The aim of this study was to explore the adaptability of 3 contemporary surgical scores (Logistic EuroSCORE [LES], EuroSCORE II [ESII], and Society of Thoracic Surgeons Predicted Risk of Mortality [STS-PROM]) for prediction of mortality after percutaneous mitral valve repair with the MitraClip system. A total of 304 patients from the multicenter Getting Reduction of mitrAl inSufficiency by Percutaneous clip implantation in ITaly registry (GRASP-IT) were stratified based on LES, ESII, and STS-PROM tertiles and analyzed by different measurements of discrimination, calibration, and global accuracy with focus on 30-day and 1-, 2-, and 3-year mortality. A statistically significant gradient in the distribution of mortality was observed at all time points with ESII, at 2 years with LES, and at 2 and 3 years with STS-PROM. ESII had the best discrimination at 30 days (C-statistic 0.80), which remained acceptable at later follow-up, being significantly superior to that of LES at each time point (p = 0.003 at 30 days, p = 0.005 at 1 year, p = 0.011 at 2 years, and p = 0.029 at 3 years). Compared with STS-PROM, ESII showed better discrimination at 30 days (C-statistic 0.80 vs 0.62, p = 0.023). All scores overpredicted the risk of mortality at 30 days and were miscalibrated at 2 and 3 years. At 1 year, there was a good agreement between the observed and predicted probabilities for ESII and STS-PROM, whereas LES remained overpredictive. ESII showed the best global accuracy at 30 days and 1 year, whereas no notable differences were noted versus LES and STS-PROM at 2 and 3 years. In conclusion, lacking specific tools for risk stratification of patients undergoing MitraClip implantation, ESII holds favorable prognostic characteristics, which makes it a valid surrogate.
Historically, the most used scoring systems to assess the risk of death in patients undergoing cardiac surgery are the Logistic EuroSCORE (LES) and the Society of Thoracic Surgeons-Predicted Risk Of Mortality (STS-PROM) score. These models are based on logistic formulas that require the imputation of a variable number of parameters and result into a crude estimation of 30-day mortality. LES has good discrimination ability but, with modern surgery, it overestimates the observed mortality. STS-PROM is based on a higher number of parameters than LES, hence it provides a more accurate risk prediction. A new algorithm named EuroSCORE II (ESII) has been recently introduced to predict early mortality after cardiac surgery, showing better calibration than LES and good discrimination ability. Notably, all these scores have been developed in the context of conventional cardiac surgery and their role for risk prediction after MitraClip therapy, to the best of our knowledge, has never been systematically investigated. The aim of this study was to evaluate the performance of LES, ESII, and STS-PROM to predict all-cause mortality in terms of discrimination, calibration, and global accuracy at early-, mid-, and long-term follow-up in patients undergoing MitraClip implantation.
Methods
The Getting Reduction of mitrAl inSufficiency by Percutaneous clip implantation in ITaly (GRASP-IT) is a registry collecting data from 4 high-volume Italian centers that performed MitraClip implantation in consecutive patients with symptomatic severe mitral regurgitation from October 2008 to October 2013. The local ethics committee at each center approved the use of clinical data for this study, and all patients provided written informed consent. The authors wrote the manuscript and are responsible for the completeness and accuracy of data gathering and analysis.
Details of the MitraClip (Abbott Vascular, Santa Clara, CA) procedure have been previously described. Briefly, after transseptal puncture, the MitraClip Delivery System was introduced into a steerable guide catheter and the MitraClip device was advanced into the left atrium, rightly positioned over the mitral orifice and advanced into the left ventricle below the mitral leaflets. It was then retracted until both leaflets were grasped. This results in permanent leaflet coaptation and a double orifice valve, with a final effect that mimics the surgical Alfieri’s technique. If necessary, the device can be reopened, the leaflets released, and the MitraClip device repositioned. A second (or third) clip was placed at operator’s discretion if further reduction of mitral regurgitation (MR) was needed, with the goal in mind of achieving MR reduction to <3+. The procedure was performed under general anesthesia and using fluoroscopic and transesophageal 2- or 3-dimensional echocardiographic guidance.
Complete data were available for the calculation of LES, ESII, and STS-PROM in all patients included in the GRASP-IT registry. The 3 scores were calculated using the respective online calculators. The study population was stratified into 3 groups (low-, mid-, and high-risk) based on each score tertiles. The outcome of interest was all-cause mortality, which was assessed at different time points (i.e., 30 days, 1 year, 2 years, and 3 years). All baseline clinical and echocardiographic data were assessed for quality and entered into a dedicated computerized database. Clinical follow-up data were prospectively collected by scheduled clinical and echocardiographic evaluations or direct phone interviews. Referring cardiologists, general practitioners, and patients were contacted wherever necessary for further information.
Statistical analysis was performed with the SPSS software, version 18 (SPSS Inc, Chicago, Illinois). For all analyses a 2 sided p <0.05 was considered significant. Continuous variables are reported as mean ± SD and compared with the analysis of variance test. Categorical data are expressed as counts and percentages and compared with the chi-square test. Cumulative rates of all-cause mortality were estimated using the Kaplan-Meier method and compared by the log-rank test.
The performance of the 3 scores was valuated in terms of discrimination, calibration, and global accuracy, as previously described. Discrimination was measured using the area under the receiver operating characteristics curve, which ranges from 0.50 (no discrimination) to 1.0 (perfect discrimination). The receiver operating characteristics curves were compared with the Delong’s method. To assess discrimination, we also investigated the Somers’ D xy rank correlation between predicted probabilities and observed responses. When D xy = 0 the model is making random prediction, and when D xy = 1 the prediction is perfectly discriminating. Calibration was measured with the Hosmer-Lemeshow method and by generating calibration plots that visually compares the prediction with the observed probability. The perfect calibrated prediction stays on the 45-degree line (intercept = 0 and slope = 1), with a plot below reflecting overestimation and a plot above reflecting underestimation. The global accuracy was calculated using the Brier’s score, defined as quadratic difference between predicted probability and observed outcome for each patient. This is an overall performance measure that consists of positive values ranging from 0 (perfect prediction) to 1 (worst prediction), with lower scores indicating a greater accuracy.
Results
A total of 304 patients (mean age 72 ± 10 years, 64% men) were analyzed. Cardiac risk factors and co-morbidities were highly prevalent, with diabetes in 35%, coronary artery disease in 54%, peripheral artery disease in 11%, and atrial fibrillation in 41%. A functional or a degenerative cause for MR was described in 79% and 21% of patients, respectively. The mean left ventricular ejection fraction was 37 ± 14%. The mean values of LES, ESII, and STS-PROM were 17 ± 5%, 7 ± 8%, and 8 ± 7%, respectively. Cut-off values for tertile separation were <7.9%, 7.9% to 17.7%, and >17.7% for LES; <3.9%, 3.9% to 9.2%, and >9.2% for ESII; and <2.9%, 2.9% to 6.6%, and >6.6% for STS-PROM.
Baseline characteristics of the study population and procedural data stratified by tertiles of LES, ESII, and STS-PROM are summarized in Table 1 . Compared with the lowest tertiles, patients in the highest tertiles of each score were generally older and unsurprisingly sicker, in that they more frequently had diabetes, chronic kidney disease, previous myocardial infarction, and heart failure. Also, they more likely presented with functional ischemic MR. There were no significant differences in key echocardiographic and procedural characteristic across tertiles for each score, with the exception of left ventricular dimensions and ejection fraction across tertiles of STS-PROM and systolic pulmonary arterial pressure across tertiles of LES and ESII.
| Variable | LES | ESII | STS-PROM | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| I tertile | II tertile | III tertile | p value | I tertile | II tertile | III tertile | p value | I tertile | II tertile | III tertile | p value | |
| (N=101) | (N=102) | (N=101) | (N=101) | (N=102) | (N=101) | (N=101) | (N=102) | (N=101) | ||||
| Baseline characteristics | ||||||||||||
| Age, (years±SD) | 67±11 | 73±9 | 76±7 | <0.001 | 67±12 | 73±9 | 75±6 | <0.001 | 64±10 | 75±7 | 77±7 | <0.001 |
| Men | 64(63%) | 70(69%) | 60(59%) | 0.39 | 68(67%) | 63(62%) | 63(62%) | 0.665 | 73(72%) | 60(59%) | 61(60%) | 0.093 |
| Diabetes mellitus | 22(22%) | 37(36%) | 46(45%) | 0.002 | 20(20%) | 31(30%) | 54(53%) | <0.001 | 24(24%) | 34(33%) | 47(46%) | 0.003 |
| Chronic kidney disease | 39(39%) | 51(50%) | 64(63%) | 0.002 | 27(27%) | 56(55%) | 71(70%) | <0.001 | 33(33%) | 50(49%) | 71(70%) | <0.001 |
| History of myocardial infarction | 25(25%) | 36(35%) | 50(49%) | 0.001 | 20(20%) | 32(31%) | 59(58%) | <0.001 | 30(30%) | 35(34%) | 46(45%) | 0.055 |
| History of heart failure | 52(51%) | 67(66%) | 80(79%) | <0.001 | 59(58%) | 61(60%) | 79(78%) | 0.003 | 61(60%) | 58(57%) | 80(79%) | 0.002 |
| History of valve surgery | 2(2%) | 9(9%) | 20(20%) | <0.001 | 0(0%) | 7(7%) | 24(24%) | <0.001 | 6(6%) | 10(10%) | 15(15%) | 0.111 |
| Chronic obstructive pulmonary disease | 17(17%) | 27(26%) | 22(22%) | 0.265 | 20(20%) | 21(21%) | 25(25%) | 0.618 | 16(16%) | 23(23%) | 27(27%) | 0.162 |
| Mitral regurgitation etiology | ||||||||||||
| Functional ischemic | 26(26%) | 48(47%) | 62(61%) | <0.001 | 25(25%) | 41(40%) | 70(69%) | <0.001 | 36(36%) | 43(42%) | 57(56%) | 0.01 |
| Functional non ischemic | 46(45%) | 32(31%) | 26(26%) | 0.009 | 46(45%) | 37(36%) | 21(21%) | 0.001 | 16(16%) | 28(27%) | 26(26%) | <0.001 |
| Non functional | 30(30%) | 22(22%) | 13(13%) | 0.016 | 31(31%) | 24(23%) | 10(10%) | 0.001 | 16(16%) | 31(30%) | 18(18%) | 0.025 |
| Echocardiographic data | ||||||||||||
| Mitral regurgitation grade 4+ | 66(65%) | 74(72%) | 72(71) | 0.462 | 66(65%) | 76(74%) | 70(69%) | 0.363 | 73(72%) | 68(67%) | 71(70%) | 0.742 |
| Left ventricular end-systolic diameter (mm±SD) | 62±12 | 63±12 | 63±11 | 0.829 | 62±13 | 64±12 | 62±10 | 0.565 | 66±12 | 60±11 | 62±10 | 0.001 |
| Left ventricular ejection fraction (%±SD) | 38±15 | 38±14 | 35±13 | 0.404 | 39±15 | 37±15 | 35±12 | 0.267 | 34±14 | 40±14 | 37±13 | 0.002 |
| Systolic pulmonary pressure (mmHg±SD) | 44±13 | 49±12 | 52±13 | <0.001 | 44±12 | 51±14 | 51±11 | <0.001 | 48±16 | 47±11 | 51±11 | 0.119 |
| Procedural data | ||||||||||||
| One clip implantation | 46(45%) | 31(30%) | 37(37%) | 0.09 | 36(36%) | 40(39%) | 38(38%) | 0.845 | 42(42%) | 34(33%) | 38(38%) | 0.446 |
| Device time (minutes±SD) | 83±49 | 80±45 | 77±40 | 0.592 | 80±47 | 82±42 | 79±45 | 0.893 | 84±46 | 77±44 | 80±43 | 0.545 |
| Procedural time (minutes±SD) | 156±79 | 159±61 | 152±61 | 0.802 | 164±77 | 154±59 | 150±64 | 0.365 | 162±76 | 157±65 | 149±60 | 0.453 |
At a mean follow-up of 447 ± 383 days, the cumulative incidence of all-cause mortality was 20.4%, with 10 deaths at 30 days, 34 at 1 year, 53 at 2 years, and 59 at 3 years. The rates of all-cause death stratified by tertiles of LES, ESII, and STS-PROM scores at 30-day and 1-, 2-, and 3-year follow-up are reported in Table 2 , with corresponding Kaplan-Meier curves depicted in Figure 1 . A statistically significant gradient in the distribution of all-cause mortality was observed at all time points with ESII, at 2 and 3 years with STS-PROM, and at 2 years with LES. All risk scores significantly stratified the 3-year incidence of mortality when comparing the lowest and highest tertiles only.
| I tertile | II tertile | III tertile | I vs II vs III tertile p value | I vs II tertile p value | II vs III tertile p value | I vs III tertile p value | |
|---|---|---|---|---|---|---|---|
| LES | |||||||
| 30 days | 0% | 5% | 5% | 0.079 | 0.024 | 0.995 | 0.024 |
| 1 year | 11% | 13% | 18% | 0.278 | 0.053 | 0.362 | 0.116 |
| 2 years | 16% | 33% | 39% | 0.041 | 0.123 | 0.293 | 0.011 |
| 3 years | 25% | 46% | 44% | 0.075 | 0.123 | 0.434 | 0.022 |
| ESII | |||||||
| 30 days | 0% | 2% | 8% | 0.005 | 0.160 | 0.053 | 0.004 |
| 1 year | 8% | 14% | 20% | 0.02 | 0.291 | 0.093 | 0.008 |
| 2 years | 16% | 24% | 52% | <0.001 | 0.214 | 0.006 | <0.001 |
| 3 years | 23% | 40% | 57% | 0.001 | 0.168 | 0.015 | <0.001 |
| STS PROM | |||||||
| 30 days | 1% | 5% | 4% | 0.290 | 0.111 | 0.717 | 0.197 |
| 1 year | 9% | 13% | 18% | 0.226 | 0.410 | 0.374 | 0.089 |
| 2 years | 12% | 37% | 37% | 0.027 | 0.024 | 0.697 | 0.008 |
| 3 years | 22% | 37% | 51% | 0.034 | 0.070 | 0.387 | 0.010 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
