dose of insulin and natriuretic peptides.
or sulfonylureas in those entering the trial with HbA1c
already tightly controlled (a strategy that may also be
ARTICLE INFORMATION
adopted in clinical practice). It is also reassuring that
Received July 25, 2019; accepted August 20, 2019.
we did not observe any events of diabetic ketoacidosis The online-only Data Supplement is available with this article at https://
(although the duration of the study was short and pa- www.ahajournals.org/doi/suppl/10.1161/circulationaha.119.042929.
tients with type-1 diabetes mellitus were excluded) or
lower limb amputations. Authors
There are several key attributes of the DEFINE-HF Michael E. Nassif, MD; Sheryl Windsor, BS; Fengming Tang, MS; Yevgeniy
Trial that deserve to be highlighted. A substantial pro- Khariton, MD; Mansoor Husain, MD; Silvio Inzucchi, MD; Darren McGuire,
MD; Bertram Pitt, MD; Benjamin Scirica, MD; Bethany Austin, MD; Mark
portion of the patients were African American, a popu- Drazner, MD; Michael Fong, MD; Michael Givertz, MD; Robert Gordon, MD;
lation traditionally under-represented in clinical trials. Rita Jermyn, MD; Stuart Katz, MD; Sumant Lamba, MD; David Lanfear, MD;
We observed a high rate of optimal guideline directed Shane LaRue, MD; JoAnn Lindenfeld, MD; Michael Malone, MD; Kenneth
Margulies, MD; Robert Mentz, MD; R. Kannan Mutharasan, MD; Michael
medical therapy, including a third of patients on ARNI, Pursley, MD; Guillermo Umpierrez, MD; Mikhail Kosiborod, MD; On behalf of
and frequent use of cardiac resynchronization therapy, the DEFINE-HF Investigators
Correspondence sultant; Company Relationship; Astra Zeneca, Lilly USA, Boehringer Ingelheim,
Merck & Co, Novo Nordisk, Metavant, Applied Therapeutics, Sanofi Aventis. Dr
ORIGINAL RESEARCH
Mikhail Kosiborod, MD, Saint Luke’s Mid America Heart Institute, 4401 Wornall
Pitt: Consultant; Company Relationship; Astra Zeneca, Bayer, Sanofi Aventis,
Rd, Kansas City, MO 64111. Email mkosiborod@saint-lukes.org
Vifor, Sarfez, KBP pharmaceuticals, scPharmaceuticals, Cereno. Stock or Stock
ARTICLE
Options; Company Relationship; Vifor, Sarfez, KBP Pharmaceuticals, scPharma-
Affiliations ceuticals, Cereno. Patents; Company Relationship; US patent # 9931412-site
specific delivery of eplerenone to the myocardium. Dr Scirica: Grant/Research
Saint Luke’s Mid America Heart Institute, Kansas City, MO (M.E.N., S.W., F.T.,
Support; Company Relationship; Dr Scirica reports research grants via Brigham
Y.K., B.A., M.K.). University of Missouri-Kansas City, MO (M.E.N., Y.K., B.A.,
and Women’s Hospital from AstraZeneca, Eisai, Novartis, and Merck, member
M.K.). Toronto General Hospital Research Institute, University Health Network,
of the TIMI Study Group which has received institutional research grant support
Toronto, Canada (M.H.). Ted Rogers Centre for Heart Research, Toronto, Can-
through Brigham and Women’s Hospital in the past 36 months from: Abbott,
ada (M.H.). University of Toronto, Canada (M.H.). Peter Munk Cardiac Centre,
Amgen, Aralez, AstraZeneca, Bayer Health, Pharmaceuticals, Inc., BRAHMS,
Toronto, Canada (M.H). Yale University School of Medicine, New Haven, CT
Daiichi-Sankyo, Eisai, GlaxoSmithKline, Intarcia, Janssen, MedImmune, Merck,
(S.I.). University of Texas Southwestern Medical Center, Dallas (D.M., M.D.).
Novartis, Pfizer, Poxel, Quark Pharmaceuticals, Roche, Takeda, The Medicines
University of Michigan School of Medicine, Ann Arbor (B.P.). Brigham and
Company, Zora B. Consultant; Company Relationship; AbbVie, Allergan, Astra
Women’s Hospital, Boston, MA (B.S., M.G.). University of Southern California,
Zeneca, Boehringer Ingelheim, Covance, Eisai, Elsevier Practice Update Cardiol-
Los Angeles (M.F.). NorthShore University, Evanston, IL (R.G.). St. Francis Hospi-
ogy, Glaxo Smith Kline, Lexicon, Medtronic, Merck, NovoNordisk, Sanofi Aventis,
tal, New York, NY (R.J.). New York University Langone Health, New York (S.K.).
Equity in Health [at] scale. Dr Mutharasan: Consultant; Company Relationship;
First Coast Cardiovascular Institute, Jacksonville, FL (S.L.). Henry Ford Hospital,
GE Healthcare. Dr Givertz: Consultant; Company Relationship; Merck. Dr Jer-
Detroit, MI (D.L.). Washington University School of Medicine, St. Louis, MO
myn: Speaker’s Bureau; Company Relationship; Novartis. Dr Lindenfeld: Grant/
(S.L.). Vanderbilt University, Nashville, TN (J.L.). Charlotte Heart Group Research
Research Support; Company Relationship; Astra Zeneca. Consultant; Company
Center, Port Charlotte, FL (M.M.). University of Pennsylvania, Philadelphia
Relationship; Abbott, Astra Zeneca, Boehringer Ingelheim, Edwards Lifesciences,
(K.M.). Duke University, Durham, NC (R.M.). Northwestern University, Chicago,
CVRx, Impulse Dynamics, Relypsa, VWave. Dr Margulies: Grant/Research Sup-
IL (R.K.M.). Heart Group of the Eastern Shore, Fairhope, AL (M.P.). Emory Uni-
port; Company Relationship; GlaxoSmithKline, Merck Sharp and Dohme, Sanofi-
versity, Atlanta, GA (G.U.). The George Institute for Global Health, Sydney, Aus-
Aventis USA. Consultant; Company Relationship; American Reagent Pharmaceu-
tralia (M.K.). University of New South Wales, Sydney, Australia (M.K.).
ticals (formerly Luitipold), MyoKardia, Inc. Dr Mentz: Grant/Research Support;
Company Relationship; National Institutes of Health (U01HL125511-01A1 and
Acknowledgments R01AG045551-01A1), Akros, Amgen, AstraZeneca, Bayer, GlaxoSmithKline,
Gilead, InnoLife, Luitpold/American Regent, Medtronic, Merck, Novartis, Sanofi.
We thank Hunter Brandt, Sharon Chavis, Cheryl Cangemi, Karry Calderon,
Honoraria; Company Relationship; Abbott, Amgen, AstraZeneca, Bayer, Boston
Stanley Cobos, Matthew Fink, Caroline Foster, Rosann Gans, Ellen Hathcock,
Scientific, Janssen, Luitpold Pharmaceuticals, Medtronic, Merck, Novartis, Sanofi.
Patricia Hendershot, Mary Hudson, Jessica Huckleberry, Carolyn Kelly, Latreina
Consultant; Company Relationship; Amgen, AstraZeneca, Luitpold, Merck, No-
Koonce, Zack Malouf, Jennifer Marshall, Robinson Mendoza, Morgan Moore,
vartis, Boehringer Ingelheim. The other authors report no conflicts.
Karen Mullinax, Navdeep Nayyar, Kelsey Neaton, Ngozi Okafor, Kimberly Pap-
rockas, Diane Peterman, Linda Pierchala, Kirsten Quiles, Melissa Ramos, Mary
Reed, Michaela Ritz, Sindy Roche, Daniel Roshevsky, Sevinj Sasunova, Rosanne
Schenks, Virginia Schmidt, Mary Sheehan, Ruth Sorg, Marlo Sierra, Emily Tem- APPENDIX
poni, Bernardo Vargas, Pamela Williams, and Jennifer Wilson.
DEFINE-HF Investigators
Downloaded from http://ahajournals.org by on October 7, 2019
3. Bui AL, Horwich TB, Fonarow GC. Epidemiology and risk profile of heart al; DAPA-HF Committees and Investigators. A trial to evaluate the effect
failure. Nat Rev Cardiol. 2011;8:30–41. doi: 10.1038/nrcardio.2010.165 of the sodium-glucose co-transporter 2 inhibitor dapagliflozin on morbid-
ORIGINAL RESEARCH
4. McMurray JJ, Packer M, Desai AS, Gong J, Lefkowitz MP, Rizkala AR, Rouleau JL, ity and mortality in patients with heart failure and reduced left ventricu-
Shi VC, Solomon SD, Swedberg K, et al; PARADIGM-HF Investigators and lar ejection fraction (DAPA-HF). Eur J Heart Fail. 2019;21:665–675. doi:
ARTICLE
13. Perkovic V, Jardine MJ, Neal B, Bompoint S, Heerspink HJL, Charytan DM, 26. Dobre D, van Jaarsveld CH, deJongste MJ, Haaijer Ruskamp FM,
Edwards R, Agarwal R, Bakris G, Bull S, et al; CREDENCE Trial Investiga- Ranchor AV. The effect of beta-blocker therapy on quality of life in heart
tors. Canagliflozin and renal outcomes in type 2 diabetes and nephropa- failure patients: a systematic review and meta-analysis. Pharmacoepide-
thy. N Engl J Med. 2019;380:2295–2306. doi: 10.1056/NEJMoa1811744 miol Drug Saf. 2007;16:152–159. doi: 10.1002/pds.1234
14. Fitchett D, Zinman B, Wanner C, Lachin JM, Hantel S, Salsali A, Johansen OE, 27. Lewis EF, Claggett BL, McMurray JJV, Packer M, Lefkowitz MP, Rouleau JL,
Woerle HJ, Broedl UC, Inzucchi SE; EMPA-REG OUTCOME® trial investiga- Liu J, Shi VC, Zile MR, Desai AS, et al. Health-related quality of life out-
tors. Heart failure outcomes with empagliflozin in patients with type 2 comes in PARADIGM-HF. Circ Heart Fail. 2017;10:e003430. doi: 10.1161/
diabetes at high cardiovascular risk: results of the EMPA-REG OUTCOME® CIRCHEARTFAILURE.116.003430
trial. Eur Heart J. 2016;37:1526–1534. doi: 10.1093/eurheartj/ehv728 28. Ekman I, Chassany O, Komajda M, Böhm M, Borer JS, Ford I, Tavazzi L,
15. Inzucchi SE, Zinman B, Fitchett D, Wanner C, Ferrannini E, Schumacher M, Swedberg K. Heart rate reduction with ivabradine and health relat-
Schmoor C, Ohneberg K, Johansen OE, George JT, et al. How does empa- ed quality of life in patients with chronic heart failure: results from
gliflozin reduce cardiovascular mortality? Insights from a mediation analy- the SHIFT study. Eur Heart J. 2011;32:2395–2404. doi: 10.1093/
sis of the EMPA-REG OUTCOME Trial. Diabetes Care. 2018;41:356–363. eurheartj/ehr343
doi: 10.2337/dc17-1096 29. Verma S, McMurray JJV. SGLT2 inhibitors and mechanisms of cardiovascu-
16. Januzzi JL Jr, Butler J, Jarolim P, Sattar N, Vijapurkar U, Desai M, lar benefit: a state-of-the-art review. Diabetologia. 2018;61:2108–2117.
Davies MJ. Effects of canagliflozin on cardiovascular biomarkers in older doi: 10.1007/s00125-018-4670-7
adults with type 2 diabetes. J Am Coll Cardiol. 2017;70:704–712. doi: 30. Flynn KE, Lin L, Moe GW, Howlett JG, Fine LJ, Spertus JA, McConnell TR,
10.1016/j.jacc.2017.06.016 Piña IL, Weinfurt KP. Relationships between changes in patient-reported
17. McMurray JJV, DeMets DL, Inzucchi SE, Køber L, Kosiborod MN, health status and functional capacity in outpatients with heart failure. Am
Langkilde AM, Martinez FA, Bengtsson O, Ponikowski P, Sabatine MS, et Heart J. 2012;163:88–94.e3. doi: 10.1016/j.ahj.2011.09.027