Article Text
PDF
Abstract
Background Echocardiographic grading of secondary mitral regurgitation (SMR) severity is challenging and involves multiple guideline-recommended parameters. We previously introduced the average pixel intensity (API) method for grading SMR. In this study, the clinical outcome in SMR based on the API method for grading MR was compared with conventional grading methods.
Methods 231 patients with systolic heart failure and reduced ejection fraction (ischaemic/non-ischaemic) and SMR were prospectively enrolled. MR was graded using all guideline-recommended parameters and the API method, which is based on the pixel intensity of the continuous wave Doppler signal. The primary outcome was MACE (major adverse cardiac event).
Results The API method was applicable in 98% of patients with SMR (n=227). During a median follow-up of 24 months, 98 patients (43%) had a MACE (cardiovascular mortality (n=50, 22%), heart failure hospitalisation (n=44, 19%), mitral valve surgery (n=11, 5%), percutaneous mitral intervention (n=12, 5%), heart transplantation (n=5, 2%)). On log-rank test, the API method was highly significant in predicting clinical outcome. On multivariable Cox proportional hazard analysis, SMR grading with the API method was an independent predictor of clinical outcome (along with NYHA class and right ventricular systolic pressure; p<0.001), increasing the event risk by 9% per 10 au API rise (p=0.001). In the same multivariable analysis, proximal isovelocity surface area (PISA)-effective regurgitant orifice area or PISA-regurgitant volume were not independent predictors of events (p=0.18 and 0.26, respectively).
Conclusion SMR grading with the API method is an independent predictor of clinical outcome and provides prognostic information in addition to clinical and other echocardiographic variables.
- mitral regurgitation
Statistics from Altmetric.com
Introduction
Secondary mitral regurgitation (SMR) occurs frequently in systolic left ventricular (LV) dysfunction and heart failure (HF).1 In the diseased LV, reduced closing forces and geometric alterations result in restrictive mitral leaflet motion and subsequent systolic regurgitation of blood into the left atrium (LA). The extent of SMR is considered a causative marker of clinical outcome,1–3 but selecting the patients that may benefit from valve intervention remains enigmatic.4
Defining severe SMR in patients with HF is challenging due to (1) the lack of a gold standard method for quantifying MR, (2) the complexity of grading SMR severity with echocardiography, (3) the difficulty in estimating LV contractility (reserve) of the diseased myocardium and (4) the dynamic and evolving nature of SMR and LV disease; (4) also, the clinical context or comorbidity of patients with HF must be taken into account when considering ‘severe SMR’; (5) finally, severe SMR must also be interpreted with respect to the specific end-point or outcome parameter(s) chosen in scientific reports.
Echocardiographic grading of SMR is recommended by integrating multiple parameters, including qualitative (eg, colour Doppler), semiquantitative (vena contracta width (VCW)) and quantitative (eg, effective regurgitant orifice area (EROA) and regurgitant volume (RV)) parameters.5 6 However, echocardiographic grading of SMR is time demanding and requires dedication and experience. Also, methodological issues on MR grading methods were claimed and cardiologists may therefore not consider quantitative assessment of SMR.1 7 Yet, grading of SMR severity in patients with HF is important for risk stratification in these patients.
We recently introduced the average pixel intensity (API) method for grading MR severity based on the measurement of the pixel intensity of the continuous wave (CW) Doppler signal of the MR flow.8 The intensity of the CW Doppler signal is based on the principle that the signal intensity of the CW envelope is proportional to the insonated blood flow and hence to the blood volume of moving scatterers (red blood cells) that pass through the mitral regurgitant orifice (ie, the regurgitant volume).9 10 The API method integrates the time-varying flow during MR in a single measure and avoids geometric assumptions or complex calculations needed for other echocardiographic MR grading parameters.5 11 Previously, we validated the API method in vitro8 and its predictive value was shown in primary MR.12 In the present study, we assess the predictive value of the API method in patients with SMR with chronic HF and we position the API method within the guideline recommended multi-parametric grading of SMR. This may be relevant for assisting in risk stratification in HF and selection of patients that may benefit from mitral valve intervention.
Methods
Patient selection and involvement
Transthoracic echocardiography was performed at Ghent University Hospital in 231 consecutive patients with HF with ejection fraction (EF) <50% (reduced/mid-range EF13). Patients had chronic stable HF and received guideline-directed medical treatment. SMR was carefully identified as tenting and restrictive motion of the mitral valve leaflet(s) due to ventricular myocardial dysfunction/remodelling. Patients with degenerative mitral valves were excluded. Patients were prospectively included between 2014 and 2018. The study was approved by the local Ethics Committee; informed consents were provided. Patients were not involved in the study design.
Echocardiographic assessment of MR
For the present study, the VIVID9 XDclear echocardiography system with M5Sc-D probe (General Electric, Waukesha, Wisconsin, USA) was used for all patients. Echocardiographic acquisitions were performed by a single operator, blinded to the patient’s clinical status or medical records. In brief, the CW Doppler envelope is acquired by carefully aligning the cursor through the vena contracta area, coaxial with the MR jet.12 The CW envelopes are then manually traced off-line, starting from the mitral closure signal to the end of the CW envelope. The traced areas are then converted into greyscale and the API value is calculated with custom-made software (figure 1).8 12 All API analyses were performed by one operator blinded to the clinical and echocardiographic characteristics of the patients. Other quantitative MR measures such as VCW, proximal isovelocity surface area (PISA)-EROA and PISA-RV were carefully assessed according to consensus recommendations.5 6
Acquisition and calculation of the API. Panel A: for the API acquisition, a continuous wave Doppler is performed over the mitral valve, well aligned with the mitral regurgitation jet. Panel B: The uncompressed image is stored and converted to greyscale. The operator manually traces the area of interest, that is, the continuous wave Doppler envelope. Panel C: using specific software, the API value is calculated and pixel intensities over time are displayed. API, average pixel intensity.
The theoretical considerations and practical aspects of the API method have been extensively described previously.12 Briefly, the machine settings that may affect the API values such as CW Doppler output power, transmission frequency of the CW Doppler system and compression were fixed as default settings and used in every patient included in the study. The gain level used for all patients (6 dB)12 was considered optimal because at this level, oversaturation or undersaturation of the CW Doppler signal intensity was avoided irrespective of MR severity, that is, keeping the greyscale intensity histograms always above 0 arbitrary units (au, unit of API) (the absolute black) and always below 255 au (the absolute white).12
Assessment of clinical outcome
Clinical outcome was assessed using patient records. Major adverse cardiac events (MACE) were prespecified and included cardiovascular mortality, hospitalisation for decompensated HF, mitral valve surgery or percutaneous intervention. The treating physician and surgical heart valve team that made the decision for mitral valve intervention were blinded to the patients’ API values.
An extensive statistical background is provided in the.
Results
Baseline characteristics
The API method could be applied in 98% (227/231 patients), which was significantly higher than other echocardiographic parameters such as PISA-based methods (75%, 173/231) and VCW (84%, 194/231) (p<0.001). In four patients (2%), API acquisition was not possible due to extreme eccentricity of the MR jet.
Table 1 shows the clinical and echocardiographic characteristics of the patient population, stratified according to occurrence of events (MACE). Patients in both groups had similar baseline characteristics such as age, sex and blood pressure. The presence of atrial fibrillation (AF) was similar in both groups. However, NYHA class was more severe in the MACE group, and a mild but statistically significant difference in EF (33 vs 36%) was observed. MACE patients had a more dilated LA and LV. Comorbidities such diabetes mellitus, pulmonary disease or renal failure were comparable; there was no difference in HF aetiology.
Clinical characteristics
As expected, MR was more severe in the MACE group and this was consistent for all grading methods, including the API method. Median PISA-EROA value was 0.22 cm² in the MACE group. Right ventricular systolic pressure (RVSP) was significantly higher in the MACE group compared with non-MACE group.
The API method had significant correlations with other MR grading methods (see).
MACE-free survival according to API method
During a median follow-up period of 24 months, 98 patients (43%) experienced at least one MACE: cardiovascular mortality (n=50, 22%), HF hospitalisation (n=45, 20%), mitral valve surgery (n=11, 5%), percutaneous mitral intervention (n=12, 5%) and heart transplantation (n=5, 2%). As 14 (6%) non-cardiac deaths occurred, overall mortality was 28% during follow-up.
Kaplan-Meier graphs (figure 2) show the MACE-free survival (first event) during the follow-up period after stratifying API values into tertiles (tertile 1:<97 au, tertile 2: 97–138 au, tertile 3:>138 au). As can be appreciated, the higher the API value/category, the higher the risk for clinical events (OR tertile 2 vs tertile 1: 3.52 (1.69–7.368); tertile 3 vs tertile 2: 2.24 (1.17–4.30); tertile 3 vs tertile 1: 9.83 (4.99–19.38); all p<0.001).
Kaplan Meier graphs show the overall MACE-free survival over time (time to first event), after stratification of API values into tertiles. The higher the API tertile, the lower the MACE-free survival (p for all<0.001). API, average pixel intensity; MACE, major adverse cardiac event.
API as an independent predictor of MACE in patients with SMR
Cox regression was performed to determine hazard ratios in univariate and multivariate model containing the following variables: MR grade expressed as API, PISA-EROA or PISA-RV (per 10 au, 0.10 cm² and 10 mL increase, respectively), age, LA volume indexed by body surface area, aetiology of cardiomyopathy (ischaemic vs non-ischaemic), presence of diabetes mellitus, AF, LV end-diastolic volume (LVEDV), NYHA class, RVSP (per 10 mm Hg increase) and EF (per 10% decrease). In univariate analysis, ventricular and atrial dimensions are predictors of MACE as well as EF, NYHA class and RVSP. For the MR grading parameters, API, PISA-EROA, PISA-RV and VCW are univariate predictors of MACE (hazard ratio (HR) 1.18 per 10 au API increase; HR 1.55 per 0.10 cm² PISA-EROA increase; HR 1.38 per 10 mL PISA-RV increase; HR 1.51 per 1 mm VCW increase). In the multivariate model, only NYHA class (HR 1.64 per class step-up), RVSP (HR 1.39 per 10 mm Hg increase), API (HR 1.095 per 10 au increase) are independent predictors of MACE. When using the same multivariate model and replacing API by PISA-EROA, PISA-RV or VCW, neither PISA-EROA nor PISA-RV were predictors of MACE, whereas VCW was an independent predictor (HR 1.32 per 1 mm increase) in this model (table 2). Correcting API or PISA-EROA for ventricular volumes1 14 did not affect the predictive value. Importantly, when considering an additional analysis that includes only patients in whom all quantitative SMR measurements could be performed (API, VCW and PISA; n=158), the same findings remain, with API being a strong and independent predictor of outcome (p=0.007), whereas PISA-based measures were not (online supplementary table 1A). In another model, using a combined endpoint of cardiovascular death and/or HF hospitalisation only (thus excluding mitral valve intervention as an outcome), NYHA, RVSP and API remained significant independent predictors, whereas other MR grading methods did not (online supplementary table 1B).
Supplemental material
Cox proportional hazards
Severity cut-off for API in SMR
Based on the outcome data and unadjusted receiver operating curves (ROC), an API value of 119 au was considered the optimal severity cut-off with a sensitivity of 74% and a specificity of 62%. All echocardiographic parameters (API, PISA-EROA, PISA-RV and VCW) have similar areas under the curve, which is around 0.7, indicating similar predictive value (online supplementary table 2). The current EROA and RV cut-offs in AHA/ACC guidelines of 0.4 cm² and 60 mL15 provide a very high specificity (both 99%), but at the expense of a very low sensitivity (9% and 3%, respectively). Considering the European Society of Cardiology (ESC) cut-offs,16 the sensitivity and specificity are more balanced. When using the Youden Index approach to calculate the optimal cut-off for PISA-EROA, PISA-RV and VCW, values of 0.18 cm², 30 mL and 5.3 mm are obtained.
In the cohort where PISA-EROA did not predict events (false negative cases—FN), the ratio of API/EROA was much higher than in the correctly predicted events (true positive cases—TP) (API/EROA 101 au/0.1 cm² (FN) vs API/EROA 54 au/0.1 cm² (TP, p<0.001) for ESC cut-off and API/EROA 79 au/0.1 cm² (FN) vs API/EROA 34 au/0.1 cm² (TP, p<0.001)) for AHA/ACC cut-off, respectively), suggesting underestimation of PISA-EROA.
Discussion
In this prospective outcome study, we show that grading SMR with the API method in patients with reduced EF identifies patients with HF at risk for relevant clinical events. Therefore, the API method may be a valuable tool to assist in the risk stratification of patients with HF in clinical practice.
Echocardiographic grading of SMR and prediction of clinical outcome in heart failure
Given the strengths and weaknesses of echocardiographic measures,1 guidelines recommend a multiparametric assessment for grading MR severity, emphasising the importance of quantitative markers such as EROA and RV.5 15 16 Theoretically, the EROA and RV concepts are appealing and rely on simple physical principles. However, reports have highlighted restrictions when considering the PISA method for assessing EROA and RV in clinical practice,1 5 17 despite its clinical validation.18 19
In the current study, API-based assessment of SMR severity is an independent predictor of MACE in patients with HF, whereas PISA-EROA and PISA-RV are not predictive in our analysis. This seems contradictory to previous SMR reports that showed independent predictive power of EROA and RV in HF.19–21 Whether a parameter may be an independent predictor or not depends on its accuracy, on its reproducibility, on the HF population studied, on the variables included in the multivariate analysis and on the specific study-endpoints. In patients with SMR, cardiovascular mortality and HF hospitalisation are determined by a complex interplay of these haemodynamic, clinical and echocardiographic variables. In previous reports on SMR outcome, EF has been included in the multivariate models to show the predictive power of EROA and RV independent of EF in patients with systolic HF.1 RVSP and NYHA class, however, have mostly not been included in multivariate models, even though these parameters seem obvious predictors of outcome in HF.19–22 When excluding NYHA and RVSP in our multivariate model, PISA-EROA and PISA-RV do have independent predictive power (HR 1.34; p=0.003), but this prognostic value does not persist when including NYHA class (but not RVSP). In contrast, the API-based SMR grading provides additional prognostic value, independent of NYHA class and RVSP.
In SMR, the altered geometry of mitral valve coaptation and the elliptical regurgitant orifice area may cause flattening of the flow convergence zone and underestimation of PISA-EROA calculation.1 23–25 Our data on the API/EROA ratio support these observations: in the cohort of patients where PISA-based measures did not predict MACE, significantly higher API/EROA ratios were observed compared with the cohort with correctly predicted events, which is predominantly attributed to PISA-underestimation of EROA. Therefore, API may unmask the PISA-based underestimation of SMR severity in these patients and may therefore be a valuable contribution to the grading of SMR ().
Severity cut-offs of SMR in patients with heart failure: a matter of endpoints?
For reasons mentioned in the introduction, defining ‘severe SMR’ in HF is challenging. With respect to echocardiographic grading of SMR, there is divergence between the ESC and AHA/ACC guidelines regarding the EROA and RV severity cut-offs: the ESC guidelines propose EROA and RV cut-offs of 0.2 cm² and 30 mL, respectively, whereas the AHA/ACC guidelines propose an EROA and RV severity cut-off of 0.4 cm² and 60 mL.15 16 Previous21 and recent SMR outcome studies15 26 on which the (updated) SMR guidelines are based, have used averaged EROA and RV values obtained from one to three different echocardiographic methods. Averaging EROA and RV values obtained from different methods is time consuming and requires multiple measures and geometric assumptions and is therefore rarely used in routine clinical practice.
When extrapolating our proposed API severity cut-off of 119 au in SMR to determine an equivalent EROA cut-off, a value of 0.18 cm² is obtained, which is in accordance with the current ESC severity cut-off of 0.2 cm².16 In contrast, when using the current AHA/ACC severity cut-offs of EROA and RV (0.4 cm² and 60 mL, respectively) in the present cohort, these cut-offs have high specificity but very low sensitivity in predicting events. Also, patients with SMR with EROA values>0.4 cm² and RV values>60 mL are rare1 11 19 in our consecutively recruited cohort (<5 %) and probably reflect very severe SMR. It is important to note that the recent AHA/ACC guidelines have revised the EROA cut-off from 0.2 to 0.4 cm² based on results from surgical intervention studies,15 26 not on clinical outcome in medically treated patients. This is relevant as many patients with HF with SMR having an EROA>0.2 cm² are eventful,21 but do not benefit from surgical intervention.15 26 As such, the upgraded AHA/ACC severity cut-off for EROA provides a limit to avoid unnecessary surgical mitral valve intervention.15 Considering the recent percutaneous mitral valve intervention studies in SMR,3 27 the definition of SMR severity appears to become more complicated, as it seems that only patients with disproportionate SMR (ie, SMR grade 3–4, but relatively smaller ventricles) may benefit from percutaneous mitral intervention.14 The term disproportionate SMR reflects a specific regurgitant fraction (RF) cut-off,14 but it is suggested that disproportionate MR may not explain all the differences between MITRA-FR and COAPT results.3 4 27 Also, a recent echocardiographic subanalysis of COAPT did not observe differences in outcome when considering a dichotomised RF of 36%,28 but this may be related to selection and limitations of echocardiographic assessments.29 In our hands, although being a crude approach for assessing disproportionate SMR, correcting API for LVEDV or LVESV did predict MACE, but no better than API as a single parameter. The API method could also be applied for selecting patients with HF that may benefit from mitral intervention. However, we currently cannot define a ‘corresponding’ API value for patients that would benefit from surgical or percutaneous mitral intervention, as the current study is not an interventional outcome study. Interestingly, when considering all the inclusion and exclusion criteria in COAPT, less than 5% of the our cohort may be eligible for MitraClip intervention.
The SMR ‘cut-off controversy’, however, does reaffirm the need for accurate and reproducible grading of MR.
Considerations on the use of API and potential limitations of the study
The API method may be valuable to grade SMR severity, complementary to the multiparametric MR grading approach recommended by international associations or guidelines. Also, because of its digital format, the API approach for grading MR severity may be considered by the expanding field of artificial intelligence in echocardiography.30
This study is monocentre, and a multicentre study is ongoing for external validation and scalability testing of the API method in patients with HF. Despite a substantial number of MACE events, this study was not powered for the individual endpoint analysis in subgroups. The general limitations of the API method have been extensively described previously.12
Key questions
What is already known on this subject?
Secondary mitral regurgitation (SMR) is a frequent consequence of left ventricular dysfunction. However, grading of SMR with echocardiography is challenging and requires dedication and expertise, and, therefore, cardiologists may not consider quantitative grading of SMR.
What might this study add?
In a cohort of 231 patients with heart failure with SMR and a median follow-up time of 24 months, the average pixel intensity (API) method was predictive of clinical events and outcome. The applicability of the API method was 98%, which was significantly higher than other MR grading parameters. On multivariable Cox proportional hazard analysis, SMR grading with the API method was an independent predictor of clinical outcome, increasing the event risk by 9% per 10 au API rise.
How might this impact on clinical practice?
The API method may be a valuable contribution to the multiparametric approach for grading SMR, which is important for risk stratification and decision making in SMR. Also, because of its simple digital format, it lends itself easily for future artificial intelligence algorithms.
References
Footnotes
Contributors VK: data collection, data analysis and manuscript writing. FT: data collection, concept, manuscript writing and editing. MEH: manuscript editing and software developer. MDB: statistical input and manuscript editing. TDB: data collection and manuscript editing.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Patient consent for publication Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement Data are available on reasonable request. NA.