Valvular and Congenital Heart Disease

Transcatheter vs surgical aortic valve replacement inintermediate-surgical-risk patients with aortic stenosis:A propensity score–matched case-control studyAzeem Latib, MB ChB, a,b,f Francesco Maisano, MD, c,f Letizia Bertoldi, MD, b Andrea Giacomini, MD, c

Joanne Shannon, MD, a Micaela Cioni, MD, c Alfonso Ielasi, MD, b Filippo Figini, MD, a,b Kensuke Tagaki, MD, a

Annalisa Franco, MD, d Remo Daniel Covello, MD, d Antonio Grimaldi, MD, d Pietro Spagnolo, MD, e

Gill Louise Buchannan, MD, b Mauro Carlino, MD, b Alaide Chieffo, MD, b Matteo Montorfano, MD, b

Ottavio Alfieri, MD, c and Antonio Colombo, MD a,b Milan, Italy

Background Limited real-world data comparing outcomes after transcatheter (TAVR) and surgical aortic valvereplacement (SAVR) in intermediate-surgical-risk patients with aortic stenosis are available.

Methods We identified 182 consecutive patients who underwent TAVR via the transfemoral (TF) route (November 2007–February 2011) and 111 moderate-to-high-risk historical case controls undergoing SAVR (August 2003–July 2008). Usingpropensity score matching based on clinical characteristics and surgical risk scores, we compared clinical outcomes in 111matched patients. Valve Academic Research Consortium definitions were applied for end point adjudication.

Results Baseline clinical characteristics, in particular Logistic European System for Cardiac Operative Risk Evaluation(23.2 ± 15.1 vs 24.4 ± 13.4) and Society of Thoracic Surgeons score (4.6 ± 2.3 vs 4.6 ± 2.6), were well matched betweengroups. Transfemoral TAVR was associated with more vascular complications (33.3% vs 0.9%, P b .001). On the other hand,acute kidney injury was more frequent after SAVR (8.1% vs 26.1%, P b .001). The rates of all-cause mortality in both TF-TAVRand SAVR groups was1.8% at 30 days (P = 1.00) and 6.4% and 8.1%, respectively, at 1 year (P = .80). At 1 year, the rate ofcerebrovascular events was similar in the 2 groups (4.6% vs 9.1%, P = .19).

Conclusions In this real-world cohort of intermediate-surgical-risk patients with aortic stenosis, TF-TAVR and SAVR wereassociated with similar mortality rates during follow-up but with a different spectrum of periprocedural complications.Furthermore, the survival rate after TF-TAVR in this group of elderly patients with intermediate Society of Thoracic Surgeonsscore was encouraging. (Am Heart J 2012;164:910-7.)

Transcatheter aortic valve replacement (TAVR) is now aviable option in the treatment of severe aortic stenosis inpatients at high risk for surgery.1 In operable high-riskpatients, cohort A of the PARTNER trial demonstratedthat TAVR is noninferior to surgical aortic valvereplacement (SAVR) for all-cause mortality at 1 year,2

thus supporting the indication for this less invasiveprocedure. These positive results and the lack of real-world data motivated us to undertake this study to further

From the aInterventional Cardiology Unit, EMO-GVM Centro Cuore Columbus, Milan, Italy,bInterventional Cardiology Unit, San Raffaele Scientific Institute, Milan, Italy, cDepartment ofCardiothoracic Surgery, San Raffaele Scientific Institute, Milan, Italy, dDepartment of CardiacAnesthesia, San Raffaele Scientific Institute, Milan, Italy, and eCentre for CardiovascularPrevention, San Raffaele Scientific Institute, Milan, Italy.fDrs Latib and Maisano contributed equally to the study and are joint first authors.Submitted May 24, 2012; accepted September 5, 2012.Reprint requests: Antonio Colombo, MD, 48 Via Buonarroti, Milan, Italy.E-mail: info@emocolumbus.it0002-8703/$ - see front matter© 2012, Mosby, Inc. All rights reserved.http://dx.doi.org/10.1016/j.ahj.2012.09.005

investigate outcomes in this field of moderate-to-high-riskpatients with aortic stenosis undergoing SAVR ortransfemoral TAVR (TF-TAVR).

MethodsPatientsWe evaluated consecutive moderate-to-high-risk patients with

severe symptomatic aortic stenosis undergoing SAVR betweenAugust 2003 and June 2008 (n = 130) and TAVR from November2007 to February 2011 (n = 276; 58.3% treated with the Edwards-SAPIEN/Edwards-SAPIEN XT (Edwards Lifesciences Inc, Irvine,CA) and 41.7%with the CoreValve (Medtronic Inc., Minneapolis,MN)). The enrollment periods for the 2 groups were sequentialbecause, as we gained experience with TAVR, this became thedefault treatment option for moderate-to-high-risk patients at ourinstitution.High surgical riskwas defined as a Logistic EuropeanSystem for Cardiac Operative Risk Evaluation (Logistic Euro-SCORE) ≥20%, or Society of Thoracic Surgeons (STS) PredictedRisk of Mortality ≥10%, or conditions not captured by the 2scores that the cardiac surgeon considered to increase the risk

Figure 1

Consort-type flowchart showing patient selection for this study.

Latib et al 911American Heart JournalVolume 164, Number 6

for standard SAVR, such as thoracic radiotherapy, coronaryartery bypass surgery with patent grafts (right internal mammaryartery, which crosses from right to left under the sternum, and/ora left internal mammary artery, which might be damaged withredo sternotomy access), porcelain aorta, liver cirrhosis (Child-Pugh classes B and C), or marked patient frailty.3-5

Patient selection for this study is summarized in Figure 1.During the SAVR study period, 145 moderate-to-high-riskpatients were evaluated: 16 were considered too high a riskfor SAVR and treated medically, and 129 underwent SAVR ofwhom 18 were uncontactable during follow-up. Of the 348patients who underwent multidisciplinary evaluation for TAVR,276 subsequently underwent the procedure and 72 wereexcluded due to the following: unsuitability for TAVR (n = 39)and thus treated medically (n = 23) or with balloon aorticvalvuloplasty (n = 16), low surgical risk and thus referred forSAVR (n = 12), no indication for aortic valve replacement (AVR)because of the absence of symptoms or significant aorticstenosis (n = 12), patient refusal (n = 3), or aortic annulus sizetoo large for currently available percutaneous valves (n = 6).Because the TF approach is the preferred access choice forTAVR in our center and to have a homogenous group ofmoderate-to-high-risk patients undergoing a less invasive proce-dure, we excluded 52 patients undergoing TAVR by non-TFapproaches (29 transaxillary, 21 transapical, and 2 transaortic).After exclusion, the cohort consisted of 111 patients undergoingSAVR and 182 patients undergoing TF-TAVR who underwentpropensity score matching to generate 111 matched pairs.We compared these propensity score–matched patients

treated with TF-TAVR and SAVR and analyzed periprocedural,

30-day, and 1-year outcomes. Clinical follow-up was performedby office visit at 30 days, 6 months, and 1 year after theprocedure or by telephonic contact in patients unable to returnfor clinical follow-up. All patients provided informed consent forboth the procedure and subsequent data collection and analysisfor research purposes.

Procedures and devicesSurgical AVR was performed according to standard surgical

practice through a standard midline sternotomy with cardiopul-monary bypass and mild systemic hypothermia. Bioprostheticvalves were implanted in all cases. Transcatheter AVR is a jointcardiology-cardiac surgery program in our institution. Allpatients undergo a thorough preprocedural screening forTAVR suitability, which includes the following: confirming theindication for AVR, evaluating vascular access to determine theimplantation approach, assessing aortic root anatomy andannulus dimensions for prosthesis type and size, definingcomorbid conditions, and estimating operative risk.4 Thedecision to perform TAVR is then evaluated by the multi-disciplinary heart team consisting of 2 interventional cardiolo-gists, 1 senior cardiac surgeon and 1 cardiac anesthetist.The technical aspects of the procedure have been described

elsewhere by our group.4,6 The device used in the TF-TAVR groupwas initially the Edwards-SAPIEN transcatheter heart valve (THV)from November 2007 until it was replaced with the SAPIEN XT inApril 2010, whereas from July 2008, both the SAPIEN THV andMedtronic CoreValve were contemporaneously available.

Table I. Baseline clinical characteristics of patients before and after propensity score matching

Characteristic TF-TAVR (n = 204) SAVR (n = 111) P TF-TAVR (n = 111) SAVR (n = 111) P

Age (y) 80.0 ± 7.6 79.4 ± 3.0 .42 80.5 ± 6.9 79.4 ± 3.0 .10Sex 108 (52.9) 49 (44.1) .14 49 (44.1) 49 (44.1) 1.00STS-PROM 7.49 ± 7.15 4.60 ± 2.63 b.001 4.57 ± 2.28 4.60 ± 2.63 .93STS morbidity and mortality 30.4 ± 16.8 23.8 ± 8.3 b.001 22.7 ± 8.1 23.8 ± 8.3 .30Logistic EuroSCORE 24.3 ± 16.8 24.4 ± 13.4 .96 23.2 ± 15.1 24.4 ± 13.4 .54Additive EuroSCORE 10.6 ± 2.8 11.0 ± 2.3 .14 10.5 ± 2.5 11.0 ± 2.3 .13Coronary artery disease 73 (35.8) 51 (45.9) .08 44 (39.6) 51 (45.9) .34Previous myocardial infarction 42 (20.6) 16 (14.4) .18 16 (14.4) 16 (14.4) 1.00Diabetes 56 (27.4) 24 (21.6) .26 21 (18.9) 24 (21.6) .62Cerebrovascular disease 29 (14.2) 20 (18.0) .37 16 (14.4) 20 (18.0) .47Left ventricular ejection fraction (%) 51.3 ± 13.3 53.6 ± 10.7 .12 53.5 ± 12.5 53.6 ± 10.7 .94Hypertension 140 (68.6) 77 (69.4) .89 78 (70.3) 77 (69.4) .88Peripheral vascular disease 48 (23.5) 38 (34.2) .04 29 (26.1) 38 (34.2) .19Chronic obstructive pulmonary disease 72 (35.3) 25 (22.5) .02 29 (26.1) 25 (22.5) .53Redo AVR 9 (4.4) 1 (0.9) .09 0 (0) 1 (0.9) .32Hemoglobin (g/dL) 12.2 ± 1.8 12.7 ± 1.5 .01 12.6 ± 1.6 12.7 ± 1.5 .53Creatinine (mg/dL) 1.36 ± 1.33 1.09 ± 0.61 .04 1.06 ± 0.38 1.09 ± 0.61 .61New York Heart Association Class III-IV 136 (66.7) 77 (69.4) .92 75 (67.6) 77 (69.4) .99Body mass index (kg/m2) 26.3 ± 4.7 25.7 ± 3.9 .28 25.5 ± 4.6 25.7 ± 3.9 .72

Values are presented as numbers (%) or mean ± SD. PROM, Predicted Risk of Mortality.

912 Latib et alAmerican Heart Journal

December 2012

Study end points and definitionsWe used the end point definitions of the Valve Academic

Research Consortium (VARC), where available.7 Events weredivided temporally into periprocedural, 30 day, and 1 year,as appropriate.

Statistical methodsContinuous variables are presented as means ± SD or medians

(interquartile range) and categorical variables as frequencies(%). Continuous variables were compared using the indepen-dent sample t test or Mann-Whitney U test, as appropriate.Categorical variables were compared using the χ2 statistic.Propensity score matching was performed to adjust for

significant differences in baseline covariates and potentialconfounders that may lead to biased estimates of treatmentoutcome.8,9 First, a propensity score for each patient wasconstructed, providing an estimate of the propensity towardbelonging to one treatment group vs the other. This was done byperforming a non-parsimonious multivariable logistic regres-sion, with the type of intervention as the end point (SAVR codedas 0, TAVR as 1). All variables that were significantly different atbaseline and that might impact on the choice of SAVR vs TAVRwere considered. The following covariates were selected tocalculate the propensity score: STS score, EuroSCORE, age, sex,left ventricular ejection fraction, diabetes, hypertension, cere-brovascular disease or prior cerebrovascular events, peripheralvascular disease, chronic obstructive pulmonary disease, priormyocardial infarction, body mass index, coronary artery disease,New York Heart Association classes 3 to 4, redo AVR, baselinecreatinine, and hemoglobin. The discrimination and calibrationability of the propensity score model were assessed by means ofthe C-statistic and the Hosmer-Lemeshow statistic. The C-statisticfor the propensity score model was 0.78, indicating gooddiscrimination. The Hosmer-Lemeshow goodness-of-fit testP value was .47, confirming good calibration and fit of themultivariable model that estimated the propensity score. Second,for the development of propensity score–matched pairs, 1:1

matching was performed using the “psmatch2” module forSTATA, with the following algorithm: nearest neighbor matchingwithout replacement and conditioningon the common support.10

The pseudo-R2, which is ameasure of overall covariate imbalance,was 0.15 (P b .0001) before matching and very low (0.027; P =.94) after matching, thus confirming the good quality of thematchand as a result the adequate balancing of covariate distributionbetween the matched groups.11

Periprocedural events were compared using the χ2 test,whereas adverse events during follow-up were estimated usingthe Kaplan-Meiermethod and compared between groupswith theuse of the log-rank test. The results of the Coxproportional hazardsanalyses stratified bymatchedpairs are reported as adjustedhazardratios (HRs) with associated 95% confidence interval (CI) and Pvalue. A P value ofb.05was considered statistically significant, andall reported P values are 2 sided. Statistical analysis was performedusing STATA 11.0 (Stata Corporation, College Station, TX).

The authors are solely responsible for the design and conductof this study, all study analyses, the drafting and editing of themanuscript, and its final contents. No extramural funding wasused to support this work.

ResultsPatientsThe initial study cohort of intermediate-to-high-risk

patients with severe aortic stenosis included 111 under-going SAVR between August 2003 and June 2008 and 182undergoing TF-TAVR from November 2007 to February2011. A comparison of baseline characteristics of theseunmatched patients showed important differences in anumber of variables including STS score, coronary arterydisease, peripheral vascular disease, chronic obstructivepulmonary disease, redo AVR, and preprocedural hemo-globin and creatinine levels (Table I). After propensityscore matching, the baseline patient characteristics were

Table II. Aortic valve prostheses type and diameter

TF-TAVI(n = 111)

SAVR(n = 111)

Prosthesis typeMedtronic CoreValve (Medtronic Inc.,Minneapolis, MN)

41 (36.9)

Edwards-Sapien XT (Edwards Lifescience,Irvine, CA)

40 (36.0)

Edwards-Sapien (Edwards Lifescience,Irvine, CA)

30 (27.0)

Carpentier-Edwards Perimount (EdwardsLifescience, Irvine, CA)

50 (45.0)

Carpentier-Edwards Perimount Magna(Edwards Lifescience, Irvine, CA)

3 (2.7)

Medtronic Mosaic (Medtronic Inc.,Minneapolis, MN)

26 (23.4)

Sorin Mitroflow (Sorin Group, Saluggia, Italy) 13(11.7)St Jude Epic (St Jude Medical,Minneapolis, MN)

16 (14.4)

St Jude Bioimplant (St Jude Medical,Minneapolis, MN)

1 (0.9)

Carbomedics (Carbomedics Inc., Austin, TX) 2 (1.8)Prosthesis size (mm)19 18 (16.2)21 37 (33.3)23 30 (27.0) 34 (30.6)25 20 (18.0)26 58 (52.2) 029 23 (20.7) 2 (1.8)

Values are presented as numbers (%).

Table III. Periprocedural and inhospital outcomes

utcomeTF-TAVI(n = 111)

SAVR(n = 111) P

ascular complications 37 (33.3) 1 (0.9) b.001Major 16 (14.4) 0 (0) b.001Minor 21 (18.9) 1 (0.9) b.001leeding complications 74 (66.7) 92 (82.9) .005Life-threatening 25 (22.5) 19 (17.1) .31Major 43 (38.7) 63 (56.8) .007Minor 6 (5.4) 10 (9.0) .30lood transfusion 51 (46.4) 59 (53.1) .31No. of units of redbloodcells transfused 1.4 ± 2.3 1.7 ± 2.7 .45KI 9(8.1) 29(26.1) b.001Stage 2 4 (3.6) 20 (18.0) .001Stage 3 5 (4.5) 9 (8.1) .27ermanent pacemaker implantation 13 (11.7) 3 (2.7) .009onversion to or urgent redocardiacsurgery

0 2 (1.8) .16

epeat procedure for valve-relateddysfunction

3 (2.7) 1 (0.9) .31

ength of stay (d), median (IQR) 6 (5-8) 7 (5-9) .07

alues are presented as numbers (%) or mean ± SD, unless otherwise specified. IQR,terquartile range.

Latib et al 913American Heart JournalVolume 164, Number 6

comparable between the study groups (Table I). Porcelainaorta was present in 16.2% (18) of the TAVR group andnone of the patients in the SAVR group because this wasan exclusion criterion for SAVR but not for TAVR.

Procedural characteristicsTransfemoral TAVRwas performedwith local anesthesia

and conscious sedation in most (84%) patients. Percutane-ous femoral access and closure was attempted in 94%(104) of patients. Failure of percutaneous closure occurredin 12% (13) patients, of whom 6 underwent surgicalclosure. An Edwards prosthesis was implanted in 63.1%and the Medtronic CoreValve in 36.9%. Overall, VARC-defined device success was 95% in the TAVR group.Reasons for device failure in 6 patients were as follows:implantation of more than 1 valve because of embolization(3 cases), moderate or severe aortic regurgitation (2 cases),and failed delivery due to impossibility to advance thevalve through the iliofemoral vessels (1 case).Surgical aortic valve replacement was successful in all

patients, with one-third (n = 39; 35.1%) undergoingconcomitant coronary artery bypass grafting. Details ofthe valve type and diameter are reported in Table II.

Periprocedural outcomesPeriprocedural outcomes are displayed in Table III. The

rates of postprocedural residual aortic regurgitation in the

O

V

B

B

A

PC

R

L

Vin

TF-TAVR and SAVR groups were as follows: none in 35% vs97%, mild in 62% vs 3%, and moderate or severe in 2% vs0%. There were no cases of coronary obstruction duringTAVR and no intraprocedural deaths in either group.Transfemoral TAVR was associated with more majorvascular complications (14.4% vs 0%, P ≤ .001) with anoticeable difference among the different THV systems:33%with Edwards-SAPIEN implantation that requires largerarterial sheath sizes (22F and 24F), 17%with the SAPIENXTthat requires 18 or 19F, and 13% with the 18F compatibleCoreValve system. VARC–defined bleeding was significant-lymore frequent after SAVR but conspicuously high in bothgroups (82.9% vs 66.7%, P = .005), with no differencebetween groups in the rates of life-threatening but a higherrate of major bleeding in the SAVR group (56.8% vs 38.7%,P = .007). At least half of patients in both groups required ablood transfusion (46% TF-TAVR vs 53% SAVR). In the TF-TAVR group, blood transfusion was related to a vascularcomplication in 47%. VARC–defined acute kidney injury(AKI) (stage 2 or 3) occurred more frequently after SAVR(26.1% vs 8.1%, P ≤ .001), in particular stage 2 AKI (18.0%vs 3.6%,P = .001), and 5.4% (6) of the SAVR group and 3.6%(2) of the TF-TAVR group required hemodialysis after theprocedure. Stage 1 AKI occurred in 27.9% (31) of patientsundergoing TF-TAVR and 31.5% (35) of patients undergo-ing SAVR. Transfemoral TAVR was associated withpermanent pacemaker implantation in 11.7%, with asignificantly higher incidence with CoreValve (24%; 10/41) as compared with Edwards (3%; 3/70). The length ofindex hospitalization was marginally shorter after TF-TAVR(6 days vs 7 days, P = .07), but most (90%) patientsundergoing TF-TAVR were discharged home, whereas all

Table IV. Clinical outcomes at 30 days and 1 year

OutcomeTF-TAVR (n = 111),

n (%)SAVR (n = 111),

n (%) PTF-TAVR (n = 111),

n (%)SAVR (n = 111),

n (%) P

Death 2 (1.8) 2 (1.8) 1.00 7 (6.4) 9 (8.1) .80Cardiovascular death 2 (1.8) 1 (0.9) 0.56 5 (4.6) 7 (6.3) .76Cerebrovascular events 4 (3.6) 9 (8.1) 0.08 5 (4.6) 10 (9.1) .19Transient ischemic attack 3 (2.7) 7 (6.4) 0.09 4 (3.7) 7 (6.4) .36Stroke 1 (0.9) 2 (1.8) 0.56 1 (0.9) 3 (2.9) .31

Myocardial infarction 0 2 (2.7) 0.24 0 2 (2.7) .24

All percentages are Kaplan-Meier estimates at the specific time point and thus do not equal the number of patients divided by the total number in the study group.

914 Latib et alAmerican Heart Journal

December 2012

patients undergoing SAVR were transferred to a cardiacrehabilitation unit.

Clinical outcomes at 30 days and 1 yearThe rate of death from any cause at 30-days was 1.8% in

both the TF-AVR and the SAVR groups (Table IV). At 1year, mortality from any cause was 6.4% in the TF-TAVRgroup as compared with 8.1% in the SAVR group (HR0.87, 95% CI 0.32-2.41, P = .80) (Figure 2A). The reasonsand timing of the 9 deaths after SAVR were as follows:cardiogenic shock (n = 2; 1.8%) on days 4 and 33,multiorgan failure (n = 1; 0.9%) on day 4, pneumonia/septicemia (n = 2; 1.8%) on days 33 and 34, cardiac arrestafter redo valve surgery (n = 1; 0.9%) on day 37,endocarditis (day 49), fatal stroke (day 92), and anoncardiovascular cause (day 183). The causes of the 7deaths in the TF-TAVR group were as follows: severe rightventricular failure within 24 hours of the procedure,hemorrhagic shock (day 7), septicemia (day 49), hemor-rhagic stroke (day 223), cardiogenic shock after perma-nent pacemaker implantation (day 250), sudden death(day 313), and multiorgan failure after hospitalization forinferior limb ischemia (day 318). The rate of anyneurologic events (ie, all strokes and transient ischemicattacks) was 3.6% in the TF-TAVR group and 8.1% in theSAVR group at 30 days (P = .08) and 4.6% and 9.1%,respectively, at 1 year (HR 0.5, 95% CI 0.17-1.46, P = .19)(Figure 2B). Aortic valve replacement was very effectivein reducing symptoms in both groups, with 90% ofpatients undergoing SAVR and 95% of patients undergo-ing TF-TAVR being in New York Heart Association class IIor lower at 1-year follow-up.As regards the VARC composite end points, 64.9%

(72) of patients undergoing TF-TAVR and 75.7% (84) ofpatients undergoing SAVR did not experience any ofthe components of the VARC combined safety endpoint at 30 days (P = .11). The lower 30-day safety inthe TF-TAVR group was accounted for predominantlyby a higher incidence of vascular complications.Furthermore, there was a trend for the transcatheterprocedure being more often associated with multiplecomplications included in the 30-day combined safetyend point (13.5% [15] vs 5.4% [6], P = .06). The rate of

the combined efficacy end point was 89% (93) and 90%(98) in the surgical and transcatheter groups, respec-tively (P = .75).

DiscussionIn this single-center retrospective comparison of

propensity score–matched patients with aortic stenosisundergoing SAVR and TF-TAVR, the main findings areas follows:

1. TF-TAVR was associated with a similar, albeit numer-ically lower, rate of death from any cause at 1 year.

2. Transcatheter and surgical AVR resulted in differentperiprocedural and inhospital complications. Therewas a lower incidence of periprocedural cerebrovas-cular events and AKI with TF-TAVR but at the expenseof more major vascular complications.

The mortality rates after TF-TAVR observed in thiscohort were lower in comparison with most publishedseries and registries of high-risk patients undergoing TAVRthat report a 1-year mortality rate that varies from 15%to 25%.5,12-14 Similarly, the 1-year mortality rate in theSAVR group was better than that reported (20%-31%) instudies involving high-risk patients or octogenarians.15-18

The lowermortality rates of patients included in this studyare probably explained by their lower surgical risk.Indeed, the average STS score was 4.6 in both groups ascompared with 11.7 and 11.8 in the TF and SAVR groupsof PARTNER.2 In this regard and if surgical risk is based onSTS score alone, the outcomes of this study could beinterpreted as those of transcatheter and surgical AVR inintermediate-risk patients. Indeed, it has been suggestedthat the STS score is more reliable and suitable in assessingperioperative mortality in high-risk patients undergoingAVR.16,19 Another important factor that should be takeninto account when interpreting the low mortality rates inthis study is the low predictive ability of surgical riskscores, especially when applied to transcatheter pro-cedures.20,21 Finally, it cannot be excluded that theremarkably lowmortality rate in the TAVR group is due toincreased operator experience, improved case selection,and a teamwork approach.

Figure 2

Kaplan-Meier time-to-event curves for death and cerebrovascular events. Time-to-event curves are shown for death from any cause (A) and anycerebrovascular event (B) in patients undergoing TF-TAVR or SAVR.

Latib et al 915American Heart JournalVolume 164, Number 6

As in the PARTNER trial, our study highlights that TAVRand SAVR are associated with different periproceduralrisks.2 This information will become important whentrying to advise patients about the risks and benefits of atranscatheter vs surgical procedure. The most frequentperiprocedural hazards after TAVR are vascular compli-cations that are inherent to a percutaneous interventionalprocedure performed with large-bore introducer sheaths.The use of smaller catheter systems such as the SAPIENXT and CoreValve almost halved the rate of majorvascular complications in this study. Indeed, we haverecently demonstrated that the reduction in sheath sizewith SAPIEN XT is associated with a 60% reduction inmajor vascular complications as compared with the largerEdwards-SAPIEN delivery system.22 Every effort shouldbe made to reduce major vascular complications, whichwere associated with life-threatening or major bleeding inalmost all affected patients in our series. In a recent studyby Gurvitch et al,23 which reported VARC-definedprocedural outcomes, 65% of VARC-defined bleedingwas directly due to a major or minor vascular complica-tion. In our study, vascular complications accounted for38% (28/74) of bleeding and 47% of blood transfusions.Therefore, there were a significant proportion of patientsin the transcatheter group in whom bleeding and therequirement for blood transfusion were not due avascular complication but rather due to preexistentanemia, an underlying condition, or where the cause ofhemoglobin drop was unclear. In addition, the initiationof dual-antiplatelet therapy before the procedure mayalso have aggravated bleeding from these aforementionedcauses. Furthermore, there was often a low threshold forblood transfusion in these elderly patients with concom-itant coronary artery disease and/or depressed leftventricular function who became symptomatic even

with mild drops in hemoglobin. The overall rates ofVARC-related bleeding were strikingly high in bothgroups, in particular in the surgical group. This doesraise questions regarding whether the VARC definition ofbleeding is overly sensitive and also to the validity ofapplying definitions conceptualized for a transcatheterprocedure to surgical valve replacement.An interesting observation from this study was the

numerically lower rate of neurologic events in thetranscatheter group at 30 days (3.6% vs 8.1%, P = .08)and 1 year (4.6% vs 9.1%, P= .19). However,Walther et al24

have similarly reported lower rates of neurologic events(3% vs 8%) and stroke (0 vs 2%) with transapical vsconventional AVR. Although, a different vascular ap-proach was used in this study, the TF and transapicalAVR has been shown to be associated with similar ratesof neurologic events.2,14,25 The neurologic assessmentof patients in this study was clinical based, and cerebralimaging was performed only in patients with symptomsor signs of a neurologic event. Thus, we cannot excludethat the rate of new cerebral defects on magneticresonance imaging, which have been shown to occur inup to 70% of patients post-TAVR, would not have beendifferent between groups.26 However, the clinicalsignificance of such magnetic resonance imaging lesionsis still uncertain.23,26

LimitationsThe most important limitation of this study is its

retrospective observational design and lack of randomi-zation. Although, we used propensity score matching toensure the comparability and similar risk profiles of thegroups, we cannot exclude that unmeasured con-founders may have influenced the study outcomes.

916 Latib et alAmerican Heart Journal

December 2012

Although the data were prospectively entered intodedicated databases, the application of the new VARCdefinitions in a retrospective manner may have led to anunderestimation of event rates.23 The exclusion of non–TF-TAVRmay be interpreted as a selection bias. However,the TF approach is the preferred approach at our center(ratio of TF/non-TF of 4:1), with patients selected fortransapical TAVR having a much higher-risk profile.4

Thus, our results are only applicable to patients who canbe treated by TAVR via the TF approach. Anotherlimitation is the sequential enrollment period andpotential for time bias. However, simultaneous enroll-ment was not possible because these moderate-to-high-risk patients are now routinely treated at our center via atranscatheter approach. Finally, the fact that 18 patientsundergoing SAVR from the initial cohort were uncontact-able for follow-up may have inadvertently biased theresults in favor of SAVR because of the potentialexclusion of patients who may have died.

ConclusionsIn this propensity-matched cohort of moderate-to-high-

risk patients with aortic stenosis, TF and surgical AVRwere associated with similar mortality at 30 days and 1year. These procedures are associated with differentperiprocedural risks, which need to be discussed withpatients when recommending either approach. Inparticular, the higher rate of major vascular complica-tions associated with transcatheter therapy, whichappears to be declining with smaller delivery systems,needs to be addressed before this less invasive approachcan become the standard of care. The excellent survivalrates after TF-TAVR in an elderly population withintermediate STS scores justifies current efforts such asthe SURTAVI trial to compare transcatheter and surgicalvalve replacement in moderate-risk patients.

References1. Webb J, Cribier A. Percutaneous transarterial aortic valve implan-

tation: what do we know? Eur Heart J 2011;32:140-7.2. Smith CR, Leon MB, Mack MJ, et al. Transcatheter versus surgical

aortic-valve replacement in high-risk patients. N Engl J Med 2011;364:2187-98.

3. Vahanian A, Alfieri O, Al-Attar N, et al. Transcatheter valveimplantation for patients with aortic stenosis: a position statementfrom the European Association of Cardio-Thoracic Surgery (EACTS)and the European Society of Cardiology (ESC), in collaboration withthe European Association of Percutaneous Cardiovascular Interven-tions (EAPCI). Eur Heart J 2008;29:1463-70.

4. Godino C, Maisano F, Montorfano M, et al. Outcomes aftertranscatheter aortic valve implantation with both Edwards-SAPIENand CoreValve devices in a single center: the Milan experience. JACCCardiovasc Interv 2010;3:1110-21.

5. Webb JG, Altwegg L, Boone RH, et al. Transcatheter aortic valveimplantation: impact on clinical and valve-related outcomes.Circulation 2009;119:3009-16.

6. Sharp AS, Michev I, Maisano F, et al. A new technique for vascularaccess management in transcatheter aortic valve implantation.Catheter Cardiovasc Interv 2010;75:784-93.

7. Leon MB, Piazza N, Nikolsky E, et al. Standardized endpointdefinitions for transcatheter aortic valve implantation clinical trials. JAm Coll Cardiol 2011;57:253-69.

8. D'Agostino Jr RB. Propensity score methods for bias reduction in thecomparison of a treatment to a non-randomized control group. StatMed 1998;17:2265-81.

9. Rubin DB. Estimating causal effects from large data sets usingpropensity scores. Ann Intern Med 1997;127:757-63.

10. Leuven E, Sianesi B. PSMATCH2: Stata module to perform fullMahalanobis and propensity score matching, common supportgraphing, and covariate imbalance testing. http://ideas.repec.org/c/boc/bocode/s432001.html. 2003 Version 4.0.4; 10 November2010.

11. Caliendo M, Kopeinig S. Some practical guidance for the imple-mentation of propensity score matching. J Econ Surveys 2008;22:31-72.

12. Tamburino C, Capodanno D, Ramondo A, et al. Incidence andpredictors of early and late mortality after transcatheter aortic valveimplantation in 663 patients with severe aortic stenosis. Circulation2011;123:299-308.

13. Zahn R, Gerckens U, Grube E, et al. Transcatheter aortic valveimplantation: first results from a multi-centre real-world registry. EurHeart J 2011;32:198-204.

14. Rodes-Cabau J,Webb JG, Cheung A, et al. Transcatheter aortic valveimplantation for the treatment of severe symptomatic aortic stenosis inpatients at very high or prohibitive surgical risk: acute and lateoutcomes of the multicenter Canadian experience. J Am Coll Cardiol2010;55:1080-90.

15. Bose AK, Aitchison JD, Dark JH. Aortic valve replacement inoctogenarians. J Cardiothorac Surg 2007;2:33.

16. Dewey TM, Brown D, Ryan WH, et al. Reliability of risk algorithms inpredicting early and late operative outcomes in high-risk patientsundergoing aortic valve replacement. J Thorac Cardiovasc Surg2008;135:180-7.

17. Thourani VH, Ailawadi G, Szeto WY, et al. Outcomes of surgicalaortic valve replacement in high-risk patients: a multiinstitutionalstudy. Ann Thorac Surg 2011;91:49-55.

18. Florath I, Albert A, Boening A, et al. Aortic valve replacement inoctogenarians: identification of high-risk patients. Eur J CardiothoracSurg 2010;37:1304-10.

19. Wendt D, Osswald BR, Kayser K, et al. Society of Thoracic Surgeonsscore is superior to the EuroSCORE determining mortality in high riskpatients undergoing isolated aortic valve replacement. Ann ThoracSurg 2009;88:468-74.

20. Gummert JF, Funkat A, Osswald B, et al. EuroSCORE overestimatesthe risk of cardiac surgery: results from the national registry of theGerman Society of Thoracic and Cardiovascular Surgery. Clin ResCardiol 2009;98:363-9.

21. Osswald BR, Gegouskov V, Badowski-Zyla D, et al. Overestimation ofaortic valve replacement risk by EuroSCORE: implications forpercutaneous valve replacement. Eur Heart J 2009;30:74-80.

22. Mussardo M, Latib A, Chieffo A, et al. Periprocedural and short-termoutcomes of transfemoral transcatheter aortic valve implantation withthe Sapien XT as compared with the Edwards Sapien Valve. J Am CollCardiol Interv 2011;4:743-50.

23. Gurvitch R, Toggweiler S, Willson AB, et al. Outcomes andcomplications of transcatheter aortic valve replacement using aballoon expandable valve according to the Valve Academic ResearchConsortium (VARC) guidelines. EuroIntervention 2011;7:41-8.

Latib et al 917American Heart JournalVolume 164, Number 6

24. Walther T, Schuler G, Borger MA, et al. Transapical aortic valveimplantation in 100 consecutive patients: comparison to propensity-matched conventional aortic valve replacement. Eur Heart J 2010;31:1398-403.

25. Wenaweser P, Pilgrim T, Roth N, et al. Clinical outcome andpredictors for adverse events after transcatheter aortic valve

implantation with the use of different devices and access routes. AmHeart J 2011;161:1114-24.

26. Rodes-Cabau J, Dumont E, Boone RH, et al. Cerebral embolismfollowing transcatheter aortic valve implantation: comparison oftransfemoral and transapical approaches. J Am Coll Cardiol 2011;57:18-28.