Hospital outcomes according to statin therapy group. Asterisk indicates P < .001, Y/N vs Y/Y; CHF, congestive heart failure; MI, myocardial infarction; N/N, patients not previously receiving statins who did not receive statins in the first 24 hours of hospitalization; VT/VF, ventricular tachycardia or ventricular fibrillation; Y/N, patients previously receiving statins who did not receive statins in the first 24 hours of hospitalization; and Y/Y, patients previously receiving statins who received statins in the first 24 hours of hospitalization.
Spencer FA, Fonarow GC, Frederick PD, Wright RS, Every N, Goldberg RJ, Gore JM, Dong W, Becker RC, French W, National Registry of Myocardial Infarction Investigators. Early Withdrawal of Statin Therapy in Patients With Non–ST-Segment Elevation Myocardial InfarctionNational Registry of Myocardial Infarction. Arch Intern Med. 2004;164(19):2162–2168. doi:10.1001/archinte.164.19.2162
There is increasing interest in the non–lipid-lowering effects of statins and their effect on outcomes in patients with acute coronary syndrome. It has been suggested that withdrawal of statin therapy during an acute coronary syndrome may attenuate any benefits of pretreatment, thereby providing indirect evidence of the importance of their non–lipid-lowering effects.
This observational study compared the demographic and clinical characteristics and hospital outcomes in patients with non–ST-segment elevation myocardial infarction enrolled in the National Registry of Myocardial Infarction 4. Comparison groups consisted of patients previously receiving statins who also received statins within 24 hours of hospital admission (n = 9001), patients previously using statins in whom therapy was discontinued (n = 4870), and patients who did not receive statins at any time before or during hospitalization (n = 54 635).
Of 13 871 patients receiving statins before hospital admission, 35.1% had treatment withdrawn during the first 24 hours of hospitalization. These patients had increased hospital morbidity and mortality rates relative to patients in whom therapy was continued, with higher rates of heart failure, ventricular arrhythmias, shock, and death. In multivariate analyses, these patients were at statistically significant increased risk of hospital death compared with those continuing statin therapy and at similar risk compared with those not receiving statins before or during hospitalization.
Withdrawal of statin therapy in the first 24 hours of hospitalization for non–ST-segment elevation myocardial infarction is associated with worse hospital outcomes. In the absence of data from randomized clinical trials, our findings suggest that statin therapy should be continued during hospitalization for myocardial infarction unless strongly contraindicated.
Clinical trial data that support the efficacy of lipid-lowering agents, particularly statins, in the primary and secondary prevention of acute coronary syndromes (ACSs) are substantial.1- 3 In addition, there is increasing evidence that these agents exert a variety of effects at the vascular and cellular levels associated with, and possibly independent of, their lipid-lowering abilities. In vitro studies4- 13 have demonstrated an effect of statin treatment on platelet aggregation, smooth muscle cell proliferation, inflammation, and endothelial cell function.
Given these diverse pathophysiologic effects, the ability of statins to favorably affect the early course of ACSs has been of increasing interest. In the MIRACL (Myocardial Ischemia Reduction With Aggressive Cholesterol Lowering) trial,14 early treatment of patients with acute myocardial infarction (AMI) with statins was associated with a small but statistically significant decrease in the primary trial composite end point. Subgroup analysis revealed that a reduction in recurrent ischemic events was responsible for most of the observed benefit. Other observational studies have demonstrated favorable long-term outcomes in patients with ACSs discharged from the hospital while receiving statin therapy.15,16 Questions remain, however, about whether this is because of a favorable effect on the acute hospital course or is secondary to enhanced long-term compliance.
Some evidence also suggests that withdrawal of statin therapy during an acute coronary event is potentially harmful. In a post hoc analysis17 of the PRISM (Platelet Receptor Inhibition in Ischemic Syndrome Management) database, risk of death and nonfatal AMI during 30-day follow-up increased almost 3-fold for patients taking statins who subsequently had them withdrawn compared with patients continuing statin therapy. In addition, initiation of statin therapy after the onset of an ACS seemed less effective than pretreatment with statins before the onset of symptoms. These data, although intriguing, need to be interpreted cautiously because the number of patients discontinuing statin therapy in this analysis was small.
Using data from a large observational database, we examined the patient and clinical characteristics and hospital outcomes associated with withdrawal of statin therapy in patients with non–ST-segment elevation MI.
The National Registry of Myocardial Infarction 4 (NRMI-4) is a cross-sectional observational database of patients hospitalized with AMI throughout the United States. The data collection process used in this study and its quality control features have been described elsewhere.18 To minimize the effect of patient transfers from registry hospitals on data quality, we restricted the study population to patients enrolled in hospitals with interventional capabilities. Data from 300 823 patients enrolled in 1230 NRMI-4 hospitals between July 1, 2000, and January 31, 2002, were used. There were 183 570 patients reported from the 459 hospitals (37.3%) with interventional capacity contributing data to the NRMI-4. Participating hospitals were larger, more likely to be affiliated with a medical school, more likely to be urban, and more likely to be certified by the Joint Commission on Accreditation of Healthcare Organizations.
Participation in the study was voluntary, and hospitals were encouraged to enter consecutive patients with AMI irrespective of treatment strategy or outcome. Individual institutions participating in the NRMI designated a local coordinator who was responsible for abstracting and forwarding data to the central registry.
To be included in the NRMI, patients must have had an AMI before hospital discharge according to the prescribed criteria,18 including a supportive clinical history, increases in serial enzyme levels, and the presence of electrocardiographic findings consistent with AMI. Of the 183 570 patients enrolled in interventional hospitals, 62 595 transferred-in patients and 4284 transferred-out patients were excluded from the analysis, as were patients with missing electrocardiograms (n = 196). From the remaining 116 495 patients, we limited our analysis to patients with non–ST-segment elevation MI (n = 78 224) to minimize the potential bias associated with inclusion of patients with ST-segment elevation MI and to enhance comparability with the PRISM analysis.17 Patients were categorized as “previously receiving statin therapy” if statins had been prescribed during the 24 hours before hospital arrival. If patients were previously receiving statin therapy but were not prescribed statins within 24 hours of hospital admission, the criteria for discontinuation of statin therapy was met (group Y/N). These patients may have received statins later during their hospital course, but for the purposes of this analysis they were still included in the Y/N group. Patients previously receiving statins who also received them early in their hospital course (within 24 hours) were classified as continued users (group Y/Y), and patients not previously receiving statins who did not receive them early during hospitalization were classified as group N/N. These patients may also have received statins after the initial 24 hours of hospitalization but continued to be included in the N/N group. Patients not previously receiving statins who began statin therapy within 24 hours of hospital admission (group N/Y; n = 9718) were excluded from the present analysis. Given the size of this database, the multiple comparisons possible, and the potential differences in the pathogenesis of “statin withdrawal” vs “early statin initiation,” this cohort will be studied and reported on separately.
Differences in the distribution of demographic, clinical, and treatment characteristics among patients in whom statin therapy was discontinued (Y/N), those continuing statin therapy (Y/Y), and those not receiving statins during either period (N/N) were examined using χ2 tests for categorical variables and analysis of variance for continuous variables. Our primary end point was hospital death. Secondary outcomes were in-hospital events of recurrent MI, congestive heart failure, cardiogenic shock, ventricular tachycardia or ventricular fibrillation, and cardiac arrest.
To control for differences in potentially confounding variables associated with hospital death rates, we performed a logistic regression analysis. In the first model, we controlled for age, sex, race, insurance provider type, hospital type (urban vs nonurban), AMI volume, geographic region, medical history variables (history of angina, MI, diabetes mellitus, congestive heart failure, smoking, percutaneous transluminal coronary angioplasty, and coronary artery bypass graft surgery), hospital admission characteristics (prehospital delay time, Killip class on initial evaluation, and systolic blood pressure), and AMI-associated characteristics (location and ejection fraction). In addition to the previously described factors, preadmission medications (angiotensin-converting enzyme inhibitors, aspirin, angiotensin II receptor blockers, antithrombin agents, β-blockers, calcium channel blockers, platelet glycoprotein IIb/IIIa receptor inhibitors, low-molecular-weight heparins, and other lipid-lowering agents) and in-hospital therapies (all of the above plus thrombolytic therapy and other reperfusion modalities) were subsequently added to the regression models.
We also developed a propensity model to identify factors associated with early statin use (or nonuse) in the hospital. Subsets of the Y/Y and Y/N therapy groups were matched according to age, sex, race, insurance provider, medical history variables (diabetes mellitus, smoking, angina, congestive heart failure, MI, percutaneous transluminal coronary angioplasty, and coronary artery bypass graft surgery), hospital admission characteristics (pulse rate, systolic blood pressure, Killip class, and chest pain), hospital characteristics (urban vs nonurban and AMI volume), geographic region, and selected preadmission medications. Similar matching was performed for subsets of the Y/N and N/N groups. For the matched analysis, differences between matched pairs were evaluated using the Wilcoxon signed rank test for continuous data and the McNemar test for binary data. Incidence rates of the primary and secondary hospital outcomes for the matched population subsets were calculated. Propensity score was then added to the logistic regression analyses to complete the final regression model.
A total of 78 224 patients with non–ST-segment elevation MI were enrolled in NRMI-4 interventional hospitals; of these patients, 13 871 (17.7%) were receiving statin therapy before hospital admission. Of patients receiving statin therapy before admission, treatment with statin agents was withdrawn in 4870 (35.1%).
Baseline characteristics of patients in whom statin therapy was discontinued were different from those in whom therapy was continued (Table 1). Patients in whom statin therapy was withdrawn were slightly older and were more likely to be female but had lower rates of most indicators of previous coronary artery disease and comorbidities. Patients who never received statins (before or during the initial 24 hours of hospitalization) were older and were more likely to be female but had fewer indicators of coronary artery disease and comorbidities than patients in the other therapy groups (Table 1).
Patients in whom statin therapy was discontinued were slightly more ill on hospital admission; an increased proportion of these patients were in an advanced Killip class, had a pulse rate greater than 100/min, had a systolic blood pressure less than 90 mm Hg, or had an anterior AMI compared with patients in whom statin therapy was continued. Patients in whom statin therapy was discontinued were slightly more likely to be treated in an urban hospital or a hospital with a large AMI volume than the other treatment groups. In addition, patients in whom statin therapy was discontinued were statistically significantly less likely to receive other effective cardiac therapies in the first 24 hours of hospital presentation.
Propensity modeling identified several variables that independently predicted the likelihood of continuing statin therapy in the hospital (or, conversely, of having it withdrawn). The model C statistic for the Y/Y vs Y/N matching was 0.59. The prevalence of each element in the matched treatment groups is given in Table 2. The Hosmer-Lemeshow goodness-of-fit statistic (P = .14) suggested that the model’s estimates fit the data at an acceptable level. A similar model was created matching patients in the Y/N and N/N groups for the same data elements (data not shown). Propensity scores resulting from this analysis were used in subsequent multivariate regression analyses.
Patients in whom statin therapy was discontinued experienced statistically significant increased rates of hospital death (12.5% vs 4.9%), heart failure or pulmonary edema, cardiogenic shock, ventricular tachycardia or ventricular fibrillation, and cardiac arrest than patients in whom statin therapy was continued (Figure). This population also had slightly increased rates of heart failure or pulmonary edema, cardiac arrest, ventricular tachycardia or ventricular fibrillation, and cardiogenic shock compared with patients who never received statins (N/N group). Analysis of these hospital outcomes in subsets of each group matched by propensity to receive statins in the first 24 hours (using the aforementioned demographic, clinical, and hospital characteristics) revealed no statistically significant attenuation of these findings (Table 3).
Patients previously receiving statin therapy who continued therapy during the first 24 hours of hospitalization had a statistically significant lower incidence of hospital death, cardiac arrest, and cardiogenic shock than patients never receiving statin therapy. After matching for the previously described variables, congestive heart failure and ventricular tachycardia or ventricular fibrillation were also statistically significantly less likely to occur in patients who continued statin therapy (data not shown).
After adjusting for potentially confounding demographic and clinical variables, patients in whom statin therapy was withdrawn continued to have a statistically significant increased likelihood of dying during hospitalization relative to patients continuing statin therapy (hazards ratio, 2.34; 95% confidence interval, 2.04-2.69) (Table 4). Further adjustment for hospital admission characteristics, receipt of prehospital and in-hospital therapies, and propensity score attenuated this risk slightly (hazards ratio, 1.83; 95% confidence interval, 1.58-2.13). Repeating these multivariate analyses after excluding patients who died within the first 24 hours of hospital admission or who had missing data on length of hospital stay and further adjusting for the receipt of any interventional procedures (cardiac catheterization, percutaneous transluminal coronary angioplasty, or coronary artery bypass graft surgery) did not appreciably change these results.
Compared with patients never receiving statins, patients pretreated with statins who had them discontinued early in the hospitalization had a slightly increased risk of dying during hospitalization after adjustment for several demographic and clinical factors (hazards ratio, 1.15; 95% confidence interval, 1.04-1.27). Additional analysis controlling for use of hospital medications and propensity score attenuated this risk (hazards ratio, 1.03; 95% confidence interval, 0.93-1.15) (Table 4). Repeating the same multivariate analyses after excluding patients who died within the first 24 hours of hospital admission or who had missing data on length of stay and further adjustment for administration of any interventional procedures did not appreciably change these results.
Of more than 75 000 patients hospitalized with non–ST-segment elevation MI, only 18% were receiving statin therapy before hospital admission. Patients who continued statin therapy had the lowest rates of hospital complications and mortality relative to other therapy groups. These findings suggest a potential benefit to pretreatment with statins. Hospital outcomes in the group in whom statin therapy was discontinued were similar to or worse than those in patients who had not received statin pretreatment. These associations remained after multivariate analysis controlling for potential confounding factors that differed between respective comparison groups.
Abrogation of the protective benefit of statin pretreatment by early withdrawal of therapy was initially described in a recent substudy analysis of the PRISM database.17 Despite a lack of effect on serum cholesterol levels during the initial 72 hours of hospitalization for AMI, patients in whom statin therapy was discontinued had an almost 3-fold higher event rate (death and nonfatal MI) than patients continuing therapy. Although limited by small sample size and few events, this association was statistically significant and seemed to be driven by an increase in the occurrence of early adverse events (0-7 days). The results of the present study, which was significantly larger and conducted using data from a more generalizable national registry, seem to confirm these findings and extend them to a broader patient population. Our data demonstrate that patients who discontinued statin use during hospitalization for AMI had an approximately 2-fold increase in hospital death rates and a statistically significant increased risk of other early clinical complications. In addition, patients who discontinued statin use had similar in-hospital mortality rates and higher rates of certain complications compared with patients who had not been pretreated with statins.
These data suggest that withdrawal of statin therapy may have a detrimental effect on the early course of patients with ACSs. Although a causal association, or even the mechanisms by which statin withdrawal may contribute to worse outcomes, cannot be determined from this study, a variety of hypotheses have been proposed. It is generally accepted that the primary benefit of these agents is exerted through their lipid-lowering effect, which can take days to weeks to occur. Indeed, data from secondary prevention studies1- 3 with a variety of lipid-lowering agents suggest that clinical benefits may not be realized for several months to years. Nevertheless, there are emerging clinical and basic science data that statin agents modulate processes beyond lipid lowering. The pleiotropic actions of these agents may include inhibition of inflammation, modulation of endothelial function, and antithrombotic effects, all of which may affect the early course of patients with ACS.4- 13
A previous study19 identified an enhanced benefit of statin therapy in patients with coronary artery disease and evidence of heightened inflammation (characterized by elevated high-sensitivity C-reactive protein [hs-CRP] levels). Although these data have generated significant interest in the potential anti-inflammatory effects of statin therapy in the early course of ACS, supportive data remain limited. In a crossover study20 of 40 men and women with high low-density lipoprotein cholesterol levels, patients were randomly assigned to treatment with simvastatin for 14 days and then to placebo for a similar duration. Patients who initially received simvastatin incurred significant decreases in hs-CRP levels during treatment as were compared with those who received placebo. Changes in hs-CRP levels from day 14 to day 28 after discontinuation of simvastatin use was not reported. In a small study21 of atorvastatin calcium therapy in 8 healthy men with reference-range cholesterol levels, a 15% decrease in hs-CRP levels was demonstrated after only 48 hours of therapy. However, levels of hs-CRP remained suppressed for 3 days after discontinuation of therapy, suggesting lack of a measurable “rebound” in inflammation.
Other alternative mechanisms by which early statin therapy and, conversely, withdrawal might affect the early course of ACS include altered platelet aggregability, inhibition of thrombus formation and growth, reduction in matrix metalloproteinase activity, and enhanced endothelial function.4- 13 The effect of early withdrawal of statin therapy on these processes remains to be clearly delineated but could contribute to further destabilization of a recently ruptured atherosclerotic plaque.
There is evidence that statins affect endothelialfunction relatively early after initiation of therapy. The RECIFE (Reduction of Cholesterol in Ischemia and Function of Endothelium) trial9 randomized patients to receive pravastatin sodium or placebo after hospitalization for non–ST-segment elevation ACS and demonstrated an improvement in endothelial function by 6 weeks after initiation of therapy. In the aforementioned study21 of 8 healthy subjects, a significant augmentation in endothelial-dependent blood flow was detected within 24 hours of statin therapy. In addition, within 1 day of cessation of therapy, endothelial-dependent blood flow returned to below baseline values despite continuously lowered cholesterol levels. In a subsequent series of experiments, these investigators22 demonstrated in a mouse model that withdrawal of statin therapy can result in marked suppression of endothelial nitric oxide production within 2 days.
The higher incidence of adverse outcomes experienced by patients in whom statin therapy was discontinued may be related to differences in baseline characteristics, clinical presentation, or lack of treatment with other beneficial therapies. In particular, patients in whom therapy was not continued had slightly higher rates of variables consistent with worse clinical presentation (tachycardia, hypotension, and increased Killip class scores). It is possible that physicians were less likely to continue statin therapy in patients who were more ill. Indeed, this same subset of patients was less likely to receive several other effective cardiac therapies during early hospitalization. However, multivariate regression analyses controlling for these and other variables did not statistically significantly attenuate our findings. In addition, our large sample size allowed us to conduct a propensity analysis, matching patients for variables predictive of continuing statin therapy within 24 hours of hospitalization, thereby minimizing the effect of such confounding. Nevertheless, it is possible that other unmeasured variables related to discontinuation of statin therapy and the hospital outcomes being studied affected our findings.
As with any retrospective analysis, the results of this observational study have several limitations that must be acknowledged. Because this study is not a randomized trial of the discontinuation of statin treatment, no firm conclusions can be drawn about a causal relationship between withdrawal of statin therapy and worse hospital outcomes. Owing to the limitations of the database, we cannot further explore differences in outcomes on the basis of statin type, dose, and timing of therapy reinstitution. It is possible that a subset of patients who had statin therapy discontinued in the first 24 hours of hospitalization had therapy restarted later in their hospital course. We cannot estimate the size of this subset, but it seems likely that inclusion of this population in the Y/N or N/N group attenuated the magnitude of observed differences in outcomes between these groups and the Y/Y cohort. Because blood samples are not collected in the NRMI, we cannot further characterize the effect of withdrawal of statin therapy on patient subsets stratified by lipid levels, markers of inflammation, or thrombosis.
In conclusion, our data suggest that more than one third of the patients with a non–ST-segment elevation MI who are already receiving statin therapy do not continue this therapy during their early hospitalization. Withdrawal of statin therapy in the first 24 hours of presentation was associated with worse hospital outcomes. This finding may reflect a loss of benefit from pretreatment with statins or a deleterious rebound of inflammatory processes, thrombosis, or endothelial dysfunction. In the absence of data from randomized clinical trials that evaluate the effect of withdrawing statin therapy, these data suggest that therapy should be continued during hospitalization for AMI unless strongly contraindicated. Further studies that evaluate physician reasons for early discontinuation of statin therapy in patients with AMI are warranted.
Correspondence: Frederick A. Spencer, MD, Division of Cardiovascular Medicine, Department of Medicine, University of Massachusetts Medical School, 55 Lake Ave N, Worcester, MA 01655 (firstname.lastname@example.org).
Accepted for Publication: December 15, 2003.
Financial Interest: Dr Fonarow has done research for, been a consultant to, and received honoraria from Pfizer Inc, Merck & Co, and Bristol-Myers Squibb.
Funding/Support: The NRMI is supported by Genentech Inc, South San Francisco, Calif.