Article Text

Download PDFPDF

Original research article
Net atrioventricular compliance is an independent predictor of cardiovascular death in mitral stenosis
  1. Maria Carmo Pereira Nunes1,2,
  2. Timothy C Tan2,
  3. Sammy Elmariah3,
  4. Lucas Lodi-Junqueira1,
  5. Bruno Ramos Nascimento1,
  6. Rodrigo do Lago3,
  7. Jose Luiz Padilha da Silva4,
  8. Rodrigo Citton Padilha Reis4,
  9. Xin Zeng2,
  10. Igor F Palacios3,
  11. Judy Hung2,
  12. Robert A Levine2
  1. 1 Hospital das Clinicas, School of Medicine, Federal University of Minas Gerais, Belo Horizonte, Brazil
  2. 2 Cardiac Ultrasound Lab, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
  3. 3 Division of Cardiology, Department of Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA
  4. 4 Department of Statistics, Federal University of Minas Gerais, Belo Horizonte, Minas Gerais, Brazil
  1. Correspondence to Professor Maria Carmo Pereira Nunes, Department of Internal Medicine, School of Medicine of the Federal University of Minas Gerais.Av. Professor Alfredo Balena, 190, Santa Efigênia, Belo Horizonte 30130 100, Minas Gerais, Brazil; mcarmo{at}


Objectives Rheumatic mitral stenosis (MS) is a progressive disease, and risk of death may persist despite relief of the obstruction. Net atrioventricular compliance (Cn) modulates the overall haemodynamic burden of the MS and may be useful in predicting cardiovascular death after percutaneous mitral valvuloplasty (PMV).

Methods A total of 427 patients (mean age 50±16 years, 84% female) with severe MS undergoing PMV were enrolled. Doppler-derived Cn was estimated at baseline using a previously validated equation. The primary endpoint was late cardiovascular death, and the secondary endpoint was a composite of all-cause mortality, mitral valve (MV) replacement or repeat PMV over a median follow-up of 31 months (IQR: 7.8–49.2 months).

Results At baseline, 209 patients (49%) were in New York Heart Association (NYHA) functional class III or IV. During follow-up, 49 patients died (41 cardiovascular deaths), 50 underwent MV replacement and 12 required repeat PMV, with an overall incidence of cardiac mortality and adverse events of 4.1 deaths and 11.1 events per 100 patient-years, respectively. Low baseline Cn was a strong predictor of both cardiac death (adjusted HR 0.70, 95% CI 0.49 to 0.86) and composite endpoint (adjusted HR 0.81, 95% CI 0.67 to 0.91) after adjusting for clinical factors, baseline pulmonary artery pressure, tricuspid regurgitation severity, right ventricular function and immediate procedural haemodynamic data. The inclusion of Cn in a model with conventional parameters resulted in improvement in 5-year cardiovascular mortality risk prediction.

Conclusions Baseline Cn is a strong predictor of cardiovascular death in patients with MS undergoing PMV, independent of other prognostic markers of decreased survival in MS, including baseline patient characteristics and postprocedural data. Cn assessment therefore has potential value in evaluation of cardiovascular mortality risk in the setting of MS.

  • Rheumatic heart disease
  • net atrioventricular compliance
  • mitral stenosis
  • percutaneous mitral valvuloplasty
  • cardiovascular death
  • mortality

Statistics from

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.


Mitral stenosis (MS) is a frequent manifestation of the rheumatic disease with a gradual progression from the initial valve damage to the onset of symptoms.1 The prognosis of patients with MS depends on the severity of symptoms, and the mortality is mainly due to heart failure or embolic events.2 However, our current understanding of the natural history of MS comes from older studies prior to the widespread use of interventional techniques for relieving MS.3

Pulmonary hypertension (PH) is related to the MS severity with impacts on symptoms, exercise capacity and outcome.2 4 However, PH has a highly variable prevalence and cannot be determined solely by the stenotic lesion. Furthermore, despite the alleviation of mitral valve (MV) obstruction after percutaneous mitral valvuloplasty (PMV), risk of adverse events, in particular late cardiac death, persists.5–8 This observation may be related to factors independent of the valve lesion per se and reflect functional and structural changes in the pulmonary circulation that may not be completely reversed by the procedure.4

There is growing recognition that the impact of MS on both PH and clinical outcome is mediated by left atrial compliance, defined as the change in pressure per unit change in volume.9–14 Net atrioventricular compliance (Cn), which expresses the compliance of the atrium and ventricle, can be quantified non-invasively by dividing the slope of the mitral deceleration into the effective MV area.15

Cn strongly contributes to the haemodynamic burden of the valve stenosis,10 11 16 17 especially contributing to the excessive elevation of the pulmonary arterial pressure in response to exercise.10 However, limited data are available regarding the prognostic value of Cn.13 18 19 We have previously shown that baseline Cn is a powerful predictor of the need for MV intervention,18 but with few events, which limited analysis of the role of Cn in predicting death and indicated the need for further studies.

Therefore, in this study, we tested the hypothesis that baseline Cn can identify patients at risk for adverse events, specifically cardiac death, in a large series of patients undergoing PMV, potentially providing additional prognostic value based on adaptation mechanisms induced by the underlying stenosis but not completely dependent on it.


Study population

All patients who underwent PMV with the Inoue balloon technique for haemodynamically significant MS between 2000 and 2012 at two centres were initially recruited for the study. Patients considered unsuitable for cardiac surgery or those with concomitant severe aortic valve disease who underwent PMV as an alternative treatment were excluded. Based on these exclusion criteria, 119 patients were not eligible, and 427 patients were included: 233 patients from Massachusetts General Hospital, USA, and 194 patients from Hospital das Clinicas of the Federal University of Minas Gerais, Brazil. The study protocol was approved by the institutional review board (IRB) of both institutions, and written informed consent was obtained from all patients after IRB approval. From both cohorts, we selected patients who had the same indications for the PMV, and the same clinical, echocardiographic and haemodynamic data were assessed. The main characteristics of the patients from USA and Brazil have been compared and are shown in online supplement table 1 in the online supplemental material.

Supplementary Material

Supplementary material 3

Patients were referred for PMV based on a combination of functional limitation, severity of obstruction and PH either at rest or during exercise.20 Comprehensive baseline comorbidity data were collected from hospital records and the Charlson comorbidity index (CCI) was calculated.21

Standard echocardiographic and Doppler examinations were performed in all patients with detailed evaluation of the valve morphology. Mitral regurgitation (MR) and tricuspid regurgitation (TR) were detected and semiquantitatively graded with colour flow imaging.22 All results were based on the average of three measurements for patients in sinus rhythm and five measurements for patients in atrial fibrillation.

Assessment of Cn

Baseline Cn was determined as previously described using the equation10 13 14: Cn (mL/mm Hg)=1270 × (planimetric MV area (cm2)/E-wave downslope (cm/s2)) before the PMV (figure 1). As a trans-septal puncture of the interatrial septum is performed during the PMV, creating an iatrogenic left-to-right shunt and impacting the LA pressure, Cn was not calculated after PMV.23

Figure 1

Calculation of Cn by Doppler echocardiography. Cn was estimated from the ratio of mitral valve area by planimetry (A) and the slope of the mitral velocity E wave by continuous wave Doppler recording from transmitral flow (B) using the equation: Cn (mL/mm Hg)=1270 × (planimetric mitral valve area (cm2)/E-wave downslope (cm/s2)). Cn, net atrioventricular compliance; LV, left ventricle; RV, right ventricle.

Cardiac catheterisation/PMV

Standard haemodynamic measurements of the left ventricular, left atrial, right ventricular and pulmonary artery pressures were recorded before and immediately after the procedure. Transpulmonary pressure gradient was calculated as the difference between mean pulmonary artery pressure (mPAP) and pulmonary artery wedge pressure. Pulmonary vascular resistance was calculated using the following formula: pulmonary vascular resistance (PVR)=(mPAP – mean PCWP)/cardiac output (CO).

PMV was performed using an anterograde trans-septal approach using Inoue technique as previously described.3 After each dilatation, a periprocedural transthoracic echocardiogram was performed to assess MV orifice area by planimetry, and the degree of MR to determine if further dilatation was required.

Endpoint definitions

The date of enrollment in the study was defined as the date on which PMV was performed. For the purpose of this study, we did not categorise procedural success as a binary outcome but assessed the results using MV area as a continuous variable and the degree of MR measured 24 hours after the procedure.

The primary endpoint was cardiovascular-related death, and the secondary endpoint was a composite of all-cause mortality, MV replacement or repeat PMV. Cardiovascular mortality was defined as death due to stroke, heart failure or sudden death. Follow-up data were obtained from clinic follow-up appointments on an outpatient basis. Additional information was also obtained by contacting family members or telephone interview of the patients.

Statistical analysis

Categorical variables, expressed as numbers and percentages, were compared using χ2 testing, whereas continuous data, expressed as mean±SD or median and IQR, were compared using Student’s unpaired and paired t-test or the Mann-Whitney U test, as appropriate.

Multiple imputation techniques were used for substitution of missing variables, especially to predict missing values of Cn that potentially biased regression coefficients and prediction. All variables had <9% missing values. Specifically, multiple imputations were generated via the multivariate imputation by chained equations procedure. Each of the 200 complete datasets generated were analysed using Cox proportional hazards models to obtain the estimates of interest. In order to prevent overfitting, the shrinkage factor of regression coefficients was applied after the models have been fitted by traditional methods.24 For each imputed dataset, adjusted (‘shrunk’) regression coefficients were calculated, and the final shrinkage factor was taken as the average over the 200 imputations.

Cox proportional hazards regression analyses were performed to identify independent predictors of cardiovascular death and combined adverse outcome, including baseline and postprocedural variables. For multivariable analyses, we used predefined Cox proportional hazards multivariable models that included covariates considered of potential prognostic impact in MS.

Regression models included cohort-stratified baseline hazard functions. In other words, we assumed that baseline risk for cardiovascular death may differ between individuals from USA and Brazil cohorts and assessed this assumption by testing for cohort–risk factor interactions. In addition, as age is strongly related to death, we have controlled for age by adding baseline age as a covariate to the Cox model.

To compare the accuracy of the model with and without Cn, receiver operating characteristics curves were generated, and the area under the curves (AUC) was calculated. Internal validity of the models was assessed with standard bootstrapping procedures.24 Net reclassification improvement (NRI) was used to evaluate the added value of Cn over the previous model for 5-year mortality.25 The continuous NRI was used as it does not require any risk categorisation and considers all changes in predicted risk for all events and nonevents.26

Statistical analysis was performed using the Statistical Package for Social Sciences for Windows, V.18.0 and R for Statistical Computing V.2.15.1 (R Foundation, Vienna, Austria). The following R packages were used: survival for the Cox’s regression models, mice for multiple imputation and rms for the validation.


Baseline characteristics of the study population

The study cohort included 361 female patients (84%); mean age was 50.3±16.1 years. The baseline characteristics of the entire study population, comparing those who survived with those who died, are summarised in table 1. The majority of patients had no comorbidities (72%), whereas 4% had CCI score of ≥3.

Table 1

Baseline clinical and echocardiographic characteristics of the entire patient population and after stratification according to the cardiovascular mortality

Patients who died had more severe and long-standing underlying disease as expressed by poor functional capacity, less-favourable valve anatomy, lower Cn, higher pulmonary artery pressure and PVR compared with patients who survived. In addition, changes in the right heart response to MS was more pronounced in non-survivor patients than survivor patients, with right atrial enlargement, right ventricle (RV) dysfunction and severe functional TR (table 1).

PMV resulted in decrease in mPAP without significant change in PVR immediately after PMV. TR severity did not change after PMV. Twenty-seven patients (6%) developed severe MR after PMV, 13 of them underwent surgery for MV replacement during the hospitalisation. The haemodynamic measurements and echocardiographic data after PMV, comparing survivors with non-survivors, are summarised in table 2.

Table 2

Haemodynamic measurements and echocardiographic data after PMV of the entire patient population and after stratification according to the cardiovascular mortality


During a median follow-up period of 31 months (IQR: 7.8–49.2 months), 49 patients (11.5%) died. The overall all-cause mortality included 41 cardiovascular deaths (30 due to right-sided heart failure, 9 due to cerebrovascular events, and two due to sudden death) and eight non-cardiovascular. The overall incidence of cardiac mortality was 4.1 deaths per 100 patient-years. In addition, 50 patients underwent subsequent MV replacement and 12 required repeat PMV with an overall incidence of all adverse events of 11.4 events per 100 patient-years. Survival rate in the patients with good immediate results after PMV at 1-year, 3-year and 5-year follow-up was 97%, 93% and 91%, respectively. Considering a combined secondary endpoint, the event-free survival rate at 1-year, 3-year and 5-year follow-up was 93%, 86% and 83%, respectively.

Predictors of cardiovascular death

Univariable Cox proportional hazards analyses identified several clinical, echocardiographic and haemodynamic parameters as predictors of cardiovascular mortality (table 3).

Table 3

Univariable Cox proportional hazards analysis for predicting cardiac death and composite endpoint

Several haemodynamic parameters of MS severity, including transmitral pressure gradients, pulmonary arterial pressure, PVR and Cn, were associated with mortality. By multivariable analysis, age (HR 1.064, 95% CI 1.037 to 1.091; NYHA functional class III/IV (HR 3.958, 95% CI 1.526 to 10.268); presence of moderate or severe TR (HR 1.914, 95% CI 1.205 to 3.036) and Cn (HR 0.70, 95% CI 0.49 to 0.86) were predictors of death (table 4). Postprocedural variables did not remain in the multivariable model that included baseline Cn.

Table 4

Multivariable Cox proportional hazards analysis for predicting cardiac death and composite endpoint (200 imputations)

The 5-year survival rate was higher for patients with NYHA functional class I or II, mild TR (online supplementary figures 1 and 2) and Cn of >4 mL/mm Hg (figure 2). In particular, survival rate at 1-year, 3-year and 5-year follow-up was 97%, 96% and 87%, respectively, in patients with Cn of >4 mL/mm Hg compared with 95%, 89% and 69%, respectively, in patients with Cn of ≤4 mL/mm Hg (p=0.010). The Kaplan-Meier curves showed a gradual increase in mortality rates, long term after the procedure, which is mainly related to the progression of the disease.

Figure 2

Kaplan-Meier survival curves according to Cn in patients with MS undergoing PMV. The selected endpoint was cardiovascular-related death. Cn, net atrioventricular compliance; MS, mitral stenosis; PMV, percutaneous mitral valvuloplasty.

Predictors of composite adverse outcome

Post-PMV MR and transvalvular gradient were important predictors of MV replacement or repeat PMV (table 3). Specifically, MR severity following the procedure was strongly associated with the composite endpoint, as early surgery was needed in most patients with procedural failure.

Cumulative incidence of adverse events was higher in patients with Cn of ≤4 mL/mm Hg (figure 3). In Cox multivariable proportional hazard analysis (table 4), the independent predictors of composite endpoint were age, Cn and TR severity at baseline and postprocedural mean transvalvular gradient and MR degree, independent of baseline or postprocedural pulmonary artery pressure. Mean transmitral gradient and MR severity immediately after PMV predicted the composite endpoint but not cardiac death.

Figure 3

Cumulative incidence of adverse events in patients with Cn of ≤4 mL/mm Hg compared with Cn of >4 mL/mm Hg. Adverse events were defined as all-cause mortality, MV replacement or repeat PMV. Cn, net atrioventricular compliance; MV, mitral valve; PMV, percutaneous mitral valvuloplasty.

Five-year mortality risk prediction model

The effect of adding Cn to the risk prediction model for 5-year cardiovascular mortality was further assessed (table 5). The NRI of including Cn in the prediction model resulted in significant improvement in model performance (NRI 0.568; 95% CI 0.186 to 0.849; p=0.002), whereas the AUC increased slightly from 0.872 (95% CI 0.799 to 0.946) to 0.883 (95% CI 0.817 to 0.950) (p=0.469).

Table 5

Risk prediction model for 5-year cardiac mortality using a model without Cn and a model with Cn


The present study identifies several prognostic factors in patients with MS undergoing valve intervention. We specifically investigate the value of Cn and other baseline and postprocedural echocardiographic and haemodynamic parameters as predictors of cardiac death in patients with rheumatic MS undergoing PMV, mainly cardiac mortality. Our large cohort therefore is representative of a spectrum of more severe stages of disease. The results show that Cn strongly predicts cardiovascular death independently of age, NYHA functional class, TR severity and immediate procedural haemodynamic data and outcome. Additionally, the inclusion of Cn resulted in significant improvement in 5-year mortality risk prediction model. Cn therefore improves our ability to predict patients who will develop adverse outcomes, especially late cardiovascular death.

The clinical progression of MS usually spans a period of several decades, characterised by progressive narrowing of the mitral orifice, with subsequent haemodynamic deterioration.2 Although surgery or percutaneous MV intervention have changed the natural course of this disease, MS remains a progressive and fatal disease. Studies describing cardiac death rates after PMV have yielded varying results, with mortality rates ranging from 3.6% to 17%, mainly 9%–17%.6–8 27 Bouleti et al 7 examined 912 relatively young patients (49±13 years) who achieved procedural success between 1986 and 1995 and showed cardiac mortality rate to be 8.7% with a median follow-up of 10.7 years. In a recent study including only patients who were successfully treated by PMV between 1991 and 2010 found 9% of late cardiovascular death.8 Similarly, in the present study, including a more contemporary cohort of patients with mean age of 50±16 years from two different countries, the overall cardiac mortality rate was 10%, with an incidence of 4.1 deaths per 100 patient-years.

Previous studies of long-term event-free survival after PMV have mainly focused on the need for surgical or percutaneous MV reintervention as an endpoint.6–8 However, MV replacement may more frequently reflect a poor result of PMV rather than progression of MV disease.3 Indeed, postprocedural MR and mean transmitral pressure gradient, both related to suboptimal immediate results, were predictors of subsequent MV intervention, but not cardiac death. In contrast to previous studies, our study examined the predictors of cardiac death in MS.

Procedural success as currently defined predicts successful valve splitting, which should be differentiated from success of the overall therapy in terms of haemodynamic improvement and survival. A cut-off value of ≥1.8 cm2 has been reported to be better predictor of event-free survival rates, especially 5 years after PMV.28 In addition, PVR may not return to normal even after MV replacement.29 Therefore, procedural success in terms of valve area may often leave the patient with irreversible haemodynamic changes that culminate in right-sided heart failure and death.

Studies assessing the prognostic value of Cn

Although Cn plays an important role as a modulator of haemodynamic consequences of MS,10 13 16 17 there are limited data regarding its prognostic implications. A previous study13 with 26 patients evaluated the relation between Cn and clinical events and found that low left atrial Cn predicted the need for MV intervention. Subsequently, we demonstrated that Cn was an important predictor of MV intervention,18 adding incremental prognostic value to clinical data and MV area. More recently, Mahfouz et al 19 showed that Cn was significantly low in MS patients who had unfavourable outcome after successful PMV, but long-term mortality was not addressed in either of the prior studies.

Rationale for Cn predicting death in rheumatic MS

Cn contributes to the overall severity of MS in several ways that may allow baseline values to predict long-term mortality. First, low Cn can lead to severe and long-lasting pulmonary vascular changes, which are independent of the stenotic valve lesion and do not resolve after relief of MS. PVR remains persistently abnormal despite adequate relief of MV obstruction in a substantial subset of MS patients.30 It is therefore reasonable that low Cn may play a role or be a marker of persistent increased PVR that leads to right-sided heart failure and ultimately death. Second, it is also conceivable that low Cn may result in an exaggerated pulmonary hypertensive response to exercise with progressive RV remodelling that result in further aggravation of TR. Finally, low Cn may reflect atrial myocardial fibrosis and stiffening. Such LA remodelling is frequently seen in patients with MS and likely accounts for the increased risk of cardioembolic events, which are another common cause of death in MS. In addition, a myocardial factor can also lead to a restricted ventricle with further impairment Cn. While these factors are beyond the scope of this study, they provide potential mechanisms by which preprocedural Cn can predict long-term mortality.

Study limitations

Although the patients at enrollment were in the same stage of progression of MS with similar immediate results after MV intervention, some geographic variations in the characteristics of patients were observed. A stratified Cox model was performed assuming that baseline risk for cardiac death may differ between USA and Brazil but that risk factors act equally on USA and Brazilian patients, and the long-term mortality is related to the disease by itself. Nevertheless, the strength of this study was the inclusion of consecutive patients from twocentres enabling us to assess the natural history of MS in the current era of PMV.

Cn data were missing in a subset of patients, but multiple imputation techniques were used for unbiased estimates. Thirty-three patients (8%) were lost to follow-up. However, there are no differences in likelihood of loss to follow-up that are related to baseline patient characteristics and outcome.

Clinical implications

This study highlights that rheumatic MS is a disease that still carries significant mortality. The major cause of death is persistent right-sided heart failure (73% of the cases). Adequate relief of obstruction, while beneficial, is not always sufficient to reverse pulmonary circulation changes.31 After PMV, PH may persist that culminate in progressive RV failure and death. Low Cn may play a role or is a marker of pulmonary vascular disease in the context of MS.

In addition, Cn is a non-invasive objective parameter that can be easily calculated using values obtained from standard echocardiographic measurements, which adds prognostic value beyond symptoms and pulmonary pressure. In particular, Cn assessment may be used in a clinical setting to identify patients who are at risk of cardiovascular death after PMV.


Baseline low Cn is a strong predictor of both cardiovascular death and adverse outcome in patients with MS undergoing PMV, after adjustment for well-known markers of decreased survival in MS, specifically taking into account the baseline patient characteristics and postprocedural haemodynamic data. Cn assessment improves the risk prediction model for 5-year cardiac mortality after PMV.

Key questions

What is already known on this subject?

Net atrioventricular compliance (Cn) is an important determinant of pulmonary hypertension in patients with mitral stenosis.

What might this study add?

Cn is associated with long-term cardiovascular mortality in patients with mitral stenosis following percutaneous mitral valvuloplasty (HR of 0.70, 95% CI of 0.49 to 0.86), which is independent of the conventional clinical parameters including age, functional class and tricuspid regurgitation severity.

How might this impact on clinical practice?

Cn is an objective parameter that can be easily calculated using values obtained from standard echocardiographic measurements, which may identify patients who are at high risk of death. In addition, Cn assessment may be used in a clinical setting to improve the risk prediction model for 5-year cardiac mortality after percutaneous mitral valvuloplasty.

Supplementary Material

Supplementary Figure 1

Supplementary Material

Supplementary Figure 2



  • Contributors All authors have read and approved the manuscript. None of them have declared conflict of interest. Each author has contributed significantly to the work.

  • Funding This study was partly supported by grants from Coordenação de Aperfeiçoamento de Pessoal deNível Superior (Brasília, Brazil); by National Institutes of Health/National Heart, Lung, and Blood Institute grants R01 HL092101 (JH), K24 HL67434 and R01 HL109506 (RAL); and by the Leducq Transatlantic Mitral Network (Paris, France; RAL).

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval The study protocol was approved by the institutional review board (IRB) of Massachusetts General Hospital (MGH), USA, and Federal University of Minas Gerais, Brazil.

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