ORIGINAL ARTICLE

Decision-making in aortic valve replacement:

bileaflet mechanical valves versus stented

bioprostheses

J.J.M. Takkenberg, J.P.A. Puvimanasinghe, L.A. van Herwerden, M.J.C. Eijkemans, E.W. Steyerberg,J.D.F. Habbema, A.J.J.C. Bogers

Background. Valve prosthesis selection for patientswho require aortic valve replacement is dependenton several interrelated factors. Often, more than onevalve type seems suitable for the indimdual patientand selection of a valve type may be difficult.Methods. The application of an evidence-basedmicrosimulation model as an objective tool tosupport the choice between a bileaflet mechanicalprosthesis and a stented bioprosthesis in theindividual patient is described. In addition, a pilotstudy investigating the effect ofknowledge gainedby this microsimulation model on prosthetic valvechoice by cardiothoracic surgeons and cardiologistsis presented for two hypothetical patients.Results. After implantation of a mechanical valve,bleeding and thromboembolism are common,especially in the elderly. After implantation of abioprosthesis, reoperation for structural failure isthe most important valve-related complication,especially in younger patients. Life expectancy afteraortic valve replacement is markedly reducedcompared with the general Dutch age-matchedpopulation, regardless of the type of valveimplanted. In the pilot study knowledge gainedby the microsimulation model caused a shift in thepreference towards a mechanical prosthesis inlinical experts.Conlusion. Microsimulation incorporating currentepidemiological data provides an objective tool toestimate prognosis for individual patients after

J.J.M. Takkenb.ig.J.P.A. Puvimanasinghe.LA. van Herwnrden.A.JJ.C. Bogers.Department of Cardiothoracic Surgery, Erasmus Medical Centre,PO Box 2040, 3000 CA Rotterdam.MJ.C. Elkemans.E.W. Styerb.rg.J.D.F. Habbema.Department of Public Health, Erasmus Medical Centre,PO Box 2040, 3000 CA Rotterdam.

Address for correspondence: J.J.M. Takkenberg.E-mail: Takkenberg@thch.azr.nI

aortic valve replacement with different valveprostheses. It may develop towards a useful clinicaldecision support system for valve prosthesisselection. (Nedi HeartJ2003;11:5-10.)

Key words: aortic valve replacement, prosthetic valvechoice, mechanical versus bioprostheses, micro-simulation, clinical decision support system

Prognosis after aortic valve replacement (AVR)depends on multiple interrelated factors associated

with the patient, the medical centre, and the type ofprosthesis used.' Given the number and complexity offactors that affect outcome afterAVR, it can be difficultto make an objective selection of the preferred valvesubstitute in the individual patient. The application ofan evidence-based microsimulation model to predictage- and gender-specific outcome after AVR isillustrated for the two most commonly used prostheticvalve types, bileaflet mechanical valves and stentedporcine bioprostheses. In addition, in a pilot study theeffect of presenting this microsimulation model tocardiothoracic surgeons and cardiologists on theirpreference for a particular valve substitute in two hypo-thetical patients is investigated.

Methods

Meta-analysis and microsimulationMeta-analyses were previously performed in order toobtain evidence on the outcome ofpatients afterAVRwith bileaflet mechanical valves (St Jude Medical) andstented porcine bioprostheses (Carpentier-Edwardsstandard or supra-annular, Hancock standard, modifiedorifice or II).23 The pooled reported estimates ofoperative mortality, valve-related events and theirconsequences (death, reoperation) were entered intoa microsimulation model.4

The basic assumption ofa microsimulation modelis that a disease follows a course in time that can beadequately characterised by a number ofdiscrete statesand the events that cause a change ofstate. After aorticvalve replacement, the patient can either die as a result

Netherlands Heart Journal, Volume 11, Number 1, January 2003 5

Decision-making in aortic valve replacement: bileaflet mechanical valves versus stented bioprostheses

ofthe procedure or stay alive. Ifthe patient stays alive,he or she remains at risk of developing valve-relatedevents for the rest of his or her life. Eventually thispatient will die of either valve-related or non-valve-related causes. A schematic representation of thesehealth states and events is given in figure 1.

From microsimulation, conclusions can be drawnfor a specific patient profile (for example, a 65-year-oldmale) by performing calculations ofrandom individuallife histories of a large number of 'cloned' patients.These calculations are repeated a number of times,thus producing a simulated or 'virtual' closed cohortof patients with identical characteristics. From thiscohort, the mean outcome can be calculated anddetailed insight can be obtained into the factors thataffect the outcome. An attractive feature of micro-simulation is that it has a memory. For example, it canadjust operative mortality, taking into account whetherthe individual patient has had previous aortic valvereplacements.

The information from the meta-analyses onoutcome after aortic valve replacement with bileafletmechanical valves and stented bioprostheses wasentered into the microsimulation model. Details onthe input ofthe microsimulation model can be foundin the appendix. Ten thousand life histories werecalculated for males at different ages (55, 65, 75 years)by randomly drawing the age ofdeath from the Dutchgeneral population life table. Since there is a highermortality rate among patients after aortic valve re-placement compared with the general population thatcannot be solely attributed to valve-related events,5'6 wemultiplied the age- and gender-specific mortalityhazard of the general population with an age- andgender-related hazard for excess mortality, based onprevious work3'79 (see appendix on page 10). In thestudy it was also assumed that operative mortalityincreased with age and with each reoperation. An earlyincreased hazard for endocarditis was employed. Afterimplantation ofa mechanical valve, the bleeding hazardderived from the meta-analysis increased with age,resulting in bleeding hazards ranging from 0.23% perpatient year in patients aged 35, 1.58% per patient yearin patients aged 60, to 4.9% per patient year in patientsaged 75.10 Mortality related to bleeding increased withage, resulting in bleeding-related annual mortality ratesvarying from 4% at age 35, 10% at age 60, to 17% atage 75.11L2 After implantation with a bioprosthesis, theoccurrence ofstructural valve failure derived from themeta-analysis was fitted into a Weibull model, whichis a generalisation ofthe exponential distribution thataccommodates a changing risk over time.'3 TheWeibull model has been used previously to describestructural valve failure in biological valves, where therisk ofstructural valve failure increases with time.'4"'5

For male patients at three different ages (55, 65and 75 years) life expectancy, event-free life expectancyand actual lifetime risk of the various valve-related'events by valve prosthesis type were calculated. To

Figure 1. Schematic representation ofdifferent health states ofapatient after aortic valve replacement.

explore the influence ofvariation in parameter estimatesa sensitivity analysis was performed. Hereto, we

doubled and halved the hazard rate ofbleeding eventsafter implantation with a mechanical prosthesis. Theeffect ofthese events on lifetime risks ofexperiencingvalve-related events was investigated. Finally, relativelife expectancy after aortic valve replacement comparedwith the general age- and gender-matched populationwas calculated, and the impact ofvalve-related eventson the reduction ofrelative life expectancy was assessed.

Validation ofthe model was done by comparing itsoutcome with the long-term outcome of aortic valvereplacement patients in a large dataset from Portland,7Oregon, US, as previously described.9

Pilot studyDuring the joint meeting of the Netherlands Societyof Cardiology and the Netherlands Association forCardiothoracic Surgery in The Hague, the Netherlandson 26 April 2002, a session on prosthetic valve choicein aortic valve replacement was organised. Precedingthe lecture on meta-analysis and microsimulation-basedestimates of outcome after AVR with a stentedbioprosthesis or a bileaflet mechanical valve, theaudience was asked to choose either a mechanical valveor a bioprosthesis for two hypothetical patients:

Case 1: A 66-year-old male, retired with an activelifestyle, presents with asymptomatic aortic valvestenosis (gradient 120 mmHg), no coronary arterydisease, normal left ventricular function, and no

concomitant morbidity. In case of mechanicalprosthesis, this patient wants to self-monitor the INRwhich is associated with low thromboembolic andbleeding risks.

Case 2: A 60-year-old female presents with dyspnoeaon exertion caused by severe aortic regurgitation after

Netherlands Heart Joumal, Volume 11, Number 1, January 2003

I "Primary AVR

6

Decision-making in aortic valve replacement: bileaflet mechanical valves versus stented bioprostheses

Table 1. Estimated occurrence of valve-related events and associated lethality by valve prosthesis type.

Mechanical valveLOR (lethality)

Stented bloprosthesisLOR (lethality)

Valve thrombosisThromboembolismBleedingEndocarditisNon-structural valve failureStructural valve failure

0.08%/year (25%)1.15%/year (7%)1.58%/year (10%)*0.45%/year (36%)0.42%/year (2%)0.00%/year (NA)

0.03%/year (67%)0.90%/year (19%)0.38%/year (21%)0.68%/year (34%)0.38%/year (5%)

Weibull function (figure 2)LOR=linearised annual occurrence rates; NA=not applicable.* In the microsimulation model an age-dependent incidence was used (see appendix)

an episode ofendocarditis. Left ventricular function isnormal-impaired; there is no coronary artery disease.The patient has non-insulin-dependent diabetesmellitus that is difficult to control (complianceproblem?). This patient is therefore at an increased riskofthromboembolism, bleeding and endocarditis.

The choice options were as follows: 1=definitelymechanical valve, 2=probably mechanical valve, 3=nopreference, 4=probably stented bioprosthesis,5=definitely stented bioprosthesis. After the lecturethe audience was confronted with the same two casesand with the evidence-based estimates of lifeexpectancy of the two cases according to themicrosimulation model. The audience was again askedto choose the preferred valve type for the twohypothetical cases. To investigate the effect of thepresentation on prosthetic valve preference in the twohypothetical cases the Wilcoxon matched-pairs signed-rank test was employed. SPSS 10.0 for Wimdows(SPSS, Chicago, Ill, USA) was used for all analyses.

Figure 2. Meta-analysis based Weibullfunction describing theactuarialfreedomfrom structural valvefailure (conditional onsurvival) after implantation ofa stented bioprosthesisforpatientsaged 55, 65 and 75years at the time ofimplantation.

Results

Meta-analysis and microsimulationAn overview of the estimated occurrence of valve-related events after implantation with a bileaflet mech-anical valve or a stented bioprosthesis from the meta-analyses is displayed in table 1. Figure 2 shows the age-dependent estimates offreedom from structural valvefailure after implantation of a stented bioprosthesisderived from the meta-analysis. Conditional onsurvival, median time to structural valve failure was13.8, 15.4, and 17.3 years for patients aged 55, 65and 75 years respectively at the time ofimplantation.

Figure 3 displays the calculated mean lifeexpectancy and event-free life expectancy after AVRwith a bileaflet valve or a bioprosthesis according tothe microsimulation model for male patients aged 55to 75 years. In figure 4 the lifetime risk ofexperiencingat least one valve-related event by valve prosthesis typeis shown for male patients aged 55, 65 and 75 years.After implantation with a mechanical valve the valve-

Figure 3. Calculatedmean andevent-free lift expectancy (LE andEFLE) afterAVR witha bileaflet valve ora bioprsthesis acordingto the microsimulation model.

Netherlands Heart Journal, Volume 11, Number 1, January 2003

e 100%-#A 80%-

E 70%-E 60%

50%

40%

30% _

20%

0%-

-75-year-old patient

65-year-old patient

55-year-old patient

5 10 15 20Time since valve replacement (years)

~20o 18 |- LE mechanical valvet LE bioprosthesis~ 16 - -x)K- EFLE mechanical valve

-A-EFLE bioprosthesis14

*64

2

male male male55 years 65 years 75 years

I .

7

Decision-making in aortic valve replacement: bileaflet mechanical valves versus stented bioprostheses

Figure4t Lifetime risk ofexperincing atlastone valve-relatedeentby valveprosthsisttpefr male patients aged 55, 65and 75ywatn.

related events that the patient encounters are mainlybleeding and thromboembolism, while after im-plantation with a bioprosthesis, structural valve failureis the most common valve-related event. The age

threshold for implantation of either valve type isapproximately 63 years. The interrupted lines in figure4 represent the effect of doubling and halving thehazard rates ofbleeding events on lifetime risks ofvalve-related events after implantation with a bileafletmechanical prosthesis. In case of a twofold increasedbleeding hazard the age threshold for implantation ofeither valve type is approximately 58 years, while thisis approximately 69 years for a twofold decreasedbleeding hazard. Figure 5 shows the relative lifeexpectancy of patients after AVR with a bileafletmechanical valve and stented bioprosthesis comparedwith the age- and gender-matched general Dutchpopulation. In addition, the relative life expectancy ofthe hypothetical case of no valve-related events, i.e.after implantation with the perfect valve substitute, isshown. The relative life expectancy after implantationwith a bileaflet mechanical valve, a stented bioprosthesisand the perfect valve substitute ranged from 75%, 69%and 81% respectively for a 55-year-old male patient,to 82%, 87% and 93% respectively for a 75-year-oldpatient.

Figure 5. Relative lift expectancy ofpatientsafterAVR comparedwith the age- andgender-matchedgeneral Dutch population.

Figure 6 shows the evidence-based microsimulationestimates of prognosis after aortic valve replacementwith either of the valve prostheses in the twohypothetical cases. Ofthe 80 questionnaires thatweredistributed among the participants of the meeting,only 31 were returned. Ofthese 31, 24 were completedcorrectly (in 7 only the first two questions were

answered). Table 2 represents the distribution of theanswers to the questionnaire. For case 1 the initialpreference was approximately 50% for a mechanicalprosthesis and 50% for a bioprosthesis. After thepresentation ofthe microsimulation model, there wasa significant shift in preference from the prostheticvalve to a mechanical prosthesis (Wilcoxon matchedpairs signed-rank test Z=-2.45; p=0.014). For case 2the initial preference was clearly in favour of a

bioprosthesis. After the presentation of the micro-simulation model no significant change in valvepreference was evident (Wilcoxon signed-rank testZ=-1.42; p=O.15) although a trend towards preferencefor a mechanical prosthesis was observed.

DiscussionMany interrelated factors play a role in the prognosisofpatients after AVR and it may therefore be hard toselect the preferred aortic valve substitute in the

Figure 6. Calculated outcomeafter AVR by prosthesis type ac-

cording to the mkrosimulationmodel.

Case 1: Total lift expectancy,

event-free liftexpectancyand re-

operation-free liftexpectancy arein favour of a mechanical

Case 2: Total life expectancyandreoperation-free lifte expectancyare better with a mechanicalprosthesis event-free lift expect-

ancyisb etterwitha bieprosthesis.

Netherlands Heart Journal, Volume 11, Number 1, January 2003

i 20 -

16 - .~~~~~~~ EF~~LE mechanical vlve-i- LE bb08"_s-1B-~~ ~ ~ ~~~~~~)-_ER. MoC.n V

I 14

12 -

°0 -

420-

male male male55 years 65 year 75 years

9:i0so-

70-60

j 50 - Healhy agmecMhd Dutch malS40- ~~The perfect valve substItuesZ30 _ Mechanical valve

20-male male male55years 65years 75yeam

8

Decision-making in aortic valve replacement: bileaflet mechanical valves versus stented bioprostheses

individual patient. Knowledge on outcome afterAVRis limited, with regard to both the length offollow-upand the number of patients, and clinical results are

usually reported on a group level which may be hardto translate to the individual patient.

This study shows that based on pooled evidencefrom clinical practice, microsimulation offers an

objective tool to support the decision for a particularaortic valve substitute in the individual patient byproviding age- and gender-specific estimates ofoutcome after AVR with stented bioprostheses andbileaflet mechanical valves. It is also a flexible tool,because it allows insight into the effect ofvariation ofthe hazard rates ofthe different valve-related events.

In addition, from a scientific point ofview micro-simulation is useful for obtaining improved knowledgeon the outcome after AVR. Microsimulation allowsdetermination of the impact of the different factorsthat cause the reduced life expectancy in patients afterAVR compared with the general population. Know-ledge ofthese factors can be used to optmise treatmentstrategies and eventually improve outcome.

This study focused on the choice between a bileafletmechanical prosthesis and a stented bioprosthesis. Afterimplantation of a mechanical valve, bleeding andthromboembolism are common, especially in the elder-ly. On the other hand, after implantation of a bio-prosthesis, reoperation for structural failure is the mostimportant valve-related complication, especially inyounger patients. The age threshold for implantingeither valve type is approximately 63 years whencomparing the lifetime risks of experiencing at leastone valve-related event. This is in good agreement withthe general recommendations ofthe American HeartAssociation and American College of Cardiology.'6However, this age threshold varies considerablydepending on the bleeding hazard, illustrating theimpact ofindividual patient characteristics and showingthat general recommendations do not always apply tothe individual patient.

Life expectancy after aortic valve replacement ismarkedly reduced compared with the general Dutchage-matched population, regardless ofthe type ofvalveimplanted. The observed reduction in life expectancycan only in part be explained by the occurrence ofvalve-related events and is more pronounced in theyounger age groups. One can currently only speculate

on the other possible causes of excess mortality afteraortic valve replacement. Sudden unexpected un-explained death and cardiac death related to valve dis-ease and cardiomyopathy appear to be the most likelycauses, but extensive further research is required.5'6With increasing age, valve-related events become moreimportant causes of mortality." "2 Especially in theseolder patient groups optimal selection of a valveprosthesis becomes increasingly important.

The pilot study described in this paper showed thatpresentation ofthe microsimulation model may cause

a shift in a clinical expert's preferences for a prostheticvalve for a particular patient. It should be noted thatthe shift seen in this pilot study, although in one casestatistically significant, was modest. The reasons forthis shift in preference were not documented, andshould be investigated in future studies on this subject.An important goal ofthe microsimulation model is toprovide the clinician with evidence-based objectiveinsight into the impact of several factors on outcomeafter aortic valve replacement. After using the micro-simulation model several times, the clinician will findhis insight into the determinants of outcome afteraortic valve replacement improves, which will facilitatethe objective selection ofthe best type ofvalve for theindividual patient.

Limitations of the data from the literature ne-cessitated certain structural assumptions. For example,a constant hazard was assumed for several valve-relatedevents, although these hazards may in fact be dependenton time and age. Another issue concerns the excessmortality that was built into the model. The age- andgender-related multiplicative hazard ratios were ob-tained from previous workon survival after implantationwith mechanical monoleaflet prostheses and stentedbioprostheses.38 We validated our excess mortalityestimates using the Portland dataset and found goodagreement.79 Therefore we are confident that it is agood reflection of reality. Finally, survival after AVRnot only depends on age and gender, but also on manyrisk factors, induding preoperative NYHAdass and thepresence of a coronary heart disease.6"7 These factorswere not taken into account in the microsimulationmodel. In this study we only considered male patientsfor the purpose of illustration. However, the model isalso capable of calculating life expectancy for femalepatients.

Nctherlands Heart Joumal, Volumc 11, Numbar 1, January 2003

Table 2. Distributions of the answers to the questionnaire (n=24).

Cas I Case I revised Case 2 Case 2 revlsed

Definitely mechanical valve 4 10 1 3Preferably mechanical valve 7 2 1 6No preference 1 5 5 1Preferably bioprosthesis 5 2 9 4Definitely bioprosthesis 7 5 8 10

9

Decision-making in aortic valve replacement: bileaflet mechanical valves versus stented bioprostheses

In conclusion, microsimulation allows for age- andgender-specific calculation of life expectancy, event-free life expectancy and lifetime risks ofthe occurrenceof valve-related events after aortic valve replacementwith both types of valve. It is both a useful clinicaldecision support system for valve prosthesis selectionand a promising tool to study factors that determineoutcome after aortic valve replacement. U

AcknowledgementsThis studywas in part supported by grant 99141 fromthe Board of Health Care Insurance (College voorZorgverzekeringen), the Netherlands. The presentationof this study at the joint meeting of the NetherlandsSociety ofCardiology and the Netherlands Society forCardiothoracic Surgery, 26 April 2002 in The Haguewas supported by St. Jude Medical BV.

References

1 Grunkemeier GL, London MR. Reliability of comparative datafrom different sources. In: Butchart EG, Bodnar E, eds. Thrombosis,embolism, and bleeding. London: ICR Publishers, 1992.

2 Takkenberg JJM, Puvimanasinghe JPA, Herwerden LA van, et al.Prognosis after aortic valve replacement with SJM bileaflet pros-theses: impact on outcome ofvarying thrombo-embolic hazard.Eur HeartJ Suppkments 2001;3(Suppl.Q):Q27-32.

3 Puvimanasinghe JP, Steyerberg EW, Takkenberg JJ, et al. Prog-nosis after aortic valve replacement with a bioprosthesis: predictionsbased on meta-analysis and microsimulation. Circulation 2001;103:1535-41.

4 Takkenberg JJM. Prognosis after autograft and allograft aortic rootreplacement. Evidence-based estimates using meta-analysis andmicrosimulation. Department of Cardio-Thoracic Surgery andDepartment of Public Health. Rotterdam: Erasmus UniversityRotterdam, 2002:187.

5 Blackstone EH. The choice of a prosthetic heart valve: how shallpatient-specific recommendations be made? J Heart Valve Dis1998;7:1-3.

6 Kvidal P, Bergstrom R, Horte LG, Stahle E. Observed and relativesurvival after aortic valve replacement.JAm Coil Cardiol2000;35:747-56.

7 Grunkemeier GL, Li HH, Starr A. Heart valve replacement: a stat-istical review of 35 years' results. JHeart Valve Dis 1999;8:466-70; discussion 470-1.

8 Steyerberg EW, Kallewaard M, GraafY van der, Herwerden LAvan, Habbema JD. Decision analyses for prophylactic replacementofthe Bjork-Shiley convexo-concave heart valve: an evaluation ofassumptions and estimates. Med DecisMaking 2000;20:20-32.

9 Takkenberg JJ, Eijkemans MJ, Herwerden LAvan, et al. Estimatedevent-free life expectancy after autograft aortic root replacement inadults. Ann Thorac Surg 2001;7:S344-8.

10 Meer FJ van der, Rosendaal FR, Vandenbroucke JP, Briet E. As-sessment ofa bleeding risk index in two cohorts ofpatients treatedwith oral anticoagulants. Thromb Haemost 1996;76:12-6.

11 Bleeding during antithrombotic therapy in patients with atrialfibrillation. The Stroke Prevention in Atrial Fibrillation In-vestigators. Arch Intern Med 1996;156:409-16.

12 Birkmeyer NJO, Birkmeyer JD, Tosteson ANA, Grunkemeier GL,Marrin CA, O'Connor GT. Prosthetic valve type for patients under-going aortic valve replacement: a decision analysis. Ann Thorac Sur,g2000;70:1946-52.

13 LawAM, KeltonWD. Simulation modeling and analysis. In: RiggsJL, ed. McGraw-Hill series in industrial engineering and manage-ment science. New York: McGraw-Hill, 1991.

14 Grunkemeier GL, Li HH, Naftel DC, Starr A, Rahimtoola SH.Long-term performance of heart valve prostheses. Curr ProblCardiol2000;25:73-154.

15 Grunkemeier GL, Chandler JG, Miller DC, Jamieson WR, StarrA. Utilization ofmanufacturers' implant card data to estimate heartvalve failure. JHeart Valve Dis 1993;2:493-503.

16 ACC/AHA guidelines for the management of patients withvalvular heart disease. A report of the American College of Car-diology/American Heart Association. Task Force on PracticeGuidelines (Committee on Management ofPatients with ValvularHeart Disease). JAm Coil Cardiol 1998;32:1486-588.

17 Stahle E, Kvidal P, Nystrom SO, Bergstrom R Long-term relativesurvival after primary heart valve replacement. EurJ CardiothoracSurg 1997;11:81-91.

Appendix: Input microsimulatlon model

Operative mortality2.7% at age 40, increasing with age (OR 1.034/year)and with each reoperation (OR 2.165)

Valve-related events after AVR with a bileafletmechanical prosthesisValve thrombosis: linearised annual occurrence rate(LOR) 0.08%/patient year, mortality rate 25%,reoperation rate 50%; thromboembolism: LOR1.15%/patient year, mortality rate 7%; bleeding:Gompertz distribution (gamma=0.076; lambda=-8.71), mortality rate increasing with age (OR1.0345/year); endocarditis: two-period risk (LOR6.6in the first six months, thereafter LOR 0.4 ), mortalityrate 36%, reoperation rate 44%; non-structural valvefailure: LOR 0.42%o/patient year, mortality rate 2%,reoperation rate 41%; structural valve failure: LOR0.0%, mortality N/A.

Valve-related events after AVR with a stentedbioprosthesisValve thrombosis: LOR 0.03%/patient year, moralityrate 67%, reoperation rate 33%; thromboembolism:LOR0.90%Yo/patient year, mortality rate 19%; bleeding:LOR 0.38%/patient year, mortality rate at age 40 14%and increasing with age (OR 1.0345/year); endo-carditis: 2-period risk (LOR 3.4 in the first 6 months,thereafter LOR 0.68), mortality rate 34%, reoperationrate 56%; non-structural valve failure: LOR 0.38%/patient year, mortality rate 5%, reoperation rate 52%;structural valve failure: Weibull distribution S(t) =e-(t/a)A, where S(t) indicates the probability of beingfree from structural valve failure at time t, and a; andj indicate the scale and shape parameters of theWeibull model. The value ofa depended on age: a =e211+00112*agc and the value of P was 3.49, mortalityrate = reoperation mortality rate.

Excess mortality

Hazard ratio for mortality for selected age groups comparedwith general Dutch population.

Age (years) HR Males HR Females

55 1.5 2.865 1.1 2.275 1 1.3

10 Nctherlands Heart Journal, Volumc 11, Number 1, January 2003