www.elsevier.com/locate/ejctsEuropean Journal of Cardio-thoracic Surgery 32 (2007) 469—474

Review

Predictors of mortality after aortic valve replacement

Yanto Sandy Tjang a,b,c,*, Yvonne van Hees d, Reiner Korfer b,Diederick E. Grobbee a,c, Geert J.M.G. van der Heijden a,e

a Julius Center of Health Sciences and Primary Care, University Medical Center, Utrecht, The NetherlandsbDepartment of Thoracic & Cardiovascular Surgery, Heart & Diabetes Center NRW, Bad Oeynhausen, Germany

cNetherlands Institutes for Health Sciences, Erasmus Medical Center, Rotterdam, The NetherlandsdHartenzorg Foundation, St. Antonius Hospital, Nieuwegein, The Netherlands

eDepartment of Cardio-Thoracic Surgery/Heart Lung Center Utrecht, University Medical Center, Utrecht, The Netherlands

Received 20 March 2007; received in revised form 1 June 2007; accepted 11 June 2007; Available online 19 July 2007

Summary

Aortic valve replacement (AVR) is recommended as a standard surgical procedure for aortic valve disease. Still the evidence for commonlyclaimed predictors of post-AVR prognosis, in particular mortality, appears scant. This systematic review reports on the evidence for predictors ofpost-AVR mortality, and may be helpful in pre-surgical risk-stratification. In PubMed, we searched for original reports of post-AVR follow-upstudies. We assessed the quality of study design and methods with a standardized checklist. Data of the reported predictors of mortality andoutcomes were extracted. Twenty-eight studies met our inclusion criteria. Sixteen studies were considered of high quality. There is strongevidence that the risk of early mortality is increased by emergency surgery, while the risk of late mortality is increased with older age andpreoperative atrial fibrillation. There is moderate evidence that the risk of early mortality is increased by older age, aortic insufficiency, coronaryartery disease, longer cardiopulmonary bypass time, reduced left ventricular ejection fraction (LV-EF), infective endocarditis, hypertension,mechanical valves, preoperative pacing, dialysis-dependent renal failure and valve size; and that the risk for late mortality is increased byemergency surgery and urgency of the operation. There is little evidence for high New York Heart Association class, concomitant coronary arterybypass graft andmany other commonly claimed risk factors for post-AVRmortality. The reported evidence on predictors of post-AVRmortality willhelp for pre-surgical risk-stratification, i.e. to discern patients at high or low risk for early and late post-AVR mortality. Future prognostic studiesshould take the evidence from this review into account and should focus on derivation of a predictive model for post-AVR survival.# 2007 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved.

Keywords: Aortic valve replacement; Mortality; Prognosis; Systematic review

1. Introduction

Aortic valve replacement (AVR) is recommended as astandard surgical procedure for most patients with sympto-matic aortic valve disease [1]. Currently, AVR accounts for13% of all adult cardiac surgery [2], and remains the mostcommon procedure among all cardiac valve operations in theUnited States [3]. The 5-year mortality of aortic valve diseasewithout surgery is estimated to range from 50% to 80% [4].After surgery, patients with severe aortic valve disease showdramatic improvement in their cardiovascular symptoms andsurvival [5,6].

Although patients undergoing AVR are at relatively highrisk for post-procedural mortality, evidence for commonly

* Corresponding author. Address: C/o Geert J.M.G. van der Heijden, JuliusCenter of Health Sciences and Primary Care, University Medical Center, PO Box85500, 3508 GA Utrecht, The Netherlands. Tel.: +31 30 250 9305;fax: +31 30 250 5485.

E-mail address: ystjang@hotmail.com (Y.S. Tjang).

1010-7940/$ — see front matter # 2007 European Association for Cardio-Thoracicdoi:10.1016/j.ejcts.2007.06.012

claimed predictors of post-AVR prognosis, in particularmortality, appears scant [7,8]. To date, a systematicsynthesis of evidence for predictors of post-AVR prognosisis lacking. Knowledge of predictors for post-AVR prognosiscould be helpful in pre-surgical risk-stratification. Therefore,we systematically reviewed the available evidence forpredictors of early and late post-AVR prognosis.

2. Material and methods

2.1. Identification and selection of publications

We searched in PubMed (up to March 2005) for evidence onpredictors of post-AVR prognosis by using search string:[(((‘Aortic Valve’[MH] OR ‘Aortic Valve Stenosis’[MAJR] OR‘Aortic Valve Prolapse’[MAJR] OR ‘Aortic Valve Insufficien-cy’[MAJR]) AND (Surgery [Subheading] OR ‘Cardiac SurgicalProcedures’[MeSH]) AND (prognos*[Title/Abstract] OR (first[Title/Abstract] AND episode [Title/Abstract]) OR cohort

Surgery. Published by Elsevier B.V. All rights reserved.

Y.S. Tjang et al. / European Journal of Cardio-thoracic Surgery 32 (2007) 469—474470

Table 1Methodological quality assessment of prognostic studies on aortic valve repla-cement

No. Study Biased Unclear

V V P C

1 Kvidal et al. [6] p2 p4 c12 Chiappini et al. [14] v3 p3 c1 c3 c53 Mistiaen et al. [15] v3 c2 c3 c5 c64 Ikonomidis et al. [16] v3 c1 c3 c5 c65 Hannan et al. [7] v2 c3 c4 c5 c66 Asimakopoulos et al. [17] v2 p2 p4 c1 c3 c57 Verheul et al. [18] v3 p2 p4 c1 c2 c3 c58 Chocron et al. [19] v3 c3 c5 c69 Aranki et al. [20] v3 c4 c6

10 Galloway et al. [21] v3 p2 p4 c1 c3 c5 c611 Nowicki et al. [22] v2 c3 c4 c5 c612 Cohen et al. [2] v2 c1 c613 He et al. [8] v3 c1 c2 c414 Lytle et al. [24] v2 c1 c3 c5 c615 Scott et al. [25] v3 c2 c4 c616 Morris et al. [23] v2 c3 c517 Gardner et al. [26] v2 v3 c1 c3 c4 c5 c618 Ruel et al. [27] v2 v3 p2 p4 c3 c5 c619 Blais et al. [28] v2 v3 c1 c3 c4 c5 c620 Florath et al. [29] v2 v3 p4 c3 c4 c521 Bloomstein et al. [30] v2 v3 c3 c4 c5 c622 Astor et al. [31] v2 v3 c1 c3 c4 c5 c623 Jamieson et al. [32] v2 v3 p2 p4 c3 c4 c5 c624 Kohl et al. [33] v2 v3 c1 c625 Tseng et al. [34] v2 v3 c1 c3 c5 c626 Gehlot et al. [35] v2 v3 c3 c5 c627 Elayda et al. [36] v2 v3 C1 c628 Hanayama et al. [37] v1 v2 v3 c3 c5 c6

Positive scores were summed to give an overall validity scores (V1, V2, V3) andincluded in the analysis.Judgment: (1) clear (2) unclear (3) biased.Checklist and explanation of quality assessment criteria:Study population: (1) C1: Clear description of inclusion and exclusion criteria:Positive if at least the following were included: age, period of study, type ofstudy population. (2) C2: Clear definition of domain of study: Positive if patientsource was described.Response: (1) V1: Positive if the response rate was �80% or loss to follow-up<20% (in terms of mortality). (2) C3: Positive if information was presentedabout patient/disease characteristics of responders and non-responders or ifthere was no selective response.Follow-up: (1) C4: Positive if the follow-up period was at least 12 months. (2)C5: Positive if demographic/clinical information (patient/disease character-istics such as age, sex and other potential prognostic predictors) was presentedfor completers and those lost to follow-up/drop-outs at the main moment ofoutcomemeasurement, or no selective drop-outs/lost to follow up, or no drop-outs/lost to follow-up.Treatment: (1) V2: Clear description of surgical procedure(s): Positive if fullydescribed or standardized.Outcome: (1) P1: Clear definition of outcome: Positive if standardized ques-tionnaires/interview or objective outcome measurements were used for eachfollow-up measurement.Predictors: (1) P2: Clear definition of predictors: Positive if baseline data werepresented.Data presentation: (1) C6: Positive if frequency, percentage or mean, median(interquartile range) and standard deviation/CI (confidence interval) werereported for the most important outcome measures. (2) C7: Positive if fre-quency, percentage or mean, median (interquartile range) and standarddeviation/CI were reported for the most important predictors. (3) P3: Positiveif the number of cases in the multivariate analysis was at least ten times thenumber of independent variables in the analysis. (4) V3: Positive if theassessment of predictors was made blind to outcome status.Validity: Item V1, V2, and V3.Precision: Item P1, P2, and P3.Comparability (combinability): Item C1, C2, C3, C4, C5, C6, and C7.

[Title/Abstract])) OR ((Aorta [Tiab] OR Aortic [Tiab]) AND(Valve [Tiab] OR Replacement [tiab] OR Surgery[tiab])) AND((prognos*[Title/Abstract] OR (first[Title/Abstract] AND epi-sode[Title/Abstract]) OR cohort[Title/Abstract])))]. To iden-tify all relevant original publications, two reviewers (YST,YvH) independently screened the titles and abstracts of allcitations harvested with our search filter. The full publicationwas screened when, after reading the abstract, doubt aboutits relevance persisted. To identify additional relevantpublications we examined the reference lists and relatedarticles in PubMed for all selected publications. A publicationwas selected for methodological quality assessment when itfulfilled the following selection criteria: report (1) early orlate outcomes after AVR, either with or without concomitantprocedures; (2) for predictors for mortality; (3) as clinicalcohort study; (4) in the English language. Studies on AVR (a)using aortic root prosthesis or minimal invasive procedure,and those (b) exclusively focused on children were excluded.

2.2. Appraisal of study design

With a standardized checklist for study design, tworeviewers (YST, YvH) independently appraised the studydesign reported in selected publications. This checklist(Table 1) [9] includes each three items on validity andprecision of methods (V1, V2, V3 and P1, P2, P3) and sevenitems on clinical aspects of the study design (criteria C1, C2,C3, C4, C5, C6, C7). The legend of Table 1 provides anexplanation of each item. Disagreements between the tworeviewers on checklist items were resolved during aconsensus discussion. In case of persisting disagreement,appraisal by a third reviewer (GvdH) resulted in a decision.

Checklist items were based on theoretical considerationsfrom the guidelines for reporting morbidity and mortalityafter cardiac valvular operations [10], and methods forprognostic studies [11]. The score on validity items (V1, V2,V3) was summed and used as methodological quality score.With a range from 0 to 6, low scores reflect poormethodological quality, i.e. they are likely to report biasedresults due to deficiencies in study methods. A priori, wegraded a score of three points or more as high quality and ascore of 1 or 2 points as low quality.

2.3. Data extraction

Details on study population, response, follow-up, out-come measurement and predictors were extracted from theincluded studies. Detailed characteristics of the includedstudies are presented on the Internet.1 For each study thereported data on the univariate, and if available, multi-variate association between predictors and outcome weretabulated. Risks ratios, odds ratios or hazard ratio above 2 orbelow 0.5 was considered clinically relevant, while 95%confidence interval excluding unity or a p-value <0.05 wasconsidered statistical significant.

To facilitate comparison and interpretation of results, thestudies were categorized according to early mortality, i.e.

1 http://www.juliuscenter.nl/appendix.pdf.

Y.S. Tjang et al. / European Journal of Cardio-thoracic Surgery 32 (2007) 469—474 471

Table 2Most frequently reported predictors of mortality after aortic valve replace-

30-days (operative) mortality or hospital mortality, and latemortality (otherwise).

2.4. Data analysis

To date, methods for statistical pooling of data fromprognostic studies are lacking. Therefore, we classifiedevidence according to four levels of evidence. Theclassification table included the number of studies reportingan increased or reduced risk of a poor prognosis for aparticular predictor, the methodological quality of thesestudies, and the consistency of the evidence across thesestudies [12,13].

2.4.1. Strong evidenceOf at least two high quality studies on the same predictor,

at least 75% report an effect in the same direction, i.e. therisk increase or decrease depending on predictor status, thatis clinically important, statistically significant, or both.Strong evidence should always be incorporated in decision-making.

2.4.2. Moderate evidenceOf at least one high quality study and one low quality study

on the same predictor, at least 75% report an effect in thesame direction, i.e. the risk increase or decrease dependingon predictor status, that is clinically important, statisticallysignificant, or both. Moderate evidence should always beconsidered in decision-making.

2.4.3. Weak evidenceAt least one high quality study reports an effect that is

clinically important, statistically significant or both; or atleast 75% of at least three low quality studies on the samepredictor report an effect in the same direction that isclinically important, statistically significant, or both. Weakevidence may be considered in decision-making.

2.4.4. Inconclusive evidenceLess than three low quality cohorts, or irrespective of

study quality inconsistent effects. Inconclusive evidenceindicates areas where research may be needed.

ment

Predictor Number of studies reporting the predictor

Early (n = 25) Late (n = 14)

NYHA III—IV 10 (40) 6 (43)NYHA IV 9 (36)Urgency 8 (32) 4 (21)Concomitant CABG 7 (28) 4 (29)Preoperative AF 7 (28) 3 (21)Emergency 7 (28) 2 (14)Increased age 5 (20)Age >70 years 5 (20)Older age 7 (50)Renal dysfunction 4 (16)CPB time >120 min 4 (16)CAD 3 (21)Male gender 3 (21)MI 4 (21)

AF: atrial fibrillation, CABG: coronary artery bypass graft, CAD: coronaryartery disease, CPB: cardiopulmonary bypass, MI: myocardial infarction,NYHA: New York Heart Association. Values are number (percentage).

3. Results

3.1. Identification and selection of the literatures

With our search we identified 1920 citations, 1711 withour PubMed search filter and 209 by screening citations andrelated articles. After assessment of title and abstracts 55titles were selected for assessment of the full publications.Subsequently, 28 articles met our selection criteria andremained for critical appraisal [2,6—8,14—37]. For duplicatepublications, only the most recent or most complete reportwas included.

3.2. Methodological quality assessment

Table 1 presents the results of the critical appraisal. Theirmethodological quality score ranked the studies: the sum of

internal validity criteria V1, V2, V3 (min—max: 0—6). Thequality score ranged from 0 to 6 points, with a median scoreof 4 points. Of the 28 studies, 16 (57%) were of high quality.Among the included 28 studies, 21 (75%) did not describeblinding (item V3,) clearly; 17 (61%) did not fully describe(standardized) surgical procedures (item V2). An additional10 (36%) did not report on follow-up over 12 months (itemC4); 6 (21%) did not report standardized assessment ofoutcome (item P1); and 6 (21%) did not include informationabout non-responders and responders (item C3).

3.3. Study characteristics

In total11 studies [2,6,14—16,19,21,24,33—35] reportedonearly and late mortality; 14 studies exclusively reported onearly mortality [7,8,20,22,23,25,26,28—32,36,37] another 3studies exclusively reported on late mortality [17,18,27].There was considerable variation among the studies withrespect to the length of follow-up (range 30 days to 187months). The outcome was denoted as operative mortality in13 studies [2,8,16,18,20,21,23,25,29,32—35], as either in-hospital mortality or hospital mortality in 7 studies[7,14,15,19,22,24,31], and as early mortality [6,17,27,30,36,37], or short-termmortality [26,28] in 8 studies. For 9 (32%)of the 28 studies [7,15,16,21,22,24,30—32] effects were notadequately reported.

Altogether, 106,660 patients were included. Sample sizevaried between 83 patients [33] and 46,397 patients [31],with age ranging from 18 to 93 years. In total, 6036 patientsdied during early follow-up period, giving an overall earlymortality risk of 0.06, ranging from 0.02 [27] to 0.18 [36].Studies reporting on late mortality [2,6,14—19,24,33—36]accumulated 48,682 patient-years of follow-up, while 2109patients died. Thereby, the late mortality rate is 4.3 per 100patient-years.

The 28 included studies reported on 91 predictors of earlymortality and 38 predictors of late mortality. Table 2 showsthe most frequently reported predictors of early and latemortality. Althoughmost of the 28 studies reported outcomes

Y.S. Tjang et al. / European Journal of Cardio-thoracic Surgery 32 (2007) 469—474472

Table 3Level of evidence for important predictors of early mortality in aortic valve replacement

Predictor Cohortassessed

Positivefindings

PositiveHQ

PositiveLQ

Negativefindings

NegativeHQ

NegativeLQ

Level ofevidence

Emergency 7 6/7 (86%) 2 4 1/7 (14%) 0 1 StrongAge >70 years 5 4/5 (80%) 1 3 1/5 (20%) 1 0 ModerateAge >80 years 3 3/3 (100%) 1 2 0 0 0 ModerateAortic insufficiency 2 2/2 (100%) 1 1 0 0 0 ModerateCoronary artery disease 3 3/3 (100%) 1 2 0 0 0 ModerateCPB time >120 min 4 3/4 (75%) 1 2 1/4 (25%) 0 1 ModerateEF < 35% 2 2/2 (100%) 1 1 0 0 0 ModerateInfective endocarditis 2 2/2 (100%) 1 1 0 0 0 ModerateHypertension 2 2/2 (100%) 1 1 0 0 0 ModerateMechanical valve 3 3/3 (100%) 1 2 0 0 0 ModeratePreoperative pacing 2 2/2 (100%) 1 1 0 0 0 ModerateDialysis-dependent RF 3 3/3 (100%) 1 2 0 0 0 ModerateValve size 2 2/2 (100%) 1 1 0 0 0 Moderate

CPB: cardiopulmonary bypass, EF: ejection fraction, HQ: high quality, LQ: low quality, RF: renal failure.Predictors with weak evidence: concomitant CABG �3 vessel, creatinine >110 mmol/l, digitalis use, unstable hemodynamic, hypercholesterolemia, LV-aortagradient >60 mmHg, male gender, renal dysfunction, respiratory failure, smoking, systolic PAP �50 mmHg, valve size 19 mm, and valve type.Predictors with inconclusive evidence: age 76—79 years, age >71 years + height <1.57 m, aortic cross-clamping, BSA <1.70 m2, BUN >25 mg/100 ml, cardioplegiause, cerebrovascular disease, concomitant CABG, concomitant MVR, concomitant surgery, COPD, creatinine >1.3 mg/l, diabetes mellitus, EF <40%, EF <45%,female + CABG + BMI >29 m2, female gender, multiple reoperation, myocardial infarction, NYHA, NYHA III/IV, NYHA IV, prosthesis-patient-mismatch (moderate-severe), postoperative arrhythmia, postoperative infection, preoperative atrial fibrillation, previous CABG, previous cardiac operation, previous percutaneousvalvuloplasty, re-AVR, salvage status, shock, time 1994—1996, time 1997—2000, triple vessel disease, and urgency.

of multivariate analyses, only 4 studies [22,26,29,32]reported on a prognostic prediction model.

3.4. Evidence summary

Tables 3 and 4 present a qualitative summary of theavailable evidence for different predictors and theirassociation with early or late mortality. Included arepredictors reported in at least one study with a clinicalrelevant effect, statistical significant effect, or both.

There is strong evidence that the risk of early mortality isincreased by emergency operations, and that the risk of latemortality is increased by older age and preoperative atrialfibrillation. In addition, there is moderate evidence that therisk of early mortality is increased by older age, aorticinsufficiency, coronary artery disease, longer cardiopulmon-ary bypass time, reduced LV-EF, infective endocarditis,hypertension, mechanical valves, preoperative pacing,dialysis-dependent renal failure and valve size. Finally, thereis moderate evidence that the risk for late mortality isincreased by emergency and urgency of the operation. HighNew York Heart Association (NYHA) class and concomitant

Table 4Level of evidence for important predictors of late mortality in aortic valve replace

Predictor Cohortassessed

Positivefindings

PositiveHQ

PosLQ

Age �71 years 2 2/2 (100%) 2 0Preoperative AF 3 3/3 (100%) 3 0Emergency 2 2/2 (100%) 1 1Urgency 3 3/3 (100%) 1 2

AF: atrial fibrillation, HQ: high quality, LQ: low quality.Predictors with weak evidence: acute pulmonal edema, age 51—60 years, age 61—70carcinoma, and race (African-American).Predictors with inconclusive evidence: aortic insufficiency, bleeding events, concomidysfunction, male gender, mechanical valve (no warfarin), myocardial infarction,operative) and valve type.

coronary artery bypass graft (CABG) were examined in a largenumber of studies, but the evidence remains inconclusive asto whether they increase the risk for early or late mortality.Other predictors were only reported once. Hence, theevidence on their predictive value remains inconclusive.

4. Discussion

This systematic review shows that there is considerableevidence for several patient and procedural characteristicsas risk factors for post-AVR mortality. The risk for early post-AVR mortality is increased with emergency operation, olderage, aortic insufficiency, coronary artery disease, longercardiopulmonary bypass time, reduced LV-EF, infectiveendocarditis, hypertension, mechanical valve, preoperativepacing, dialysis-dependent renal failure and with increasingvalve size. In addition, the risk for late post-AVR mortalityincreases with emergency operation, older age and pre-operative atrial fibrillation. There is insufficient evidence formany other commonly claimed putative predictors of post-AVR mortality, notably NYHA class and concomitant CABG.

ment

itive Negativefindings

NegativeHQ

NegativeLQ

Level ofevidence

0 0 0 Strong0 0 0 Strong0 0 0 Moderate0 0 0 Moderate

years, age >75 years, age >80 years, diabetes mellitus, EF, EF <35%, previous

tant CABG, concomitant cardiac operation, COPD, coronary artery disease, LV->2 comorbidity, NYHA IIIB/IV, peak gradient >75 mmHg, renal failure (post-

Y.S. Tjang et al. / European Journal of Cardio-thoracic Surgery 32 (2007) 469—474 473

To our knowledge, this is the first systematic review onpredictors of mortality after aortic valve replacement. Weidentified and scrutinized 28 prognostic cohort studies. Themajority suffered from flaws in the design and conduct. Themethods of only 16 of the 28 studies were of sufficient quality.Therewas considerable heterogeneity regarding study design;notably studypopulation,definitionofpredictors,definitionofoutcome measures and length of follow-up. Many studiesfocusedonunivariate analyses,while some reportedoutcomesofmultivariateanalyses.Only4 studies [22,26,29,32] reportedon a developed prognostic model. For a relatively largenumber of studies outcomemeasures were not clearly definedor outcome data were not clearly reported.

4.1. Review methods

Because publication bias is likely in every systematicreview, the reported risk estimates turn away from a nullresult, and so are likely to be optimistic [38,39]. Aftercomprehensively searching PubMed, we thoroughly checkedrelated articles and reference lists of included publications.Since we only included full English language publications, wemight have missed studies. Therefore, possible language andretrieval bias that may contribute to overestimation of riskcannot be excluded [40—42]. This strengthens the conclusionthat there is little evidence for NYHA class and concomitantCABG and many other commonly claimed putative predictorsof post-AVR mortality.

We used an explicit and transparent method to summarizethe available evidence. With a cut-off point of 40% for highquality study methods, instead of 50%, the evidence forincreased early mortality risk with older age, aorticinsufficiency, coronary artery disease, valve size, and foran increased late mortality risk with emergency operation,changed from moderate to strong.

4.2. Implications for care

The reported evidence on predictors of post-AVRmortalitywill help pre-surgical risk-stratification, i.e. to discernpatients at high or low risk of early and late post-AVRmortality. The results from this systematic review may servethe development of alternative management strategies, inparticular for the patients at very high-risk [43]. In addition,physicians should bear in mind that there is hardly anyevidence for many commonly claimed predictors of post-AVRmortality.

4.3. Implications for research

Sufficiently large and well-designed prognostic studies onpost-AVR mortality are needed. Such future studies shouldtake the evidence from this review into account. They shouldin particular focus on adequate definition of outcomemeasures, assert blinding of subjective endpoints, eitheror both standardize or accurately describe the surgicalprocedures and assure completeness of follow-up atsufficient length. Instead of risk ratios, odds ratios or hazardratios, both the numerator (number of events) and thedenominator (total number of patients included, theiraccrued person-time of follow-up, or both) should be

reported accurately [44]. A multivariable analysis ofpredictors may contribute to stratification of patients athigh or low risk for early or late post-AVR mortality.Therefore, future prognostic studies should focus on deriva-tion of a predictive model for post-AVR survival. Moreover, asthe long-term consequences of AVR extend beyond post-operative complications or survival, improvement in qualityof life is receiving more attention as an important goal ofsurgery. Relatively little is known about changes in post-AVRquality of life. Therefore, quality of life assessment should beincluded in the future post-AVR outcome studies.

References

[1] Braunwald E. Aortic valve replacement: an update at the turn of themillennium. Eur Heart J 2000;21:1032—3.

[2] Cohen G, David TE, Ivanov J, Armstrong S, Feindel CM. The impact of age,coronary artery disease, and cardiac comorbidity on late survival afterbioprosthetic aortic valve replacement. J Thorac Cardiovasc Surg1999;117:273—84.

[3] Sedrakyan A, Hebert P, Vaccarino V, Paltiel AD, Elefteriades JA, Mattera J,Lin Z, Roumanis SA, Krumholz HM. Quality of life after aortic valvereplacement with tissue and mechanical implants. J Thorac CardiovascSurg 2004;128:266—72.

[4] Waszyrowski T, Kasprzak JD, Krzeminska-Pakula M, Dziatkowiak A,Zaslonka J. Early and long-term outcome of aortic valve replacementwith homograft versus mechanical prosthesis—8-year follow-up study.Clin Cardiol 1997;20:843—8.

[5] Kvidal P, Bergstrom R, Malm T, Stahle E. Long-term follow-up of morbidityand mortality after aortic valve replacement with a mechanical valveprosthesis. Eur Heart J 2000;21:1099—111.

[6] Kvidal P, Bergstrom R, Horte LG, Stahle E. Observed and relative survivalafter aortic valve replacement. J Am Coll Cardiol 2000;35:747—56.

[7] Hannan EL, Racz MJ, Jones RH, Gold JP, Ryan TJ, Hafner JP, Isom OW.Predictors of mortality for patients undergoing cardiac valve replace-ments in New York State. Ann Thorac Surg 2000;70:1212—8.

[8] He GW, Acuff TE, Ryan WH, Douthit MB, Bowman RT, He YH, Mack MJ.Aortic valve replacement: determinants of operative mortality. AnnThorac Surg 1994;57:1140—6.

[9] Kuijpers T, van der Windt DA, van der Heijden GJ, Bouter LM. Systematicreview of prognostic cohort studies on shoulder disorders. Pain 2004;109:420—31.

[10] Edmunds Jr LH, Clark RE, Cohn LH, Grunkemeier GL, Miller DC, Weisel RD.Guidelines for reporting morbidity and mortality after cardiac valvularoperations. Ad Hoc Liaison Committee for Standardizing Definitions ofProsthetic Heart Valve Morbidity of The American Association for ThoracicSurgery and The Society of Thoracic Surgeons. J Thorac Cardiovasc Surg1996;112:708—11.

[11] Altman DG. Systematic reviews of evaluations of prognostic variables. BrMed J 2001;323:224—8.

[12] Sackett DL, Straus SE, Richardson WS. Evidence-based medicine. How topractice and teach EBM. Guidelines. Edinburgh: Churchill Livingstone;2000.

[13] Ariens GA, van Mechelen W, Bongers PM, Bouter LM, van der Wal G.Physical risk factors for neck pain. Scand J Work Environ Health 2000;26:7—19.

[14] Chiappini B, Camurri N, Loforte A, Di Marco L, Di Bartolomeo R, MarinelliG. Outcome after aortic valve replacement in octogenarians. Ann ThoracSurg 2004;78:85—9.

[15] Mistiaen W, Van Cauwelaert P, Muylaert P, Wuyts F, Harrisson F, Bortier H.Risk factors and survival after aortic valve replacement in octogenarians.J Heart Valve Dis 2004;13:538—44.

[16] Ikonomidis JS, Kratz JM, Crumbley 3rd AJ, Stroud MR, Bradley SM, SadeRM, Crawford Jr FA. Twenty-year experience with the St Jude Medicalmechanical valve prosthesis. J Thorac Cardiovasc Surg 2003;126:2022—31.

[17] Asimakopoulos G, Edwards MB, Taylor KM. Aortic valve replacement inpatients 80 years of age and older: survival and cause of death based on1100 cases: collective results from the UK Heart Valve Registry. Circula-tion 1997;96:3403—8.

Y.S. Tjang et al. / European Journal of Cardio-thoracic Surgery 32 (2007) 469—474474

[18] Verheul HA, van den Brink RB, Bouma BJ, Hoedemaker G, Moulijn AC,Dekker E, Bossuyt P, Dunning AJ. Analysis of risk factors for excessmortalityafter aortic valve replacement. J Am Coll Cardiol 1995;26:1280—6.

[19] Chocron S, Etievent JP, Viel JF, Etievent JP, Viel JF, Meneveau N, ClementF, Schipman N, Taberlet C, Alwan K. Aortic valve replacement in theelderly: a comparative assay of potential risk factor modification. J HeartValve Dis 1995;4:268—73.

[20] Aranki SF, Rizzo RJ, Couper GS, Adam DH, Collins Jr JJ, Gildea JS, KinchlaNM, Cohn LH. Aortic valve replacement in the elderly. Effect of genderand coronary artery disease on operative mortality. Circulation 1993;88:II17—23.

[21] Galloway AC, Colvin SB, Grossi EA, Bauman FG, Sabban YP, Esposito R,Ribakove GH, Culliford AT, Slater JN, Glassman E. Ten-year experiencewith aortic valve replacement in 482 patients 70 years of age or older:operative risk and long-term results. Ann Thorac Surg 1990;49:84—91.discussion 91—3.

[22] Nowicki ER, Birkmeyer NJ, Weintraub RW, Leavitt BJ, Sanders JH, DaceyLJ, Clough RA, Quinn RD, Charlesworth DC, Sisto DA, Uhlig PN, OlmsteadEM, O’Connor GT, Northern New England Cardiovascular Disease StudyGroup and the Center for Evaluative Clinical Sciences Dartmouth MedicalSchool.. Multivariable prediction of in-hospital mortality associated withaortic and mitral valve surgery in Northern New England. Ann Thorac Surg2004;77:1966—77.

[23] Morris JJ, Schaff HV, Mullany CJ, Rastogi A, McGregor CG, Daly RC, FryeRL, Orszulak TA. Determinants of survival and recovery of left ventricularfunction after aortic valve replacement. Ann Thorac Surg 1993;56:22—9.discussion 29—30.

[24] Lytle BW, Cosgrove DM, Taylor PC, Goormastic M, Stewart RW, Golding LA,Gill CC, Loop FD. Primary isolated aortic valve replacement. Early andlate results. J Thorac Cardiovasc Surg 1989;97:675—94.

[25] Scott WC, Miller DC, Haverich A, Dawkins K, Mitchell RS, Jamieson SW,Oyer PE, Stinson EB, Baldwin JC, Shumway NE. Determinants of operativemortality for patients undergoing aortic valve replacement. Discriminantanalysis of 1479 operations. J Thorac Cardiovasc Surg 1985;89:400—13.

[26] Gardner SC, Grunwald GK, Rumsfeld JS, Cleveland Jr JC, Schooley LM,Gao D, Grover FL, McDonald GO, Shroyer AL. Comparison of short-termmortality risk factors for valve replacement versus coronary artery bypassgraft surgery. Ann Thorac Surg 2004;77:549—56.

[27] Ruel M, Kulik A, Lam BK, Rubens FD, Hendry PJ, Masters RG, Bedard P,Mesana TG. Long-term outcomes of valve replacement with modernprostheses in young adults. Eur J Cardiothorac Surg 2005;27:425—33.discussion 433.

[28] Blais C, Dumesnil JG, Baillot R, Simard S, Doyle D, Pibarot P. Impact ofvalve prosthesis-patient mismatch on short-term mortality after aorticvalve replacement. Circulation 2003;108:983—8.

[29] Florath I, Rosendahl UP, Mortasawi A, Bauer SF, Dalladaku F, Ennker IC,Ennker JC. Current determinants of operative mortality in 1400 patientsrequiring aortic valve replacement. Ann Thorac Surg 2003;76:75—83.

[30] Bloomstein LZ, Gielchinsky I, Bernstein AD, Parsonnet V, Saunders C,Karanam R, Graves B. Aortic valve replacement in geriatric patients:determinants of in-hospital mortality. Ann Thorac Surg 2001;71:597—600.

[31] Astor BC, Kaczmarek RG, Hefflin B, Daley WR. Mortality after aortic valvereplacement: results from a nationally representative database. AnnThorac Surg 2000;70:1939—45.

[32] JamiesonWR, Edwards FH, Schwartz M, Bero JW, Clark RE, Grover FL. Riskstratification for cardiac valve replacement. National Cardiac SurgeryDatabase. Database Committee of The Society of Thoracic Surgeons. AnnThorac Surg 1999;67:943—51.

[33] Kolh P, Lahaye L, Gerard P, Limet R. Aortic valve replacement in theoctogenarians: perioperative outcome and clinical follow-up. Eur J Car-diothorac Surg 1999;16:68—73.

[34] Tseng EE, Lee CA, Cameron DE, Stuart RS, Greene PS, Sussman MS,Watkins L, Gardner TJ, Baumgartner WA. Aortic valve replacement inthe elderly. Risk factors and long-term results. Ann Surg 1997;225:793—802. discussion 802—4..

[35] Gehlot A, Mullany CJ, Ilstrup D, Schaff HV, Orzulak TA, Morris JJ, SalyRC. Aortic valve replacement in patients aged eighty years and older:early and long-term results. J Thorac Cardiovasc Surg 1996;111:1026—36.

[36] Elayda MA, Hall RJ, Reul RM, Alonzo DM, Gillette N, Reul Jr GJ, Cooley DA.Aortic valve replacement in patients 80 years and older. Operative risksand long-term results. Circulation 1993;88:II11—6.

[37] Hanayama N, Fazel S, Goldman BS, Mitoff PR, Sever J, Fremes SE.Contemporary trends in aortic valve surgery: a single centre 10-yearclinical experience. J Card Surg 2004;19:552—8.

[38] Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR. Publication bias inclinical research. Lancet 1991;337:867—72.

[39] Burdett S, Stewart LA, Tierney JF. Publication bias and meta-analyses: apractical example. Int J Technol Assess Health Care 2003;19:129—34.

[40] Moher D, Fortin P, Jadad AR, Juni P, Klassen T, Le Lorier J, Liberati A, LindeK, Penna A. Completeness of reporting of trials published in languagesother than English: implications for conduct and reporting of systematicreviews. Lancet 1996;347:363—6.

[41] Moher D, Pham B, Klassen TP, Schulz KF, Berlin JA, Jadad AR, Liberati A.What contributions do languages other than English make on the results ofmeta-analyses? J Clin Epidemiol 2000;53:964—72.

[42] Egger M, Zellweger-Zahner T, Schneider M, Junker C, Lengeler C, Antes G.Language bias in randomised controlled trials published in English andGerman. Lancet 1997;350:326—9.

[43] Moons KG, Harrel FE, Steyerberg EW. Should scoring rules be basedon odds ratios or regression coefficients? J Clin Epidemiol 2002;55:1054—5.

[44] Christakis GT, Weisel RD, Fremes SE, Teoh KH, Skalenda JP, Tong CP,Azuma JY, Schwartz L, Mickleborough LL, Scully HE. Can the results ofcontemporary aortic valve replacement be improved? J Thorac Cardio-vasc Surg 1986;92:37—46.