Article Text

Download PDFPDF

Original article
Impact of prosthesis–patient mismatch after mitral valve replacement: a multicentre analysis of early outcomes and mid-term survival
  1. William Y Shi1,
  2. Cheng-Hon Yap2,3,
  3. Philip A Hayward1,
  4. Diem T Dinh2,
  5. Christopher M Reid2,
  6. Gilbert C Shardey2,4,5,
  7. Julian A Smith4,5
  1. 1Department of Cardiac Surgery, Austin Hospital, University of Melbourne, Melbourne, Australia
  2. 2Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia
  3. 3Department of Cardiac Surgery, Bristol Heart Institute, Bristol, UK
  4. 4Department of Cardiothoracic Surgery, Monash Medical Centre, Melbourne, Australia
  5. 5Department of Surgery, Monash University, Melbourne, Australia
  1. Correspondence to Dr Cheng-Hon Yap, Department of Cardiac Surgery, Bristol Heart Institute, Malborough Street, Bristol Royal Infirmary, Bristol BS28HW, UK; cheng-hon.yap{at}


Background Prosthesis–patient mismatch (PPM) is characterised by the effects of inadequate prosthesis size relative to body surface area (BSA). It is uncertain whether PPM after mitral valve replacement impacts upon clinical outcome. This was examined in an Australian population.

Methods From 2001 to 2009, 1006 mechanical and bioprosthetic mitral valves were implanted across 10 institutions. Effective orifice areas (EOA) were obtained from a literature review of in vivo echocardiographic data. Absent, moderate and severe PPM was defined as an indexed EOA (EOA/BSA) of >1.20 cm2/m2, >0.90 to ≤1.20 cm2/m2 and ≤0.9 cm2/m2, respectively. Early outcomes and 7-year survival were compared between these three groups.

Results PPM was absent in 34%, moderate in 53% and severe in 13% of patients. Patients with PPM were more likely to be male (42% vs 52% vs 62%, p<0.0001) and obese (14% vs 20% vs 56%, p<0.0001). Postoperatively there was similar 30-day mortality (5% vs 5% vs 6%, p=0.83) and early any mortality/morbidity (24% vs 27% vs 29%, p=0.40). Seven-year survival was similar between groups (72±4.1% vs 76±3.2% vs 69±10.3%, p=0.76). PPM did not predict adverse events after logistic and Cox regressions with and without propensity score adjustment. Subgroup analyses of those with isolated mitral valve surgery, patients with preoperative congestive heart failure and non-obese patients failed to show an association between PPM and mid-term mortality.

Conclusions Overall, PPM was not associated with poorer early outcomes or mid-term survival. Oversizing valves may be technically hazardous and do not yield superior outcomes. Easier implantation by appropriate sizing appears justified.

  • Surgery-valve
  • prosthetic heart valve
  • prosthesis-patient mismatch
  • mitral valve
  • mitral regurgitation
  • mitral stenosis
View Full Text

Statistics from


Prosthesis–patient mismatch (PPM) is a clinical entity characterised by the physiological effects of an inadequate effective orifice area (EOA) of the cardiac valve prosthesis relative to body surface area (BSA), resulting in a mismatch between EOA and flow requirements.1 2

PPM has been widely reported after aortic valve replacement (AVR). It is associated with deleterious effects on left ventricular remodelling, regression of mitral regurgitation, functional status as well as early and late survival.3–10

While aortic root enlargement has been described for avoidance of PPM in AVR,11 similar techniques are not available for the mitral valve, although strategies for optimising prosthesis size12 including upsizing by implantation in the left atrial free wall have been described and utilised.13 Nonetheless, there has been increasing debate as to whether PPM after mitral valve replacement indeed predicts adverse short- and long-term events. The mechanisms by which PPM may affect survival have been described in detail previously.14 Potential physiological disturbances—such as high left atrial pressures, pulmonary venous and pulmonary arterial hypertension, right ventricular strain—may translate to an increased risk of subsequent mortality.14 15 Despite this, single-centre studies of mitral PPM have reported conflicting results,12 14 16 17 with a recent large study suggesting no clinical effect.18

The impact of PPM is clinically pertinent as any potential association with early or late adverse outcomes might influence surgeons' operative strategy. If it exists, it might offer one explanation for superior outcomes after mitral valve repair compared with replacement, with larger orifice areas in general preserved by repair.

In this study we aimed to examine the impact of PPM after mitral valve replacement on 30-day mortality, early morbidity and mid-term mortality by review of an Australian multicentre database.


Data collection

We performed a retrospective review of a multicentre database containing all adult cardiac procedures performed from 1 July 2001 to 31 December 2009 in 10 institutions. Data were prospectively compiled as part of the Australasian Society of Cardiac and Thoracic Surgeons database project, which records adult cardiac surgery procedures in the state of Victoria with mandatory participation of all six government-funded adult cardiac surgery units. Selected interstate units have joined in recent years.

During the study period the participating institutions were six Victorian public hospitals, one Victorian private hospital and three interstate public hospitals. Patients were excluded from analysis if they were undergoing cardiac reoperation, concomitant aortic valve procedure, thoracic aortic surgery and other non-cardiac procedures.

Overall, 1006 mechanical and bioprosthetic mitral valves were implanted during the study period and included for analysis.

The database records patient demographics, preoperative risk factors, operative details, postoperative hospital course and morbidity and mortality outcomes.19 Data were collected prospectively (using an agreed data set and definitions) by surgeons, perfusionists, resident medical officers and database managers. Thirty-day mortality information is obtained by telephone contact with patients, family members or the medical practitioner. The database is regularly subject to external audit measures with recently reported overall accuracy of data of 97.4%.20 Mid-term survival status of patients was obtained from the Australian National Death Index up to 18 March 2010.

Individual patient consent was waived for this study as the institutional review board of each participating unit had previously approved the use of this database for research.

Definition of prosthesis-patient mismatch

Valve prostheses EOA was obtained from a literature review of individual studies reporting in vivo echocardiographic data. Where more than one source was present, the mean value was calculated. Values were excluded if derived from less than five patient studies. We did not include EOAs from manufacturers' data or from reports in which the sample size was not clear. Valve EOA is presented in table 1. The EOA index (EOAI) was calculated using the equation EOAI = EOA/BSA. We defined absent, moderate and severe PPM as an EOAI of >1.20 cm2/m2, ≤1.20 cm2/m2 and ≤0.9 cm2/m2, respectively, in line with previous reports.12 14 17 18

Table 1

Effective orifice area (EOA) values obtained from literature review

Study end points

Our primary study end points were 30-day and 7-year mortality. In a secondary analysis we compared a number of important 30-day adverse events including prolonged ventilation (greater than 24 h), return to the operating theatre (for bleeding or valve dysfunction), myocardial infarction (MI), stroke, renal replacement therapy, as well as 30-day hospital readmissions for congestive heart failure (CHF) and valve dysfunction. We also examined a composite end point of ‘any mortality/morbidity’ which encompasses the events listed above.

Postoperative MI was defined as at least two of the following: cardiac enzyme elevation (creatinine kinase-myocardial band >30 U/l or troponin >20 mg/l), new wall motion abnormalities and new Q waves on at least two serial ECGs. CHF was defined by at least two of paroxysmal nocturnal dyspnoea, dyspnoea on exertion due to heart failure, chest x-ray showing pulmonary congestion or patient having received treatment for CHF with ACE inhibition, diuretics, carvedilol or digoxin.

Subgroup analyses

We examined the impact of PPM on mid-term mortality in three subgroups. An analysis was performed for patients who underwent isolated mitral valve replacement with no concomitant surgery (n=500) to examine for any impact of PPM not apparent in the entire cohort.

Analysis was repeated in those with pre-existing CHF (defined above, n=563). This was performed as these patients may be more vulnerable to any subtle effects of PPM masked by analysis of the entire cohort. Pulmonary arterial pressures were not uniformly collected by the participating units and therefore could not be analysed.

The presence of obesity may influence the classification of PPM. A recent study showed that, after AVR, PPM negatively impacts upon survival in non-obese patients only35 as the use of BSA to calculate may overestimate the prevalence and severity of PPM in obese patients.35 Furthermore, stroke volume and cardiac output have been shown to be more closely correlated with fat-free body mass than adipose mass, BMI and other variables.36 We therefore performed a subgroup analysis for patients with a BMI of <30 kg/m2 (n=776) to verify the findings in the entire cohort where a potential exists for the misclassification of PPM in obese patients.

Statistical analysis

Categorical variables were expressed as frequencies and compared using the Fisher exact and χ2 tests. Continuous variables were expressed as mean±SD and compared using the unpaired t test and one-way analysis of variance for the three PPM categories, while the Bonferroni post hoc test was used for multiple pairwise comparisons.

The Kaplan–Meier method and log-rank test were used to compare survival. Multivariable logistic regression and Cox regression, both performed in a backward elimination fashion, were used to identify independent predictors of early and mid-term outcomes, respectively. We also modelled EOAI as a continuous variable in regression analyses. The backward elimination method was used so as to include all potential predictors of end points in the initial models and subsequently eliminate covariates in an iterative process to create a final model. Inclusion of all potentially important variables in the initial model allows their joint predictive behaviour to be initially evaluated, which is important given that a set of variables may exhibit predictive capability even if a subset does not.

In order to further elicit any subtle impact of severe PPM, we performed multivariable adjustment with the propensity score. The propensity score is a measure of the likelihood that a patient will be classified as having ‘severe PPM’ based on the patient's covariate scores. To generate propensity scores we fitted a logistic regression model in which the outcome was the patient having severe PPM after MVR, and entered all demographic, preoperative clinical and investigative variables that clinically could potentially and plausibly affect early outcomes and mid-term mortality as well as patient classification into the severe PPM group. These variables included age, gender, hypertension, hypercholesterolaemia, diabetes mellitus, obesity, BSA, renal failure (preoperative creatinine >200 μmol/l), cerebrovascular disease, peripheral vascular disease, chronic obstructive pulmonary disease (COPD), coronary artery disease, history of acute MI, infective endocarditis, congestive cardiac failure, atrial fibrillation, New York Heart Association class, urgency of case, inotrope, anticoagulant and/or immunosuppressive medication use at time of surgery, previous percutaneous coronary intervention, mitral valve pathology, left-main coronary disease, number of diseased coronary systems, left ventricular function, concomitant coronary artery bypass grafting (CABG), concomitant tricuspid valve repair, prosthesis size and bioprothesis implantation. The c-statistic was calculated to evaluate discriminative ability of the model. Once generated, the propensity scores and the severe PPM variable were forced into backward elimination logistic and Cox regressions for early outcomes and mid-term mortality, respectively, and the adjusted odds and hazard ratios for severe PPM calculated.

To enable a two-group propensity score comparison, the analysis was initially performed upon exclusion of the ‘moderate PPM group’ as we reasoned that its inclusion would provide too heterogeneous a group with which to compare patients in the severe PPM category, thus masking any potentially significant findings. Subsequently, the same analysis was repeated for the entire cohort comparing patients with severe PPM to those with moderate or insignificant PPM.

We elected not to use propensity score matching as the marked group differences between PPM groups would have resulted in only a small number of matched patient pairs.


Of the 1006 study patients, PPM was severe in 133 (13%), moderate in 532 (53%) and absent in 341 (34%). The clinical characteristics of the patients are shown in table 2. Patients with PPM were more likely to be male (42% vs 52% vs 62%, p<0.0001), obese (14% vs 20% vs 56%, p<0.0001), have >2+ mitral valve regurgitation (85% vs 86% vs 93%, p=0.054) and more likely to have had a previous MI (14% vs 19% vs 29%, p=0.001). PPM was more likely to occur with implantation of a bioprosthesis (30% vs 37% vs 62%, p<0.0001). Despite this, there were no statistically significant differences in additive (p=0.75) or logistic (p=0.79) EuroSCORE between the three groups.

Table 2

Preoperative clinical characteristics

Intraoperative and early outcomes are shown in table 3. Thirty-day mortality was similar between groups (5% vs 5% vs 6%, p=0.83). There were similar rates of all other morbidities between the three groups including the composite end point of any mortality/morbidity.

Table 3

Intraoperative and early postoperative outcomes

Logistic regression identified several independent predictors of 30-day mortality and early morbidity. However, the category of PPM was not an independent predictor (table 4).

Table 4

Logistic and Cox regressions for independent predictors of early and late outcomes

Seven-year survival was similar between groups (72±4.1% vs 76±3.2% vs 69±10.3%, p=0.76). Kaplan-Meier survival curves are displayed in figure 1. After multivariable Cox regression, moderate and severe PPM did not predict mid-term mortality (table 4).

After multivariable adjustment with the propensity score (c statistic=0.96 for analysis with moderate PPM group excluded; c statistic=0.95 for analysis including moderate PPM group) in logistic and Cox regressions, severe PPM was still not an independent predictor of adverse events (table 5). This was the case regardless of whether or not the moderate PPM group was included for analysis. Furthermore, EOAI—when modelled as a continuous variable—was not a predictor of early any mortality/morbidity or mid-term mortality.

Table 5

Impact of severe patient–prosthesis mismatch (PPM) after multivariable adjustment with propensity scores

An analysis was performed for the subgroup of patients who underwent isolated mitral valve replacement with no concomitant surgery (n=500). Thirty-day mortality (4 (2.1%) vs 8 (3.1%) vs 1 (1.6%), p=0.72) and 30-day any mortality/morbidity (32 (17%) vs 47 (18%) vs 14 (23%), p=0.62) were comparable between groups. Furthermore, no statistical significance was observed in 7-year survival (74±5.4% vs 87±3.0% vs 53±20%, p=0.47). Cox regression did not show an association between moderate (p=0.25, HR 0.67, 95% CI 0.34 to 1.62) and severe PPM (p=0.27, HR 0.55, 95% CI 0.19 to 1.61) and mid-term mortality.

Figure 1

Kaplan–Meier survival curve for the three categories of prosthesis–patient mismatch (PPM) in (A) the entire cohort (p=0.76, log-rank test), (B) a subgroup undergoing isolated mitral valve replacement (p=0.47) and (C) a subgroup of patients with pre-existing congestive heart failure (p=0.19). p=NS for comparisons of absent PPM versus moderate PPM, absent versus severe and moderate versus severe for all Kaplan–Meier analyses.

The analysis was repeated in those with pre-existing CHF (n=563). Thirty-day mortality (11 (5.6%) vs 16 (5.4%) vs 6 (8.2%), p=0.65) and 30-day any mortality/morbidity (56 (29%) vs 86 (29%) vs 28 (38%), p=0.26) were comparable between groups. Again, no statistical significance was observed in 7-year survival (68±5.2% vs 75±4.3% to 65±11%, p=0.19). Cox regression also failed to show an association between moderate (p=0.16, HR 0.64, 95% CI 0.38 to 1.07) and severe PPM (p=0.99, HR 1.0, 95% CI 0.49 to 2.03) and mid-term mortality.

In non-obese patients (n=776), 30-day mortality (15 (5.1%) vs 22 (5.2%) vs 4 (6.9%), p=0.85) and 30-day any mortality/morbidity (75 (26%) vs 120 (28%) vs 16 (28%), p=0.71) were comparable between groups. No statistical significance was observed in 7-year survival (71±4.4% vs 76±3.4% vs 66±19%, p=0.96). Cox regression showed no association between moderate (p=0.30, HR 0.79, 95% CI 0.50 to 1.25) and severe PPM (p=0.76, HR 0.87, 95% CI 0.37 to 2.08) and mid-term mortality.

Only a single reference set of in vivo EOA data could be obtained for the ATS valve (ATS Medical, Minneapolis, Minnesota, USA) as one37 of the two sources with published EOAs specifically acknowledged that they were derived from the pressure half time equation. The source we used for the haemodynamic characteristics of the ATS valve, while obtained in vivo, did not specify that the EOA values were calculated using the continuity equation. No other EOA data for the ATS valve could be obtained from the literature for verification. The data we have employed suggest a larger EOA than the other mechanical prostheses in this study. It is possible that the published EOA is an overestimate, reducing the prevalence of PPM among recipients. If so, this might have masked a PPM effect in view of the proportion of the whole cohort occupied by the ATS group.

We therefore repeated the same analysis involving all subgroups as described after exclusion of the 163 (16%) patients who received the ATS valve. This verified the findings from the entire cohort with neither category of PPM nor EOAI being predictive of any early adverse events or mid-term mortality in all statistical models and subgroup analyses (entire cohort without ATS valve: Kaplan–Meier 58±8.1% vs 76±3.2% vs 69±10.3%, p=0.35; Cox regression p=0.44 for severe PPM).


This study is the first to report multicentre data on the impact of PPM after mitral valve replacement. This analysis reflects a multicentre ‘real-world’ experience, thus theoretically minimising potential biases associated with surgical techniques unique to individual units or subspecialist surgeons.

Intraoperatively, aortic cross clamp and cardiopulmonary bypass pump times were significantly longer in patients with severe PPM, which is consistent with a previous report.14 This probably reflects the greater proportion of patients with left main and triple vessel coronary disease who received concomitant CABG. We detected a trend (p=0.057) towards a greater need for haemofiltration between the absent and severe PPM groups. This may again be related to the prevalence of coronary disease as well as greater proportion preoperative left ventricular (LV) dysfunction with increasing PPM.

In our analysis, no impact of PPM was detected either in the entire cohort or subgroup analyses. Our findings are consistent with the larger tri-centre analysis performed by Jamieson et al18 which found no impact of mitral valve PPM on mortality in 2440 patients after 15 years of follow-up. However, PPM was found to be associated with poorer long-term survival in patients with pre-existing pulmonary hypertension,18 a phenomenon we were unable to verify as pulmonary arterial pressures were not uniformly available in our dataset.

A comprehensive analysis of 765 consecutive procedures by Aziz et al12 found PPM to be associated with deleterious outcomes and recommended avoidance of PPM in patients undergoing mechanical valve replacement and in older patients receiving bioprosthetic valves. However, the study population differed from ours in that 22% of cases were re-operations, of which half were of the mitral valve. Additionally, 23% of patients underwent concomitant AVR, where similar undersizing may have also influenced longer term outcomes. Furthermore, 31% of patients received the Hancock Standard, Hancock II and Medtronic Hall prostheses, none of which were implanted in our series. The study by Magne et al14 also differs from ours in its population. There, 85% of patients received mechanical prostheses while 53% received a prosthesis ≤27 mm and 16% one which was ≤25 mm. Again, cardiac re-operations, including those of the mitral valve were included in the sample. The role of PPM in patients receiving small prostheses (≤27 mm) may warrant further investigation, given the findings of an association of PPM with late mortality. We did not perform this subgroup analysis because of the relatively small number of patients (25%) receiving such valves. Indeed, the impact of PPM may potentially lie in the complexity of the case or be influenced by the heterogeneity of the sample population.

Lam et al previously reported a greater rate of late readmissions for CHF in PPM patients,16 while an association between PPM and subsequent postoperative pulmonary arterial hypertension has also been reported.15 16 However, in this study, data were not available to investigate these end points. The lack of difference in long-term mortality may in part be due to improvements in the medical management of CHF which may be preventing the physiological effects of PPM from impacting on survival. We performed a subgroup analysis in patients with pre-existing CHF in order to investigate the impact of PPM when clinical effects of elevated left atrial pressure are already present. However, our negative finding may be secondary to the poor correlation between the definition of CHF in our database and the presence or absence of pulmonary arterial hypertension.

We did not identify any impact of PPM in any of our subgroup analyses. Analysis after exclusion of the ATS valve confirmed the findings in the entire cohort.

Our results also show that bioprosthesis use is independently associated with late mortality, consistent with reports from Jamieson et al18 and Lam et al.16 It is possible that degenerative changes in the bioprostheses are responsible for this observation, although further studies are required to investigate this finding.

In this study a greater degree of PPM was associated with a higher proportion of comorbidities, which were identified as independent predictors of mid-term mortality including COPD, concomitant CABG and bioprosthesis implantation. Despite this, PPM was not associated with adverse early outcomes or mid-term survival even in unadjusted analyses. While there are a number of potential statistical explanations for this—such as heterogeneity of data and immeasurable covariates exerting opposing effects—it cannot be discounted that PPM might potentially be beneficial in the diseased heart by limiting preload at a level with which the impaired LV can cope, and the resultant pulmonary venous hypertension being well tolerated, either by chronic pulmonary vascular changes in long-standing mitral valve disease38 or by being non-progressive and remaining within limits with which the right ventricle (RV) can cope and may already be well attuned to. If this is true, in a proportion of patients at least, this may mask any discernible negative effect of PPM in unadjusted analyses.

We speculate that there may be subgroups in which PPM exerts different effects. In the failing LV with preserved RV function, it may prove protective by limiting preload as described above. In patients with preoperative mitral stenosis with RV impairment, valve replacement even with a degree of PPM may prove neutral by improving RV afterload to tolerable levels while the limitation of left heart output is insignificant relative to the limitations of pulmonary flow imposed by chronic pulmonary vascular changes. Finally, PPM may be harmful in gross biventricular impairment where it maintains elevated RV afterload at intolerable levels and limits LV preload below the critical level required by the failing myocardium attuned to high filling pressures.

These varying effects in a spectrum of clinical scenarios may explain why, overall, no discerning negative effect of PPM can be found in our unadjusted and adjusted analyses such as Kaplan–Meier and regression estimates of early outcomes and mid-term survival.

As alluded to previously, the prostheses used in this study are different from those in previously published series. In particular, popular prostheses such as the Medtronic-Hall, Advantage, Intact, Hancock Standard and Hancock II were not used in our study, which may be contributing to the differences in results between this and earlier studies.

Limitations of the study

We acknowledge the limitations inherent in this study. As a retrospective analysis, hidden biases may influence findings despite multivariable correction. In addition, the EOAs of the prostheses were obtained from published in vivo echocardiographic data collected from various centres, with the EOAs of some prostheses being derived from series with relatively small patient numbers. Furthermore, the relatively lower number of patients remaining at the later years of follow-up limits the ability to make robust statistical conclusions, especially in subgroup analyses where patient numbers are small.


In this multicentre analysis of 1006 patients, PPM after mitral valve replacement was not associated with 30-day adverse events or mid-term mortality. Surgeons may therefore select mitral valve prostheses for ease of implantation without prejudicing these early and mid-term outcomes. Manoeuvres to upsize the prosthesis are potentially hazardous39—particularly in bioprostheses where struts may impinge on the left ventricular free wall—and does not appear to be warranted by these findings.


The authors acknowledge all surgeons who contributed to the operations studied in this paper. The following investigators, data managers and institutions currently participate in the ASCTS database: Alfred Hospital: A Pick, J Duncan; Austin Hospital: S Seevanayagam, M Shaw; Cabrini Health: G Shardey; Geelong Hospital: M Morteza, C Bright; Flinders Medical Centre: J Knight, R Baker, J Helm; Jessie McPherson Private Hospital: J Smith, H Baxter; John Hunter Hospital: A James, S Scaybrook; Lake Macquarie Hospital: B Dennett, M Jacobi; Liverpool Hospital: B French, N Hewitt; Mater Health Services: AM Diqer, J Archer; Monash Medical Centre: J Smith, H Baxter; Prince of Wales Hospital: H Wolfenden, D Weerasinge; Royal Melbourne Hospital: P Skillington, S Law; Royal Prince Alfred Hospital: M Wilson, L Turner; St George Hospital: G Fermanis, C Redmond; St Vincent's Hospital, VIC: M Yii, A Newcomb, J Mack, K Duve; St Vincent's Hospital, NSW: P Spratt, T Hunter; The Canberra Hospital: P Bissaker, K Butler; Townsville Hospital: R Tam, A Farley; Westmead Hospital: R Costa, M Halaka. We also wish to acknowledge Dr Peter C Austin of the University of Toronto for his statistical advice.


View Abstract


  • This paper was presented on 22 March 2011 at the Annual Meeting of the Society for Cardiothoracic Surgery in Great Britain and Ireland, Excel London, London, UK.

  • Funding The Australasian Society of Cardiac and Thoracic Surgeons (ASCTS) National Cardiac Surgery Database is funded by the Department of Human Services, Victoria and the Health Administration Corporation (GMCT) and the Clinical Excellence Commission (CEC), New South Wales, Australia.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.