Congestive Heart Failure

Class effects of statins in elderly patients withcongestive heart failure: A population-based analysisStephane Rinfret, MD, MSc,a Hassan Behlouli, PhD,b Mark J. Eisenberg, MD, MPH,c,d Karin Humphries, DSc,e

Jack V. Tu, MD, PhD,f and Louise Pilote, MD, MPH, PhDb,g Vancouver, British Columbia, Toronto, Ontario, andMontreal, Quebec, Canada

Background Long-term treatment with statins reduces mortality in patients with congestive heart failure (CHF). Whetherstatin agents exert a class effect is unknown.

Methods We analyzed long-term mortality in Canadian patients aged ≥65 years who were discharged from hospitalwith a diagnosis of CHF from January 1998 to December 2002. Administrative data from Quebec, Ontario, and BritishColumbia were merged. We compared patients prescribed with atorvastatin, simvastatin, pravastatin, and lovastatin.

Results A total of 15368 patients hospitalized with a diagnosis of CHF fulfilled the inclusion criteria for this study. In thisfinal dataset, 6670 (43.4%) filled a prescription for atorvastatin, 4261 (27.7%) for simvastatin, 3209 (20.9%) for pravastatin,and 1228 (8.0%) for lovastatin. Clinical characteristics and proportion of days covered with a statin prescription were similaracross groups. Drug dosages were relatively low, with 82% of patients who received the agent at a dose of ≤20 mg. Althoughcontrolling for time-dependent covariates representing current use and dosage, as well as for age, sex, coronary arterydisease, and several other comorbidities, treatment with pravastatin (adjusted hazards ratio [HR] 0.94, 95% CI 0.83-1.07),lovastatin (adjusted HR 1.02, 95% CI 0.88-1.17), or simvastatin (adjusted HR 0.92, 95% CI 0.83-1.01) had a similareffectiveness to prevent mortality compared to atorvastatin (reference in this analysis) in this population with CHF. Time-dependent exposure to a statin was highly protective against mortality.

Conclusions Statins exert a class effect in patients with CHF, when used at a relatively low dose. The favorable effectsappear largely independent of drug dosage. (Am Heart J 2008;155:316-23.)

There has been increasing evidence over the last fewyears of the benefits of 3-hydroxy-3-methylglutarylcoenzyme A reductase inhibitors, or statins, in patientswith congestive heart failure (CHF).1,2 Although theirbeneficial effects on the primary and secondary preven-tion of cardiovascular events and mortality are explained

From the aDivision of Cardiology, Department of Medicine, Centre Hospitalier dl'Université de Montréal, University of Montreal, Montreal, Quebec, Canada, bDivisionof Clinical Epidemiology, McGill University Health Center, McGill University, MontreaCanada, Divisions of cCardiology and dClinical Epidemiology, Department oMedicine, Jewish General Hospital, McGill University, Montreal, Canada, eDivision oCardiology, University of British Columbia and Centre for Health Evaluation andOutcome Science, Vancouver, British Columbia, fInstitute for Clinical EvaluativSciences, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, Canadaand gDivision of Internal Medicine, McGill University Health Center, McGill UniversityMontreal, Canada.Supported by grants from the Canadian Institutes of Health Research (#MOP53181MOP19671).Submitted April 20, 2007; accepted September 13, 2007.Reprint requests: Louise Pilote, MD, MPH, PhD, Division of Internal Medicine and ClinicaEpidemiology, Montreal General Hospital, 1650 Cedar Avenue, Montreal, QuebecCanada H3G 1A4.E-mail: louise.pilote@mcgill.ca0002-8703/$ - see front matter© 2008, Mosby, Inc. All rights reserved.doi:10.1016/j.ahj.2007.09.004

e

l,ff

e,,

,

l,

by their power to reduce low-density lipoproteincholesterol (LDL-C),3 the favorable effects of statin in CHFare believed to be independent of the power of statinagents to reduce LDL-C.1 Multiple pleiotropic effects ofstatins or cholesterol-independent mechanisms havebeen described,1,2,4,5 including antithrombotic and anti-proliferative effects in patients with CHF that may lead toimprovement of symptoms and clinical outcomes.2,6

Statin agents differ in multiple characteristics, includingliver and renal metabolism, half-life, effect on other serumlipid components, bioavailability, and potency to reduceLDL-C.7,8 No randomized controlled trial (RCT) to datehas compared statin agents in a population with CHF.Drugs in the same class are generally thought to betherapeutically equivalent because of similar mechanismsof action, which is also called the class effect. In patientswith CHF, the relative importance of the potency of thedrug to reduce LDL-C compared to those pleiotropiceffects, which might differ between agents, is unknown.These differences could potentially influence the extentto which one given statin drug is beneficial in patientswith CHF. We conducted a population-based study toexamine the relative effectiveness of different statins inpatients with CHF. The objective of the study was toevaluate whether statins exert a class effect on long-term

Rinfret et al 317American Heart JournalVolume 155, Number 2

survival in elderly patients who were hospitalized with aprimary diagnosis of CHF in 3 Canadian provinces.

MethodsData sourcesData on the treatment and clinical outcomes of all patients

aged≥65 years who were admitted for CHF in Quebec, Ontario,and British Columbia (BC) (the 3 most populous provinces inCanada) between January 1, 1998, and December 31, 2002,were obtained from government administrative databases ineach province and merged for this analysis.The hospital discharge summary databases used to identify

patients with CHF in Quebec, Ontario, and BC are respectivelycalled the following: Maintenance et Exploitation des Donnéespour l'Étude de la Clientèle Hospitalière (Med-Echo), CanadianInstitute for Health Information (CIHI), and the BC PatientHospitalization Data Base. Using encrypted provincial healthinsurance numbers, these databases were linked to theprovincial physician and drug claims databases, which containinformation on Quebec patients' in- and out-patient diagnosticand therapeutic procedures, as well as drug prescriptions (Régiede l'Assurance Maladie du Québec [RAMQ] in Quebec, OntarioDrug Benefit Plan database in Ontario, and PharmaCare [N65years] in BC).Complete survival data were obtained for Quebec patients

by using data from both the Med-Echo and RAMQ databases.The mortality variable is only recorded in the Med-Echodatabase if a patient dies in the hospital, whereas the RAMQmortality variable is based on death certification that occursout of hospital. In Ontario, the Ontario Registered PersonsDatabase (RPDB) and CIHI database were used to obtainmortality data, as it contains information on the vital status ofall residents covered under the health insurance plan. In BC,linkage with Deaths Registry of the BC Vital Statistics Agencyprovided data on mortality. Survival status was available up toMarch 2004 for Quebec, December 2003 for Ontario, andDecember 2002 for BC patients. The accuracy of survival dataafter myocardial infarction (MI) has been confirmed forpatients with acute MI in Ontario and Québec.9,10 Therefore,we considered likely that similar estimates can be extrapolatedto patients with CHF.

Study populationFor this analysis, we included all elderly patients, aged

≥65 years, who were admitted with presumed new onsetCHF. Given that the objective of the analysis was to assessthe class effect of statin agents in patients with CHF, andgiven that there would have been major issues of selectionbias going into the comparison of users and nonusers aswell as issues of survival bias, we excluded patients notexposed to any statin during follow-up. These patients wereadmitted to hospital with a most responsible dischargediagnosis of CHF (International Classification of Diseases–9threvision, code 428.x).Several exclusion criteria were applied to increase the

specificity of the new onset of CHF diagnosis, as follows: (1)previous admission for CHF within the past 3 years (toensure that patients were as homogeneous as possible); (2)admitted to a long-term care hospital; (3) residence outside

province; (4) invalid health care number; (5) CHF as acomplication; (6) transferred from another acute care hospital(to avoid counting patients twice); (7) discharged to a long-term care institution, a rehabilitation center, or moved outof province (as information on out-patient medication use isnot available).

Outcome of interest, exposure variable, and covariatesThe outcome of interest was all-cause mortality. Exposure

was the type of statin agent the patient filled after discharge.Follow-up for each patient was from the time of the firststatin prescription (time 0) to death or the end of the studyperiod. On the basis of the first statin prescribed, 4 statingroups were formed (atorvastatin, pravastatin, simvastatin,lovastatin). Use of fluvastatin was too infrequent in thispopulation and was not included. Rosuvastatin was notavailable at the time of administrative data collection. Forstatin usage patterns, we recorded the number of patientswho switched or stopped the initially prescribed statintreatment. Stopping treatment was defined as discontinuationof the initial statin or the absence of a prescription for theinitial statin or more days after the end of the previousprescription. To indicate patient adherence on the treatment,we calculated the ratio of the total number of days suppliedfor the initial statin divided by the total number of follow-updays (proportion of days covered). To assess the impact ofstatin dose, we adjusted for the initial daily dose of eachstatin by creating a binary variable “at or above mostfrequently used dose.” We determined this dose by referringto the dose most frequently tested in the large-scale RCTs ofeach statin for long-term cardiovascular prevention, whichwas evaluated to be 10 mg for atorvastatin and 40 mg for theother statins.11 We also considered the relative strength ofeach statin agents to reduce LDL cholesterol in sensitivityanalyses using Quebec data. Atorvastatin 10 mg, lovastatin40 mg, and simvastatin 40 mg can all reduce LDL-C by 37%.12

Atorvastatin 20 mg, lovastatin 80 mg, and simvastatin 80 mgcan further reduce LDL-C by 42% to 45%. Pravastatin isweaker and 80 mg will be required to reduce LDL by 33%,similar to atorvastatin 10 mg.12 Therefore, for this sensitivityanalysis, we considered equivalent dosages: 10 mg ofatorvastatin, 40 mg of simvastatin, 40 mg of lovastatin, and80 mg of pravastatin.Patient demographic characteristics and comorbidities at

discharge were determined from the hospital dischargedatabases, including coexisting cardiovascular and lung dis-eases, chronic kidney, or liver conditions (including renal andliver failure), as well as diabetes mellitus, dementia, andmalignant disease. Concurrent use of major cardiac medicationswas also recorded, like β-blockers, angiotensin-convertingenzyme inhibitors, calcium-channel blockers, digoxin, diuretics,warfarin, and clopidogrel. Use of acetylsalicylic acid was notavailable from BC and Ontario data (because of availability overthe counter) and thus ignored. Use of statins during the yearbefore the index hospitalization was included as a baselinecovariate. Information was obtained regarding the inhospitalprocedure performed (catheterization, percutaneous coronaryintervention [PCI], or coronary artery bypass graft [CABG]),length of hospital stay, time to first statin prescription, year ofthe CHF hospitalization, specialty of the treating physician(cardiologist, internist, general practitioner, or other specialist),

Table I. Demographic and clinical characteristics of patients with CHF prescribed statins

Characteristic Lovastatin Pravastatin Simvastatin Atorvastatin

No. of patients (% of total) 1228 (8) 3209 (21) 4261 (28) 6670 (43)Median follow-up (y) 1.9 2.0 2.1 2.2Median age (y) 75.9 75.1 75.2 74.7Males (%) 50 57 55 53Baseline comorbidities (%)Previous MI 23 28 28 26Previous PCI 1.2 2.8 2.4 3.0Previous CABG 0.7 1.6 1.9 2.7Hypertension 29 33 33 35Diabetes 32 36 36 42Cardiac dysrhythmia 25 29 26 25Cerebrovascular disease 7.3 7.7 6.6 6.6Renal failure 11 13 12 13COPD 21 22 22 21Malignancy 2.3 2.9 2.7 2.5

Other prescriptions at discharge (%)ACE Inhibitor 63 62 64 62ARB 7.2 9.0 8.3 9.1Digoxin 33 31 28 29Nitrates 50 51 52 50β-Blockers 34 40 41 45All diuretics 82 83 83 82Spironolactone 13 15 15 18Loop diuretics 80 81 80 79Warfarin 25 27 25 24Calcium-channel blockers 29 29 30 30Amiodarone 11 12 13 11

Specialty of treating physician (%)GP 44 37 38 38Cardiologist 23 33 31 29Internist 25 23 25 24

Hospital characteristics (%)University affiliation 17 18 19 20Hospitalization for CHF volumeHigh volume 75 80 77 76Medium volume 21 17 20 20Low volume 3.6 2.5 2.4 3.6

ARB, Angiotensin receptor blocker; COPD, chronic obstructive pulmonary disease; GP, general practitioner.

318 Rinfret et alAmerican Heart Journal

February 2008

type of hospital (teaching or not), hospital volume, hospitallocation (urban or rural).

Statistical analysesDescriptive statistics were used to compare baseline patient

characteristics between statin groups. Kaplan-Meier curveswere generated for each separate province and all-causemortality by the type of statin used compared using the log-ranktest. After pooling data from the 3 provinces, a multivariate Coxproportional hazards model was used to assess the associationsbetween type of statin used and survival. The proportionalhazard assumption was assessed by a plot of log (−log(survivalfunction)). Adjusted hazard ratios (HRs) for each statincompared with the reference statin (atorvastatin, the mostfrequently used) and 95% CIs were reported, with adjustmentmade for baseline characteristics and potential confounders. Inaddition, we also used 2 binary time-dependent exposurevariables, one to identify periods of exposure and the other toaccount for dosage. The variable for exposure indicated currentexposure by assigning a value of 1 to all periods during whichthe patient had a filled prescription and 0 to other periods. The

variable for dose assigned the value of 1 to those periods whenthe current dosage was at or above the most frequently useddosage in randomized trials and the value of 0 to periods whendosage was below that. In survival analyses, patients werecensored at the time of switching or stopping the initial statin.We used time-dependent modeling, with a nonexposed periodconsidered default. Therefore, if a patient was not exposed for agiven period, a 0 was assigned to the exposure variable.However, if the patient failed to fill the prescription for 7 days orlonger, the patient was censored. Forced-entry regression wasused to include all these variables in all multivariable models toadjust the between-drug comparisons for potential confounders.We performed sensitivity analyses on the Quebec sub-dataset.

Those analyses were performed (a) to account for using adifferent method for the effect of dosing and also (b) tosubsequently restrict the analysis to patients without overtevidence of coronary heart disease. In the first analysis, wemodeled dosage using the relative “atorvastatin-equivalent”dosage. To do that, the lovastatin dose was divided by 4, thesimvastatin dose by 4, and the pravastatin dose by 8. This“atorvastatin-equivalent” dosage, in milligrams, was used as a

Table II. Prescription characteristics for patients prescribed a statin at discharge

Characteristic Lovastatin Pravastatin Simvastatin Atorvastatin

No. of patients (% of total) 1228 (8) 3209 (21) 4261 (28) 6670 (43)Use of statin before index hospitalization (%) 97 86 83 80Time delay to fill the statin prescription after discharge (median and IQR, in days) 18 (1-41) 13 (1-35) 11 (0-35) 11 (0-35)Duration of statin use in the first year (median in days) 316 328 336 333Adherence of filled prescriptions (%)Mean adherence (PDC) ⁎ 0.81 0.83 0.83 0.84High adherence (%) y 69 73 75 75

Persistence (%) z 75 81 84 84Statin switch (%)During first year 13 9.3 6.1 2.8During follow-up 25 19 12 5.2

Mean dosages (mg) 25 24 20 18Median dosages (mg) (IQR) 20 (20-22) 20 (20-30) 20 (10-20) 10 (10-20)Dose distribution (%)10 mg 2.2 9.6 35 5320 mg 77 65 47 3440 mg 19 24 14 9.980 mg 1.0 0.3 0.7 1.2

Dose changed during follow-up (%)Increased 15 20 23 23Decreased 9.5 12 14 15

Statin at-or-above most frequently used dosage § 97 25 62 99

IQR, Interquartile range (25th, 75th percentiles).⁎Adherence is defined as the proportion of days for which a patient was covered (PDC or proportion of days covered) by prescriptions over the year after discharge or until death ithe patient died in the year after discharge.yAdherence of 80% of days covered (0,80) or higher.zPersistence is defined as being on any statin drug 1 year after initial drug prescription filling.§10 mg for atorvastatin and 40 mg for the other statins, in most RCT.

Rinfret et al 319American Heart JournalVolume 155, Number 2

continuous variable in the Cox model. Finally, to try to mitigatethe effect of coronary artery disease and confounding byindication, we excluded in one final sensitivity analysis allpatients with previous MI, CABG, and PCI, to verify the stabilityof our findings, again using similar time-dependent Cox models.The sensitivity analyses were only performed in the Quebeccohort because of data accessibility. All statistical analyses wereperformed using the SAS version 8.2 statistical software (SASInstitute Inc, Cary, NC).

ResultsStudy sampleThe initial merged dataset was composed of 206423

patients hospitalized with newly diagnosed CHF. Thefinal sample of 15368 patients was selected after theexclusion of 84396 patients with previous CHF admis-sions, 19376 patients aged b65 or N105 years, 13758who died in hospital, 4459 who were discharged tolong-term care hospitals, 3930 who transferred fromanother hospital, 895 who were not admitted to anacute care hospital, 896 who presented CHF as acomplication, 874 nonprovince residents, 717 patientswith invalid health card number, 503 duplicates, 53591without any statin prescription during follow-up, and7660 patients who filled their statin prescriptionN90 days after discharge. In this final dataset, 6670

f

(43.4%) filled a prescription for atorvastatin, 4261(27.7%) for simvastatin, 3209 (20.9%) for pravastatin,and 1228 (8.0%) for lovastatin.

Patient, physician, and hospital characteristicsThe study sample was composed of 15368 elderly

individuals, with a median age of 75 years and a medianfollow-up of 2.1 years (Table I). Patients who filled aprescription for one of the 4 study statin drugs sharedsimilar demographic and clinical characteristics, exceptthat patients prescribed atorvastatin were more likely tobe diabetics (42% vs 35%) and to be receiving β-blockers (45% vs 40%) than the rest of the studypopulation treated with another statin. Patients receiv-ing lovastatin were less likely to be male (50% vs 55%)or to have been treated by a cardiologist (23% vs 31%)than the rest of the study population treated withanother statin. Evidence of coronary artery disease(prior MI, prior PCI, or CABG) was balanced betweengroups. We did not observe a pattern of preferentialprescribing of a given statin to sicker or healthierpatients. Patient characteristics were similar across the3 provinces (data not shown).

Prescription characteristicsPrescription characteristics are presented in Table II.

Most patients were already treated with a statin when

Figure 1

Survival rates by statin agent used and by provinces.

320 Rinfret et alAmerican Heart Journal

February 2008

hospitalized for their first episode of CHF. Patientsdischarged on atorvastatin were least likely to havebeen on a statin before the index hospitalization.Compliance was similar across statin groups. Patientsin our sample had 83% of days covered with theirstatin drug during follow-up (until death or switch),and 74% had a high adherence (N80% of dayscovered) to their drug regimen. Persistence was alsohigh and similar across groups, with 83% of patientsfilling a statin prescription within 60 days of the1-year follow-up time point. Patients prescribedatorvastatin were the least likely to be switched toanother statin during the follow-up period. Lovastatinand pravastatin users were more likely to switchduring follow-up.Average dosage was relatively low in all groups, with

≤82% of patients receiving the agent at a dose of≤20 mg.Very few patients (≤1%) were prescribed the highestdose (80 mg) of each statin. The proportion of patientswho changed doses during follow-up was similaracross groups.

Survival after hospitalization for CHFFigure 1 shows the cumulative probability of survival in

the different statin groups with unadjusted Kaplan-Meiercurves, by province. Mortality rates were similar across all4 statin groups (all log-rank P values N .05).We used a multivariate Cox model to adjust for the

baseline characteristics listed in Table I and accountfor the relative dosage as “at or above most frequentlyused dosage” and model statin agents as time-dependent exposure variables. In this analysis, treat-ment with pravastatin (adjusted HR 0.94, 95% CI 0.83-1.07), lovastatin (adjusted HR 1.02, 95% CI 0.88-1.17),or simvastatin (adjusted HR 0.92, 95% CI 0.83-1.01)had a similar effectiveness to prevent mortalitycompared to atorvastatin (reference in this analysis) inthis population with CHF (Table III). Time-dependentexposure to a statin was highly protective againstmortality. However, a drug dosage N10 mg foratorvastatin or 40 mg for all the others was notassociated with further protection. The provincevariable was not predictive of adverse outcomes (datanot shown).Sensitivity analyses were done with the Quebec data,

which comprised 43% of the total population (Table IV).After modeling the drug dosage as “atorvastatin-equiva-lent” instead of “at or above most frequently useddosage,” we observed again a similar effectiveness acrossall 4 agents. Exposure to a statin was highly protective(adjusted HR 0.16, 95% CI 0.14-0.19), whereas a higherdosage (“atorvastatin-equivalent”) of the drug was not.Each increase of 1 mg of the “atorvastatin-equivalent”dose was associated with a borderline significant 1%relative increase in the risk of mortality (adjusted HR 1.01,95% CI 1.00-1.01) (Table IV, section A). Finally, excluding

Table III. Time-dependent multivariable ⁎ Cox model for all-cause mortality: all patients

Variable HR

95%Confidence

limits

Atorvastatin (reference) – – –Simvastatin 0.92 0.83 1.01Pravastatin 0.94 0.83 1.07Lovastatin 1.02 0.88 1.17Time-dependent exposure to a statin 0.35 0.31 0.40At-or-above most frequent dosage y 0.94 0.84 1.05

⁎Controlling for all variables listed in Table I.y10 mg for atorvastatin and 40 mg for the other statins.

Table IV. Time-dependent multivariable ⁎ Cox model forall-cause mortality ⁎⁎

HR

95%Confidence

limits

Section A. Modeling dosage as “atorvastatin-equivalent” dosage(n = 6612)Atorvastatin (reference) – – –Simvastatin 1.02 0.86 1.20Pravastatin 1.13 0.94 1.35Lovastatin 1.23 0.92 1.63Time-dependent exposure to a statin 0.16 0.14 0.19Atorvastatin-equivalent dosage y 1.01 1.00 1.01

Section B. Excluding patients with previous MI, CABG, or PCI(n = 3044)Atorvastatin (reference) – – –

Rinfret et al 321American Heart JournalVolume 155, Number 2

Simvastatin 1.07 0.83 1.37Pravastatin 1.34 1.03 1.74Lovastatin 1.38 0.96 2.00Time-dependent exposure to a statin 0.15 0.12 0.19Atorvastatin-equivalent dosage y 1.01 1.00 1.02

patients with a previous MI, PCI, or CABG within the3 years before or during the index hospitalization did nothave any impact on the direction of the findings(Table IV, section B).

zAtorvastatin/1, Simvastatin/4, Pravastatin/8, Lovastasin/4.12⁎Controlling for all variables in Table I.⁎⁎Quebec data only.y10 mg for atorvastatin and 40 mg for the other statins.11

DiscussionIn this elderly population with newly diagnosed

CHF, we observed no difference in effectivenessbetween 4 commonly used statin agents (atorvastatin,simvastatin, lovastatin, pravastatin) in the preventionof mortality. These observations were robust andindependent of the way the drug dosage variablewas used in models. We also observed no differentialeffectiveness across all 4 statin agents when restrict-ing the dataset to patients without definite evidenceof CAD (previous MI, PCI, or CABG), suggestingsome benefits in patients with nonischemic cardio-myopathy. The favorable effects of statin in ourpatients were found to be independent of drugdosage in multivariable models. Time-dependentexposure to a statin was associated with animportant and significant 65% relative risk reductionin mortality.The class effect of statin therapy in CHF had

never been studied before. Our analysis is the firstto investigate the relative role of commonly usedstatin agent and dosages on mortality in patientswith CHF. Our findings suggest that, if a favorableeffect of statins exists in CHF, it is likely indepen-dent of their lipid-lowering effects, and that statinagents might exert a class effect for this indication.Although we are still waiting evidence from ongoinglarge RCTs,2 observational data have shown thatstatin therapy reduces the need for urgent cardiactransplant13 and is associated with improved survivalin patients with CHF, irrespective of its cause, atleast in the elderly.14 Although further reduction inthe LDL-C levels are clearly associated with improvedoutcomes in patients with established CAD,3 thefavorable effect of statins on CHF appears to be

independent of their lipid-lowering power. In severalsmall RCTs conducted mostly in patients withnonischemic cardiomyopathy, treatment with selectedstatin agents was associated with short-term choles-terol-independent effects, including improvedendothelial function,15-17 reduced inflammation (levelsof TNF-α, IL-6, C-reactive protein),15,17,18 reducedprothrombotic agents (PAI-1, antithrombin III, proteinC, factor V),18,19 but also improved symptoms, qualityof life, and ejection fraction compared to placebo.17,18

Also, patients who developed chronic CHF during theScandinavian Simvastatin Survival Study follow-up hadsimilar changes in their lipoprotein level during follow-up compared to patients who did not present CHF.Yet, a significant 2% absolute reduction in theincidence of CHF after 1 year was observed with statintherapy in this study.20 Moreover, lower LDL-C can bedeleterious in patients with CHF, increasing mortal-ity,21,22 through the so-called lipoprotein-endotoxinhypothesis.23 Statins reduce lipopolysaccharide-bindingprotein synthesis. This lipoprotein binds to lipopoly-saccharide, a pro-inflammatory endotoxin, that trans-locates in the circulation through the intestinal tractreflecting the effect.23 We also observed an unfavor-able trend for the effect of higher doses of statin onmortality in our population, although the type of datawe analyzed can only be used to generate hypothesesand is certainly not definitive. Nevertheless, in themultivariable model, every increase by 1 mg of the“atorvastatin-equivalent” dose was associated with a 1%relative increase in the risk of mortality (adjusted HR

322 Rinfret et alAmerican Heart Journal

February 2008

1.01, 95% CI 1.00-1.01) in the Quebec sub-dataset.Although LDL-C levels are not available from the typeof data we analyzed, higher doses of statin areexpected to further reduce LDL-C.12

Our study has, however, several potential limitations.The first limitation is the observational and adminis-trative nature of the dataset, from which we cannotdistinguish precisely nonischemic to ischemic cardio-myopathy. Therefore, we cannot rule out preferentialprescribing of a given statin, especially atorvastatin, forsome higher risk patients with ischemic cardiomyo-pathy. However, no difference in baseline prevalenceof previous MI, PCI, or CABG was observed whencomparing patients treated with the different statinagents. Moreover, those factors were controlled for inthe multivariable model. Also, restricting the popula-tion to those patients without overt evidence of CADdid not change the direction of our results. Second,we have no ability with this type of administrativedata to assess whether patients were hospitalized forsystolic or diastolic heart failure. Third, the type ofdata did not allow us to test whether lower LDL-Cserum levels were predictive of increased mortality.Actually, an alternative explanation for the highermortality observed with higher atorvastatin-equivalentdosage in the model is confounding by indication,where the sicker patients with CAD would receive thehighest dosages. However, this association remainedafter excluding patients with overt CAD. Fourth, thisstudy is really about the class effect of relatively lowdosages of statin in CHF. The percentage of patientswho received high dosages was very small (b1%received 80 mg or higher doses). Fifth, although ouranalysis suggests a time-dependent protective effect ofstatin therapy, we cannot exclude potential con-founding by indication (statin discontinued because ofterminally ill heart failure, for example) or misclassi-fication biases (patients admitted to a nursing home,thus considered nonusers and censored). Therefore,conclusions cannot be drawn from this analysis aboutthe protective effect of statin agents in patients withCHF. However, this would not affect our observationsabout the class effect of the different agents in thispopulation. Finally, there is limited correlationbetween administrative diagnostic codes and someclinical disease, especially for some of the comorbid-ities used in the model.In conclusion, our study demonstrates that statins

exert a class effect in patients with CHF, at relativelylow dose. The favorable effects appear largelyindependent of drug dosage. Our results suggest thatnon–lipid-lowering effects of statins might be moreimportant than potency to reduce LDL-C in patientswith CHF. We believe such results, given the costdifferences among agents, have potential importanthealth policy implications. Moreover, to the best of

our knowledge, there are no head-to-head trials andthere may never be trials in patients with CHF, giventhat most of the statin agents are going off patentsoon or have gone off patent. Therefore, observationaldata may well remain the only evidence for a classeffect of statins for this indication.

References1. Kurian KC, Rai P, Sankaran S, et al. The effect of statins in heart

failure: beyond its cholesterol-lowering effect. J Card Fail 2006;12:473-8.

2. Raina A, Pickering T, Shimbo D. Statin use in heart failure: a causefor concern? Am Heart J 2006;152:39-49.

3. Cannon CP, Steinberg BA, Murphy SA, et al. Meta-analysis ofcardiovascular outcomes trials comparing intensive versus moderatestatin therapy. J Am Coll Cardiol 2006;48:438-45.

4. Liao JK, Laufs U. Pleiotropic effects of statins. Annu Rev PharmacolToxicol 2005;45:89-118.

5. Yildirir A, Muderrisoglu H. Non-lipid effects of statins: emergingnew indications. Curr Vasc Pharmacol 2004;2:309-18.

6. Liao JK. Statin therapy for cardiac hypertrophy and heart failure.J Investig Med 2004;52:248-53.

7. Lennernas H, Fager G. Pharmacodynamics and pharmacokineticsof the HMG-CoA reductase inhibitors. Similarities and differences.Clin Pharmacokinet 1997;32:403-25.

8. Jones P, Kafonek S, Laurora I, et al. Comparative dose efficacy studyof atorvastatin versus simvastatin, pravastatin, lovastatin, andfluvastatin in patients with hypercholesterolemia (the CURVES study).Am J Cardiol 1998;81:582-7.

9. Austin PC, Daly PA, Tu JV. A multicenter study of the coding accuracyof hospital discharge administrative data for patients admitted tocardiac care units in Ontario. Am Heart J 2002;144:290-6.

10. Levy AR, Tamblyn RM, Fitchett D, et al. Coding accuracy of hospitaldischarge data for elderly survivors of myocardial infarction.Can J Cardiol 1999;15:1277-82.

11. Baigent C, Keech A, Kearney PM, et al. Efficacy and safety ofcholesterol-lowering treatment: prospective meta-analysis of datafrom 90,056 participants in 14 randomised trials of statins. Lancet2005;366:1267-78.

12. Law MR, Wald NJ, Rudnicka AR. Quantifying effect of statinson low density lipoprotein cholesterol, ischaemic heart disease,and stroke: systematic review and meta-analysis. BMJ 2003;326:1423.

13. Horwich TB, MacLellan WR, Fonarow GC. Statin therapy isassociated with improved survival in ischemic and non-ischemicheart failure. J Am Coll Cardiol 2004;43:642-8.

14. Ray JG, Gong Y, Sykora K, Tu JV. Statin use and survival outcomesin elderly patients with heart failure. Arch Intern Med 2005;165:62-7.

15. Tousoulis D, Antoniades C, Bosinakou E, et al. Effects ofatorvastatin on reactive hyperemia and inflammatory process inpatients with congestive heart failure. Atherosclerosis 2005;178:359-63.

16. Strey CH, Young JM, Molyneux SL, et al. Endothelium-amelioratingeffects of statin therapy and coenzyme Q10 reductions in chronicheart failure. Atherosclerosis 2005;179:201-6.

17. Node K, Fujita M, Kitakaze M, et al. Short-term statin therapyimproves cardiac function and symptoms in patients withidiopathic dilated cardiomyopathy. Circulation 2003;108:839-43.

Rinfret et al 323American Heart JournalVolume 155, Number 2

18. Laufs U,Wassmann S, Schackmann S, et al. Beneficial effects of statinsin patients with non-ischemic heart failure. Z Kardiol 2004;93:103-8.

19. Tousoulis D, Antoniades C, Bosinakou E, et al. Effects of atorvastatinon reactive hyperaemia and the thrombosis-fibrinolysis system inpatients with heart failure. Heart 2005;91:27-31.

20. Kjekshus J, Pedersen TR, Olsson AG, et al. The effects of simvastatinon the incidence of heart failure in patients with coronary heartdisease. J Card Fail 1997;3:249-54.

Correct

The list of authors for the article entitled “Rationale and DAfter Fibrinolysis to Enhance Reperfusion in Acute MyocarHeart J 2008;155:19-25 is incorrect. The following author wInternal Medicine, University of Manitoba, Winnipeg, ManitoWarren J. Cantor, MD; David Fitchett, MD; Bjug BorgundvaLaurie J. Morrison, MD; John Ducas, MD; Anatoly Langer,Schwartz, MD; Vladimir Dzavik, MD; and Shaun G. Goodma

21. Horwich TB, Hamilton MA, Maclellan WR, et al. Low serum totalcholesterol is associated with marked increase in mortality inadvanced heart failure. J Card Fail 2002;8:216-24.

22. Rauchhaus M, Clark AL, Doehner W, et al. The relationship betweencholesterol and survival in patients with chronic heart failure. J AmColl Cardiol 2003;42:1933-40.

23. Rauchhaus M, Coats AJ, Anker SD. The endotoxin-lipoproteinhypothesis. Lancet 2000;356:930-3.

ion

esign of the Trial of Routine ANgioplasty and Stentingdial Infarction (TRANSFER-AMI)” that appeared in Amas omitted by mistake: John Ducas, MD, Department ofba, Canada. The correct order of authors is as follows:ag, MD; Michael Hefferman, MD; Eric A. Cohen, MD;MD; Shamir Mehta, MD; Charles Lazzam, MD; Briann, MD.