Metformin Use and Mortality in Ambulatory Patients With

Diabetes and Heart Failure

David Aguilar, MD; Wenyaw Chan, PhD; Biykem Bozkurt, MD;
Kumudha Ramasubbu, MD; Anita Deswal, MD, MPH

Background—Despite the common coexistence of diabetes and heart failure (HF), the optimal medial treatment of diabetes
in HF patients has not been well studied. We sought to compare the association between metformin use and clinical
outcomes in a cohort of ambulatory patients with diabetes and established HF.
Methods and Results—Using propensity score–matched samples, we examined the association between metformin use and
the risk of death or risk of hospitalization in a national cohort of 6185 patients with HF and diabetes treated in
ambulatory clinics at Veteran Affairs medical centers. In this cohort, 1561 (25.2%) patients were treated with metformin.
At 2 years of follow-up, death occurred in 246 (15.8%) patients receiving metformin and in 1177 (25.5%) patients not
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018

receiving metformin (P⬍0.001). In the propensity score–matched analysis (n⫽2874), death occurred in 232 (16.1%)
patients receiving metformin compared with 285 (19.8%) patients not receiving metformin (hazard ratio, 0.76; 95%
confidence interval, 0.63 to 0.92; P⬍0.01). In propensity score–matched analyses, HF hospitalization or total
hospitalization rates were not significantly different between individuals treated with metformin compared with those
not treated with metformin (hazard ratio, 0.93; 95% confidence interval, 0.74 to 1.18; and hazard ratio, 0.94; 95%
confidence interval, 0.83 to 1.07, respectively).
Conclusions—Metformin therapy was associated with lower rates of mortality in ambulatory patients with diabetes and
HF. Future prospective studies are necessary to define the optimal therapy for diabetic patients with HF. (Circ Heart
Fail. 2011;4:53-58.)
Key Words: diabetes 䡲 heart failure 䡲 metformin 䡲 survival 䡲 prognosis

D iabetes mellitus and heart failure (HF) commonly coex-

ist in the same patient. In a recent community based
survey in Olmstead County, investigators determined that
20% of individuals presenting with a first episode of HF had
a prior diagnosis of diabetes.1 In addition, the investigators
demonstrated that prevalence of diabetes in patients with HF
had increased markedly over a 20-year period (3.8% per year)
without signs of abatement. In hospitalized patients with decom-
pensated HF, the prevalence of diabetes appears even greater
than community-based surveys and may extend to 44%.2 Im-
portantly, both community-based studies1 and clinical trials3,4
have demonstrated that the presence of diabetes in patients with
HF is associated with increased risk of death compared with HF
patients without diabetes. Therefore, efforts to adequately treat
these patients have become increasingly important.
Clinical Perspective on p 58
The optimal diabetic treatment strategy in patients with
diabetes and HF has not been well defined, and some studies
have suggested potential harm associated with several antidi-
abetic medications.5 Metformin is a biguanide that decreases
hepatic glucose production, improves glucose uptake and
utilization, and improves insulin sensitivity.6 Until recently,
metformin has been contraindicated in patients with HF
because of concerns regarding risk of lactic acidosis.7,8
Subsequent retrospective, nonrandomized studies have dem-
onstrated that metformin appears safe and may be associated
with lower morbidity and mortality rates in diabetic patients
with established HF when compared with other diabetic
therapy.9,10 These nonrandomized studies are subject to in-
herent limitations of potential selection bias associated with
metformin use. For example, it is possible that providers did
not prescribe metformin in elderly patients with more ad-
vanced disease or renal insufficiency due to safety concerns.
Although randomized controlled clinical trials are ideal to
reduce potential selection bias, an initial attempt to perform a
randomized controlled study of metformin use in diabetic
patients with HF was terminated because of feasibility con-

Received April 7, 2010; accepted October 1, 2010.

From the Winters Center for Heart Failure Research and Section of Cardiology (D.A., B.B., K.R., A.D.), Department of Medicine, Baylor College of
Medicine, Houston, Tex; the Section of Cardiology (D.A., B.B., K.R., A.D.), Michael E. DeBakey VA Medical Center, Houston, Tex; the University of
Texas Health Science Center School of Public Health (W.C.), Houston, Tex; and Houston Center for Quality of Care and Utilization Studies (A.D.),
Michael E. DeBakey VA Medical Center, Houston, Tex.
Correspondence David Aguilar, MD, Cardiovascular Division, Baylor College of Medicine, 1709 Dryden St, BCM 620, Suite 500, Box 13, Houston,
TX 77030. E-mail daguilar@bcm.edu
© 2011 American Heart Association, Inc.
Circ Heart Fail is available at http://circheartfailure.ahajournals.org DOI: 10.1161/CIRCHEARTFAILURE.110.952556

53
 54 Circ Heart Fail January 2011

cerns.11 Propensity score methods are statistical techniques Table 1. Patient Characteristics of the Study Cohort (Before
developed to balance covariates in 2 groups in observational Propensity Match)
studies and therefore to reduce potential bias.12 Therefore, we Metformin No Metformin
sought to evaluate the relationship between metformin and (n⫽1561) (n⫽4624) P Value
mortality in a cohort of patients with diabetes and established Age, y 67.3⫾9.3 70.0⫾9.3 ⬍0.001
HF using propensity score–matched analyses. Male, % 92.2 93.9 0.02
Race, % ⬍0.001
Methods White 76.8 74.8
Study Design and Sample Black 8.3 12.8
We performed an observational study of a national cohort of veterans Other/unknown 14.9 12.5
with HF treated in ambulatory clinics at Veteran Affairs (VA) BMI, kg/m2 32.6⫾7.0 31.3⫾6.6 ⬍0.001
medical centers using the VA External Peer Review Program (EPRP)
data between October 2000 and September 30, 2002. The VA EPRP SBP, mm Hg 133⫾21 131⫾22 0.001
was created to assess and improve the quality of care for VA patients LVEF, % ⬍0.001
and has been described previously.13,14 The sampling pool of Normal or mildly reduced 46.6 44.3
outpatients for EPRP included ambulatory patients with common (LVEF ⱖ40%)
chronic diseases such as HF, diabetes, ischemic heart disease (prior Moderate or severely reduced 33.2 39.4
myocardial infarction), and chronic obstructive pulmonary disease
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018

(LVEF ⬍40%)
identified by specific ICD-9 codes. Experienced data abstractors then
reviewed electronic medical records for validation of sample selection Unknown 20.2 16.3
criteria, including a documentation of a diagnosis of HF in the outpatient Diabetes with complications, % 52.3 59.3 ⬍0.001
charts for the EPRP HF cohort,14 For further validation, 70 patients in Peripheral vascular disease, % 20.1 24.8 ⬍0.001
EPRP cohort from the Houston VA were reviewed; documentation of a
Hypertension, % 77.0 75.0 0.11
diagnosis of HF by a clinician in the electronic medical records was
confirmed in 96% of cases. Patient-level data from the EPRP HF cohort Atrial fibrillation, % 25.7 30.0 0.001
were linked with 5 existing national VA databases to obtain further Past myocardial infarction, % 32.9 36.0 0.03
demographic, laboratory, pharmacy, and outcome data. Prior HF hospitalization within 11.0 21.0 ⬍0.001
Individuals from the EPRP HF cohort who had diabetes as 2 y, %
identified in the EPRP data (n⫽8842) and who were prescribed
COPD, % 25.6 27.6 0.11
hypoglycemic medications in the pharmacy database (n⫽7147) were
included in this study. Diabetic therapy was ascertained using Cancer, % 13.6 17.8 ⬍0.001
pharmacy data and was based on prescriptions filled 90 days before HbA1C, % 7.8⫾1.6 7.6⫾1.6 ⬍0.001
the index outpatient visit or 30 days after the index outpatient visit. GFR, mL/min/1.73 m2 67.1⫾20.6 52.0⫾22.5 ⬍0.001
In addition to diabetic status and therapy, baseline demographics and
BUN, mg/dL 22.0⫾9.9 30.3⫾16.5 ⬍0.001
concomitant cardiac medications were assessed at the index visit.
The most recent laboratory data within 1 year before the index visit Hemoglobin, mg/dL 13.6⫾1.6 13.2⫾1.8 ⬍0.001
and up to 2 weeks after the index visit were used. Because of the Cholesterol, mg/dL 175⫾41 171⫾42 ⬍0.01
important potential for confounding between metformin use and Sodium, mEq/L 139⫾3.3 139⫾3.5 0.41
renal dysfunction, patients with missing creatinine values were
Medications
excluded from these analyses (n⫽962), leaving a total of 6185
patients in the total cohort. Other missing laboratory values were Insulin, % 33.5 53.1 ⬍0.001
imputed using the median value of the study cohort for that Sulfonylurea, % 59.1 58.8 0.82
parameter and a dummy variable was used to indicate replacement of TZD, % 11.6 11.8 0.84
missing data. Glomerular filtration rate (GFR) was calculated using
ACE/ARB, % 86.5 82.2 ⬍0.001
the 4-variable Modification of Diet in Renal Disease (MDRD)
equation.15 Covariates that reflected diabetes severity included he- Spironolactone 24.2 26.1 0.11
moglobin A1C and a variable that documented the presence of a ␤-blocker, % 59.5 61.4 0.18
diabetic complication including neuropathy, nephropathy, retinopa- Statin, % 59.7 55.5 ⬍0.01
thy, or peripheral vascular disease.
Data expressed as mean⫾SD or percentage.
BMI indicates body mass index; SBP, systolic blood pressure; LVEF, left
Primary and Secondary Outcomes
ventricular ejection fraction; COPD, chronic obstructive pulmonary disease; HbA1C,
The primary outcome was time to death over 2 years of complete
hemoglobin A1C; BUN, blood urea nitrogen; TZD, thiazolidinedione; and ACE/ARB,
follow-up after the index visit. Secondary outcomes included time to
angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker.
hospitalization for HF (based on diagnosis-related group) and time to
any hospitalization.
propensity model consists of 29 baseline variables: age; sex; race;
Statistical Analysis body mass index; systolic and diastolic blood pressures; left ventric-
Continuous variables are presented as means with standard devia- ular ejection fraction; history of hypertension, peripheral vascular
tions and categorical variables are presented with percentages. disease, cerebrovascular disease, atrial fibrillation, myocardial in-
Differences in baseline demographics between groups were ascer- farction, cancer, chronic obstructive pulmonary disease, and diabetic
tained using ␹2 tests for categorical variables and t tests for complications; HF hospitalization within the last 2 years; hemoglo-
continuous variables. Two-sided probability values ⬍0.05 were bin; glycosylated hemoglobin (hemoglobin A1C); sodium; blood urea
considered significant. nitrogen; GFR; total cholesterol; triglycerides; treatment with
Propensity scores, defined as the conditional probability of being ␤-blockers, sulfonylurea, thiazolidinedione, insulin, statins, and angio-
treated given the covariates, were used to adjust for possible bias due tensin-converting enzyme inhibitor or angiotensin II receptor blockers.
to nonrandom assignments of patients to different treatments.12 A We then performed a 1-to-1 matched analysis without replacement
propensity score was created for each individual using logistic on the basis of the estimated propensity score of each patient. The
regression, in which the outcome variable was metformin use. This nearest available Mahalanobis metric matching method was used to
 Aguilar et al Metformin and Mortality in Diabetic Heart Failure 55

Table 2. Patient Characteristics After Propensity Score Matching

Metformin No Metformin % Standardized
(n⫽1437) (n⫽1437) P Value Difference
Age, y 67.6⫾9.2 67.8⫾10.2 0.53 ⫺2.1
Male, % 92.3 92.9 0.52 ⫺2.2
Race, % 0.75
White 77.1 78.3 ⫺2.9
Black 8.9 8.4 1.8
Other/unknown 14.0 13.3 2.0
BMI, kg/m2 32.5⫾7.0 32.6⫾7.2 0.76 ⫺1.4
SBP, mm Hg 132⫾21 132⫾22 0.89 0.5
LVEF, % 0.69
Normal or mildly reduced (LVEF ⱖ40%) 46.1 44.9 2.4
Moderate or severely reduced (LVEF ⬍40%) 34.3 35.8 ⫺3.1
Unknown 19.6 19.3 0.8
Diabetes with complications, % 53.5 52.1 0.46 2.8
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018

Peripheral vascular disease, % 20.6 20.9 0.85 ⫺0.7

Hypertension, % 76.6 76.1 0.76 1.1
Atrial fibrillation, % 26.1 26.9 0.64 ⫺1.8
Past myocardial infarction, % 33.6 32.9 0.69 1.5
Prior HF hospitalization within 2 y, % 11.8 13.0 0.34 ⫺3.6
COPD, % 25.5 26.4 0.58 ⫺2.1
Cancer, % 14.1 15.6 0.27 ⫺4.2
HbA1C, % 7.8⫾1.6 7.8⫾1.8 0.26 ⫺4.1
GFR, L/min/1.73 m2 66.0⫾20.1 65.1⫾24.7 0.28 4.0
BUN, mg/dL 22.5⫾10.1 23.1⫾11.6 0.10 ⫺5.5
Hemoglobin, mg/dL 13.6⫾1.6 13.6⫾1.7 0.62 ⫺1.8
Cholesterol, mg/dL 174⫾41 174⫾42 0.84 0.7
Sodium, mEq/L 139⫾3 139⫾3 0.62 2.0
Medications
Insulin, % 36.4 36.5 0.97 ⫺0.2
Sulfonylurea, % 63.4 65.1 0.35 ⫺3.5
TZD, % 12.0 12.6 0.65 ⫺1.8
ACE/ARB, % 86.2 86.2 1.0 0
Spironolactone 25.5 24.1 0.39 3.2
␤-blocker, % 60.0 58.1 0.31 3.9
Statin, % 59.6 56.9 0.15 5.5
Data expressed as mean⫾SD or percentage.
BMI indicates body mass index; SBP, systolic blood pressure; LVEF, left ventricular ejection fraction; COPD, chronic obstructive
pulmonary disease; HbA1C, hemoglobin A1C; BUN, blood urea nitrogen; TZD, thiazolidinedione; and ACE/ARB, angiotensin-converting
enzyme inhibitor or angiotensin II receptor blocker.

perform the matching and to select the final data set for analysis. It
has been shown that this method is better than other propensity
matching methods in balancing the covariates between 2 treatment
groups.12 After propensity score matching, baseline characteristics
were compared with ␹2 tests for categorical variables and t tests for
continuous variables. In addition, we assessed the success of pro-
pensity score matching to balance covariates in the 2 groups using
standardized differences. Standardized differences of less than 10%
support the assumption of balance between the 2 groups.16
Cox proportional hazards models stratified on matched pairs were
used to assess the relationship between metformin use and the
outcome of interest, which included time to death, time to heart
failure hospitalization, and time to any hospitalization. The propor-
tional hazards assumption was met using Schoenfield residuals.
Exploratory tests for potential interaction between metformin use
and outcomes in certain subgroups (obese patients, renal insuffi-
ciency, and previous myocardial infarction) were performed by
adding a cross-product term of these variables in the Cox model.
Results
Baseline Characteristics
Of the 6185 patients in this cohort, 1561 (25.2%) patients
were receiving metformin therapy. Baseline characteristics
for the entire cohort and for the propensity score–matched
cohort are demonstrated in Table 1 and Table 2, respectively.
Before matching, patients receiving metformin were more
likely to be younger and have a higher body mass index,
systolic blood pressure, glycosylated hemoglobin, cholesterol,
and GFR than patients not receiving metformin. Patients
receiving metformin were also less likely to have peripheral
vascular disease, prior myocardial infarction, hospitalization
for HF within the previous 2 years, diabetic complications,
and cancer. Patients receiving metformin were less likely to
receive insulin therapy but were more likely to have received
 56 Circ Heart Fail
1.00
0
mates
0.95
vival Estim
0.90
Kaplan--Meier Surv
N M
No Metformin
tf
0.85
0.80
75
0.7
0 100 200
Figure 1. Kaplan–Meier survival curves and metformin treatment
in the propensity score–matched cohort. Probability value⫽0.01
(log-rank).
statins or angiotensin-converting enzyme inhibitors or angio-
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018
tensin II receptor blockers.
Unadjusted Clinical Outcomes
At 2 years of follow-up, 246 (15.8%) of patients receiving
metformin and 1177 (25.5%) of patients not receiving met-
formin had died (unadjusted hazard ratio [HR], 0.58; 95%
confidence interval [CI], 0.57 to 0.67; P⬍0.001). Similarly,
the proportions of patients hospitalized with HF (11.5%
versus 16.4%, P⬍0.001) and hospitalized for any cause
(40.9% versus 47.9%, P⬍0.001) were lower in patients
receiving metformin compared with those not receiving
metformin over the 2 years of follow-up.
Propensity Score–Matched Models
After propensity score matching, patient characteristics were
similar in patients receiving metformin versus those not receiv-
ing metformin. In the propensity score–matched cohort
(n⫽2874), death occurred in 232 (16.1%) patients receiving
metformin compared with 285 (19.8%) patients not receiving
metformin (Figure 1 and Table 3). Using Cox proportional
hazard models, metformin therapy was associated with a reduced
risk of time to death (HR, 0.76; 95% CI, 0.63 to 0.92; P⬍0.01).
Additional exploratory subgroup analyses did not demon-
strate a statistical interaction between metformin use and mor-
tality in the subgroups of obese patients (body mass index ⬎30
kg/m2), renal insufficiency (GFR ⬍60 mL/min/1.73 m2), and
those with a previous myocardial infarction (Figure 2).
Hospitalization events for the propensity score–matched
cohort are shown in Table 3. In propensity score–matched
analyses, there was no statistically significant difference in
time to HF hospitalization or time to all-cause hospitaliza-
Table 3. Outcomes in Propensity Score–Matched Cohort
Outcome
Death, n (%)
HR (95% CI)
HF hospitalization, n (%)
HR (95% CI)
Total hospitalization, n (%)
HR (95% CI)
January 2011
tions between individuals treated with metformin compared
with those not treated with metformin (HR, 0.93; 95% CI,
Metformin 0.74 to 1.18; P⫽0.56; and HR, 0.94; 95% CI, 0.83 to 1.07;
P⫽0.35, respectively).
i
Discussion
In a national cohort of ambulatory patients with diabetes and
HF, metformin was prescribed for diabetic treatment in
approximately 25% of patients. Using propensity score–
matched analyses, we demonstrate that metformin therapy
300 400 500 600 700 was associated with a survival benefit over 2 years of
Time (days) follow-up. This benefit was consistent across subgroups of
diabetic patients with HF, including those patients with renal
insufficiency, prior myocardial infarction, and obesity. In
addition, there were no statistically significant differences in
time to HF hospitalization or time to all-cause hospitalization
over 2 years of follow-up between those patients receiving
metformin compared with those not receiving metformin.
Until recently, the use of metformin has been contraindicated
in patients with HF because of concerns regarding the potential
risk of lactic acidosis.7 These concerns originally stemmed from
reports of an association between phenformin, another bigua-
nide, and several cases of lactic acidosis.17 In addition, in the first
year of postmarketing surveillance after metformin was intro-
duced in the United States, 47 patients treated with metformin
were reported to have lactic acidosis (approximately 5 cases per
100 000 patients treated), 18 of whom had history of HF.7
Despite these initial reports, subsequent analyses revealed that
the risk of lactic acidosis associated with metformin use was
very low in patients with type 2 diabetes and may not be higher
than diabetic patients not receiving metformin therapy.18,19
Two previous retrospective cohort studies have demon-
strated that metformin use appears safe and may be poten-
tially beneficial in diabetic patients with established HF. In
one of these studies, which included 16 417 Medicare bene-
ficiaries with diabetes who were recently discharged from the
hospital with a primary diagnosis of HF, metformin use was
associated with a 13% lower risk of death and 8% lower risk
of readmission for HF compared with diabetic patients not
receiving insulin-sensitizing diabetic therapy at 1 year of
follow-up.10 Furthermore, readmissions for metabolic acido-
sis were similar among patients not treated with an insulin
sensitizer (2.6%) compared with those discharged with a
prescription for metformin (2.3%, P⫽0.40) in the Medicare
population. Similarly, in a retrospective analysis of 1833 new
users of oral diabetic therapy with HF, fewer deaths occurred
in subjects receiving metformin monotherapy or combination
therapy with metformin and sulfonylurea when compared
with sulfonylurea therapy alone.9
Despite the strengths of these previous retrospective cohort
Metformin No Metformin
studies, several limitations exist. Multivariable modeling
(n⫽1437) (n⫽1437) techniques were used in previous studies9,10 to control for
232 (16.1) 285 (19.8)
potential confounding, but residual confounding may persist.
Specifically, selection bias, in that physicians may prescribe
0.76 (0.63–0.92, P⬍0.01) Reference
metformin to patients with less severe disease, may contribute
171 (11.9) 180 (12.5)
to the improved outcomes previously associated with met-
0.93 (0.74–1.18, P⫽0.56) Reference
formin use. Propensity score methods are statistical tech-
599 (41.7) 626 (43.6)
niques used to reduce the impact of treatment selection bias
0.94 (0.83–1.07, P⫽0.35) Reference
often seen in the estimation of treatment effects in nonran-
 Aguilar et al
Subgroup No. of Patients
GFR ≥ 60 1628
GFR < 60 1246
Past Myocardial Infarction 956
No Past Myocardial Infarction 1918
BMI > 30 kg/m2 1693
BMI ≤ 30 kg/m2 1181
0.4 0.6 0.8 1.0 1.2 1.4
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018
Hazard Ratio
domized, observational data.12,16 In the study of new users of
oral diabetic therapy, propensity score regression analyses
were performed, but the study lacked important potential
confounding variables such as renal function, left ventricular
function, and glycemic control.9
Our study extends the previous literature by using propen-
sity score–matched analyses to examine the association be-
tween metformin use and outcomes in a national cohort of
ambulatory patients with diabetes and HF. The sample size of
the cohort and the large number of demographic and clinical
variables available at baseline, including data on renal func-
tion, allow for propensity score matching on multiple vari-
ables and subsequent adequate balance of baseline character-
istics between treated and untreated patients. The improved
survival associated with metformin use persisted after pro-
pensity score–matched analyses and remained consistent with
previously published estimates.9,10
There are several potential mechanisms beyond glycemic
control through which metformin may improve outcomes in
HF patients.6 Compared with several other diabetic treatment
options, metformin is associated with a reduction in weight
gain and improvements in measures of insulin resistance. In
addition, metformin has been associated with improvements
in endothelial function, lipoprotein metabolism, oxidative
stress, and abnormalities of coagulation.6 These potential
mechanisms may have contributed to the reduction in mac-
rovascular events observed in a subset of overweight patients
treated with metformin in the United Kingdom Prospective
Diabetes Study (UKPDS) 3420 and in a recent study of
patients with type 2 diabetes in which metformin was added
to a background of insulin therapy.21
In addition to potential prevention of macrovascular
events, metformin has been associated with improvements in
cardiac function in both animal models of HF and models of
diabetic cardiomyopathy. For example, metformin treatment
is associated with improvements in measures of cardiac
performance in streptozotocin-diabetic rats.22 More recently,
metformin treatment led to improvements in survival and in
left ventricular function in murine models of ischemia-
Metformin and Mortality in Diabetic Heart Failure 57
P-value for
interaction
0.09
0.62
Figure 2. Hazard ratio for mortality associated with
metformin treatment in selected subgroups.
0.16
induced HF.23 Similarly, metformin therapy prevented pro-
gression of HF in a canine model of pacing-induced HF.24
These beneficial cardiac effects appear to be mediated
through activation of AMP-activated protein kinase and its
downstream mediators, including endothelial nitric oxide.23,24
Despite the strength of this study, several limitations should
be noted. Importantly, the incidence of lactic acidosis was not
obtained. It is reassuring that total hospitalizations in the entire
cohort and in the propensity score–matched sample were not
increased in patients receiving metformin therapy, but, given the
low reported incidence of potential lactic acidosis, our sample is
not suitable to address the issue of safety. Nonetheless, our
study, when combined with previously published studies in HF
patients,10,11 adds to a growing body of literature that metformin
use appears safe in diabetic patients with HF. Additionally,
although propensity score–matched analyses were performed to
minimize potential bias, residual bias and confounding may
persist. For example, New York Heart Association classification
was not available in this data set; we have attempted to address
this limitation by using other important prognostic variables in a
HF population, such as previous HF hospitalization and left
ventricular ejection fraction. In addition, medication use was
assessed only at baseline, and changes in medications over the
time of follow-up are not available. Despite the total number of
variables used to create the propensity score, the lack of these
and other key variables may contribute to residual confounding.
In conclusion, in a cohort of patients with diabetes and
established HF, prescription of metformin was associated with a
survival benefit over 2 years of follow-up. Given the current
burden and expected growth in the number of patients with
diabetes and HF, it is critically important that future studies
assess the optimal treatment strategy for glycemic control in this
population. Addressing these issues will require carefully de-
signed prospective observational studies to confirm safety and
randomized controlled clinical trials to assess efficacy. Although
these trials may be difficult to perform, they are required if we
are to reduce the growing burden of diabetes and HF.
Acknowledgments
We thank the Office of Quality and Performance of the Veterans Health
Administration for Providing External Peer Review Program data. The
 58 Circ Heart Fail January 2011

views expressed in this article are those of the authors and do not
necessarily represent those of the Department of Veteran Affairs.
Sources of Funding
This study was supported in part by VA Health Services Research
and Development Service grant IIR 02-082-1 (to Dr Deswal). Dr
Aguilar is a recipient of a NIH Mentored Career Development
Award (5K01-HL092585-02).
Disclosures
None.
References
1. From AM, Leibson CL, Bursi F, Redfield MM, Weston SA, Jacobsen SJ,
Rodeheffer RJ, Roger VL. Diabetes in heart failure: prevalence and
impact on outcome in the population. Am J Med. 2006;119:591–599.
2. Adams KF Jr, Fonarow GC, Emerman CL, LeJemtel TH, Costanzo MR,
Abraham WT, Berkowitz RL, Galvao M, Horton DP. Characteristics and
outcomes of patients hospitalized for heart failure in the United States:
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018
10. Masoudi FA, Inzucchi SE, Wang Y, Havranek EP, Foody JM, Krumholz
HM. Thiazolidinediones, metformin, and outcomes in older patients with
diabetes and heart failure: an observational study. Circulation. 2005;111:
583–590.
11. Eurich DT, Tsuyuki RT, Majumdar SR, McAlister FA, Lewanczuk R,
Shibata MC, Johnson JA. Metformin treatment in diabetes and heart
failure: when academic equipoise meets clinical reality. Trials. 2009;
10:12.
12. D’Agostino RB Jr. Propensity score methods for bias reduction in the
comparison of a treatment to a non-randomized control group. Stat Med.
1998;17:2265–2281.
13. Jha AK, Perlin JB, Kizer KW, Dudley RA. Effect of the transformation
of the Veterans Affairs Health Care System on the quality of care. N Engl
J Med. 2003;348:2218 –2227.
14. Deswal A, Petersen NJ, Urbauer DL, Wright SM, Beyth R. Racial
variations in quality of care and outcomes in an ambulatory heart failure
cohort. Am Heart J. 2006;152:348 –354.
15. Levey AS, Coresh J, Greene T, Stevens LA, Zhang YL, Hendriksen S,
Kusek JW, Van Lente F. Using standardized serum creatinine values in
the modification of diet in renal disease study equation for estimating
glomerular filtration rate. Ann Intern Med. 2006;145:247–254.
16. Austin PC. Balance diagnostics for comparing the distribution of baseline

rationale, design, and preliminary observations from the first 100,000

cases in the Acute Decompensated Heart Failure National Registry
(ADHERE). Am Heart J. 2005;149:209 –216.
3. Aguilar D, Solomon SD, Kober L, Rouleau JL, Skali H, McMurray JJ,
Francis GS, Henis M, O’Connor CM, Diaz R, Belenkov YN, Varshavsky
S, Leimberger JD, Velazquez EJ, Califf RM, Pfeffer MA. Newly
diagnosed and previously known diabetes mellitus and 1-year outcomes
of acute myocardial infarction: the VALsartan In Acute myocardial
iNfarcTion (VALIANT) trial. Circulation. 2004;110:1572–1578.
4. MacDonald MR, Petrie MC, Varyani F, Ostergren J, Michelson EL,
Young JB, Solomon SD, Granger CB, Swedberg K, Yusuf S, Pfeffer MA,
McMurray JJ. Impact of diabetes on outcomes in patients with low and
preserved ejection fraction heart failure: an analysis of the Candesartan in
Heart failure: Assessment of Reduction in Mortality and morbidity
(CHARM) programme. Eur Heart J. 2008;29:1377–1385.
5. Eurich DT, McAlister FA, Blackburn DF, Majumdar SR, Tsuyuki RT,
Varney J, Johnson JA. Benefits and harms of antidiabetic agents in
patients with diabetes and heart failure: systematic review. BMJ. 2007;
335:497.
6. Kirpichnikov D, McFarlane SI, Sowers JR. Metformin: an update. Ann
Intern Med. 2002;137:25–33.
7. Misbin RI, Green L, Stadel BV, Gueriguian JL, Gubbi A, Fleming GA.
Lactic acidosis in patients with diabetes treated with metformin. N Engl
J Med. 1998;338:265–266.
8. Inzucchi SE, Masoudi FA, McGuire DK. Metformin in heart failure.
Diabetes Care. 2007;30:e129.
9. Eurich DT, Majumdar SR, McAlister FA, Tsuyuki RT, Johnson JA.
Improved clinical outcomes associated with metformin in patients with
diabetes and heart failure. Diabetes Care. 2005;28:2345–2351.
covariates between treatment groups in propensity-score matched
samples. Stat Med. 2009;28:3083–3107.
17. Misbin RI. Phenformin-associated lactic acidosis: pathogenesis and
treatment. Ann Intern Med. 1977;87:591–595.
18. Misbin RI. The phantom of lactic acidosis due to metformin in patients
with diabetes. Diabetes Care. 2004;27:1791–1793.
19. Salpeter SR, Greyber E, Pasternak GA, Salpeter EE. Risk of fatal and
nonfatal lactic acidosis with metformin use in type 2 diabetes mellitus:
systematic review and meta-analysis. Arch Intern Med. 2003;163:
2594 –2602.
20. Effect of intensive blood-glucose control with metformin on compli-
cations in overweight patients with type 2 diabetes (UKPDS 34). UK
Prospective Diabetes Study (UKPDS) Group. Lancet. 1998;352:
854 – 865.
21. Kooy A, de Jager J, Lehert P, Bets D, Wulffele MG, Donker AJ,
Stehouwer CD. Long-term effects of metformin on metabolism and
microvascular and macrovascular disease in patients with type 2 diabetes
mellitus. Arch Intern Med. 2009;169:616 – 625.
22. Verma S, McNeill JH. Metformin improves cardiac function in isolated
streptozotocin-diabetic rat hearts. Am J Physiol. 1994;266:H714 –H719.
23. Gundewar S, Calvert JW, Jha S, Toedt-Pingel I, Ji SY, Nunez D, Ram-
achandran A, Anaya-Cisneros M, Tian R, Lefer DJ. Activation of AMP-
activated protein kinase by metformin improves left ventricular function
and survival in heart failure. Circ Res. 2009;104:403– 411.
24. Sasaki H, Asanuma H, Fujita M, Takahama H, Wakeno M, Ito S, Ogai A,
Asakura M, Kim J, Minamino T, Takashima S, Sanada S, Sugimachi M,
Komamura K, Mochizuki N, Kitakaze M. Metformin prevents pro-
gression of heart failure in dogs: role of AMP-activated protein kinase.
Circulation. 2009;119:2568 –2577.

CLINICAL PERSPECTIVE
Diabetes mellitus and heart failure (HF) commonly coexist in the same patient, and, importantly, the presence of diabetes
in HF patients is associated with increased morbidity and mortality. Therefore, efforts to adequately treat diabetes have
become increasingly important. Historically, metformin has been contraindicated in HF patients because of concerns of
lactic acidosis. Recent animal studies and retrospective studies of patients with diabetes and HF have suggested that
metformin may in fact be beneficial. One important limitation of these previous human studies is the potential for selection
bias, in that patients with more severe illness may be less likely to have been prescribed metformin. Propensity score
methods are statistical techniques developed to balance covariates in 2 groups in observational studies and therefore reduce
potential selection bias. In the present study of a national cohort of ambulatory patients with diabetes and established HF,
we used propensity score–matched methods to assess the association between metformin and outcomes. Using these
techniques, we demonstrate that metformin use was associated with lower rates of mortality at 2-year follow-up, without
differences in HF hospitalizations or total hospitalizations between the 2 groups. Future prospective studies are needed to
confirm the potential benefits and safety of metformin in diabetic patients with HF. Until these data are available, our study
adds to a growing body of literature suggesting that metformin may be beneficial in this population.
 Metformin Use and Mortality in Ambulatory Patients With Diabetes and Heart Failure
David Aguilar, Wenyaw Chan, Biykem Bozkurt, Kumudha Ramasubbu and Anita Deswal
Downloaded from http://circheartfailure.ahajournals.org/ by guest on January 14, 2018

Circ Heart Fail. 2011;4:53-58; originally published online October 15, 2010;
doi: 10.1161/CIRCHEARTFAILURE.110.952556
Circulation: Heart Failure is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX
75231
Copyright © 2010 American Heart Association, Inc. All rights reserved.
Print ISSN: 1941-3289. Online ISSN: 1941-3297

The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://circheartfailure.ahajournals.org/content/4/1/53

Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Circulation: Heart Failure can be obtained via RightsLink, a service of the Copyright Clearance Center, not
the Editorial Office. Once the online version of the published article for which permission is being requested is
located, click Request Permissions in the middle column of the Web page under Services. Further information
about this process is available in the Permissions and Rights Question and Answer document.

Reprints: Information about reprints can be found online at:

http://www.lww.com/reprints

Subscriptions: Information about subscribing to Circulation: Heart Failure is online at:

http://circheartfailure.ahajournals.org//subscriptions/