Are biventricular and biatrial function truly preserved after arterial switch operation?

Abstract

Arterial switch operation is currently the standard procedure when correcting for transposition of the great arteries (TGA) with a good longevity of the intervention. While aortic and coronary artery complications as well as mortality are low, rightsided lesions such as right ventricular (RV) outflow and neopulmonary valve dysfunction may occur and require reintervention. Patients undergoing cardiac reinterventions in childhood are more likely to have residual lesions and have a higher risk of valvular and ventricular dysfunction as well as arrhythmia. In TGA, left ventricular (LV) function is reported normal at longterm followup, 5 but RV dysfunction is present at midterm after arterial switch operation. Systemic ventricular function has been suggested as a prognostic marker of outcome in TGA after atrial switch and in congenital corrected TGA. Even if echocardiography is the firstline imaging modality for cardiovascular evaluation of patients with congenital heart disease, cardiac magnetic resonance (CMR) is considered gold standard for ventricular and atrial volumes and function owing to high accuracy, precision and reproducibility. As most studies of patients with TGA and arterial switch operation are small cohorts presenting either on a ventricular or atrial functional results, whole heart assessment including all four cardiac chambers in a comprehensive manner has been lacking. A larger population of patients with TGA and arterial switch operation also allows a meaningful investigation of whether any ventricular and atrial functional alterations could be caused by differences in simple and complex TGA morphology or by usual and unusual coronary artery anatomy. The study by Schuwerk et al addresses these issues and assesses biventricular and biatrial function in a large (n=103) cohort of patients with TGA after arterial switch operation using CMR in comparison with healthy controls (n=77). The authors found several atrial and ventricular measures of function decreased in patients with TGA in the univariate analysis when compared with controls, but when using a statistical machine learning and logistic regression approach, only RV longitudinal and left atrial (LA) function were impaired at median 16 years after arterial switch operation. Biventricular and biatrial function and deformation were not substantially different between the subgroups (simple vs complex TGA and usual vs unusual coronary anatomy). The use of a statistical machine learning approach with multiple logistic regression to identify functional parameters that are different between patients and controls is novel but could possibly be problematic. To have valid results from a logistic regression, colinearity between parameters should be avoided if possible and a dichotomous or categorical outcome variable is needed. In this study, there are apparent multiple colinearities between the functional parameters. Most evidently is ejection fraction (EF) that is derived from enddiastolic and endsystolic volume, and that mitral annular plane systolic excursion (MAPSE) is part of the long axis strain calculations. Moreover, ventricular strain and MAPSE are expected to correlate to EF. When it comes to the outcome variable in this study, it is to distinguish a patient with TGA from a healthy person. The findings that only tricuspid annular plane systolic excursion (TAPSE) and LA strain are different between the groups are surprising and one could wonder if this is explained by the statistical method using many colinear parameters that would dilute the effect of each parameter. As such, the highly significant difference in LVEF between patients and controls (p<0.0001) is lost when so many parameters of LV function are included in the logistic regression analysis. How the more commonly used and simpler approach of a ttest with Bonferroni correction for multiple comparisons would have turned out, would be interesting to know. Possibly, the conclusions that only TAPSE and LA longitudinal function are impaired and that there is a largely preserved biventricular and biatrial function after arterial switch operation compared with controls would differ. Describing the results in more detail, both ventricles had lower function in patients than controls (LVEF 53% vs 59.5%, MAPSE 9.9 mm vs 13.7 mm, LV global longitudinal strain (GLS) −20.7% vs −23.7%, RVEF 52.4% vs 55%, and TAPSE 10.2 mm vs 20 mm, however with no difference in RVGLS). Hereby, the LV measures are supported by each other, while the RVGLS stands in stark contrast to the mildly reduced RVEF and markedly reduced TAPSE. To exclude apical motion, the authors chose an extracardiac reference point for the TAPSE measurements. If apical rocking is present and the LV is ‘pulled’ by a wellfunctional RV in a translational matter, an extracardiac reference point could induce a low measure of TAPSE while RVGLS is just below normal. Another intrinsic methodological consideration one should keep in mind when using RVGLS is whether to include the septal region to the assessment or not. RV free wall strain is not as interventricular dependent as RVGLS that includes the septum. As noted by the authors, altered TAPSE is found postoperatively 10 and surgery might alter the atrioventricular (AV) coupling caused by interference of the pericardial integrity. It would be interesting to know if this can explain the low TAPSE 16 years after surgery. The authors suggest that the altered atrial function may be owed to diastolic ventricular dysfunction, and this is a compelling explanation. The atrial pump physiology is however also highly dependent on systolic ventricular pumping and mechanistically assessing atrial function without addressing the AV dependence is truncating the pump physiological explanations to such findings. When the ventricular myocardial contraction is pulling the AV plane towards the apex, the atrium is lengthened. Likewise, when the ventricle relaxes, the atrium is shortened in length. A reduced atrial strain could therefore be expected depending on decreased ventricular pulling and relaxation of the AV plane. One can therefore wonder how much of a reduced atrial strain is truly an impaired atrial function or if this could be a result Department of Clinical Sciences Lund, Clinical Physiology, Lund University and Skåne University Hospital, Lund, Sweden Laboratory of Clinical Physiology, National Heart Lung and Blood Institute, National Institutes of Health, Bethesda, Maryland, USA