Myocardial deformation and acute cellular rejection after heart transplantation: Impact of inter‐vendor variability in diagnostic effectiveness

A gentler approach to monitor for heart transplant rejection

Despite developments in circulating biomarker and imaging technology in the assessment of cardiovascular disease, the surveillance and diagnosis of heart transplant rejection has continued to rely on histopathologic interpretation of the endomyocardial biopsy. Increasing evidence shows the utility of molecular evaluations, such as donor-specific antibodies and donor-derived cell-free DNA, as well as advanced imaging techniques, such as cardiac magnetic resonance imaging, in the assessment of rejection, resulting in the elimination of many surveillance endomyocardial biopsies. As non-invasive technologies in heart transplant rejection continue to evolve and are incorporated into practice, they may supplant endomyocardial biopsy even when rejection is suspected, allowing for more precise and expeditious rejection therapy. This review describes the current and near-future states for the evaluation of heart transplant rejection, both in the settings of rejection surveillance and rejection diagnosis. As biomarkers of rejection continue to evolve, rejection risk prediction may allow for a more personalized approach to immunosuppression.

Heart Transplantation

Heart transplantation (HTx) remains the last therapeutic resort for patients with advanced heart failure. The present work is a clinically focused review discussing current issues in heart transplantation. Several factors have been associated with the outcome of HTx, such as ABO and HLA compatibility, graft size, ischemic time, age, infections, and the cause of death, as well as imaging and laboratory tests. In 2018, UNOS changed the organ allocation policy for HTx. The aim of this change was to prioritize patients with a more severe clinical condition resulting in a reduction in mortality of people on the waiting list. Advanced heart failure and resistant angina are among the main indications of HTx, whereas active infection, peripheral vascular disease, malignancies, and increased body mass index (BMI) are important contraindications. The main complications of HTx include graft rejection, graft angiopathy, primary graft failure, infection, neoplasms, and retransplantation. Recent advances in the field of HTx include the first two porcine-to-human xenotransplantations, the inclusion of hepatitis C donors, donation after circulatory death, novel monitoring for acute cellular rejection and antibody-mediated rejection, and advances in donor heart preservation and transportation. Lastly, novel immunosuppression therapies such as daratumumab, belatacept, IL 6 directed therapy, and IgG endopeptidase have shown promising results.

Prediction of heart transplant rejection from routine pathology slides with self-supervised deep learning

Aims

One of the most important complications of heart transplantation is organ rejection, which is diagnosed on endomyocardial biopsies by pathologists. Computer-based systems could assist in the diagnostic process and potentially improve reproducibility. Here, we evaluated the feasibility of using deep learning in predicting the degree of cellular rejection from pathology slides as defined by the International Society for Heart and Lung Transplantation (ISHLT) grading system.

Methods and results

We collected 1079 histopathology slides from 325 patients from three transplant centres in Germany. We trained an attention-based deep neural network to predict rejection in the primary cohort and evaluated its performance using cross-validation and by deploying it to three cohorts. For binary prediction (rejection yes/no), the mean area under the receiver operating curve (AUROC) was 0.849 in the cross-validated experiment and 0.734, 0.729, and 0.716 in external validation cohorts. For a prediction of the ISHLT grade (0R, 1R, 2/3R), AUROCs were 0.835, 0.633, and 0.905 in the cross-validated experiment and 0.764, 0.597, and 0.913; 0.631, 0.633, and 0.682; and 0.722, 0.601, and 0.805 in the validation cohorts, respectively. The predictions of the artificial intelligence model were interpretable by human experts and highlighted plausible morphological patterns.

Conclusion

We conclude that artificial intelligence can detect patterns of cellular transplant rejection in routine pathology, even when trained on small cohorts.

The Role of Left Atrial Longitudinal Strain in the Diagnosis of Acute Cellular Rejection in Heart Transplant Recipients

Our aim was to investigate the role of left atrial longitudinal strain (LALS) in the non-invasive diagnosis of acute cellular rejection (ACR) episodes in heart transplant (HTx) recipients. Methods: We performed successive echocardiographic exams in 18 consecutive adult HTx recipients in their first year after HTx within 3 h of the routine surveillance endomyocardial biopsies (EMB) in a single center. LALS parameters were analyzed with two different software. We investigated LALS association with ACR presence, as well as inter-vendor variability in comparable LALS values. Results: A total of 147 pairs of EMB and echo exams were carried out. Lower values of LALS were significantly associated with any grade of ACR presence. Peak atrial longitudinal strain (PALS) offered the best diagnostic value for any grade of ACR, with a C statistic of 0.77 using one software (95% CI 0.68−0.84, p < 0.0005) and 0.64 with the other (95% CI 0.54−0.73, p = 0.013) (p = 0.02 for comparison between both curves). Reproducibility between comparable LALS parameters was poor (intraclass correlation coefficients were 0.60 for PALS, 95% CI 0.42−0.73, p < 0.0005; and 0.42 for PALS rate, 95% CI −0.13−0.68, p < 0.0005). Conclusions: LALS variables might be a sensitive marker of ACR in HTx recipients, principally discriminating between those studies without rejection and those with any grade of ACR. Inter-vendor variability was significant.

Global longitudinal strain in heart transplantation recipients using different vendors: reliability and validity in a tertiary hospital in Colombia

Global Longitudinal Strain (GLS) is a useful tool to follow-up heart transplant (HT) recipients. Important inter-vendor variability of GLS measurements has been reported in healthy subjects and different conditions, but there is still limited evidence among HT patients. We assessed the reliability and validity of GLS using two vendors (General Electric and Philips) in a group of consecutive and stable adult HT recipients. Patients underwent two concurrent GLS analyses during their echocardiographic follow-up. We evaluated GLS inter-vendor reliability using Bland-Altman's limits of agreement (LOA) plots, computing its coverage probability (CP) and the intraclass correlation coefficient (ICC). Validity was assessed though receiver operating characteristics (ROC) curves, predictive values, sensitivity and specificity of GLS for each vendor to detect a normal left ventricle function. 78 pairs of GLS studies in 53 stable HT patients were analyzed. We observed a modest inter-vendor reliability with a broad LOA (less than 50% of values falling out our CP of 2% and an ICC of 0.49). ROC analyses (areas under the curve of 0.824 Vs. 0.631, p < 0.05) and diagnosis test indices (Sensitivity of 0.73 Vs. 0.64; and Specificity of 0.79 Vs. 0.50) favored GE over Philips. Inter-vendor variability for GLS analysis exceeded clinically acceptable limits in HT recipients. GLS from GE software seemed to show higher validity as compared to Philips'. The present study provides evidence to consider caution for the interpretation of GLS for clinical management in the follow-up of HT patients, especially when GLS is measured by different vendors.

Tissue Doppler velocities for ruling out rejection in heart transplant recipients in the context of myocardial strain imaging: a multivariate, prospective, single-center study

To investigate the value of tissue Doppler velocities for ruling out treatment-requiring acute cellular rejection (TR-ACR), in the context of myocardial deformation analysis performed by means of speckle tracking echocardiography. We performed serial echocardiograms in 37 heart transplant recipients in their first year post-transplantation within 3 h of the routine surveillance endomyocardial biopsies (EMB). The association of the sum of lateral mitral annulus systolic (s') and early diastolic (e') velocities, in absolute values, measured by tissue Doppler echocardiography (s'+ e'), with TR-ACR (ACR grade ≥ 2R) was investigated by multivariate analysis, including classic echocardiographic parameters and myocardial deformation variables. A total of 251 pairs of EMB and echo exams were performed, 35 (14%) with rejection grade ≥ 2R (TR-ACR). s' + e' was independently associated to TR-ACR (OR 0.80, 95%CI 0.72-0.89, p < 0.0005), with a C statistic of 0.79 (95%CI 0.71-0.87, p < 0.0005) by ROC curve analysis. An  s'+ e' value ≥ 23 cm/s, present in 43% of studies, had a negative predictive value of 98% for ruling out TR-ACR. Moreover, in the same patients, s'+ e' significantly decreased when TR-ACR occurred after a study without this condition (- 3.7 ± 3.3 cm/s, p = 0.003), but it was similar when rejection status was the same in the present versus the previous study. A drop in s'+ e' value < 2.7 cm/s from the previous echocardiogram, had a 99% negative predictive value for ruling out TR-ACR. Tissue Doppler velocities, a widely available echo parameter, were found to be a valuable marker for ruling out TR-ACR in this multivariate study which included myocardial deformation variables.