Dataset Information

Machine Learning Prediction of Response to Cardiac Resynchronization Therapy: Improvement Versus Current Guidelines.

ABSTRACT: BACKGROUND:Cardiac resynchronization therapy (CRT) has significant nonresponse rates. We assessed whether machine learning (ML) could predict CRT response beyond current guidelines. METHODS:We analyzed CRT patients from Cleveland Clinic and Johns Hopkins. A training cohort was created from all Johns Hopkins patients and an equal number of randomly sampled Cleveland Clinic patients. All remaining patients comprised the testing cohort. Response was defined as ?10% increase in left ventricular ejection fraction. ML models were developed to predict CRT response using different combinations of classification algorithms and clinical variable sets on the training cohort. The model with the highest area under the curve was evaluated on the testing cohort. Probability of response was used to predict survival free from a composite end point of death, heart transplant, or placement of left ventricular assist device. Predictions were compared with current guidelines. RESULTS:Nine hundred twenty-five patients were included. On the training cohort (n=470: 235, Johns Hopkins; 235, Cleveland Clinic), the best ML model was a naive Bayes classifier including 9 variables (QRS morphology, QRS duration, New York Heart Association classification, left ventricular ejection fraction and end-diastolic diameter, sex, ischemic cardiomyopathy, atrial fibrillation, and epicardial left ventricular lead). On the testing cohort (n=455, Cleveland Clinic), ML demonstrated better response prediction than guidelines (area under the curve, 0.70 versus 0.65; P=0.012) and greater discrimination of event-free survival (concordance index, 0.61 versus 0.56; P<0.001). The fourth quartile of the ML model had the greatest risk of reaching the composite end point, whereas the first quartile had the least (hazard ratio, 0.34; P<0.001). CONCLUSIONS:ML with 9 variables incrementally improved prediction of echocardiographic CRT response and survival beyond guidelines. Performance was not improved by incorporating more variables. The model offers potential for improved shared decision-making in CRT (online calculator: http://riskcalc.org:3838/CRTResponseScore ). Significant remaining limitations confirm the need to identify better variables to predict CRT response.

SUBMITTER: Feeny AK

PROVIDER: S-EPMC6588175 | biostudies-literature | 2019 Jul

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Machine Learning Prediction of Response to Cardiac Resynchronization Therapy: Improvement Versus Current Guidelines.

Feeny Albert K AK Rickard John J Patel Divyang D Toro Saleem S Trulock Kevin M KM Park Carolyn J CJ LaBarbera Michael A MA Varma Niraj N Niebauer Mark J MJ Sinha Sunil S Gorodeski Eiran Z EZ Grimm Richard A RA Ji Xinge X Barnard John J Madabhushi Anant A Spragg David D DD Chung Mina K MK

Circulation. Arrhythmia and electrophysiology 20190620 7

<h4>Background</h4>Cardiac resynchronization therapy (CRT) has significant nonresponse rates. We assessed whether machine learning (ML) could predict CRT response beyond current guidelines.<h4>Methods</h4>We analyzed CRT patients from Cleveland Clinic and Johns Hopkins. A training cohort was created from all Johns Hopkins patients and an equal number of randomly sampled Cleveland Clinic patients. All remaining patients comprised the testing cohort. Response was defined as ≥10% increase in left v ...[more]

PMID: 31216884

Similar Datasets

Project description:BackgroundCardiac resynchronization therapy (CRT) improves heart failure outcomes but has significant nonresponse rates, highlighting limitations in ECG selection criteria: QRS duration (QRSd) ≥150 ms and subjective labeling of left bundle branch block (LBBB). We explored unsupervised machine learning of ECG waveforms to identify CRT subgroups that may differentiate outcomes beyond QRSd and LBBB.MethodsWe retrospectively analyzed 946 CRT patients with conduction delay. Principal component analysis (PCA) dimensionality reduction obtained a 2-dimensional representation of preCRT 12-lead QRS waveforms. k-means clustering of the 2-dimensional PCA representation of 12-lead QRS waveforms identified 2 patient subgroups (QRS PCA groups). Vectorcardiographic QRS area was also calculated. We examined following 2 primary outcomes: (1) composite end point of death, left ventricular assist device, or heart transplant, and (2) degree of echocardiographic left ventricular ejection fraction (LVEF) change after CRT.ResultsCompared with QRS PCA Group 2 (n=425), Group 1 (n=521) had lower risk for reaching the composite end point (HR, 0.44 [95% CI, 0.38-0.53]; P<0.001) and experienced greater mean LVEF improvement (11.1±11.7% versus 4.8±9.7%; P<0.001), even among patients with LBBB with QRSd ≥150 ms (HR, 0.42 [95% CI, 0.30-0.57]; P<0.001; mean LVEF change 12.5±11.8% versus 7.3±8.1%; P=0.001). QRS area also stratified outcomes but had significant differences from QRS PCA groups. A stratification scheme combining QRS area and QRS PCA group identified patients with LBBB with similar outcomes to non-LBBB patients (HR, 1.32 [95% CI, 0.93-1.62]; difference in mean LVEF change: 0.8% [95% CI, -2.1% to 3.7%]). The stratification scheme also identified patients with LBBB with QRSd <150 ms with comparable outcomes to patients with LBBB with QRSd ≥150 ms (HR, 0.93 [95% CI, 0.67-1.29]; difference in mean LVEF change: -0.2% [95% CI, -2.7% to 3.0%]).ConclusionsUnsupervised machine learning of ECG waveforms identified CRT subgroups with relevance beyond LBBB and QRSd. This method may assist in objective classification of bundle branch block morphology in CRT.

Project description:AimsA waiting period of more than 3 months is recommended for patients before undergoing cardiac resynchronization therapy (CRT). However, due to an anticipated high mortality rate, early implementation of CRT might be beneficial for some patients. We aimed to evaluate the rate and the probability of left ventricular (LV) function improvement and their predictors in patients with heart failure (HF) with indications for CRT.Methods and resultsFrom March 2011 to February 2014, a total of 5625 hospitalized patients for acute HF were consecutively enrolled in 10 tertiary hospitals. Among them, we analysed 1792 patients (mean age 63.96 ± 15.42 years, female 63.1%) with left ventricular ejection fraction (LVEF) ≤ 35% at the baseline echocardiography and divided them into three groups: 144 with left bundle branch block (LBBB), 136 with wide QRS complexes without LBBB, and 1512 not having these findings (control). We compared and analysed these three groups for improvement of LV function at follow-up echocardiography. In patients who met CRT indications (patients with LBBB or wide QRS complexes without LBBB), logistic regression was performed to identify risk factors for no improvement of LV. No improvement of LV was defined as LVEF ≤ 35% at follow-up echocardiography or the composite adverse outcomes: death, heart transplantation, extracorporeal membrane oxygenation, or use of a ventricular assist device before follow-up echocardiography. A classification tree was established using the binary recursive partitioning method to predict the outcome of patients who met CRT indications. In a median follow-up of 11 months, LVEF improvement was observed in 24.3%, 15.4%, and 40.5% of patients with LBBB, wide QRS complexes without LBBB, and control, respectively. Patients meeting CRT indications had higher 3 month mortality rates than the control (24.6% vs. 17.7%, P = 0.002). Multivariable logistic regression analysis revealed that large LV end-systolic dimension [odds ratio (OR) 1.10, 95% confidence interval (CI) 1.05-1.15, P < 0.001], low LVEF (OR 0.92, 95% CI 0.87-0.98, P = 0.006), diabetes requiring insulin (OR 6.49, 95% CI 2.53-19.33, P < 0.001), and suboptimal medical therapy (OR 6.85, 95% CI 3.21-15.87, P < 0.001) were significant factors predictive of no improvement. A decision tree analysis was consistent with these results.ConclusionsPatients with CRT indications had higher mortality during their follow-up compared with control. LV function improvement was rare in this population, especially when they had some risk factors. These results suggest that the uniform waiting period before CRT implantation could be reconsidered and individualized.

Project description:Objectives: To test whether (1) electromechanical dyssynchrony induces region-specific alterations in the myocardial transcriptome and (2) dyssynchrony-induced gene expression changes can be corrected by cardiac resynchronization (CRT). Background: To date, CRT is the only heart failure treatment that can both acutely and chronically increase systolic function and prolong survival, something not yet achieved by a drug therapy. However, the mechanisms underlying the benefits of CRT remain elusive. Methods: Adult dogs underwent left bundle branch ablation (LBBB) and right atrial pacing at 200 bpm for either 6 weeks (dyssynchronous heart failure, DHF, n=12) or 3 weeks followed by 3 weeks of resynchronization by bi-ventricular pacing at the same pacing rate (CRT, n=10). Control animals without LBBB were not paced (NF, n=14). Echocardiography and invasive hemodynamic measurements were performed at 3 and 6 weeks. At 6 weeks, RNA was isolated from the anterior and lateral LV walls and hybridized onto canine-specific 44K microarrays. Results: In DHF, transcriptional changes consistent with re-expression of a fetal gene program were primarily observed in the anterior LV, resulting in increased regional heterogeneity of gene expression within the left ventricle. Dyssynchrony-induced region-specific expression changes in 1050 transcripts were reversed by CRT to levels of NF hearts (false discovery rate <5%). CRT remodeled transcripts with metabolic and cell signaling function and greatly reduced regional heterogeneity of gene expression compared with DHF. Conclusions: Our results demonstrate a profound effect of electromechanical dyssynchrony on the regional cardiac transcriptome, causing gene expression changes primarily in the anterior LV wall. CRT corrected the alterations in gene expression in the anterior wall by reversing the fetal gene expression pattern, supporting a global effect of biventricular pacing on the ventricular transcriptome that extends beyond the pacing site in the lateral wall.

Project description:AimsCardiac resynchronization therapy (CRT) may exert its beneficial haemodynamic effect by improving ventricular synchrony and improving atrioventricular (AV) timing. The aim of this study was to establish the relative importance of the mechanisms through which CRT improves cardiac function and explore the potential for additional improvements with improved ventricular resynchronization.Methods and resultsWe performed simulations using the CircAdapt haemodynamic model and performed haemodynamic measurements while adjusting AV delay, at low and high heart rates, in 87 patients with CRT devices. We assessed QRS duration, presence of fusion, and haemodynamic response. The simulations suggest that intrinsic PR interval and the magnitude of reduction in ventricular activation determine the relative importance of the mechanisms of benefit. For example, if PR interval is 201 ms and LV activation time is reduced by 25 ms (typical for current CRT methods), then AV delay optimization is responsible for 69% of overall improvement. Reducing LV activation time by an additional 25 ms produced an additional 2.6 mmHg increase in blood pressure (30% of effect size observed with current CRT). In the clinical population, ventricular fusion significantly shortened QRS duration (Δ-27 ± 23 ms, P < 0.001) and improved systolic blood pressure (mean 2.5 mmHg increase). Ventricular fusion was present in 69% of patients, yet in 40% of patients with fusion, shortening AV delay (to a delay where fusion was not present) produced the optimal haemodynamic response.ConclusionsImproving LV preloading by shortening AV delay is an important mechanism through which cardiac function is improved with CRT. There is substantial scope for further improvement if methods for delivering more efficient ventricular resynchronization can be developed.Clinical trial registrationOur clinical data were obtained from a subpopulation of the British Randomised Controlled Trial of AV and VV Optimisation (BRAVO), which is a registered clinical trial with unique identifier: NCT01258829, https://clinicaltrials.gov.

Dataset Information

Machine Learning Prediction of Response to Cardiac Resynchronization Therapy: Improvement Versus Current Guidelines.

Publications

Machine Learning Prediction of Response to Cardiac Resynchronization Therapy: Improvement Versus Current Guidelines.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets