Impact on In-Hospital Outcomes With Drug-Eluting Stents Versus Bare-Metal Stents (from 665,804 Procedures)

aDeparbDepartmeland, Newpital CentePittsburgheDepartmefDepartmencine, Miampital, Newmanuscript

Drs. Bshare equa

No stuare solelydrafting an

See pa*CorreE-mail

0002-9149http://dx.do

Impact on In-Hospital Outcomes With Drug-Eluting StentsVersus Bare-Metal Stents (from 665,804 Procedures)

Apurva O. Badheka, MDa,*, Shilpkumar Arora, MDa, Sidakpal S. Panaich, MDa,Nileshkumar J. Patel, MDb, Nilay Patel, MDa, Ankit Chothani, MDc, Kathan Mehta, MDd,

Abhishek Deshmukh, MDe, Vikas Singh, MDf, Ghanshyambhai T. Savani, MDf,Kanishk Agnihotri, MDa, Peeyush Grover, MDf, Sopan Lahewala, MDg, Achint Patel, MDg,Chirag bambhroliya, MDa, Ashok Kondur, MDa, Michael Brown, MDa, Mahir Elder, MDa,

Amir Kaki, MDa, Tamam Mohammad, MDa, Cindy Grines, MDa, and Theodore Schreiber, MDa

Contemporary large-scale data, regarding in-hospital outcomes depending on the types of

tment ont of CYork;r, WasMedicant of Cat of Ci, FlorYork,receiveadheka,l contridy specrespond editinge 1636spondinaddres

/14/$ -i.org/1

stent used for percutaneous coronary intervention (PCI) is lacking.Wequeried theHealthcareCost and Utilization Project’s Nationwide Inpatient Sample from 2006 to 2011 using theInternationalClassification ofDiseases,NinthRevision,ClinicalModificationprocedure code36.06 (bare-metal coronary artery stent, BMS) or 36.07 (drug-eluting coronary artery stent,DES) for PCI. All analyses were performed using the designated weighting specified to theNationwide Inpatient Sample database to minimize bias. Primary outcome was in-hospitalmortality. Wald’s chi-square test was used for categorical variables. We built a hierarchical2 level model adjusted for multiple confounding factors, with hospital identification incor-porated as random effects in the model and propensity match analyses were used to adjustconfounding variables. A total of 665,804 procedures were analyzed, which were represen-tative of 3,277,884 procedures in the United States. Use of bare-metal stents (BMS) wasassociated with greater occurrence of in-hospital mortality compared with that of drug-eluting stents (DES; 1.4% vs 0.5%, p <0.001). The association stayed significant afteradjustment of various possible confounding factors (odds ratio forDES versus BMS0.59 [0.54to 0.64, p <0.001]) and also in propensity matched cohorts (1.2% vs 0.7%, p <0.001). Theresults continued to be similar in the following high-risk subgroups: diabetes (0.57 [0.50 to0.64, <0.001]), acute myocardial infarction and/or shock (0.53 [0.49 to 0.57, <0.001]), age >80(0.66 [0.58 to 0.74, <0.001]), and multivessel PCI (0.55 [0.46 to 0.66, <0.001]). In conclusion,DES use was associated with lesser in-hospital mortality compared with BMS. This outcomebenefit was seen across subgroups in various subgroups including elderly, diabetics, and acutemyocardial infarction as well as multivessel interventions. � 2014 Elsevier Inc. All rightsreserved. (Am J Cardiol 2014;114:1629e1637)

Although drug-eluting stents (DES) have been purportedfor their effectiveness in reducing in-stent restenosis and long-term target vessel revascularization rates, their impact oncardiovascular mortality beyond the use of bare-metal stents

f Cardiology, Detroit Medical Center, Detroit, Michigan;ardiology, Staten Island University Hospital, Staten Is-cDepartment of Cardiology, MedStar Washington Hos-hington, DC; dDepartment of Cardiology, University ofl Center Shadyside Hospital, Pittsburgh, Pennsylvania;rdiology, University of Arkansas, Little Rock, Arkansas;ardiology, University of Miami Miller School of Medi-ida; and gDepartment of Cardiology, Mount Sinai Hos-New York. Manuscript received June 1, 2014; revisedd and accepted August 19, 2014.Arora, Panaich, Nileshkumar J. Patel, and Nilay Patel

bution to this manuscript.ific funding was used to support this work. The authorssible for the study design, conduct and analyses, andg of the manuscript and its final contents.for disclosure information.g author: Tel: (þ408) 324-4516; fax: (þ203) 737-2437.s: [email protected] (A.O. Badheka).

see front matter � 2014 Elsevier Inc. All rights reserved.0.1016/j.amjcard.2014.08.033

(BMS) has been debatable.1e3 Additionally, some authorshave suggested higher stent thrombosis rates with some DESespecially over long term questioning their usefulness overBMS.4 Most of the randomized control trials (RCTs) havebeen limited by their inclusion of patients with stable coronaryartery disease (CAD)5,6 and further by limited sample sizes7,8

affecting their ability to draw meaningful conclusionsregarding mortality differences between various stent sub-types. Indeed, recentmeta-analyses of theseRCTswith greatersample sizes have indicated significant mortality benefit ofDES.2 Nonetheless, data regarding the postprocedural out-comes and benefits of using DES over BMS in real-worldclinical practice are sparse. The major objective of our studywas to study the difference in postprocedural mortality afterpercutaneous coronary intervention (PCI) between DES andBMS in a broad range of patient population from the largestpublicly available inpatient database in the United States.

Methods

The study cohort was derived from the NationwideInpatient Sample (NIS) database, a subset of the Healthcare

www.ajconline.org

Figure 1. Data selection algorithm.

1630 The American Journal of Cardiology (www.ajconline.org)

Cost and Utilization Project sponsored by the Agency forHealthcare Research and Quality, from 2006 to 2011. TheNIS is the largest publicly available all-payer inpatient caredatabase in the United States, including data on approxi-mately 7 to 8 million discharges per year, and it is a strat-ified sample designed to approximate a 20% sample of UScommunity (nonfederal, short term, general, and specialty)hospitals.9 National estimates are produced using samplingweights provided by the sponsor. The details regarding theNIS data have been previously published.10 Annual dataquality assessments of the NIS are performed, which guar-antee the internal validity of the database. The NIS databaseresults have been shown to correlate well with otherhospitalization discharge databases in the United States11

and have also been used to explain trends in otheracute medical and surgical conditions.12

We queried the NIS database using the InternationalClassification of Diseases, Ninth Revision, Clinical Modifi-cation (ICD-9-CM) procedure code of 36.06 (bare metalcoronary artery stent) or 36.07 (drug eluting coronary artery

stent) for PCI if present in primary or secondary procedurefield. After excluding patients with age <18 years; traumaadmissions or observations with missing data for age,gender, or death; and procedures where both stents typeswere used, the final study sample size included 665,804procedures (representative of 3,277,884 procedures in theUnited States; Figure 1).

In-hospital mortality was defined if a patient died duringhospitalization. Preventable procedural complications wereidentified by patient safety indicators (PSIs), which havebeen established by the Agency for Healthcare Research andQuality to monitor preventable adverse events during hos-pitalization. These indicators are based on ICD-9-CM codesand Medicare severity diagnosis-related groups and eachPSI has specific inclusion and exclusion criteria13 PSI in-dividual measure technical specifications, Version 4.4,March 2012 (Agency for Healthcare Research and Quality(US)) was used to identify and define preventable compli-cations.14 Other procedure related complications, whichincluded postprocedure hemorrhage requiring blood

Table 1Baseline characteristics of the study population, according to the type of stent utilized

Baseline

Variable BMS DES OVERALL P value

Total no. of PCI (unweighted NO.) 154421(23.19) 511383(76.81) 665804Total no. of PCI (weighted no.) 760594( 23.20) 2517290(76.80) 3277884Age(Years) 64.9�0.03 64.1�0.02 64.3�0.01 <0.001Male 67.0% 65.9% 66.2% <0.001Female 33.0% 34.1% 33.8% <0.001Race* <0.001White 63.0% 62.7% 62.8%Non-white 15.7% 15.1% 15.3%Missing 21.3% 22.1% 22.0%Charlson/deyo comorbidity index† <0.0010 16.0% 24.1% 22.2%1 39.6% 39.7% 39.7%>¼2 44.4% 36.2% 38.1%Obesity 12.0% 12.0% 12.0% 0.8Hypertension 68.9% 71.3% 70.8% <0.001Diabetes Mellitus 30.5% 33.5% 32.8% <0.001Heart failure 0.6% 0.3% 0.4% <0.001Chronic pulmonary disease 17.3% 14.4% 15.1% <0.001Peripheral vascular disease 11.0% 10.1% 10.3% <0.001Renal failure 10.9% 8.7% 9.2% <0.001Neurological disorder or paralysis 3.5% 2.5% 2.7% <0.001Anemia or coagulopathy 12.0% 7.8% 8.8% <0.001Hematological or oncological malignancy 2.7% 1.0% 1.4% <0.001Rheumatoid arthritis or other collagen vascular Disease 1.9 % 1.7 % 1.7 % <0.001Depression, psychosis or substance abuse 11.3% 8.1% 8.9% <0.001Median household income category for patient’s zip

codez(Percentile)<0.001

1. 0-25th 28.8% 25.8% 26.5%2. 26-50th 27.8% 27.3% 27.4%3. 51-75th 24.5% 25.0% 24.9%4. 76-100th 18.8% 22.0% 21.2%

Primary Payer <0.001Medicare / Medicaid 58.3% 54.8% 55.6%Private including HMO 29.8% 38.0% 36.1%Self pay/no charge/other 11.9% 7.2% 8.3%Hospital bed size <0.001Small 8.6% 7.7% 7.9%Medium 20.4% 19.9% 20.0%Large 71.0% 72.5% 72.1%Hospital Location <0.001Rural 6.0% 5.7% 5.8%Urban 94.0% 94.3% 94.2%Hospital Region <0.001Northeast 17.9% 21.6% 20.8%Midwest or North Central 29.5% 26.4% 27.1%South 42.3% 41.9% 42.0%West 10.2% 10.1% 10.%Hospital Teaching status <0.001Non-teaching 44.1% 42.5% 42.9%Teaching 55.9% 57.5% 57.1%Admission types <0.001Emergent/Urgent 76.1% 69.1% 70.7%Elective admission 23.9% 30.9% 29.3%Admission day <0.001Weekdays 82.3% 86.2% 85.3%Weekend 17.7% 13.8% 14.7%Length of stay (Days) 3.15�0.01 2.54�0.01 2.68�0.01 <0.001Total charges($) 16823 � 27 18050 � 14 17768 � 12 <0.001Disposition <0.001

(continued)

Coronary Artery Disease/Impact of DES Versus BMS 1631

Table 1(continued)

Baseline

Variable BMS DES OVERALL P value

Home 94.6% 97.4% 96.8%Facility/others 5.4% 2.6% 3.2%Death 1.4% 0.5% 0.7%ProceduresFraction Flow Reserve 0.3% 0.4% 0.4% <0.001Intravascular Ultrasound 4.4% 5.5% 5.2% <0.001Intra Aortic Balloon Pump 0.5% 0.2% 0.3% <0.001Co-DiagnosisAcute Myocardial Infarction 53.1% 37.0% 40.7% <0.001Shock 2.2% 0.8% 1.1% <0.001Unstable (ShockþVentilator) 2.2% 0.8% 1.1% <0.001

AHRQ ¼ Agency for Healthcare Research and Quality; AMA ¼ American Medical Association; HMO ¼ health maintenance organization; ICD-9-CM ¼International Classification of Diseases, Ninth Revision, Clinical Modification.* Race was missing in 22.0% of the study population.† Charlson/Deyo comorbidity index was calculated as per Deyo classification.1 Comorbidities were identified by ICD-9-CM code Mentioned in any of the

diagnostic fields.z This represents a quartile classification of the estimated median household income of residents in the patient’s zip code. These Values are derived from zip

code-demographic data obtained from Claritas. The quartiles are identified by values of 1 to 4, indicating the poorest to wealthiest populations. Because theseestimates are updated annually, the value ranges vary by year (http://www.hcup-us.ahrq.gov/db/vars/zipinc_qrtl/nisnote.jsp).

1632 The American Journal of Cardiology (www.ajconline.org)

transfusion, iatrogenic cardiac complications (includingpostprocedural myocardial infarction (MI) and post-procedural need for revascularization), pericardial compli-cations, requiring open heart surgery, other iatrogenicrespiratory complications (which included ventilator asso-ciated pneumonia, postprocedure aspiration pneumonia, andother respiratory complications not elsewhere classified),postprocedural stroke or transient ischemic attack, and othervascular complications were identified using ICD-9-CMcodes (listed in Supplementary Table 1) in any secondarydiagnosis field. Vascular complications were defined as PSIcode for accidental puncture or ICD-9-CM codes for injuryto blood vessels, creation of arteriovenous fistula, injury toretroperitoneum, vascular complications requiring surgery,and other vascular complications not elsewhere classified.“Any complications” was defined as occurrence of �1postprocedure complications listed in SupplementaryTable 1. Similar method has been used before.15 The ICD-9-CM codes were used to identify each of these diagnosesand procedures.

NIS variables were used to identify patients’ de-mographic characteristics including age, gender, and race(Table 1). We defined severity of co-morbid conditionsusing Deyo’s modification of Charlson co-morbidity index(CCI).16 This index contains 17 co-morbid conditions withdifferential weights. The score ranges from 0 to 33, withgreater scores corresponding to greater burden of co-morbiddiseases (Supplementary Table 2). Facilities were consid-ered to be teaching hospitals if they had an AmericanMedical Association approved residency program, were amember of the Council of Teaching Hospitals, or had afulltime equivalent interns and residents to patients’ ratio of0.25 or greater. Annual hospital volume was determined ona year-to-year basis using the unique hospital identificationnumber to calculate the total number of procedures per-formed by a particular institution in a given year.

The Healthcare Cost and Utilization Project NIS containsdata on total charges for each hospital in the databases,which represent the amount that hospitals are billed forservices. To calculate the estimated cost of hospitalizations,the NIS data were merged with cost-to-charge ratios avail-able from Healthcare Cost and Utilization Project. Using themerged data elements from the cost-to-charge ratio files andthe total charges reported in the NIS database, we convertedthe hospital’s total charge data to cost estimates by simplymultiplying total charges with the appropriate cost-to-chargeratio. These costs are in essence standardized and can bemeasured across hospitals and are used in remainder ofreport. Adjusted cost for each year was calculated in termsof the 2010 cost, after adjusting for inflation according to thelatest consumer price index data released by the US gov-ernment on January 16, 2013.17

Stata IC 11.0 (StataCorp, College Station, Texas) andSAS 9.2 (SAS Institute Inc, Cary, North Carolina) wereused for analyses. Weighted values of patient level obser-vations were generated to produce a nationally representa-tive estimate of the entire US population of hospitalizedpatients. Differences between categorical variables weretested using the chi-square test and differences betweencontinuous variables were tested using the Student t test. A pvalue <0.05 was considered significant.

Hierarchical models or multilevel models are designed toanalyze data with nested observations. The NIS data set isinherently hierarchical, namely, the data have group-specific(i.e., hospital) attributes and within each group (i.e., hospi-tal) there are patients who contribute patient-specific attri-butes to the data. Hierarchical models take intoconsideration the effect of nesting (e.g., patient-level effectsnested within hospital-level effects). Hence, it is superior tosimple regression modeling for the available data set. Hi-erarchical mixed-effects logistic regression models wereused for categorical dependent variables such as in-hospital

Table 2Baseline characteristics and outcomes of the study population afterpropensity matching

Baseline table(Propensity Score Match Cohort)

Demographic variable BMS DES P-Value

Total no. of PCI(unweighted no.) 67344 67344Age (Years) 65.1�0.05 65.4�0.05 <0.01Gender 0.04Male 67.3% 67.8%Female 32.7% 32.2%Diabetes Mellitus 29.9% 30.2% 0.24Myocardial Infarction or Shock 46.9% 47.1% 0.52Charlson/deyo comorbidity index 0.910 19.3% 19.3%1 40.1% 40.0%>¼2 40.6% 40.7%Primary Payer <0.01Medicare / Medicaid 58.2% 59.0%Private including HMO 30.7% 30.7%Self pay/no charge/other 11.1% 10.3%Hospital characteristicsHospital Teaching status 0.5Non-teaching 46.0% 45.9%Teaching 54.0% 54.1%Admission types 0.27Emergent/Urgent 72.5% 72.7%Elective admission 27.5% 27.3%Operator Volume 0.871st quartile(1-14) 29.8% 29.8%2nd quartile(15-40) 27.2% 27.4%3rd quartile(41-94) 23.7% 23.6%4th quartile(95-718) 19.3% 19.2%Hospital Volume 0.671st quartile(1-350) 21.8% 21.9%2nd quartile(351-635) 24.7% 24.4%3rd quartile(636-1128) 26.7% 26.7%4th quartile(1129-3608) 26.9% 26.9%OutcomesDeath 1.2% 0.6% <0.001Any complication 6.6% 5.4% <0.001Length of Stay (Days) 2.91 � 0.01 2.72�0.01 <0.001Cost of Hospitalization($) 15692�34 18153�36 <0.001

HMO ¼ health maintenance organization.

Coronary Artery Disease/Impact of DES Versus BMS 1633

mortality and procedural complications, and hierarchicalmixed-effects linear regression models were used forcontinuous dependent variables such as cost of hospitali-zation and length of stay (LOS). Two-level hierarchicalmodels (with patient-level factors nested within hospital-level factors) were created with the unique hospital identi-fication number incorporated as random effects within themodel. In all multivariate models, we included hospital-level variables such as hospital region (Northeast, South,and Midwest with West as referent); teaching versusnonteaching hospital; hospital procedure volume; andpatient-level variables such as age, gender, acute MI, shock,Deyo modification of Charlson co-morbidity index, occur-rence of procedural complications, admission over theweekend, primary payer (with Medicare and/or Medicaidconsidered as referent), and in addition to type of stentsused. All interactions were thoroughly tested. Multi-collinearity, defined as a perfect linear relation or very high

correlation between 2 or more predictor (independent) var-iables was assessed using variance inflation factor, withvariance inflation factor >20 suggestive of multicollinearity.

We used propensity-scoring method to establish matchedcohorts to control for imbalances of patients’ and institu-tional characteristics between the 2 different stent groupsthat might influence treatment outcome. A propensity score,which was assigned to each hospitalization, was based onmultivariate logistic regression model that examined theimpact of 10 variables (patient demographics, co-morbidities, and hospital characteristics; Table 2) on thelikelihood of treatment assignment. Patients with similarpropensity score in the 2 treatment groups were matchedusing a 1 to 1 scheme without replacement using greedyalgorithm.18

Results

Table 1 lists baseline characteristics of the study popu-lation. A total of 665,804 procedures were analyzed, whichwere representative of 3,277,884 procedures in the UnitedStates. The mean age was 64.9 � 0.05 years (BMS) and64.1 � 0.02 years (DES). Men constituted 66.2% of theoverall population with 62.8 being whites. Hypertensionwas the most common co-morbidity present in 70.8 of thepatients, whereas diabetes was present in 32.8. Furthermore,40.7 of the patients who underwent PCI had a diagnosis ofMI. A total of 57% of stents were placed in teaching hos-pitals and the rest in nonteaching hospitals. Approximately70.7 of the procedures were emergent. Medicare and/orMedicaid (58.3% for BMS and 54.8% for DES) was theprimary payer.

In univariate analysis, the overall postprocedural mor-tality (Table 1) was 0.49% in the cohort receiving DES and1.43% for BMS. The results continued to be similar in thefollowing subgroups: diabetes (1.4% vs 0.5%, p <0.001),acute myocardial infarction (AMI) and/or shock (2.5% vs1.1%, p <0.001), age >80 (3.4% vs 1.5%, p <0.001), andmultivessel PCI (1.8% vs 0.5%, p <0.001; Figure 2).Among relevant complications (Table 3), vascular compli-cations were significantly lesser with use of DES (1.7% forDES vs 2.1% for BMS; p value <0.001), whereas signifi-cant hemorrhage (hemorrhage required transfusion)occurred in only 0.4% of the patients receiving DES versus0.7% with the use of BMS (p value <0.001).

In multivariate analysis, the overall postprocedural mor-tality (Table 4) was higher in the cohort receiving DES(odds ratio [OR] 0.59, 95% confidence interval [CI] 0.54 to0.64, p value <0.001). Similar results were found infollowing subgroups (Table 5): diabetes (OR 0.57, 95% CI0.50 to 0.64, p value <0.001), AMI and/or shock (OR 0.53,95% CI 0.49 to 0.57, p value <0.001), age >80 (OR 0.66,95% CI 0.58 to 0.74, p value <0.001), multivessel PCI (OR0.55, 95% CI 0.43 to 0.72, p value <0.001). In addition,increasing burden of co-morbidities as represented bygreater Charlson co-morbidity index scores was associatedwith higher in-hospital mortality (OR 2.70, 95% CI 2.13 to3.43, p value <0.001; Table 4). Age (OR 1.05, 95% CI 1.05to 1.06, p value <0.001) and female gender (OR 1.19, 95%CI 1.11 to 1.29, p value <0.001) were also significantpredictors of postprocedural mortality. Likewise, MI (OR

Figure 2. In-hospital mortality according to type of stent used.

Table 3Procedural outcomes in drug eluting vs bare metal stent groups

Complications BMS DES p value

Any complications 6.9 4.6 <0.001Death D Any complications 7.6 4.9 <0.001Death D Vascular D Stroke D

Cardiac D Requirement of Open heartsurgery

5.3 3.5 <0.001

Death D Vascular D Stroke D

Cardiac D Requirement of Open heartsurgery D Renal D DVT D Infectious

6.3 4.2 <0.001

Vascular complications 2.1 1.7 <0.0011)Postop-hemorrhage requiring transfusion 0.7 0.4 <0.0012)Vascular injury 1.5 1.3 <0.001Cardiac complications 1.8 1.3 <0.001Iatrogenic cardiac complications 1.7 0.1 <0.001Respiratory complications (Post-op resp

failure)2.0 1.0 <0.001

1)Acute renal failure requiring dialysis 0.3 0.2 <0.0012)Acute severe metabolic derangement 0.5 0.3 <0.001Postoperative DVT/PE 0.6 0.4 <0.001

DVT ¼ Deep Venous Thrombosis; PE ¼ Pulmonary Embolism.

1634 The American Journal of Cardiology (www.ajconline.org)

3.83, 95% CI 3.38 to 4.34, p value <0.001) and shock (OR15.26, 95% CI 13.73 to 16.95, p value <0.001) predictedpoor postprocedural outcomes. These factors were predic-tive of primary outcome irrespective of whether operatorvolume was included in the final model (Table 4 andSupplementary Table 3). A greater operator (OR 0.74, 95%CI 0.61 to 0.90, p value <0.001 for the highest quartile) andhospital volume (OR 0.71, 95% CI 0.58 to 0.86, p value<0.001 for the highest quartile) were both predictive oflesser in-hospital mortality.

Table 2 lists the baseline characteristics in a propensity-matched cohort. After propensity score matching 67,344matched pairs were found. Patient demographics, co-morbidities, and hospital characteristics were similaramong the 2 groups. In the propensity-matched cohort, useof BMS was associated with a greater occurrence of in-hospital mortality compared with that of DES (1.2% vs0.6%, p value <0.001). Likewise, BMS was associated withlonger LOS (2.9 � 0.01 days [BMS] vs 2.7 � 0.01 days[DES], p value <0.001). In contrast, DES use was alsoassociated with higher costs ($18,153� 36 vs $15,692� 34,

p value<0.001) and lower complication rate (5.0% vs 6.6%,p value <0.001).

Overall, DES was associated with shorter LOS (2.5 �0.01 days [BMS] vs 3.2 � 0.01 days [DES], p value<0.001). In contrast, DES use was also associated withhigher costs ($18,050 � 14 vs $16,823 � 27, p <0.001).After adjusting with confounders (Table 6), LOS of DESstays shorter (LOS �0.26 days, 95% CI �0.28 to �0.25days, p value <0.001) and cost of care of DES stays higher(cost $2181, 95% CI $2,132 to $2,229, p value <0.001)compare with BMS. Higher age, AMI, shock, femalegender, higher Charlson/Deyo score were associated withlonger LOS, whereas elective admission, private insurance,and greater hospital volume quartile were associated withshorter LOS. Higher cost is associated with higher age,AMI, shock, higher Charlson/Deyo, teaching hospitals,weekend admission, and highest hospital volume quartile,whereas elective admission and private insurance wereassociated with lower cost of care.

Discussion

The key finding of our study was the significant differ-ence in terms of in-hospital mortality between patientsreceiving DES and BMS. The patients receiving DES hadboth lower in-hospital mortality and combined end point ofpostprocedural mortality and complications compared withpatients who got BMS. These results corroborate the recentreports on improved safety of DES, documented in obser-vational registries and meta-analyses.19,20

Although DES have been widely considered as beingeffective in reducing in-stent restenosis rates in publishedliterature, data on their mortality benefit over BMS have beenquestionable at best.1,3 This lack of difference in survivaloutcomes in RCTs have been partly attributed to the artifactof their restricted low-risk profile study populations andrelatively small sample sizes.5,6,8,21 Moreover, severalstudies have either not included the latest generation of DES.With improving newer generation DES and operator expe-rience, PCI is being expanded for more complex lesions andpatient populations. Extrapolating clinical trials outcome datato clinical practice is also fraught with issues regardingdiverse clinical use of DES in real-world including off-labelindications in up to 60% to 70% of the procedures as persome reports.21,22 Our real-world analysis derived from the

Table 4Multivariate analysis of primary and secondary outcome

Primary Outcome (In-hospital Mortality) Secondary Outcomes (In-hospital mortality þ any complications)

Odds Ratio lower Limit Upper Limit P-Value Odds Ration Lower Limit Upper Limit P-Value

Age 1.05 1.05 1.06 <0.001 1.02 1.01 1.02 <0.001Female 1.19 1.11 1.29 <0.001 1.35 1.31 1.39 <0.001AMI 3.83 3.38 4.34 <0.001 1.98 1.90 2.06 <0.001SHOCK 15.26 13.73 16.95 <0.001 11.79 10.97 12.67 <0.001Any Complication 5.85 5.34 6.42 <0.001Type of StentBare Metal Stent Referent Referent Referent Referent Referent ReferentDrug Eluting stent 0.59 0.54 0.64 <0.001 0.79 0.76 0.82 <0.001Primary PayerMedicare / Medicaid Referent Referent Referent Referent Referent ReferentPrivate including HMO 0.86 0.76 0.97 0.01 0.87 0.84 0.90 <0.001Self pay/no charge/other 1.19 1.00 1.42 0.05 0.85 0.80 0.90 <0.001Admission typesEmergent/Urgent Referent Referent Referent Referent Referent ReferentElective admission 0.63 0.55 0.73 <0.001 1.01 0.95 1.07 0.78Charlson/deyo comorbidity index0 Referent Referent Referent Referent Referent Referent1 1.66 1.30 2.11 <0.001 1.20 1.15 1.27 <0.001>¼2 2.70 2.13 3.43 <0.001 2.11 1.98 2.24 <0.001Hospital Teaching statusNon-teaching Referent Referent Referent Referent Referent ReferentTeaching 0.92 0.83 1.02 0.10 0.97 0.90 1.04 0.38Admission dayWeekdays Referent Referent Referent Referent Referent ReferentWeekend 1.17 1.07 1.27 <0.001 1.05 1.02 1.09 <0.001Hospital Volume1st quartile(1-350) Referent Referent Referent Referent Referent Referent2nd quartile(351-635) 0.94 0.84 1.06 0.33 0.98 0.92 1.04 0.413rd quartile(636-1128) 0.83 0.72 0.95 0.01 1.06 0.96 1.16 0.274th quartile(1129-3608) 0.77 0.66 0.90 <0.001 0.92 0.83 1.03 0.15c-Statistics* 0.91 0.71

* Model discrimination was determine by c-Statistics.

Table 5Adjusted odds ratio in different subgroups

Primary Outcome P value

Overall population 0.59(0.54-0.64) <0.001SubgroupsAge>80 years 0.66(0.58-0.74) <0.001Uninsured 0.56(0.43-0.72) <0.001Charlson’s comorbidity index >¼2 0.62(0.57-0.67) <0.001Diabetes 0.57(0.50-0.64) <0.001AMI or Shock 0.53(0.49-0.57) <0.001Multivessel PCI 0.55(0.46-0.66) <0.001

Coronary Artery Disease/Impact of DES Versus BMS 1635

largest all-comer nationwide publicly available databaseincluded all generations of DES used during the study periodacross the nation’s PCI centers. Moreover, our study includedPCI done for stable coronary artery diseases as well as acuteMI and/or cardiogenic shock and also patient populationswith different baseline co-morbidities allowing for plausiblybetter real-world clinical implications. Furthermore, wereport DES to be associated with improved mortality out-comes even in complex multivessel and bifurcation PCI inour study. Besides mortality outcomes, reduced bleeding andvascular complications in patients who received DES further

reiterated their safety profile over BMS in the postoperativeperiod.

Our subgroup analysis revealed significant mortalitybenefit of DES in study cohorts including diabetics, elderly(>80 years of age) as well patients with MI and/or shock,which in turn were all associated with worse mortality out-comes in our study. Age has been a significant predictor ofpost-PCI outcomes 23 with previous studies demonstratinglesser use of evidence-based therapies in elderly patients.Similar to the results of recently published studies,19,20 wefound DES to be associated with lower postproceduralmortality in patients >80 years of age. In an additionalsubgroup analysis, DES were predictive of better outcomesin even uninsured patients. This is especially importantbecause insurance status has been widely published to pre-dict access to healthcare and even the use of DES.24 Indeed,patients with private insurance had better outcomes thanthose with Medicaid and/or Medicare in our study, furtherreinforcing the need to improve access to potentially bettertreatment strategies irrespective of insurance status.

Diabetics with a greater propensity for neointimal hyper-plasia in relatively smaller caliber vessels with diffuseatherosclerotic burden tend to have worse outcomes poststentimplantation.19,25 Previous studies have been largely

Table 6Multivariate analysis of length of stay and cost of care

Multivariate analysis of LOS Multivariate analysis of cost of care

LOS LL UL p-value Cost LL UL p-value

Age 0.02 0.02 0.02 <0.001 29 27 31 <0.001Female 0.26 0.25 0.28 <0.001 28 -14 70 0.19AMI 0.63 0.62 0.65 <0.001 2271 2222 2320 <0.001SHOCK 3.95 3.89 4.01 <0.001 13602 13400 13804 <0.001Type of StentBare Metal Stent Referent Referent Referent Referent Referent ReferentDrug Eluting stent -0.26 -0.28 -0.25 <0.001 2181 2132 2229 <0.001Primary PayerMedicare / Medicaid Referent Referent Referent Referent Referent ReferentPrivate including HMO -0.20 -0.21 -0.18 <0.001 -241 -291 -190 <0.001

Self pay/no charge/other -0.14 -0.16 -0.11 <0.001 -257 -338 -176 0.00Admission types (%)Emergent/Urgent Referent Referent Referent Referent Referent ReferentElective admission -1.03 -1.04 -1.01 <0.001 -2226 -2276 -2176 <0.001

Charlson/deyo comorbidity index0 Referent Referent Referent Referent Referent Referent1 0.16 0.14 0.17 <0.001 369 314 423 <0.001>¼2 1.14 1.12 1.15 <0.001 2405 2347 2462 <0.001Hospital Teaching statusNon-teaching Referent Referent Referent Referent Referent ReferentTeaching 0.00 -0.05 0.04 0.98 519 297 740 <0.001

Admission day (%)Weekdays Referent Referent Referent Referent Referent ReferentWeekend 0.53 0.51 0.55 <0.001 1249 1191 1307 <0.001

Hospital Volume1st quartile(1-350) Referent Referent Referent Referent Referent Referent2nd quartile(351-635) -0.03 -0.06 0.01 0.18 192 47 337 0.033rd quartile(636-1128) -0.14 -0.18 -0.10 0.00 -41 -221 139 0.674th quartile(1129-3608) -0.14 -0.19 -0.08 0.00 1369 1150 1588 <0.001

1636 The American Journal of Cardiology (www.ajconline.org)

inconsistent regarding the selection of stents in this patientpopulation.26 Our study, which included a sizable proportionof diabetics (30%), showed significant mortality reductionwith DES than BMS in the postprocedural period in thissubgroup extending the findings of our colleagues 27 to newergeneration stents in a real-world population-based cohort.

Patients presenting with acute MI have higher rates ofplatelet activation and thrombus burden with consequentrisk of stent thrombosis.28 Most RCTs have excluded thesehigher risk patients, being underpowered to detect low-frequency end points such as death or MI. Recently, Pal-merini et al demonstrated reduced cardiac mortality withnewer everolimus-eluting stents than BMS in a networkmeta-analysis of ST Segment Elevation Myocardial Infarc-tion RCTs.29 Indeed, evolving stent technology includingdurable and/or bioabsorbable polymers, newer drugs, andstent platforms has resulted in improved efficacy and safetyoutcomes in some recent studies.2,30 Our study observationsregarding better postprocedural outcomes in a high-riskpopulation subgroup of patients with MI and/or shock addto the growing literature on the safety of DES.

The limitations of our study relate to post hoc analysis ofan administrative database, which include lack of long-termfollow-up data, clinical information (use of antiplateletagents, detailed angiographic data, and so on) and possibleoversights in coding. Furthermore, we cannot draw anytemporal and/or causal associations for the worse outcomes

seen in patients receiving BMS. Because our study includedstents over several years, some of the earlier generationstents with relatively poorer outcome profile could haveplausibly skewed the results. Although the NIS samplingdesign has been extensively used for research previously,reliance on administrative databases could be fraught withunder-reporting of secondary or co-morbid diagnosis andmisrepresentation of procedure volume. However, sucherrors are expected to be equally distributed among studygroups. Furthermore, if some operators perform proceduresin hospitals outside the NIS database, the operator volumecould be underestimated. These limitations aside, the NISoffers the largest publicly available database with a sizablesample and has extensively contributed clinically relevantliterature on periprocedural outcomes in the past.15

Acknowledgment: None declared.

Disclosures

The authors have no conflicts of interest to disclose.

Supplementary Data

Supplementary data associated with this article can be found,in the online version, at http://dx.doi.org/10.1016/j.amjcard.2014.08.033.

Coronary Artery Disease/Impact of DES Versus BMS 1637

1. Stettler C, Wandel S, Allemann S, Kastrati A, Morice MC, Schömig A,Pfisterer ME, Stone GW, Leon MB, de Lezo JS, Goy JJ, Park SJ,Sabaté M, Suttorp MJ, Kelbaek H, Spaulding C, Menichelli M,Vermeersch P, Dirksen MT, Cervinka P, Petronio AS, Nordmann AJ,Diem P, Meier B, Zwahlen M, Reichenbach S, Trelle S, Windecker S,Jüni P. Outcomes associated with drug-eluting and bare-metal stents: acollaborative network meta-analysis. Lancet 2007;370:937e948.

2. Bangalore S, Kumar S, Fusaro M, Amoroso N, Attubato MJ, Feit F,Bhatt DL, Slater J. Short- and long-term outcomes with drug-elutingand bare-metal coronary stents: a mixed-treatment comparison anal-ysis of 117 762 patient-years of follow-up from randomized trials.Circulation 2012;125:2873e2891.

3. Kaiser C, Galatius S, Erne P, Eberli F, Alber H, Rickli H, Pedrazzini G,Hornig B, Bertel O, Bonetti P, De Servi S, Brunner-La Rocca HP,Ricard I, Pfisterer M. Drug-eluting versus bare-metal stents in largecoronary arteries. N Engl J Med 2010;363:2310e2319.

4. McFadden EP, Stabile E, Regar E, Cheneau E, Ong AT, Kinnaird T,Suddath WO, Weissman NJ, Torguson R, Kent KM, Pichard AD,Satler LF, Waksman R, Serruys PW. Late thrombosis in drug-elutingcoronary stents after discontinuation of antiplatelet therapy. Lancet2004;364:1519e1521.

5. Moses JW, Leon MB, Popma JJ, Fitzgerald PJ, Holmes DR, O’Sh-aughnessy C, Caputo RP, Kereiakes DJ, Williams DO, Teirstein PS,Jaeger JL, Kuntz RE. Sirolimus-eluting stents versus standard stents inpatients with stenosis in a native coronary artery. N Engl J Med2003;349:1315e1323.

6. Stone GW, Ellis SG, Cox DA, Hermiller J, O’Shaughnessy C, MannJT, Turco M, Caputo R, Bergin P, Greenberg J, Popma JJ, Russell ME.A polymer-based, paclitaxel-eluting stent in patients with coronaryartery disease. N Engl J Med 2004;350:221e231.

7. Menichelli M, Parma A, Pucci E, Fiorilli R, De Felice F, Nazzaro M,Giulivi A, Alborino D, Azzellino A, Violini R. Randomized trial ofSirolimus-Eluting Stent Versus Bare-Metal Stent in Acute MyocardialInfarction (SESAMI). J Am Coll Cardiol 2007;49:1924e1930.

8. Spaulding C, Henry P, Teiger E, Beatt K, Bramucci E, Carrié D,Slama MS, Merkely B, Erglis A, Margheri M, Varenne O, Cebrian A,Stoll HP, Snead DB, Bode C. Sirolimus-eluting versus uncoatedstents in acute myocardial infarction. N Engl J Med 2006;355:1093e1104.

9. Quality AfHRa. Healthcare Cost and Utilization Project (HCUP). July2008. Rockville: HCUP Databases; 2008. Available at: http://www.hcupus.ahrq.gov/nisoverview.jsp. Accessed on September, 2014.

10. Steiner C, Elixhauser A, Schnaier J. The healthcare cost and utilizationproject: an overview. Eff Clin Pract 2002;5:143e151.

11. Whalen D, Houchens R, Elixhauser A. 2004 HCUP NationwideInpatient Sample (NIS) Comparison Report. HCUP Methods SeriesReport # 2007-03 Online. U.S. Agency for Healthcare Research andQuality, 2007.

12. Epstein AJ, Polsky D, Yang F, Yang L, Groeneveld PW. Coronaryrevascularization trends in the United States, 2001-2008. JAMA2011;305:1769e1776.

13. McDonald KM, Romano PS, Geppert J, Davies SM, Duncan BW,Shojania KG, Hansen A. Measures of Patient Safety Based on HospitalAdministrative Data - The Patient Safety Indicators 2002.

14. All Patient Safety Indicators Technical Specifications, Version 4.4.Agency for Healthcare Research and Quality (US).; Available at: http://www.qualityindicators.ahrq.gov/modules/PSI_TechSpec.aspx. Accessedon September, 2014.

15. DeshmukhA1, PatelNJ, Pant S, ShahN,ChothaniA,MehtaK,Grover P,SinghV,Vallurupalli S, SavaniGT,BadhekaA, Tuliani T,DabhadkarK,DibuG,ReddyYM, SewaniA,KowalskiM,Mitrani R, PaydakH,Viles-Gonzalez JF. In-hospital complications associated with catheter ablationof atrial fibrillation in the United States between 2000 and 2010: analysisof 93 801 procedures. Circulation 2013;128:2104e2112.

16. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidityindex for use with ICD-9-CM administrative databases. J ClinEpidemiol 1992;45:613e619.

17. http://www.usinflationcalculator.com. Accessed on September, 2014.18. Brunt ME, Egorova NN, Moskowitz AJ. Propensity score-matched

analysis of open surgical and endovascular repair for type B aorticdissection. Int J Vasc Med 2011;2011:364046.

19. Douglas PS, Brennan JM, Anstrom KJ, Sedrakyan A, Eisenstein EL,Haque G, Dai D, Kong DF, Hammill B, Curtis L, Matchar D, BrindisR, Peterson ED. Clinical effectiveness of coronary stents in elderlypersons: results from 262,700 Medicare patients in the AmericanCollege of Cardiology-National Cardiovascular Data Registry. J AmColl Cardiol 2009;53:1629e1641.

20. Groeneveld PW, Matta MA, Greenhut AP, Yang F. Drug-elutingcompared with bare-metal coronary stents among elderly patients. J AmColl Cardiol 2008;51:2017e2024.

21. Kirtane AJ, Gupta A, Iyengar S, Moses JW, Leon MB, Applegate R,Brodie B, Hannan ED, Harjai K, Jensen LO, Park SJ, Perry R, Racz M,Saia F, Tu JV, Waksman R, Lansky AJ, Mehran R, Stone GW. Safetyand efficacy of drug-eluting and bare metal stents: comprehensivemeta-analysis of randomized trials and observational studies. Circu-lation 2009;119:3198e3206.

22. Austin D, Oldroyd KG, McConnachie A, Slack R, Eteiba H, FlapanAD, Jennings KP, Northcote RJ, Pell AC, Starkey IR, Pell JP. Drug-eluting stents versus bare-metal stents for off-label indications: a pro-pensity score-matched outcome study. Circ Cardiovasc Interv 2008;1:45e52.

23. Peterson ED, Dai D, DeLong ER, Brennan JM, Singh M, Rao SV,Shaw RE, Roe MT, Ho KK, Klein LW, Krone RJ, Weintraub WS,Brindis RG, Rumsfeld JS, Spertus JA. Contemporary mortality riskprediction for percutaneous coronary intervention: results from 588,398procedures in the National Cardiovascular Data Registry. J Am CollCardiol 2010;55:1923e1932.

24. Gaglia MA Jr1, Torguson R, Xue Z, Gonzalez MA, Collins SD, Ben-Dor I, Syed AI, Maluenda G, Delhaye C, Hanna N, Wakabayashi K,Kaneshige K, Suddath WO, Kent KM, Satler LF, Pichard AD,Waksman R. Insurance type influences the use of drug-eluting stents.JACC Cardiovasc Interv 2010;3:773e779.

25. Abizaid A1, Kornowski R, Mintz GS, Hong MK, Abizaid AS, MehranR, Pichard AD, Kent KM, Satler LF, Wu H, Popma JJ, Leon MB. Theinfluence of diabetes mellitus on acute and late clinical outcomesfollowing coronary stent implantation. J Am Coll Cardiol 1998;32:584e589.

26. Bangalore S1, Kumar S, Fusaro M, Amoroso N, Kirtane AJ, ByrneRA, Williams DO, Slater J, Cutlip DE, Feit F. Outcomes with variousdrug eluting or bare metal stents in patients with diabetes mellitus:mixed treatment comparison analysis of 22,844 patient years of follow-up from randomised trials. BMJ 2012;345:e5170.

27. Stettler C, Allemann S, Wandel S, Kastrati A, Morice MC, Schomig A,Pfisterer ME, Stone GW, Leon MB, de Lezo JS, Goy JJ, Park SJ,Sabate M, Suttorp MJ, Kelbaek H, Spaulding C, Menichelli M, Ver-meersch P, Dirksen MT, Cervinka P, De Carlo M, Erglis A, Chechi T,Ortolani P, Schalij MJ, Diem P, Meier B, Windecker S, Juni P. Drugeluting and bare metal stents in people with and without diabetes:collaborative network meta-analysis. BMJ 2008;337:a1331.

28. Merlini PA1, Bauer KA, Oltrona L, Ardissino D, Cattaneo M, Belli C,Mannucci PM, Rosenberg RD. Persistent activation of coagulationmechanism in unstable angina and myocardial infarction. Circulation1994;90:61e68.

29. Palmerini T1, Biondi-Zoccai G, Della Riva D, Mariani A, Sabaté M,Valgimigli M, Frati G, Kedhi E, Smits PC, Kaiser C, Genereux P,Galatius S, Kirtane AJ, Stone GW. Clinical outcomes with drug-elutingand bare-metal stents in patients with ST-segment elevation myocardialinfarction: evidence from a comprehensive network meta-analysis.J Am Coll Cardiol 2013;62:496e504.

30. Bangalore S1, Toklu B, Amoroso N, Fusaro M, Kumar S, Hannan EL,Faxon DP, Feit F. Bare metal stents, durable polymer drug elutingstents, and biodegradable polymer drug eluting stents for coronaryartery disease: mixed treatment comparison meta-analysis. BMJ2013;347:f6625.