Stenestrand U, Wallentin L, for the Swedish Register of Cardiac Intensive Care (RIKS-HIA) . Early Statin Treatment Following Acute Myocardial Infarction and 1-Year Survival. JAMA. 2001;285(4):430-436. doi:10.1001/jama.285.4.430
Author Affiliations: Department of Cardiology, University Hospital of Linköping, Linköping, Sweden (Dr Stenestrand); Department of Cardiology, University Hospital of Uppsala, Uppsala, Sweden (Dr Wallentin).
Context Randomized trials have established statin treatment as secondary prevention
in coronary artery disease, but it is unclear whether early treatment with
statins following acute myocardial infarction (AMI) influences survival.
Objective To evaluate the association between statin treatment initiated before
or at the time of hospital discharge and 1-year mortality after AMI.
Design and Setting Prospective cohort study using data from the Swedish Register of Cardiac
Intensive Care on patients admitted to the coronary care units of 58 Swedish
hospitals in 1995-1998. One-year mortality data were obtained from the Swedish
National Cause of Death Register.
Patients Patients with first registry-recorded AMI who were younger than 80 years
and who were discharged alive from the hospital, including 5528 who received
statins at or before discharge and 14 071 who did not.
Main Outcome Measure Relative risk of 1-year mortality according to statin treatment.
Results At 1 year, unadjusted mortality was 9.3% (1307 deaths) in the no-statin
group and 4.0% (219 deaths) in the statin treatment group. In regression analysis
adjusting for confounding factors and propensity score for statin use, early
statin treatment was associated with a reduction in 1-year mortality (relative
risk, 0.75; 95% confidence interval, 0.63-0.89; P
= .001) in hospital survivors of AMI. This reduction in mortality was similar
among all subgroups based on age, sex, baseline characteristics, previous
disease manifestations, and medications.
Conclusions Early initiation of statin treatment in patients with AMI is associated
with reduced 1-year mortality. These results emphasize the importance of implementing
the results of randomized statin trials in unselected AMI patients.
The Scandinavian Simvastatin Survival Study,1
the Cholesterol And Recurrent Events trial,2
and the Long-term Intervention with Pravastatin in Ischaemic Disease trials3 have demonstrated that treatment with 3-hydroxy-3-methylglutaryl
coenzyme A reductase inhibitors (statins) initiated 3 to 6 months after acute
myocardial infarction (AMI) reduces mortality in patients with elevated cholesterol
levels. Among high-risk individuals with high cholesterol levels but without
previous manifestations of coronary artery disease, statin treatment as primary
prevention reduces the risk of subsequent coronary events.4,5
Even though there are indications that early statin treatment initiated during
the hospital stay for acute coronary syndromes would be beneficial,6- 9 the only
clinical trial that has studied early statin intervention in acute coronary
syndromes is the recently completed but not yet published Myocardial Ischemia
Reduction with Aggressive Cholesterol Lowering study.10
Thus, the proportion of patients with AMI who are treated with statins while
in the hospital varies among hospitals. Therefore, we investigated the relationship
between 1-year mortality and statin treatment initiated before or at the time
of hospital discharge in a large cohort of unselected patients with AMI admitted
to the coronary care units of Swedish hospitals during 1995-1998.
The Swedish Register of Cardiac Intensive Care, also known as the Register
of Information and Knowledge About Swedish Heart Intensive Care Admissions
(RIKS-HIA), registers all patients admitted to the coronary care units of
participating hospitals. Information is reported on case record forms that
include 100 variables. On admission, 30 variables are recorded, including
information on age, sex, smoking status, hypertension, diabetes mellitus,
hyperlipidemia, previous angina pectoris, previous MI, previous coronary revascularization,
previous medications, symptoms, and electrocardiogram changes at entry and
specified time points (previous refers to events
occurring or medication started before the current admission) (Table 1). During the hospital stay, another 37 variables are recorded
regarding biochemical markers, echocardiography, reperfusion treatment, pharmacological
treatment, interventional procedures, major arrhythmias, and other major complications.
At discharge, 33 variables are recorded, including complications, diagnosis,
and outcomes during the hospital stay; risk assessment with stress test, coronary
angiography, and revascularization procedures; and medications at discharge.
The criteria for the diagnosis of AMI are standardized and identical
for all participating hospitals using World Health Organization criteria11 and a measurement indicating twice the upper limit
of normal of an appropriate biochemical marker (usually creatine kinase–MB
protein concentration) as the biochemical criterion. Electrocardiograms are
evaluated for presence or development of Q waves, ST-segment changes, T-wave
inversions, or left bundle-branch block.
Source data verification was continuously performed by comparison of
the registry information to the hospital's patient records in 50 randomly
selected patients in 10 hospitals every year by an external monitor. In the
first 1004 computer forms from 21 hospitals comprising 92 368 variables,
there was 94% agreement overall between the registered information and the
source data in the patient records.
One-year mortality data were obtained by merging the RIKS-HIA database
with the National Cause of Death Register, which includes the vital status
of all Swedish citizens in 1995-1999.
All patients for whom data were entered into the RIKS-HIA were informed
of their participation in the registry (patients could request to be excluded
from the registry) and the long-term follow-up. The registry and the merging
with registries were approved by the National Board of Health and Welfare
and the Swedish Data Inspection Board.
Comparisons between different patient strata and different categories
of hospitals were analyzed by χ2 tests for categorical variables
and by the t test for continuous variables. Bivariate
analyses and multiple covariate Cox regression analyses12
were used to identify any variable with a significant influence on mortality.
To avoid the influence of early mortality before the opportunity of exposure
to the studied treatment, the primary analysis was performed only in patients
who were alive at discharge. To identify any possible hazard associated with
the studied treatment, the analysis was also performed in all admitted patients.
The analyses were also performed for 30-, 60-, and 90-day survivors to allow
for even longer periods of early mortality in patients who could have been
perceived to have a too-short life expectancy to benefit from statin treatment.
In all analyses, statin treatment was defined as statins prescribed at the
time of discharge from the hospital, and no statin treatment was defined as
no statins prescribed at discharge.
Propensity analysis13 was performed regarding
the probability of statin use. For each patient, a propensity score indicating
the likelihood of having statins prescribed at discharge was calculated by
forward logistic regression analysis and included 42 covariates: age, sex,
smoking status, previous MI, previous percutaneous coronary intervention or
coronary artery bypass graft surgery, history of diabetes mellitus, history
of hypertension, circulatory arrest at arrival, medications before study entry
(including angiotensin-converting enzyme inhibitors, anticoagulants, aspirin, β-blockers,
calcium channel inhibitors, digitalis, diuretics, long-acting nitrates, and
statins), acute reperfusion treatment, intravenous β-blockers, intravenous
or subcutaneous anticoagulants, intravenous nitroglycerin, atrial fibrillation,
congestive heart failure, reinfarction, stress test administration, echocardiography,
coronary angiography, medications at hospital discharge (including angiotensin-converting
enzyme inhibitors, oral anticoagulants, aspirin, β-blockers, calcium
channel inhibitors, digitalis, diuretics, and long-acting nitrates), revascularization
before discharge, type of hospital (primary, secondary, or tertiary), hospital
size (number of AMI admissions per year <100, 100-199, 200-399, or ≥400),
teaching hospital status, presence of catheterization laboratory in the hospital,
admission year, and each hospital's statin prescription rate (divided into
<25%, 25%-35%, and >35%) among AMI patients younger than 80 years old.
Goodness of fit of the propensity score was evaluated by the c statistic and the Hosmer-Lemeshow test. Cholesterol levels were not
among the compulsory variables denoted in the RIKS-HIA registry and therefore
could not be included.
All these variables together with individual propensity scores were
forced into the multiple covariate Cox regression analyses evaluating the
association of statins at discharge with 1-year mortality. The same model
was also used in subgroups based on age and dichotomized risk factors. All
statistical analyses were performed using SPSS Version 10.0 software (SPSS
Inc, Chicago, Ill).
Nineteen hospitals participated in the registry in 1995, 32 hospitals
participated in 1996, 46 hospitals in 1997, and 58 hospitals in 1998. Data
for this study were collected for 1995-1998 and included 137 262 admissions
to 58 coronary care units. Of these admissions, 30 240 were patients
with a first recorded admission for AMI during the registration period (considered
the index event for follow-up) and had information recorded about statin treatment
at discharge. Patients who died before hospital discharge (n = 1303) were
excluded from the main analysis. Because the rate of statin treatment at discharge
among patients aged 80 years or older (n = 5983) was only 4% (n = 210) and
because of the increased risk of concomitant disease among these elderly patients,
we excluded patients aged 80 years or older from the analyses. An additional
3355 patients were excluded because of incomplete data. The remaining 19 599
hospital discharge survivors of a first registered AMI who were younger than
80 years and had complete data constituted the patient population for this
Baseline characteristics of patients with and without statin treatment
at discharge are shown in Table 1.
Statin-treated patients were younger and less likely to have diabetes mellitus
or be taking medications indicative of previous or in-hospital heart failure.
Patients who received statins were more often smokers and had more previous
MIs, more previous percutaneous coronary interventions, and more coronary
artery bypass graft surgery, as well as more aspirin and β-blocker medication
(indicative of more previous manifestations of coronary artery disease). Statin
treatment was also associated with a higher rate of acute reperfusion, as
well as intravenous or subcutaneous anticoagulant treatment (Table 2).
The proportion of patients in each hospital who received different treatments
or who developed complications (eg, congestive heart failure) in the hospital
that were known to influence or were associated with mortality varied substantially
among the 58 participating hospitals (data not shown). For instance, the median
(10th-90th percentile) rates were 38% (32%-45%) for acute reperfusion, 29%
(14%-50%) for intravenous β-blocker use, 7% (2%-33%) for coronary angiography,
and 2% (0.3%-29%) for revascularization before discharge. For statin treatment
at discharge, the median rate among hospitals was 27%, and the proportion
of treated patients was 12% for the lowest vs 48% for the highest decile.
In the propensity score analysis,13 variables
most strongly correlated with prescription of statins at discharge were (in
descending strength of association) statin treatment before study entry, later
admission year, younger age, higher statin prescription rate at the hospital, β-blocker
treatment at discharge, predischarge administration of stress test, nonteaching
hospital status, predischarge echocardiogram, absence of atrial fibrillation,
larger hospital, and performance of coronary angiography. The goodness of
fit of the propensity score was evaluated by the c
statistic (area under the receiver operating characteristic curve, 0.81),
and the Hosmer-Lemeshow test (χ2 = 21.87; P = .005). The propensity score itself was associated with improved
survival (P<.001) and, hence, identified patients
with lower risk. When the means and SDs of the propensity scores were compared
within quintiles between patients receiving and not receiving statins at discharge,
the groups were comparable within the lower 4 strata (Table 3).
Among the 14 071 patients without statin treatment, the unadjusted
1-year mortality was 9.3% (n = 1307) compared with 4.0% (n = 219) among the
5528 patients with statin treatment. In Cox regression analysis, adjusting
for the 43 covariates including the propensity score for statin use, statin
treatment at discharge was associated with a reduction in 1-year mortality
(3.7% vs 5.0%; relative risk [RR], 0.75; 95% confidence interval [CI], 0.63-0.89; P = .001) (Figure 1, Table 4). When the propensity score variable
was excluded from analysis, the RR was 0.73 (95% CI, 0.62-0.87; P<.001). Adjusting for the propensity score only, statin treatment
was associated with improved 1-year survival (RR of 1-year mortality, 0.78;
95% CI, 0.67-0.91; P = .001). Adjusting for propensity
score and all other covariates for the 15 680 patients in the lower 4
quintiles, the RR was 0.71 (95% CI, 0.58-0.86; P<.001).
In the individual propensity score quintiles, the adjusted RR for 1-year mortality
was between 0.66 and 0.83, although it did not reach statistical significance
in all subgroups.
Mean (SD) hospitalization time was 7.06 (5.25) days and the median was
6 days. The absence of any possible hazards associated with early treatment
with statins was supported by the even larger relative benefit observed when
all patients from the time of admission were included (Table 5). The consistency of the result, after exclusion of patients
possibly perceived to have a too-short life expectancy to indicate initiation
of statin treatment, was supported by similar reductions in RR when the analysis
was restricted to survivors at 30, 60, and 90 days (Table 5).
The benefit associated with early initiation of statin treatment was
also supported by the results of the Cox regression analyses in subgroups
based on the most important factors related to mortality. The reduction in
1-year mortality was most pronounced in the group aged 60 to 69 years (RR,
0.50; 95% CI, 0.36-0.69) and was similar but less prominent in older and younger
patients. Furthermore, statin treatment was associated with lower 1-year mortality
regardless of sex, diabetes mellitus, prior MI, congestive heart failure,
ST-segment or non–ST-segment elevation MI, treatment with digitalis,
diuretics, β-blockers, and subcutaneous or intravenous anticoagulants,
and whether the patient was administered a predischarge exercise test (data
To ensure that inclusion of patients with missing data would not alter
the results, missing values for any variable were substituted with "no" and
"yes," respectively, in 2 separate analyses of all 22 952 AMI patients
younger than 80 years old. Statin treatment at discharge was associated with
decreased 1-year mortality in both alternatives (RR, 0.76; 95% CI, 0.65-0.88; P<.001 and RR, 0.79; 95% CI, 0.68-0.91; P = .002, respectively).
Finally, RIKS-HIA does not indicate what type of statin is prescribed.
Instead, data on all drugs supplied from wholesalers to all Swedish pharmacies
were obtained from the National Corporation of Swedish Pharmacies, which holds
a monopoly on pharmaceutical drugs in Sweden. At the time of the analysis,
in 1995-1998, the sale of statins, measured in daily doses, in Sweden were:
simvastatin, 74%; pravastatin, 14%; atorvastatin, 7%; and fluvastatin, 5%
(Karolina Antonov, PhD, written communication, November 11, 2000).
The patient cohort in this study included unselected consecutive survivors
of a first recorded MI from a large number of different hospital types. The
only exclusion criterion used was age 80 years and older, since the recorded
information focused on coronary artery disease and, therefore, data on other
important comorbidities influencing survival in the elderly population might
not be available. There were no exclusions due to presence or absence of specific
risk factors, comorbidities, anticipated adverse effects, participation in
clinical trials, or contraindications to certain medications. The representativeness
of the cohort was also strengthened by the inclusion of all patients with
MI from the general population at centers with different levels of care, including
two thirds of the hospitals within an entire country. Compared with the National
Registry of Myocardial Infarction14 in the
United States, the Swedish registry does not focus on thrombolytic therapy
but includes all types of MI patients and a wider selection of background
characteristics and treatments, which allows for adjustment of a large number
of confounding factors.
The reason for the choice of patients who were alive at hospital discharge
and the evaluation of patients who survived at 30, 60, and 90 days after discharge
was to avoid the bias of overloading the nontreated group with patients who
died before they had the opportunity to receive statin treatment or were perceived
to have a too-short life expectancy to initiate treatment. Another consideration
is whether early statin treatment might be associated with increased hospital
mortality, which would not be apparent among those who were alive at discharge.
However, considering that the results concerning the reduced risk associated
with early statin treatment also were consistent when we evaluated all admitted
patients as well as 30-, 60-, and 90-day survivors, we believe these possible
sources of bias are unlikely to invalidate the major result.
The large variations in certain therapies among hospitals within 1 country,
despite similar treatment guidelines, is similar to previous reports of differences
among hospitals in different countries.15,16
There is consensus concerning some therapies, such as use of reperfusion for
patients with ST-segment elevation and use of aspirin and β-blockade
at discharge, that were used in fairly similar proportions in all hospitals
in this study. However, regarding therapies for which the evidence is weaker
and guidelines less clear, such as starting statin treatment in the hospital,
there were considerably larger variations than can be explained by different
patient populations in different hospitals. These large differences provided
an opportunity to evaluate association with outcome of the almost random variation
of some treatments among hospitals in combination with a multivariate adjustment
for differences in other factors.
Despite evaluation of the large variation in indications for statin
treatment among hospitals and exclusion of early deaths, a selection bias
remained, as clinicians seemed to have avoided use of statins in patients
perceived to have a short life expectancy. This was illustrated by the differences
in baseline characteristics, in which statin-treated patients were younger
and were less likely to have diabetes and signs of or medications for congestive
heart failure but more likely to have prior MI, revascularization, and treatment
for angina. Therefore, the difference in mortality by early statin treatment
needed to be documented by a combination multiple covariate propensity and
Cox regression analysis forcing all available possibly confounding factors
into the models to compensate for these baseline differences.
The propensity analyses were included to minimize confounding by physician
preferences for statin treatment in lower-risk patients. The comparison of
patient groups stratified by or with similar propensity scores should provide
matched cohorts concerning recorded covariates. Because statin treatment remained
associated with lower RR of 1-year mortality within strata with equal probability
of statin prescription, the hypothesis that early statin treatment is associated
with improved survival after AMI is supported. The propensity score was also
included in all multiple covariate Cox regression survival analyses to reduce
confounding by factors associated with statin treatment as well as with improved
survival. Using this approach, the main results were similar in all subgroup
analyses but 1, further suggesting that the improved survival is related to
statin treatment rather than patient selection.
This study has several limitations. First, although these results come
from a large cohort and adjustment was performed for a large number of confounding
covariates, a registry study cannot adjust for all confounders and, hence,
cannot replace a randomized controlled trial. Thus, the effects of in-hospital
initiation of statin treatment vs initiation several months later need to
be demonstrated in the ongoing randomized trials of early vs later start of
statin treatment in patients with AMI.
Second, the indication for start of statin treatment was highly variable
among and within hospitals, and lipid measurements were not part of the compulsory
data set. However, the recommended indication for statin treatment during
the study was elevated cholesterol level (>5.2 mmol/L [200 mg/dL]) or low-density
lipoprotein cholesterol level (>3.0 mmol/L [115 mg/dL]).17,18
Accordingly, we believe the study illustrates the potential benefits of early
statin use in patients with an appropriate indication for long-term secondary
prevention with statin treatment rather than use of statins in any patient
with AMI regardless of lipid levels.
Third, we were unable to determine whether patients who were prescribed
statins at hospital discharge continued their medication during the following
year, or whether patients who did not receive statins at discharge were given
them during the subsequent months. If, as demonstrated in randomized trials,
statin treatment reduces mortality, both of these considerations (discontinuation
of statins among treated patients and institution of statins among those not
treated at hospital discharge) would most likely reduce the estimated benefit
of statin treatment that we observed.
Fourth, we evaluated all-cause mortality rather than cardiovascular
mortality because cause-of-death data were only available for 1995-1997, not
for 1998-1999. However, when the analysis was performed including only patients
with available data on cause of death, the statin and no-statin groups had
similar rates of mortality due to cardiac death (77% and 73%, respectively).
These results are the first to our knowledge to indicate that initiation
of statin treatment at or before hospital discharge in survivors of AMI is
associated with reduced mortality during the subsequent year. The relative
reduction in mortality is in accordance with previous experiences in randomized
trials,1- 3 with
an apparent early reduction of the event rates in the statin treatment group.
Some studies have suggested that the effects of statins, such as improved
endothelial function,7 plaque regression, and
less major cardiovascular events, occur within 6 months.6
Statin treatment might also be associated with a reduction in inflammatory
activity, as indicated by a decrease in C-reactive protein level,19,20 which might contribute to a lower
risk of subsequent events.21,22
Although it is possible that our findings are related to an early plaque-stabilizing
effect of statin treatment that is most effective in patients with a recent
acute event, the timing of this benefit needs to be verified in randomized
trials that focus on a comparison between early and later initiation of statin
treatment (eg, the Myocardial Ischemia Reduction with Aggressive Cholesterol
However, considering that the potential benefits seem substantial and
adverse effects few, our findings suggest that initiation of statin treatment
before or at the time of hospital discharge should be recommended for AMI
survivors with total cholesterol or low-density lipoprotein cholesterol levels
above the current guideline levels for statin treatment as secondary prevention.17,18