HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to Personal Folders Download to citation manager Author home page(s): Eugene H. Blackstone Delos M. Cosgrove John H. Lemmer, Jr D. Craig Miller Eric G. Butchart Giulio Rizzoli Magdi Yacoub Permission Requests Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Blackstone, E. H. Articles by Chai, A. Search for Related Content PubMed PubMed Citation Articles by Blackstone, E. H. Articles by Chai, A. Related Collections Valve disease

J Thorac Cardiovasc Surg 2003;126:783-793
© 2003 The American Association for Thoracic Surgery

Surgery for acquired cardiovascular disease

Prosthesis size and long-term survival after aortic valve replacement

^a Department of Thoracic and Cardiovascular Surgery, The Cleveland Clinic Foundation, Cleveland, Ohio, USA,
^b Department of Biostatistics and Epidemiology, The Cleveland Clinic Foundation, Cleveland, OhioUSA
^c The University of British Columbia, Vancouver, British Columbia, Canada
^d Dartmouth Medical School, Dartmouth-Hitchcock Medical Center, Lebanon, NHUSA
^e Northwest Surgical Associates, Portland, OreUSA
^f Stanford University, Stanford, CalifUSA
^g University Hospital, Cardiff, United Kingdom
^h Instituto Chirurgia Cardiovascolare, University of Padova, Padova, Italy
ⁱ Harefield Hospital, London, United Kingdom
^j Edwards Lifesciences Corporation, Irvine, CalifUSA

Read at the Eightieth Annual Meeting of The American Association for Thoracic Surgery, Toronto, Ontario, Canada, April 30–May 3, 2000.

Received for publication April 28, 2000; revisions received November 27, 2002; accepted for publication February 26, 2003.

^* Address for reprints: Eugene H. Blackstone, MD, The Cleveland Clinic Foundation, 9500 Euclid Ave, Desk F25, Cleveland, OH 44195, USA
blackse{at}ccf.org

	Abstract

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
OBJECTIVE: This study was undertaken to quantify the relationship between prosthesis size adjusted for patient size (prosthesis-patient size) and long-term survival after aortic valve replacement.

METHODS: Data from nine representative sources on 13,258 aortic valve replacements provided 69,780 patient-years of follow-up (mean 5.3 ± 4.7 years), with reliable survival estimates to 15 years. Prostheses included 5757 stented porcine xenografts, 3198 stented bovine pericardial xenografts, 3583 mechanical valves, and 720 allografts. Manufacturers’ labeled prosthesis size was 19 mm or smaller in 1109 patients. Expressions of prosthesis-patient size assessed were indexed internal prosthesis orifice area (in centimeters squared per square meter of body surface area) and standardized internal prosthesis orifice size (Z, the number of SDs from mean normal native aortic valve size). Multivariable hazard domain analysis with balancing score and risk factor adjustment quantified the association of prosthesis-patient size with survival.

RESULTS: Prosthesis-patient size down to at least 1.1 cm²/m² or -3 Z did not adversely affect intermediate- or long-term survival (P > .2). However, 30-day mortality increased 1% to 2% when indexed orifice area fell below 1.2 cm²/m² (P = .002) or standardized orifice size fell below -2.5 Z (P = .0003). The increased early risk affected fewer than 1% of patients receiving bioprostheses but about 25% of those receiving mechanical devices.

CONCLUSIONS: Aortic prosthesis-patient size down to 1.1 cm²/m² or -3 Z did not reduce intermediate- or long-term survival after aortic valve replacement. However, patient-prosthesis size under 1.2 cm²/m² or -2.5 Z was associated with a 1% to 2% increase in 30-day mortality. Prosthesis-patient sizes this small or smaller were rarely implanted in patients receiving bioprostheses.

The study has two unique features. First, it uses an analytic method that permits independent assessment of the association of prosthesis-patient size with long-, intermediate-, and short-term survivals. Second, it uses a large multi-institutional group of patients to achieve sufficient statistical power to make these assessments. Purposes of the investigation were to quantify the relationship of prosthesis-patient size to long-, intermediate-, and short-term survivals and from this relationship to develop a clinically useful algorithm for identifying a prosthesis size below which survival might be affected.

	Patients and methods

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
Patients
Individual patient data on 13,258 AVRs were assembled from nine sources (Table 1). Included were adults at least 18 years old for whom prosthesis type, model, and labeled size were recorded. Excluded were patients with a preexisting valve prosthesis in a location other than aortic, those undergoing a multiple valve operation, those with native or prosthetic valve endocarditis, those undergoing emergency AVR, and those undergoing concomitant procedures other than coronary artery bypass grafting. Patients undergoing aortic root enlargement were not excluded. Half the operations were performed before 1990 (Table 1).

Prostheses
Prostheses were of four varieties: stented porcine xenografts (n = 5757, designated porcine), stented bovine pericardial xenografts (n = 3198, designated pericardial), mechanical devices (n = 3583, of which 2125 were bileaflet and 1458 were monoleaflet), and allografts (n = 720). Patient characteristics varied according to variety of prosthesis received (Tables 2A and 2B).

Prosthesis size was defined in terms of geometric dimensions of prostheses, not in vitro or in vivo functional dimensions. These properties included the manufacturer’s labeled size (in millimeters), which refers inconsistently to diameter of external sewing ring (mechanical prostheses), mounting ring (stented xenografts), or internal orifice (allografts), and the geometric internal orifice diameter (in millimeters), a consistent physical dimension (Appendix Table 1). ⁵ They did not include patient’s tissue annulus diameter, a proposed ISO standard (ISO/CD 5840 working decrement) that was not recorded in any source. Patient size was represented by BSA, calculated from height and weight.⁶

View larger version (22K): [in this window] [in a new window]   Figure 1. Cumulative distribution of physical internal valve orifice dimension referenced to patients’ BSA (prosthesis-patient size). Distribution is stratified according to variety of prosthesis: mechanical, stented bovine pericardial (pericardial), stented porcine xenograft (porcine), and allograft. For orientation, median size is value on horizontal axis that corresponds to value of 50% on vertical axis; half the sizes were smaller than this and half larger. Value on horizontal axis corresponding to 10% on vertical axis is the 10th percentile, meaning that 10% of valves were smaller than this and 90% larger. A, Indexed orifice area. B, Standardized orifice size.

Data analysis
Survival estimates
Overall survivals were 95.8% at 30 days and 91%, 79%, 58%, and 37% at 1, 5, 10, and 15 years (Figure 2, A). Instantaneous risk of death peaked immediately after operation, fell rapidly during the first 6 postoperative months, and then rose steadily (Figure 2, B). These represent three additive hazard phases labeled early, constant, and late. To determine the independent influence of prosthesis-patient size on long-, intermediate-, and short-term survivals, a multiphase hazard decomposition method was used (available for use with the SAS system at http://www.clevelandclinic.org/heartcenter/hazard). Each hazard phase (Figure 2, B) incorporated an independent set of risk factors, permitting quantification of the effect of prosthesis-patient size within each time frame determined by the data themselves, not arbitrarily, without assuming proportional hazards.⁹

View larger version (15K): [in this window] [in a new window]   Figure 2. Overall survival after AVR. A, Survival. Parametric estimates (solid line) are enclosed within confidence limits equivalent to 1 SE (68%). Numbers in parentheses represent numbers of patients still being followed at 1, 5, 10, and 15 years. Faintly visible at these same intervals are vertical bars representing confidence limits for corresponding nonparametric Kaplan-Meier estimates. B, Hazard function (instantaneous risk of death). Solid curve enclosed within asymmetric 68% confidence limits represents parametric overall hazard estimates. Individual hazard components are labeled early, constant, and late; they sum to yield overall hazard. Rapidly falling early phase of risk lasted approximately 6 months (short-term survival), was dominated by operative mortality, and had effective statistical power of 774 events. Constant hazard phase across all time dominated between 6 months and 5 years (intermediate-term survival) and had statistical power of 1654 events. Late hazard phase, rising steadily from time zero, dominated beyond 5 years (late-term survival) and had statistical power of 1470 events. Variables in multivariable analyses modulated area beneath early hazard phase, raised or lowered constant hazard, and tilted late hazard component.

Prosthesis-patient size and survival
Separate analyses were performed of the relation of risk-adjusted mortality to continuous values of indexed orifice area and standardized orifice size (Z). The specific linearizing transformations of scale for these (and other continuous variables) for each hazard phase were selected by bootstrap bagging (Appendix 2).^13,14 Whenever computatively possible, effects of both large and small prosthesis-patient size were estimated (Appendix 3).

Clinical algorithm
An internal prosthesis orifice diameter was calculated across the spectrum of BSA below which increased risk of mortality might be anticipated. For this, formulas for indexed orifice area and standardized orifice size were rearranged and solved at fixed values of prosthesis-patient size.

Presentation
Survival curves are presented with both parametric and Kaplan-Meier estimates. Confidence limits are asymmetric and equivalent to 1 SE (68%). To illuminate multivariable findings, nomograms were constructed by solving the multivariable equations for a 70-year-old woman of 1.9 m² BSA with aortic stenosis, in New York Heart Association functional class III, and undergoing primary isolated AVR in the current era.

	Results

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
Prosthesis size and non–risk-adjusted survival
Small, nonsystematic survival differences were observed with respect to both indexed orifice area (Figure 3) and standardized orifice size (Figure 4). Systematic survival decrements were found with decreasing labeled size (Appendix Figure 1, A) and according to variety of prosthesis (Appendix Figure 1, B). However, these differences were largely accounted for by differing patient characteristics according to prosthesis size and variety (Table 2B and Appendix 2).

View larger version (18K): [in this window] [in a new window]   Figure 3. Non–risk-adjusted survival and indexed orifice area. A, Time-related survival stratified by indexed orifice area. B, Kaplan-Meier estimates of 1-, 5-, and 10-year survivals in finely grouped strata of indexed orifice area.

View larger version (18K): [in this window] [in a new window]   Figure 4. Non–risk-adjusted survival and standardized orifice size (Z). A, Time-related survival stratified by Z. B, Kaplan-Meier estimates of 1-, 5-, and 10-year survivals in finely grouped strata of Z.

View larger version (19K): [in this window] [in a new window]   Appendix Figure 1 Non–risk-adjusted survival. A, Stratified by labeled size. B, Stratified by variety of aortic prosthesis inserted: allograft, mechanical, stented bovine pericardial (pericardial), or stented porcine xenograft (porcine).

View larger version (29K): [in this window] [in a new window]   Figure 5. Magnitude and shape of effect of continuous prosthesis-patient size on 30-day and 1-, 5-, 10-, and 15-year survivals. Depictions are nomograms of multivariable equations solved for a typical patient with aortic stenosis (see Methods) with valve size varied. A, Indexed orifice area. B, Standardized orifice size.

View larger version (16K): [in this window] [in a new window]   Figure 6. Magnitude and shape of effect of continuous prosthesis-patient size on 30-day mortality. Depictions are nomograms of multivariable equations solved for a typical patient with aortic stenosis (see Methods) with valve size varied. A, Indexed orifice area. B, Standardized orifice size.

By substituting either 1.2 cm²/m² or -2.5 Z into their respective formulas, a simple nomogram reveals the internal orifice size of a prosthesis at these thresholds for any BSA (Figure 7). The nomogram can be used in the operating room by drawing a vertical line from a patient’s BSA to the solid line and a horizontal line from there to the internal orifice diameter. Appendix Table A1 is then used to find the next largest labeled-size prosthesis that provides either indexed orifice area or standardized orifice size above these threshold values.

View larger version (13K): [in this window] [in a new window]   Figure 7. Minimum internal prosthesis orifice size necessary to achieve indexed orifice area of at least -1 (1.6 cm2/m2), -2 (1.3 cm2/m2), and -2.5 (1.2 cm2/m2) SDs (Z) below mean normal aortic valve size.

	Discussion

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

First, the natural history of mild, fixed native aortic valve stenosis is unknown. Progressive aortic disease tends to become apparent in elderly persons; mild nonprogressive stenosis may be well tolerated. Second, patients who undergo AVR have lower long-term survival than an age-, sex-, and race-matched general population (Appendix Figure 2). ⁶ Mortality is affected by complications of warfarin anticoagulation (particularly its variability²¹) among patients with mechanical devices and by reoperation for structural valve degeneration among those with biologic prostheses. Yet in this study and others,²² it has been difficult to demonstrate a survival difference according to prosthesis variety (Appendix 3). Thus, the multifactorial nature of reduced survival after AVR may mask subtle individual components of it, such as small prosthesis-patient size. Third, few prostheses, and almost no bioprostheses, had a prosthesis-patient size smaller than the 95% confidence limits of normal native human aortic valves (-2 Z). Thus, AVR as practiced in centers represented in this study is such that prosthesis-patient size is rarely excessively small.²³ Fourth, degenerative aortic valve stenosis is a disease of elderly patients with already limited lifespans. Although we searched for interactions of age with prosthesis-patient size and found none, elderly patients simply may not live long enough to manifest a survival decrement from small prosthesis-patient size. Fifth, the 1% to 2% increase in short-term mortality found among patients with small prosthesis-patient size may serve as a biologic selection process. However, our cohort of patients with such prostheses (1109 patients with a labeled size of 19 mm or smaller) was sufficiently large, short-term mortality sufficiently low, and survival sufficiently long to suggest that this is not an important explanation.

View larger version (18K): [in this window] [in a new window]   Appendix Figure 2 Non–risk-adjusted survival stratified by age group. Of 1685 patients 50 years old or younger, 1065 were alive and traced at 5 years, 539 at 10 years, and 160 at 15 years. Of 8224 between 50 and 75 years old, 4010 were alive and traced at 5 years, 1725 at 10 years, and 303 at 15 years. Of 2636 who were 75 years old or older; 698 were alive and traced at 5 years, 140 at 10 years, and 10 at 15 years. For each age group, age-, race-, and ethnicity-matched population life table curve is shown as dot-dash line. Note that younger patients show more marked departure from normal life expectancy.

Issues
At the time of its original presentation, in subsequent discussions of its findings, and among ourselves, this study raised several methodologic issues.

Statistical power and risk adjustment
Confounding reliable assessment of the relation of prosthesis size to mortality are differences in characteristics of patients receiving each size and variety of prosthesis, necessitating risk adjustment; inconsistencies between manufacturer’s labeled size for various types of prosthesis, necessitating adoption of consistent expressions of prosthesis size; limitations of single-institution studies, particularly regarding number of patients receiving small prostheses, necessitating a multi-institutional study; and low operative mortality, necessitating a large study with a sufficient number of deaths to achieve adequate statistical power. By amassing a large AVR database with a sufficient number of events (3898 deaths), we were available to achieve statistical power to identify even a small survival impact of prosthesis-patient size and to localize it to a specific follow-up time frame while adjusting simultaneously for many important patient variables, device models, data sources, and selection factors.

Prosthesis size
Prosthesis size may be based on physical dimensions or functional performance. Physical dimensions include labeled size and internal orifice size. Because manufacturers’ conventions for labeling prosthesis size differ among devices, we selected a common metric: geometric internal orifice diameter.⁵ Functional size includes in vitro and in vivo effective orifice areas (EOAs). The former may be static at a variety of steady flow rates or dynamic with a variety of pulsatile waveforms and flow rates³³; the latter is estimated clinically under a range of incompletely controlled conditions in patients by echocardiography according to various formulas.

We have not included information about in vitro EOA, because methodology has been difficult to standardize and results are difficult to reproduce. We had no individual patient data on functional in vivo EOA. In vivo EOA varies from moment to moment with patient activity,^34,35 cardiac output and blood pressure, and dynamics of the left ventricular outflow tract, as well as intrinsic prosthesis properties, although it is well correlated with geometric internal orifice area.³⁴ It is therefore possible that one or more individual patients had functional prosthesis-patient size mismatch that we did not evaluate. However, in vivo EOA is unavailable at the time of selecting a prosthesis at AVR.

Pibarot and colleagues³⁶ and Dumesnil and coworkers³⁴ have suggested that rather than using geometric prosthesis dimensions as a fixed reference for prosthesis size, a fixed referent value of in vivo EOA should be used, which they termed projected EOA. Referent values were obtained from informal meta-analysis of literature sources. To the extent that in vivo EOA differs importantly from a linear relation to geometric orifice area, substituting projected EOA for geometric size is advantageous. However, this strategy has several drawbacks: It suffers from flow dependency, large scatter in the data, rest versus exercise differences, and limited availability of data for each prosthesis size and model. Also, when fixed referent values are used for intraoperative decision making, they acquire the same limitations as the simpler geometric values used in this study. For intraoperative decision making, we therefore recommend using the relation of survival to prosthesis geometric orifice dimensions.

Patient size
Native aortic valve size is related to somatic growth, as is cardiac output. The most consistent allometric relationship, then, is BSA.⁶

Normalization
Normalization of any expression of prosthesis size can be made to body size. Use of indexed valve area assumes a linear relation between valve area and BSA, which is nearly, but not completely, accurate.⁶ Use of dimensionless standardized values (Z) is based on the variability of native annulus size among similarly sized healthy individuals. Rimoldi and Lev³⁷ introduced the idea that normality and abnormality can be expressed in terms of number of SDs away from normal size given a patient’s cardiac structures measure. Thus, only 2.5% of the normal population have a valve size 2 SDs or more below mean normal size (-2 Z). Both mean normal value and its SD are estimated by regression analysis of normal patient data.⁷ Then, given BSA and valve size, the regression equation is solved to yield estimated mean normal valve size and its SD, from which Z is calculated.¹⁵

There is a logarithmic relation between indexed and standardized prosthesis-patient size. Standardized size expands the scale of small values compared with indexed values (contrast parts A and B of Figure 5).

Limitations
Despite the large AVR database used in this analysis, small prosthesis-patient size was rare. Only 28 patients had an indexed orifice area smaller than 0.85 cm²/m². Nevertheless, we believe the information about prosthesis-patient size and survival is reliable down to an indexed orifice area of approximately 1.1 cm²/m² or 3 SDs below native aortic valve size.

An insurmountable limitation is that neither institution nor valve variety, model, or size was chosen randomly for a given patient. Nonrandom allocation introduces bias. Use of both saturated model techniques and balancing scores provides adjustment for known factors but no explicit protection against factors unavailable across all data sources or unrecorded factors (Appendix 2).¹¹

A study like this relies on existing databases that contain a common set of identically defined and relevant variables, similar outcome assessment, and complete reporting. Not all known or purported risk factors for mortality after AVR were available in all data sets. Among the missing variables were ones related to ventricular morphology, measured ventricular function, and noncardiac comorbidities. These variables may not have been distributed uniformly across institutions and may interact with choice of valve prosthesis and size.

All-cause mortality was the end point. It is noninterpretative and was chosen despite the claim that valve-specific mortality is more sensitive.¹⁶ In fact, reported causes of death are untrustworthy, particularly for elderly patients.⁸ Nevertheless, it is a limitation that we do not know who died suddenly. Although reliable follow-up for survival estimates extended to 15 years, the mean was only 5.3 years. Long-term information requires inclusion of patients operated on in earlier eras, when operative mortality was higher than it is today; we attempted to adjust for this by including date of operation in all analyses.

A strength of this study is representativeness. Databases were solicited from investigators well known for heart valve investigation in North America and Europe and from two organizations responsible for multi-institutional heart valve registries. We also included the spectrum of valve varieties; however, a limitation is the absence of stentless xenografts. Except from one manufacturer, internal orifice size of stentless valves was claimed not to be available, although Rao and colleagues³² have reported that the internal orifice size of a Toronto stentless prosthesis of labeled size 25 is equivalent to that of a pericardial stented prosthesis of labeled size 21. Allografts were therefore considered to represent stentless devices, despite their use primarily in young and middle-aged patients (only 10% of recipients were older than 67 years).³⁸

Inferences
Surgical
Within the constraints of prudent AVR in these institutions, we identified no relationship between prosthetic-patient size and intermediate- and long-term survivals. However, we found a small increase in early mortality among patients who received prostheses with small valve orifices in relationship to body size. Although we recognize that outcome after AVR is related to many factors,³⁹ with currently available prostheses few patients should require aortic root enlargement with its attendant complexity, prolonged operative time, and risks of bleeding, heart block, and mortality,²⁹ particularly if a bioprosthesis is used.

Manufacturing
This study challenges the value of concentrating on better hemodynamic performance. Rather, to improve long-term survival, a durable nonthrombogenic prosthesis is needed that would permit operation much earlier in the natural history of aortic valve disease, before secondary left ventricular remodeling takes place and the patient reports important symptoms.

	Discussion

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
Dr David H. Adams (Boston, Mass). Dr Blackstone, I, like most people in the audience, continue to learn from you. By combining a pool of nine multi-institutional databases, you are in a good position to answer some of the important clinical questions associated with aortic valve prosthesis size and operative mortality. Your conclusion that small aortic prostheses increased operative mortality by 1% in 10% of patients is statistically elegant. In the manuscript version that I read, you concluded rather strongly that "this study settles the debate about whether smaller aortic prosthesis size in the range used clinically adversely influences early, intermediate, and late survival." In this regard, we have a few questions.

First, the preoperative risk factors available for analysis were somewhat limited. Aren’t you concerned about the potential influence of additional variables, such as recent myocardial infarction, congestive heart failure, diabetes, hypertension, and particularly low ejection fraction in the setting of aortic stenosis?

Second, neither valve type nor institution nor size was chosen randomly, and not all prostheses were used in all institutions. Thus, operative survival could have been influenced by institutional differences instead of valve size. Can you tell us whether there were differences in operative mortality among the nine participating centers?

Finally, do you recommend on the basis of these results that we abandon our pursuit of more efficient valves in patients undergoing AVR? I congratulate you on your study and thank the Association for the privilege of the floor.

Dr Blackstone. Thank you, Dr Adams. One of the strengths of meta-analysis of a large series such as this is that one can look at a few pinpointed questions. However, one has the disadvantage of not having a large number of risk-adjusting factors in common, with common definitions across institutions. So you are quite correct that some of the risk factors, perhaps some of the more subtle risk factors, could not be looked at in this study.

I remain rather surprised that valve size makes so little difference. I think we don’t know as much as we should about what affects mortality after AVR. Let me give you an example of this. I have looked for all 13,000 of these patients at what the expected survival should have been had they been part of the regular population. What I found is the that the survival of the old patients followed the trajectory for old patients in the normal population. That is surprising, because these were highly selected patients. The younger the patients were, the more they deviated from expected survival. So there are some powerful factors that increase mortality in patients with aortic valve disease, and I am not sure that we have discovered those yet (see Appendix Figure 2).

You ask whether there were institutional and other differences in the data. The institutional differences, the differences according to a variety of prostheses, were all surprisingly very small. Should one abandon them, looking at more efficient valves? My feeling, as I don’t think we said in the manuscript you read, was not that one should abandon looking for more efficient valves. It is just that the implication that the more efficient the valves, the better the survival, may not be a direct link. I think we know less about mortality after AVR than factors such as efficiency of valve and size.

Dr Vivek Rao (Toronto, Ontario, Canada). Dr Blackstone, I enjoyed that presentation. This was an impressive number of aortic valves across a number of institutions. As you know, the Toronto and Vancouver group presented a combined result at last year’s American Heart Association meeting, where we did demonstrate a significant effect of prosthesis size on late survival. And to compare a couple of differences between your study and ours, first of all, the mean follow-up in your study was about 5.5 years, compared with almost 10 years in our study, and in fact the differences that we observed didn’t become apparent until 7 years of follow-up. That is, the survival curves were parallel until 7 years. In fact, looking at your slide at the indexed in vitro effective orifice areas, the curves do tend to separate at about that time point. So my first question is, do you think with additional follow-up that you will be able to demonstrate a significant difference?

Second, I would like your comments on the concept of valve-related mortality, because, again, that was the significant end point in our study. You have looked at overall survival in your study. Would looking at the more sensitive end point of valve-related mortality have affected your results?

Finally, I think that the concept of patient-prosthetic mismatch is more important in younger patients. With the significant number of patient data that you have, if you restricted an analysis to those younger patients, could you find an increased risk of sudden death from patient-prosthesis mismatch?

Dr Blackstone. Dr Rao, of course we have been interested in your concerns in this for a long time. I think I have definitive answers to all three of your questions.

As you notice, the effective orifice area in vitro does show what appears to be a separation at, say, 5 or so years. This completely vanished the moment we did a multivariable analysis and took into account patient risk factors. So I think that what you are seeing on the slide is not risk adjusted, and that it vanishes once adjustment is made.

You have asked whether a difference would be seen if one were to look more sensitively at valve-related death. In this particular analysis and, again, coming from multiple institutions, it is hard to look at exactly the mode of death. You know from a publication that we have from the Cleveland Clinic with Dr Eric Topol that we are philosophically opposed to looking at more than the hard point of all-cause mortality. I think that is a very hard point. It is the point that the patients are most concerned with, not why they die but the fact that they do die. It may be more sensitive, but it should be revealed if it is really there. Only about half the patients are being followed up longer than 5.5 years, but we had thousands of patients alive at 10 years. So proportionately we may not have many patients alive at 10 years, but in actual numbers, we have thousands.

Your last point is whether there may be some differences across young age. We have done an extensive interaction analysis looking at absolutely every variable in the data set in relationship to size to see whether there is a gradient that is different for younger patients, for older patients, and so on, and we found no significant interactions.

Dr Guo-Wei He (Hong Kong, China). Congratulations, Dr. Blackstone, on your wonderful study with this large size sample. A while ago we did a study with Gary Grunkemeier, mainly looking at Portland data, which is a part of your study, after 30 years’ follow-up for AVR in small aortic root and published in the Annals of Thoracic Surgery. We found that the small size of aortic root, which only fit 19- and 21-mm prostheses, carries a higher late mortality, I mean more than 10 years, when the BSA is greater than 1.6 m².

In your study, the mean follow-up was 6.5 years and the maximum follow-up was 20 years; in our study, maximum follow-up was about 30 years. And the differences between your study and ours therefore may be related to the period of the follow-up. Is there a tendency that with the increase of follow-up period, the influence of small aortic valve and aortic valve root on the late survival likewise increases?

Dr Blackstone. Thank you, Dr He. Yes, you have been active in this area. We have looked in the following way to try and answer your question. These 13,000 patients were subjected to 4000 bootstrap resampling analyses. This is a technique where you sample the patients as if you had created, in this case, 4000 separate analyses, and we looked at the possibility that there was a late effect of any kind. What we found is that in studies of this size there was a 25% chance of finding a late effect.

Dr Denton A. Cooley (Houston, Tex). Thank you for the opportunity to discuss this excellent analysis. I am quite surprised with your conclusions, because they are against current conventional thinking. I think from a practical standpoint that the surgeon is more guilty of oversizing the valve with a stented valve than of undersizing the valve, and I am reassured to find that those of us who tend to undersize the valve have some reason for doing so, especially in patients undergoing double valve replacement. Thank you for this nice presentation.

	Appendix 1: variables available for analysis

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
Demographic

• Age in years as a continuous variable
  
• Gender
  
• Height, weight, and BSA in square meters as continuous variables
  

Clinical status

• New York Heart Association functional class (ordinal and as individual classes)
  

Pathophysiology

• Aortic stenosis with or without regurgitation
  
• Aortic regurgitation with or without stenosis
  
• Pure aortic stenosis
  
• Pure aortic regurgitation
  
• Mixed aortic stenosis and regurgitation
  

Procedure

• Previous AVR
  
• Concomitant coronary artery bypass grafting
  

Experience

• Date of operation (continuous count of days since January 1, 1969, expressed in years)
  
• Binary variable representing each source of data
  

Prosthesis

• Each prosthesis model as an individual variable
  
• Prosthesis variety (mechanical, stented bovine pericardial xenograft, stented porcine xenograft, allograft)
  

	Appendix 2: details of risk adjustment

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
Methods
Scaling
The scales on which to express the continuous variables age, BSA, date of operation, and prosthesis-patient size were suggested by univariable, hazard-phase–specific decile analysis and formalized by bootstrap bagging, selecting those linearizing transformations of scale that appeared with P < .05 in 50% or more of 1000 bootstrap resampling models.^13,14,40

Selection bias
To adjust for biases in choice of prosthesis size that may have confounded outcome, balancing scores were constructed by linear regression, forcing into each model all available variables (Appendix 1) to predict prosthesis-patient size.⁴¹ From these regression equations, values for each expression of prosthesis-patient size were predicted for each patient; these predicted values constituted the balancing scores. The balancing score corresponding to each expression of prosthesis-patient size was included as a further adjustment with all other variables in each hazard phase. To illustrate the magnitude of bias between prosthesis size and patient, procedural, and valve variety factors, a logistic regression analysis was performed to identify factors associated with the use of prostheses of labeled size 19 mm or smaller.

Results
Risk adjustment
The statistical significance and direction of the association of risk-adjustment variables with survival are presented in Appendix Table 2. Table 2. Associations include the strong effect of patient factors such as advanced age (Appendix Figure 2), advanced symptoms, aortic regurgitation, previous AVR, and ischemic heart disease. Operative risk was higher earlier in the experience. Small and generally statistically nonsignificant differences in mortality were observed across data sources and among models of aortic prosthesis, with the exception of higher early risk in patients receiving allografts.

Illustration of bias
Compared with other patients, the 1109 who received a prosthesis size 19 mm or less were older, smaller, more likely to have aortic stenosis, and were more likely to have undergone a previous AVR and concomitant coronary artery bypass grafting (Appendix Table 3).

	Appendix 3: details of prosthesis-patient size analysis

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 
Methods
Missing values
Sporadic missing values were imputed during analysis by the mean value of nonmissing data. Indicator variables for each variable with missing values were generated and included in regressions that generate balancing scores, and in the saturated multivariable models of survival.

Dichotomous prosthesis-patient size variables
Although we concentrated on continuous measurements, in the multivariable analyses we also examined binary variables for indexed orifice area less than 1.25 cm²/m² and standardized orifice size less than -2.5 Z, representing the lowest 10th percentile of patients, as suggested by Rao and colleagues.¹⁶

Interactions
After prosthesis-patient size was entered into each hazard phase of the multivariable analyses, assessments of interactions were performed. These included 2-way interactions of patient-prosthesis size with each risk-adjustment variable and 3-way interactions with age and gender.⁴²

Results
Reliability of findings
Bootstrap bagging with 1000 resampled models was used to assess the reliability of association of prosthesis-patient size with mortality at a significance level of at least P = .05. In the early hazard phase, there was a 90% probability that smaller indexed orifice area was associated with higher mortality, including an 8% probability that both smaller and larger size were associated. For standardized orifice size, the corresponding probabilities were 90% and 60%.

In contrast, in the constant hazard phase there was a 20% probability that smaller indexed orifice area was associated with higher mortality, including a 6% probability that both smaller and larger size were associated. For standardized orifice size, the corresponding probabilities were 46% and 4.5%.

In the late hazard phase there was a 25% probability that smaller indexed orifice area was associated with mortality, including a 2% probability that both smaller and larger size were associated. For standardized orifice size, the corresponding probabilities were 24% and 25%.

Search for modulating factors
Interactions between patient factors and expressions of prosthesis-patient size that might modulate these estimates of magnitude of risk were not statistically significant (P > .1). In particular, the pattern of risk in elderly men with aortic stenosis receiving small prostheses was no different from that determined for the entire study group (P = .9).⁴² Interactions between prosthesis-patient size and valve models did not modulate survival.

	Acknowledgments

 
We thank Tess Knerik for editorial assistance and data managers at each source of information.

	References

Top Abstract Patients and methods Results Discussion Discussion Appendix 1: variables available... Appendix 2: details of... Appendix 3: details of... References

 

1. Rahimtoola SH. The problem of valve prosthesis-patient mismatch. Circulation. 1978;58:20–24[Abstract/Free Full Text]
   
2. Rahimtoola SH, Murphy E. Valve prosthesis-patient mismatch: a long-term sequela. Br Heart J. 1981;45:331–335[Abstract/Free Full Text]
   
3. Shepard RB, Simpson DC, Sharp JF. Energy equivalent pressure. Arch Surg. 1966;93:730–740[Medline]
   
4. Berry DA, Stangl DK. Meta-analysis in medicine and health policy. New York: Marcel Dekker; 2000.
   
5. Christakis GT, Buth KJ, Goldman BS, Fremes SE, Rao V, Cohen G, et al. Inaccurate and misleading valve sizing: a proposed standard for valve size nomenclature. Ann Thorac Surg. 1998;66:1198–1203[Abstract/Free Full Text]
   
6. Kirklin JW, Barratt-Boyes BG. Cardiac surgery. 2nd ed. New York: Churchill Livingstone; 1993.
   
7. Capps SB, Elkins RC, Fronk DM. Body surface area as a predictor of aortic and pulmonary valve diameter. J Thorac Cardiovasc Surg. 2000;119:975–982[Abstract/Free Full Text]
   
8. Lauer MS, Blackstone EH, Young JB, Topol EJ. Cause of death in clinical research: time for a reassessment? J Am Coll Cardiol. 1999;34:618–620[Free Full Text]
   
9. Blackstone EH, Naftel DC, Turner ME Jr. The decomposition of time-varying hazard into phases, each incorporating a separate stream of concomitant information. J Am Stat Assoc. 1986;81:615–624
   
10. Steyerberg EW, Eijkemans MJ, Harrell FE Jr, Habbema JD. Prognostic modelling with logistic regression analysis: a comparison of selection and estimation methods in small data sets. Stat Med. 2000;19:1059–1079[Medline]
    
11. Blackstone EH. Comparing apples and oranges. J Thorac Cardiovasc Surg. 2002;123:8–15[Free Full Text]
    
12. Drake C, Fisher L. Prognostic models and the propensity score. Int J Epidemiol. 1995;24:183–187[Abstract/Free Full Text]
    
13. Blackstone EH. Breaking down barriers: helpful breakthrough statistical methods you need to understand better. J Thorac Cardiovasc Surg. 2001;122:430–439[Free Full Text]
    
14. Breiman L. Bagging predictors. Machine Learning. 1996;26:123–140
    
15. Medalion B, Blackstone EH, Lytle BW, White J, Arnold JH, Cosgrove DM. Aortic valve replacement: is valve size important? J Thorac Cardiovasc Surg. 2000;119:963–974[Abstract/Free Full Text]
    
16. Rao V, Jamieson WR, Ivanov J, Armstrong S, David TE. Prosthesis-patient mismatch affects survival after aortic valve replacement. Circulation. 2000;102(19 Suppl 3):III5–9
    
17. He GW, Grunkemeier GL, Gately HL, Furnary AP, Starr A. Up to thirty-year survival after aortic valve replacement in the small aortic root. Ann Thorac Surg. 1995;59:1056–1062[Abstract/Free Full Text]
    
18. Hanayama N, Christakis GT, Mallidi HR, Joyner CD, Fremes SE, Morgan CD, et al. Patient prosthesis mismatch is rare after aortic valve replacement: valve size may be irrelevant. Ann Thorac Surg. 2002;73:1822–1829[Abstract/Free Full Text]
    
19. Frapier JM, Rouviere P, Razcka F, Aymard T, Albat B, Chaptal PA. Influence of patient-prosthesis mismatch on long-term results after aortic valve replacement with a stented bioprosthesis. J Heart Valve Dis. 2002;11:543–551[Medline]
    
20. Pibarot P, Dumesnil JG. Hemodynamic and clinical impact of prosthesis-patient mismatch in the aortic valve position and its prevention. J Am Coll Cardiol. 2000;36:1131–1141[Abstract/Free Full Text]
    
21. Butchart EG, Payne N, Li HH, Buchan K, Mandana K, Grunkemeier GL. Better anticoagulation control improves survival after valve replacement. J Thorac Cardiovasc Surg. 2002;123:715–723[Abstract/Free Full Text]
    
22. McGiffin DC, O’Brien MF, Galbraith AJ, McLachlan GJ, Stafford EG, Gardner MA, et al. An analysis of risk factors for death and mode-specific death after aortic valve replacement with allograft, xenograft, and mechanical valves. J Thorac Cardiovasc Surg. 1993;106:895–911[Abstract]
    
23. Izzat MB, Kadir I, Reeves B, Wilde P, Bryan AJ, Angelini GD. Patient-prosthesis mismatch is negligible with modern small-size aortic valve prostheses. Ann Thorac Surg. 1999;68:1657–1660[Abstract/Free Full Text]
    
24. Connolly HM, Oh JK, Schaff HV, Roger VL, Osborn SL, Hodge DO, et al. Severe aortic stenosis with low transvalvular gradient and severe left ventricular dysfunction: result of aortic valve replacement in 52 patients. Circulation. 2000;101:1940–1946[Abstract/Free Full Text]
    
25. Pibarot P, Honos GN, Durand LG, Dumesnil JG. The effect of prosthesis-patient mismatch on aortic bioprosthetic valve hemodynamic performance and patient clinical status. Can J Cardiol. 1996;12:379–387[Medline]
    
26. Abdelnoor M, Nitter-Hauge S, Trettli S. Relative survival of patients after heart valve replacement. Eur Heart J. 1990;11:23–28[Abstract/Free Full Text]
    
27. He GW, Acuff TE, Ryan WH, Douthit MB, Bowman RT, He YH, et al. Aortic valve replacement: determinants of operative mortality. Ann Thorac Surg. 1994;57:1140–1146[Abstract]
    
28. Morris JJ, Schaff HV, Mullany CJ, Rastogi A, McGregor CG, Daly RC, et al. Determinants of survival and recovery of left ventricular function after aortic valve replacement. Ann Thorac Surg. 1993;56:22–29[Abstract]
    
29. Sommers KE, David TE. Aortic valve replacement with patch enlargement of the aortic annulus. Ann Thorac Surg. 1997;63:1608–1612[Abstract/Free Full Text]
    
30. Castro LJ, Arcidi JM Jr, Fisher AL, Gaudiani VA. Routine enlargement of the small aortic root: a preventive strategy to minimize mismatch. Ann Thorac Surg. 2002;74:31–36[Abstract/Free Full Text]
    
31. Medalion B, Lytle BW, McCarthy PM, Stewart RW, Arheart KL, Arnold JH, et al. Aortic valve replacement for octogenarians: are small valves bad? Ann Thorac Surg. 1998;66:699–705[Abstract/Free Full Text]
    
32. Rao V, Christakis GT, Sever J, Fremes SE, Bhatnagar G, Cohen G, et al. A novel comparison of stentless versus stented valves in the small aortic root. J Thorac Cardiovasc Surg. 1999;117:431–436[Abstract/Free Full Text]
    
33. Blais C, Pibarot P, Dumesnil JG, Garcia D, Chen D, Durand LG. Comparison of valve resistance with effective orifice area regarding flow dependence. Am J Cardiol. 2001;88:45–52[Medline]
    
34. Dumesnil JG, Honos GN, Lemieux M, Beauchemin J. Validation and applications of indexed aortic prosthetic valve areas calculated by Doppler echocardiography. J Am Coll Cardiol. 1990;16:637–643[Abstract]
    
35. Barratt-Boyes BG. The timing of operation in valvular insufficiency. J Card Surg. 1987;2:435–452[Medline]
    
36. Pibarot P, Dumesnil JG, Cartier PC, Metras J, Lemieux MD. Patient-prosthesis mismatch can be predicted at the time of operation. Ann Thorac Surg. 2001;71(5 Suppl):S265–268[Abstract/Free Full Text]
    
37. Rimoldi H, Lev M. A note on the concept of normality and abnormality in quantitation of pathologic findings in congenital heart disease. Pediatr Clin North Am. 1963;10:589–591
    
38. Jin XY, Zhang ZM, Gibson DG, Yacoub MH, Pepper JR. Effects of valve substitute on changes in left ventricular function and hypertrophy after aortic valve replacement. Ann Thorac Surg. 1996;62:683–690[Abstract/Free Full Text]
    
39. Ghosh PK. Small aortic valve prostheses in octogenarians. Ann Thorac Surg. 1999;67:1543–1544[Free Full Text]
    
40. Altman DG, Andersen PK. Bootstrap investigation of the stability of a Cox regression model. Stat Med. 1989;8:771–783[Medline]
    
41. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55[Abstract/Free Full Text]
    
42. Adams DH, Chen RH, Kadner A, Aranki SF, Allred EN, Cohn LH. Impact of small prosthetic valve size on operative mortality in elderly patients after aortic valve replacement for aortic stenosis: does gender matter? J Thorac Cardiovasc Surg. 1999;118:815–822[Abstract/Free Full Text]
    

This article has been cited by other articles:

				E. R. Nowicki, G. B. Pettersson, N. G. Smedira, E. E. Roselli, E. H. Blackstone, and B. W. Lytle Aortic Allograft Valve Reoperation: Surgical Challenges and Patient Risks Ann. Thorac. Surg., September 1, 2008; 86(3): 761 - 768. [Abstract] [Full Text] [PDF]

L. G. Svensson, T. Dewey, S. Kapadia, E. E. Roselli, A. Stewart, M. Williams, W. N. Anderson, D. Brown, M. Leon, B. Lytle, et al. United States feasibility study of transcatheter insertion of a stented aortic valve by the left ventricular apex. Ann. Thorac. Surg., July 1, 2008; 86(1): 46 - 55. [Abstract] [Full Text] [PDF]