Letters to the Editor

References

1. Aoyagi S, Nishimi Y, Kawano H, Tayama E,Fukunaga S, Hayashida N, et al. Obstructionof St. Jude Medical valves in the aortic posi-tion: significance of a combination of cinera-diography and echocardiography. J ThoracCardiovasc Surg. 2000;120:142-7.

2. Aoyagi S, Nishimi Y, Tayama E, FukunagaS, Hayashida N, Akashi H, et al. Obstructionof St. Jude Medical valves in the aortic posi-tion: a consideration for pathogenic mecha-nism of prosthetic valve obstruction. J Car-diovasc Surg. 2002;10:339-44.

3. Teshima H, Hayashida N, Enomoto N,Aoyagi S, Okuda K, Uchida M. Diagnosis ofprosthetic valve dysfunction by multidetec-tor-row computed tomography. Ann ThoracSurg. 2003;75:1631-3.

doi:10.1016/j.jtcvs.2003.11.067

Prosthesis size and prosthesis-patient size are unrelated toprosthesis-patient mismatchTo the Editor:We read with interest the report of Black-stone and colleagues1 published in the Sep-tember 2003 issue of the Journal, as well asthe editorial of Gillinov and coworkers2 inthe August 2003 issue. The implicit con-clusion of these two articles was that pros-thesis-patient mismatch (PPM) is a rareoccurrence after aortic valve replacementand that it has a negligible impact withregard to postoperative outcomes. To de-fine PPM and analyze its consequences,both sets of authors chose, however, to usean indexed area based on the internal geo-metric dimension of the prosthesis dividedby the patient’s body surface area, ratherthan the indexed effective orifice area,which is the physiologic parameter mostoften used to define PPM. They justifiedtheir choice of parameter on the basis thatgeometric measures “are determined beforeimplantation, have much less variability,and are independent of hemodynamicstate.”1

The physiologic and clinical relevanceof the indexed internal geometric area asused by these authors must, however, bechallenged. Indeed, it has never beenshown that this parameter can be related, inany significant manner, to transvalvularpressure gradients; in particular, it has beendemonstrated that the indexed internal geo-metric orifice area cannot be used to predictwhich patients will have high postoperative

3

gradients. Inherent to the pathophysiology

1852 The Journal of Thoracic and Cardiov

of valve PPM is the concept that too smalla prosthesis in too large a patient will pro-duce abnormally high gradients and thushave potentially detrimental consequencessuch as might occur with a native aorticstenosis. Thus if the indexed internal geo-metric area cannot be related to postoper-ative gradients, we do not see how it canlogically be used to identify PPM or tocharacterize its severity.

In contrast, and despite its inherent lim-itations, the indexed effective orifice area isthe only parameter that has consistentlybeen shown to correlate with postoperativegradients, as well as being highly predic-tive of adverse outcomes.3-5 Indeed, whenthe definition of PPM is based on this pa-rameter, the phenomenon has been shownto be highly prevalent (19%-70%, depend-ing on series4,5) and to be associated withless symptomatic improvement, worse he-modynamics at rest and during exercise,less regression of left ventricular hypertro-phy, and more cardiac events after opera-tion.4,5 A recent report from our own lab-oratory has clearly shown that PPM has amajor impact on early mortality, particu-larly in patients with poor left ventricularfunction and that, in contrast to other riskfactors, it can easily be prevented by use ofa simple strategy at the time of operation.5

Such a strategy was recently used by Cas-tro and associates6; as a result, the inci-dence of moderate-severe PPM in theirpopulation was only 2.5%, instead of the17% that would have occurred had thisprospective strategy not been used,whereas operative mortality remained low(1%). Extrapolating these findings to thetotal number of aortic valve replacementsbeing performed each year in North Amer-ica, it is estimated that approximately 1000operative deaths could potentially beavoided through use of such a strategy.

In this context, the conclusion of thesetwo articles with regard to the prevalenceof PPM and its consequences cannot beaccepted at face value, because the param-eter used to define PPM is not valid tocharacterize postoperative hemodynam-ics.3 To the contrary, we still believe in the“value of concentrating on better hemody-namic performance”1 and that researchaimed at properly identifying PPM, as wellas preventing it, can significantly contrib-ute to improved outcomes after aortic valve

surgery.

ascular Surgery ● June 2004

Jean G. Dumesnil, MD, FRCPC, FACC

Philippe Pibarot, DVM, PhD, FACC

Research Center of Laval Hospital/Quebec

Heart Institute

Laval University

Sainte-Foy, Quebec, Canada

References

1. Blackstone EH, Cosgrove DM, JamiesonWR, Birkmeyer NJ, Lemmer JH, Miller DC,et al. Prosthesis size and long-term survivalafter aortic valve replacement. J Thorac Car-diovasc Surg. 2003;126:783-96.

2. Gillinov AM, Blackstone EH, Rodriguez LL.Prosthesis-patient size: measurement andclinical implications. J Thorac CardiovascSurg. 2003;126:313-6.

3. Pibarot P, Dumesnil JG, Cartier PC, Metras J,Lemieux M. Patient-prosthesis mismatch canbe predicted at the time of operation. AnnThorac Surg. 2001;71(Suppl 5):S265-8.

4. Pibarot P, Dumesnil JG. Hemodynamic andclinical impact of prosthesis-patient mis-match in the aortic valve position and itsprevention. J Am Coll Cardiol. 2000;36:1131-41.

5. Blais C, Dumesnil JG, Baillot R, Simard S,Doyle D, Pibarot P. Impact of prosthesis-patient mismatch on short-term mortality af-teraorticvalvereplacement.Circulation.2003;108:983-8.

6. Castro LJ, Arcidi JM, Fisher AL, GaudianiVA. Routine enlargement of the small aorticroot: a preventive strategy to minimize mis-match. Ann Thorac Surg. 2002;74:31-6.

doi:10.1016/j.jtcvs.2003.11.073

Reply to the Editor:We have long respected the excellent andprolific work of Dumesnil and Pibarot, so itis not surprising that we referred to theirpublications in our articles. However, theyare unhappy in part with the way we haveinterpreted (or misinterpreted) theirwork,1,2 and in part with the surgical rec-ommendations we based on survival datafrom our multicenter meta-analysis,1 whichdiffer from their own, which were basedlargely on functional prosthesis perfor-mance data.

What We StudiedSurvivalWe studied the relation of geometric pros-thesis size to time-related survival withnine sources of data and nearly 70,000 pa-tient-years of follow-up among 13,258 pa-tients who underwent aortic valve replace-

ment with mechanical prostheses, stent-

Letters to the Editor

mounted bioprostheses, and allografts atmultiple institutions on two continents.1

Long-term survival is a relevant andimportant goal of aortic valve replace-ment,3 although how it is associated withefficiency of prosthetic devices remainscontroversial. Use of small label-size pros-theses has been presumed by many sur-geons to reduce survival after valve re-placement. When patients were stratifiedby label size in our large study, however,prosthesis size was found to have little im-pact on survival (in contrast to substantialimpact of patient factors, such as age). Un-like all previous studies, ours had the sta-tistical power to detect even small differ-ences in survival, with 1109 prostheses oflabel size 19 mm or smaller and 2984 oflabel size 21 mm.

To increase sensitivity, we focused oursurvival analysis on the geometric dimen-sion of the internal orifice of the prosthesisnormalized to patient size (prosthesis-pa-tient size), a concept familiar in congenitalheart disease but less so in adult heart dis-ease.4 Stratifying patients by prosthesis-patient size revealed no impact on long-term or intermediate survival. However, inrisk-adjusted and balancing score–adjustedanalyses, when geometric orifice size wasless than about the lower 95% confidencelimit for normal human aortic valve size,which is uncommon in prudent aortic valvereplacement (only 85 patients had prosthe-sis-patient size smaller than 0.85 cm2/m2),early survival was affected to a small ex-tent (1% to 2%).

Adjusting Prosthesis-Patient Sizefor Prosthesis ModelPhysics of blood flow dictates that as pros-thesis geometric orifice size decreases, en-ergy loss at a given flow rate increases.However, inertia, turbulence, other distur-bances of blood flow velocity profile, andmany other factors also affect energy loss.Therefore, as described in the multi-insti-tutional article,1 we searched for a relationof prosthesis-patient size to survival thatwas specific for each model of prosthesisby means of interaction terms. We foundnone of these to be statistically significant,indicating similarity of the relationship ofgeometric prosthesis-patient size to sur-

vival.

Th

Surgical InferencesSurgical inferences drawn concerning pros-thesis size selection were based on survivaldata within the context of prudent valvereplacement in these numerous representa-tive institutions. We had no functionalprosthesis performance data after operationfor these patients, which might have furtherincreased sensitivity of the survival analy-sis, thereby further refining prognosis andpossibly surgical recommendations.

What We Did Not StudyContrary to the letter, we did not study“postoperative outcomes” in general. Wefocused on one end point only: time-relatedmortality from any cause after aortic valvereplacement. It is possible that either geo-metric or functional prosthesis-patient sizemight be more strongly related to other endpoints—such as sudden death, death inheart failure, or various measures of qualityof life—that do not translate into a detect-able survival difference.

We also studiously avoided the termprosthesis-patient mismatch. No doubt aclinical syndrome of prosthesis-patientmismatch occurs, as described by Rahim-toola5 and Rahimtoola and Murphy.6 How-ever, through the years its definition hasceased being a clinical diagnosis and hasbecome arbitrarily inferred from postoper-ative echocardiographic velocity measure-ments. We had no postoperative echocar-diographic data in the assembled databasesto address this issue, so we cannot discussit further.

Alternative ProposalFunctional Projected Orifice SizeMore interesting is the alternative proposedby Dumesnil and Pibarot: that rather thangeometric orifice size, we should use func-tional projected orifice size.7 They claimthat projected size is more highly corre-lated with pressure gradient across a pros-thesis than is geometric size.8 Thus, logi-cally, it should be more sensitive todifferences in time-related survival. Thismight be the case.

PrerequisitesBefore accepting this alternative, severalthings need to be in place and clarified. Justas we had available an equation relatingpatient size to normal aortic valve size de-rived from 6801 measurements,9 so we

need an equation relating patient size to

e Journal of Thoracic and Cardiovascular S

functional prosthesis size that is based on asubstantial number of measurements. Forthis, high-quality individual patient postop-erative Doppler velocity data for each labelsize of current aortic valve prostheseswould need to be assembled, along withheight and weight, and an appropriate anal-ysis would need to be performed to obtainthe equation. However, accurate (unbi-ased), precise (reproducible), and extensive(power) functional estimates of prosthesissize are elusive. They are related to manypatient characteristics, left ventricular out-flow complex factors, time, measurementvariability, interobserver variability,method of computation, and myriad otherfactors. It is sobering that even in vitrolaboratory estimates of hemodynamicproperties of prostheses are fraught withimportant institutional variation (as muchas 200% to 400%)!10 Thus, it is likely thatthe in vivo relation will have substantialscatter. Then an association should besought between survival and functionalprosthesis-patient size, with reference tothe normalizations suggested in the arti-cles.1,2

Two Notes of CautionCorrelation Between VariablesDerived From the Same DataOne premise of the letter is that transpros-thesis pressure gradients are more closelycorrelated with functional prosthesis sizethan with geometric prosthesis size (andimplicitly that pressure gradient is relatedto survival).8 It is not appreciated in suchcomparisons that the same Doppler veloc-ities are used to compute both pressuregradient and functional prosthesis size.This guarantees a close correlation betweenthe two; they are not independent measure-ments! This obligatory correlation leadsDumesnil and Pibarot to state that “ indexedinternal geometric orifice area cannot beused to predict which patients will havehigh postoperative gradients.” This is, inour opinion, stretching the statistics.

Avoiding 1000 Deaths per YearThe second note of caution is the notionthat if a policy of liberal aortic root en-largement accompanying aortic valve re-placement were adopted, then “each yearin North America...approximately 1000operative deaths could potentially be

avoided.” This too is a stretch of statistics.

urgery ● Volume 127, Number 6 1853

Letters to the Editor

The extrapolation is based on Castro andcolleagues’ observation11 of a single deathamong 114 patients (70% confidence limits0.1%-3.0% mortality) in the concurrent set-ting of 543 patients deemed not in need ofaortic root enlargement; the latter sustaineda mortality of 4.1% with 70% confidencelimits of 3.2% to 5.1%.11 These confidencelimits and the context caution against con-cluding that widespread application of theproposed algorithm is likely to lower oper-ative mortality.

Small sets of data pose two dangers,identifying spurious relations in small sub-groups and failing to identify relations be-cause of lack of statistical power. Theseconcerns motivated collaborators in themulti-institutional study to contribute morethan 13,000 patients with 70,000 patientyears of follow-up.

DisclaimerIn responding to this letter, we must ac-knowledge that we reside within the insti-tution with the highest volume of heartvalve operations on the North Americancontinent. The practice at The ClevelandClinic Foundation has not been to avoidusing small-sized prostheses in small aorticroots. Yet hospital mortality for 881 pri-mary isolated aortic valve replacements inthe last 5 years was 1.2%, and for 996primary combined aortic valve replace-ments with coronary artery bypass graftingit was 2.0%. In some instances, left ven-tricular outflow tract myectomy is per-formed for obstruction at that level, butrarely is aortic root enlargement per-formed.

We remain interested in performance ofprostheses, but we believe that availableevidence suggests that other factors havemore impact on long-term all-cause mor-tality than does prosthesis-patient size,however expressed.

Eugene H. Blackstone, MDA. Marc Gillinov, MD

Delos M. Cosgrove, MDDepartment of Thoracic and Cardiovascular

SurgeryThe Cleveland Clinic Foundation

Cleveland, OH 44195

References

1. Blackstone EH, Cosgrove DM, JamiesonWR, Birkmeyer NJ, Lemmer JH, Miller DC,

et al. Prosthesis size and long-term survival

1854 The Journal of Thoracic and Cardiov

after aortic valve replacement. J ThoracCardiovasc Surg. 2003;126:783-96.

2. Gillinov AM, Blackstone EH, RodriguezLL. Prosthesis-patient size: measurementand clinical implications. J Thorac Cardio-vasc Surg. 2003;126:313-6.

3. Rahimtoola SH. The next generation ofprosthetic heart valves needs a proven trackrecord of patient outcomes at �15 to 20years. J Am Coll Cardiol. 2003;42:1720-1.

4. Kouchoukos NT, Blackstone EH, Doty DB,Hanley FL, Karp RB. Cardiac surgery. 3rded. Philadelphia: Churchill Livingstone;2003, p. 3-65.

5. Rahimtoola SH. The problem of valve pros-thesis-patient mismatch. Circulation. 1978;58:20-4.

6. Rahimtoola SH, Murphy E. Valve prosthe-sis-patient mismatch. A long-term sequela.Br Heart J. 1981;45:331-5.

7. Blais C, Dumesnil JG, Baillot R, Simard S,Doyle D, Pibarot P. Impact of valve pros-thesis-patient mismatch on short-term mor-tality after aortic valve replacement. Circu-lation. 2003;108:983-8.

8. Pibarot P, Dumesnil JG, Cartier PC, Me-tras J, Lemieux M. Patient-prosthesis mis-match can be predicted at the time ofoperation. Ann Thorac Surg. 2001;71(Suppl 5):S265-8.

9. Capps SB, Elkins RC, Fronk DM. Bodysurface area as a predictor of aortic andpulmonary valve diameter. J Thorac Car-diovasc Surg. 2000;119:975-82.

10. Marquez S, Hon RT, Yoganathan AP. Com-parative hydrodynamic evaluation of bio-prosthetic heart valves. J Heart Valve Dis.2001;10:802-11.

11. Castro LJ, Arcidi JM, Fisher AL, Gaudi-ani VA. Routine enlargement of the smallaortic root: a preventive strategy to mini-mize mismatch. Ann Thorac Surg. 2002;74:31-6.

doi:10.1016/j.jtcvs.2003.12.040

Epidemiologic contrast of predictors’trends for outcomes of coronaryartery bypass grafting: Heart failureversus ventricular function versusleft main diseaseTo the Editor:I read with interest the article by Davier-wala and colleagues1 in the November2003 issue of the Journal. The work ana-lyzed differential change in predictors of inhospital mortality after coronary artery by-pass grafting. The study elegantly demon-strated the diminishing statistical signifi-cance for left ventricular function greaterthan 20% as a predictor for mortality dur-ing a 12-year period. Several explanationsare provided related to patient comorbidity,interventional cardiology, surgeon experi-

ence, and intensive patient care. It is well

ascular Surgery ● June 2004

established that one of the major determi-nants of morbidity during and after coro-nary artery bypass grafting is low left ven-tricular ejection fraction.2 The results ofnumerous coronary artery bypass graftingtrials performed in the 1970s and 1980sshow that despite this increased morbidity,the benefits of this procedure for patientswith multivessel coronary artery diseaseand low left ventricular systolic function inmany cases outweigh the risks.3 The articledoes not contain data for the actual causesof death in this large group of patients,which would be epidemiologically rele-vant. Given the data provided, however, itis difficult to observe a contrast in thetrends between congestive heart failure(CHF) and left ventricular function as pre-dictor variables. Davierwala and col-leagues1 also stated in their discussion thatfrom studying the data again in Table 1 it isclear to the reader that CHF held an in-creasing proportion as morbidity in the pa-tient cohort (7.8% vs 9.4% vs 9.4%) duringthe interval (1990-1993 vs 1994-1997 vs1998-2001). Moreover, according to theoriginal article’s Table 3, after a multivar-iate analysis CHF showed an increasingtrend for odds ratio by year group (1.9 vs3.6).1 Given this contrast of predictor out-come trends, the authors should have in-cluded other explanations. In simple terms,the CHF increasing trend could reflect anincrease in diastolic heart failure, assumingthat the proportions of new patients in eachtime interval were significant. In otherwords, the decline of the predictor value inleft ventricular dysfunction was not relatedto interyear group patient death. The au-thors stated that left main disease in a“somewhat counterintuitive finding” was“unmasked” in the last time cohort, with anodds ratio of 1.7 in Table 3 of the originalarticle.1 From their data, both CHF and leftmain disease increased in prevalence, yetworsening ventricular dysfunction de-clined. This constellation may reflect achange in the ventricular dysfunction func-tion from systolic to diastolic biometri-cally.

Furthermore, the typical risk factors andcomorbidities of female gender, diabetes,and hypertension together increased signif-icantly during the entire period of theirstudy. This association has now been foundto be the same with diastolic dysfunction.4

As we know now, there is growing appre-

ciation of diastolic heart failure as a distinct