Masoudi FA, Wang Y, Inzucchi SE, Setaro JF, Havranek EP, Foody JM, Krumholz HM. Metformin and Thiazolidinedione Use in Medicare Patients With Heart Failure. JAMA. 2003;290(1):81-85. doi:10.1001/jama.290.1.81
Author Affiliations: Division of Cardiology, Department of Medicine, Denver Health Medical Center (Drs Masoudi and Havranek), and Divisions of Cardiology (Drs Masoudi and Havranek) and Geriatric Medicine (Dr Masoudi), Department of Medicine, University of Colorado Health Sciences Center, Denver; Colorado Foundation for Medical Care, Aurora (Drs Masoudi, Havranek, and Krumholz); Sections of Cardiovascular Medicine (Drs Setaro, Foody, and Krumholz and Mr Wang) and Endocrinology (Dr Inzucchi), Department of Medicine, and Section of Health Policy and Administration, Department of Epidemiology and Public Health (Dr Krumholz), Yale University School of Medicine, and Center for Outcomes Research and Evaluation, Yale-New Haven Hospital (Dr Krumholz), New Haven, Conn.
Context According to package inserts, metformin is contraindicated in diabetic
patients receiving drug treatment for heart failure therapy, and thiazolidinediones
are not recommended in diabetic patients with symptoms of advanced heart failure.
Little is known about patterns of use of these antihyperglycemic drugs in
diabetic patients with heart failure.
Objective To determine the proportions of patients hospitalized with heart failure
and concomitant diabetes treated with metformin or thiazolidinediones.
Design Serial cross-sectional measurements using data from retrospective medical
Setting Nongovernmental acute care hospitals in the United States.
Patients Two nationally representative samples of Medicare beneficiaries hospitalized
with the primary diagnosis of heart failure and concomitant diabetes between
April 1998 and March 1999 and between July 2000 and June 2001.
Main Outcome Measures The prescription of either metformin or a thiazolidinedione at hospital
Results In the 1998-1999 sample (n = 12 505), 7.1% of patients were discharged
with a prescription for metformin, 7.2% with a prescription for a thiazolidinedione,
and 13.5% with a prescription for either drug. In the 2000-2001 sample (n
= 13 158), metformin use increased to 11.2%, thiazolidinedione use to
16.1%, and use of either drug to 24.4% (P<.001
for all comparisons). Similar increases were seen among patients of all age
groups, all races, and both sexes.
Conclusions The use of metformin and thiazolidinediones is common and has increased
rapidly in Medicare beneficiaries with diabetes and heart failure in direct
contrast with explicit warnings against this practice by the Food and Drug
Administration. Further studies to establish the safety and effectiveness
of this practice are needed to ensure optimal care of patients with diabetes
and heart failure.
Although diabetes and heart failure commonly coexist, the treatment
of diabetes in patients with both conditions is complicated because the popular
oral insulin sensitizers are not recommended for treatment of patients with
moderate-to-severe heart failure. According to the package insert, metformin
is contraindicated in all patients with "heart failure requiring pharmacologic
treatment" because of the increased risk of potentially lethal lactic acidosis.1 Thiazolidinediones are not recommended for patients
with "New York Heart Association class III or IV status" because these agents
expand intravascular volume and may exacerbate heart failure.2,3 Despite
these strong admonitions, limited data suggest that metformin is often prescribed
in patients with contraindications.4- 6
Our objective was to determine the rates of prescription of these medications
in a contemporary national sample of Medicare patients discharged after hospitalization
for heart failure. Given the high prevalence of diabetes among populations
with heart failure, this represents an unexplored area with significant potential
implications for the safety of patients with both conditions.
The National Heart Care Project is a Centers for Medicare & Medicaid
Services initiative to improve the quality of care for Medicare beneficiaries
hospitalized with heart failure. This project included the construction of
2 cross-sectional samples of Medicare fee-for-service beneficiaries hospitalized
with a principal discharge diagnosis of heart failure (International Classification of Diseases, Ninth Revision, Clinical Modification codes 402.01, 402.11, 402.91, 404.01, 404.91, or 428). This method
of case identification is considered specific for heart failure.7
The first sample was obtained between April 1998 and March 1999 and
the second between July 2000 and June 2001. All identified discharges during
a 6-month period within each of the sampling frames within each of the 50
states, Washington, DC, and Puerto Rico were sorted by age, sex, race, and
treating hospital. In each state, up to 800 records were selected; a census
of records was obtained in states with fewer than 800 eligible records. Records
were reviewed in central clinical data abstraction centers for demographic
characteristics, medical history, laboratory evaluations, and medications
taken at admission and prescribed at discharge. Data quality was ensured through
the use of trained reviewers, electronic abstraction instruments, and record
The initial samples included 39 477 records from 1998-1999 and
39 405 records from 2000-2001. Patients without valid social security
numbers, those undergoing long-term hemodialysis, and those transferred to
another hospital or who left against medical advice were excluded. Only patients
with the diagnosis of diabetes in the hospital record were included. Patients
younger than 65 years and those not surviving hospitalization were excluded.
After applying these criteria, samples of 12 505 records from 1998-1999
and 13 158 records from 2000-2001 were analyzed.
A complete list of admission and discharge medications was abstracted
from the records. Patients prescribed insulin-sensitizing agents at discharge
were categorized in nonexclusive groups, including treatment with (1) metformin
(Glucophage and Glucophage XR; Bristol-Myers Squibb, New York, NY), (2) thiazolidinediones,
or (3) either agent. The only thiazolidinedione available on the US market
during the first sampling period was troglitazone (Rezulin; Parke-Davis, a
division of Warner-Lambert Co, Morris Plains, NJ). Although the manufacturer
withdrew this medication from the US market because of fatalities due to hepatic
failure in March 2000, this was between the sampling periods. By the beginning
of the second sampling period, both rosiglitazone (Avandia; GlaxoSmithKline,
Research Triangle Park, NC, approved in May 1999) and pioglitazone hydrochloride
(Actos; Takeda, Lincolnshire, Ill, approved in July 1999) were available for
use in the United States.
Medication use was assessed in subgroups of patients by demographic
characteristics and cardiovascular history. Because renal insufficiency (defined
in the metformin package insert as a serum creatinine level of ≥1.5 mg/dL
[≥133 µmol/L] in men or ≥1.4 mg/dL [≥124 µmol/L] in women)
is also considered an absolute contraindication to metformin use,1 prescription patterns in patients with and without
renal insufficiency using this definition were also assessed. The creatinine
measurement closest to discharge was used to define renal insufficiency.
Other drug prescriptions at admission and discharge were also examined.
To determine the proportion of new prescriptions for metformin and thiazolidinediones,
admission use of both medications was assessed. Because metformin is contraindicated
in patients with heart failure requiring drug therapy,1 patterns
of prescription of drugs commonly used to treat heart failure were assessed.
Because fluid retention with thiazolidinediones is more common in patients
also treated with insulin,2,3 use
of oral agents was also assessed according to concomitant insulin prescription
The proportions of patients treated with different classes of antihyperglycemic
agents were calculated for both samples. To account for the National Heart
Care Project sampling design and to estimate the national prevalence of medication
use, probability weights, based on inverse sampling fraction for each state,
were applied. Bivariate comparisons between treatment proportions between
the 2 samples were calculated using the Wald χ2 test for categorical
variables. All statistical analyses were performed with SAS 8.2 statistical
software (SAS Institute Inc, Cary, NC).
The characteristics of the patients in the 2 samples are shown in Table 1. In 1998-1999, 7.1% of diabetic
patients were treated with metformin at hospital discharge (Table 2). By 2000-2001, metformin was prescribed to 11.2% of diabetic
patients, a relative increase of 61% (P<.001).
Use in 2000-2001 increased significantly in patients of all age groups and
other strata studied (Table 2).
Although physicians prescribed metformin less frequently for patients with
elevated serum creatinine levels, metformin was used in 4.4% of patients with
renal insufficiency in 1998-1999; this increased to 6.7% by 2000-2001.
Patterns of use of thiazolidinediones during 1998-1999 were similar
to those of metformin. In the remeasurement sample, however, thiazolidinedione
prescriptions increased to 16.1%, representing a relative increase of 124%
(P<.001). By 2000-2001, the use of these agents
increased significantly in all patient strata.
By 2000-2001, 24.4% of all patients and more than 20% of patients in
nearly all strata were treated with either medication. Exceptions included
patients older than 85 years (16.6% treated with either drug) and those with
severe left ventricular systolic dysfunction (18.8%).
Most patients discharged with a prescription for metformin were taking
this medication at admission (88% in 1998-1999 and 87% in 2000-2001, Table 3). At discharge, 98% of patients
treated with metformin were also prescribed at least one drug commonly used
for the treatment of heart failure. More than half of these patients treated
were also discharged with a prescription for a sulfonylurea in both periods.
Similarly, more than 80% of patients discharged with a thiazolidimedione
prescription were taking a medication in this class at admission. Although
concomitant insulin use decreased between sampling frames, 41% of patients
discharged with a thiazolidinedione prescription were also discharged with
an insulin prescription in 2000-2001. In contrast, concomitant sulfonylurea
prescription at discharge among these patients increased from 35% in 1998-1999
to 40% in 2000-2001.
In this study of nationally representative samples of hospitalized heart
failure patients with diabetes, the use of antihyperglycemic agents not recommended
in the package insert in this setting was widespread. This practice nearly
doubled between 1998-1999 and 2000-2001, reflecting increasing popularity
and familiarity with these agents by prescribing physicians. Use of these
drugs in patient populations at particularly high risk for adverse events
(ie, metformin with renal dysfunction or thiazolidinediones with insulin)
According to the prescribing information approved by the US Food and
Drug Administration (FDA), metformin is contraindicated in patients with heart
failure requiring drug therapy and thiazolidinediones are not recommended
for patients with advanced heart failure.1- 3 There
are several possible explanations for the discordance between these recommendations
and the findings of this study. First, some physicians may not be aware of
the potential risks of prescribing these medications to patients with heart
failure. Second, clinicians may believe that the importance of glucose control
overrides the potential risks or that the risks in well-compensated heart
failure are reduced. In the case of metformin, this perception may be in part
attributable to the "black box" warning, which states that "patients with
congestive heart failure requiring pharmacologic management, in particular
those with unstable or acute congestive heart failure who are at risk of hypoperfusion
and hypoxemia, are at increased risk for lactic acidosis."1 Finally,
physicians may believe that the risks of these drugs are overestimated. Ultimately,
however, the assessment of drug safety should emerge from rigorous studies.
Therefore, additional data are urgently needed to clarify the optimal approach
to the treatment of diabetes in patients with heart failure.
The safety concerns surrounding metformin result from its association
with life-threatening lactic acidosis. Although the risk of this complication
is reportedly low in general populations (between 2 and 10 per 100 000
person-years of treatment), these precautions for metformin likely reflect
reports of higher risks in patients with heart failure and renal insufficiency.8- 13
A recent systematic review by the Cochrane Collaboration13 of
176 clinical studies with more than 35 000 patient-years of follow-up
identified no cases of lactic acidosis. Although some of the studies in this
review permitted the inclusion of patients with cardiovascular conditions
and renal insufficiency, there was inadequate information to estimate the
risk of lactic acidosis in these high-risk subgroups. The current recommendations
for metformin use may stem not only from the uncertainty surrounding the risk
of lactic acidosis in patients with heart failure but also from experience
with phenformin hydrochloride, which was withdrawn from the market due to
a relatively high risk of lactic acidosis.
The recognition of the potential hazards of fluid retention due to thiazolidinediones
in patients with heart failure has increased over time. The original package
insert for troglitazone included a statement that use in patients with advanced
heart failure symptoms should be considered only if the perceived benefits
of treatment outweighed the potential risks.14 The
package inserts for rosiglitazone and pioglitazone were changed in 2000 to
include a specific caution to avoid use in patients with class III and IV
heart failure unless the benefit was judged to outweigh the risk.15 In response to postmarketing events, the FDA strengthened
this precaution to a warning in April 2002.16
Because fluid retention with thiazolidinedione use is not accompanied
by decreased left ventricular systolic performance,17 the
clinical relevance of this phenomenon is not clear. A recent series from a
single heart failure clinic found that although 17% of patients taking thiazolidinediones
developed significant fluid retention after initiation of therapy, this was
manifested as peripheral edema and usually resolved promptly with discontinuation
of use of the drug.18
Both metformin and thiazolidinediones exert positive effects on cardiovascular
risk factors.17,19,20 Metformin
also decreases the risk of macrovascular events in overweight patients with
newly diagnosed diabetes.21 Despite these potential
advantages, however, the FDA does not support the current patterns of use
of insulin sensitizers in patients with diabetes and coexisting heart failure.
Until further data clarify the safety and effectiveness of these medications
in patients with heart failure, clinicians must practice contrary to explicit
recommendations or use a more limited armamentarium in the treatment of diabetes
in patients with heart failure.
National guidelines do not focus extensively on the treatment of diabetes
in heart failure patients. The guidelines for diabetes treatment by the American
Diabetes Association mention that both metformin and thiazolidinediones are
not recommended in many patients with heart failure.22 The
American College of Cardiology/American Heart Association guidelines for the
treatment of heart failure mention that thiazolidinediones should be "used
with caution" in patients with heart failure but do not note the metformin
contraindication.23 Future guidelines should
provide more explicit guidance in defining the safe treatment of diabetes
in patients with heart failure. Greater clinical safety data would provide
a stronger basis for such recommendations.
Because all patients in this study were hospitalized, we cannot characterize
patterns of prescription of insulin sensitizers in a broader range of patients
with heart failure and diabetes. Patients hospitalized with heart failure,
however, are likely at higher risk for complications from these medications
than are ambulatory patients.
Because New York Heart Association classification is not recorded consistently
and reliably in hospital records, it is not possible to determine the clinician's
assessment of the patient's functional status, which is relevant to the use
of thiazolidinediones. However, patients with New York Heart Association class
I or II symptoms are unlikely to be hospitalized for heart failure.
In conclusion, this study demonstrates that elderly, diabetic patients
hospitalized with heart failure are commonly treated with antihyperglycemic
drugs that are not recommended for patients with moderate-to-severe heart
failure. There is a need for all parties interested in safe diabetes treatment
to increase awareness of the recommended approach to diabetes in patients
with heart failure. Further studies are needed to establish whether these
drugs can be used safely and effectively in patients with heart failure. Other
important questions for future study include whether patients with heart failure
and normal systolic function are at the same risk for adverse events as those
with reduced systolic function and whether the optimization of heart failure
therapy may allow the safe use of these agents in some patient groups.