Left Ventricular Dysfunction Following Repair of Ventricular Septal Defects in Infants

Background Left ventricular systolic dysfunction (LVSD) is frequently observed following repair of ventricular septal defects (VSD), although little is known about its incidence, time course, or risk factors. Among infants undergoing VSD repair, for postoperative LVSD, we sought to determine (1) incidence, (2) predictors, and (3) time to resolution. Methods We queried our institution’s surgical database for infants who underwent repair of isolated VSDs from November 2001 through January 2019. The primary outcome was postoperative LVSD, which was de�ned as a shortening fraction of < 26% by M-mode. Postoperative echocardiograms were reviewed and measurements were made using standard methods. Receiver operating characteristic analysis was generated to determine the preoperative left ventricular internal dimension (LVIDd) z-score most predictive of LVSD. Multivariable analysis was conducted to determine associations with LVSD; covariates in the model were weight percentile, genetic syndrome, preoperative diuretic, VSD type, and preoperative LVIDd z-score.


Introduction
Ventricular septal defects (VSDs) are one of the most common forms of congenital heart disease (CHD), and surgical closure of these defects is one of the most commonly performed congenital heart surgeries.
[1, 2] Therefore, a thorough understanding of the pre-and postoperative hemodynamics of these lesions is crucial for family counseling, preoperative decision-making, surgical planning, and postoperative management.
Closure of VSDs is often preemptive in an effort to mitigate the long-term sequelae of a ventricular level left-to-right shunt; these sequelae include failure to thrive, the development of pulmonary vascular obstructive disease, pulmonary over circulation, and/or heart failure .[3][4][5] Although a large left-to-right shunt is not expected to cause systolic dysfunction preoperatively, left ventricular systolic dysfunction (LVSD) is sometimes observed following VSD repair.[6][7][8][9][10][11] However, there has been scant literature focused on the incidence of this dysfunction; even fewer studies have attempted to determine preoperative variables associated with LVSD, and the timeframe for its resolution.[6,8,11] Our primary objectives in this study of a large cohort of infants undergoing surgical closure of moderate and large VSDs were to (1) determine the incidence of postoperative LVSD, (2)

Methods
The Institutional Review Board at the University of Arkansas for Medical Sciences approved this retrospective study and waived the need for individual patient consent (IRB 203877, approved 1/07/2015).The study period was from November 2001 through January 2019.We queried the congenital cardiac surgery and echocardiography databases at Arkansas Children's Hospital and identi ed patients who underwent surgical closure of hemodynamically signi cant VSDs (i.e., clinical manifestations of pulmonary overcirculation such as tachycardia, tachypnea, pulmonary congestion, failure-to-thrive, and need for anticongestive therapies) at less than one year of age.Video 1 shows a standard transatrial VSD patch closure.Patients were excluded if they had another structural cardiac anomaly other than patent foramen ovale or atrial septal defect.With the exception of pulmonary artery banding, patients with prior cardiac surgery were also excluded.
Medical records were reviewed and all relevant demographic and clinical data were collected, including anatomic type of VSD (I-IV, Type I: conal, supracristal, infundibular, subarterial, outlet, doubly committed; Type II: perimembranous, membranous, conoventricular; Type III: inlet/atrioventricular canal; Type IV: muscular).[12] For each subject, the most contemporaneous preoperative echocardiogram was selected for review.Two pediatric cardiologists (EHB and JAD) reviewed the study echocardiograms and independently measured left ventricular internal dimension on M-mode images in the parasternal long-axis view during both systole (LVIDs) and diastole (LVIDd) and calculated the preoperative shortening fraction (SF).Left ventricular systolic function was graded by SF: ≥ 26% = normal, 19-25% = mildly diminished, 14-18% = moderately diminished, and ≤ 13% = severely diminished.Left ventricular size was assessed based on LVIDd z-score.[13] The rst postoperative study echocardiogram was de ned as the rst complete transthoracic echocardiogram performed ≥ 48 hours after surgery and yet within 2 weeks of surgery.Study measurements of ventricular size and function were made independently.The time period of 48 hours was selected to avoid sampling subjects who may have had low cardiac output syndrome.For all patients with postoperative LVSD, all available follow-up echocardiograms were reviewed and measured until LVSD resolved.

Statistical Methods
Summary statistics are presented using mean ± standard deviation (SD) for normally distributed variables and median and inter quartile range (IQR) for non-normally distributed variables.Categorical variables are presented using proportions (%).Continuous variables were compared using Mann-Whitney U tests for non-normally distributed variables and t-tests for normally distributed variables.Categorical variables were evaluated using χ 2 tests, unless cells sizes were small, in which case a Fisher's exact test was used.Univariable analyses of all clinical variables were performed to determine association with postoperative LV dysfunction.Those in the univariable analysis with a p-value of ≤ 0.2 were considered for the multivariable model.Multicollinearity was assessed using both clinical knowledge and variance in ation factor.Odds ratios with 95% con dence intervals and p-values were estimated for all variables included in the nal multivariable logistic regression model, as well as their respective univariable logistic regression models.
A receiver operative characteristic (ROC) curve was generated in order to evaluate a threshold for preoperative LVIDd z-score postoperative LVSD.P-values < 0.05 were considered statistically signi cant.All statistical analyses were performed using R Core Team (2020).R: A language and environment for statistical computing.R Foundation for Statistical Computing, Vienna, Austria.URL https://www.R-project.org/.

Results
There were 164 patients who met inclusion criteria.Table I  Preoperatively, 79 (48%) of the patients met criteria for failure to thrive.Of the 164 patients, 143 (87%) were on at least one diuretic prior to surgery.A total of 58 patients (35%) had a genetic syndrome, with the majority of those (44 of the 58, 27% of the entire cohort) being diagnosed with Trisomy 21.There were six patients in the cohort who had a history of PA band prior to VSD closure.Of the six, only one had preoperative LV dilation, and none had postoperative LV dysfunction.
Table II demonstrates comparisons between patients with and without postoperative LVSD.Following repair, 62 (38%) patients had LVSD.Follow-up echocardiograms were available in 60 (97%) of those patients.The median time of rst postoperative echocardiogram was 6 days (IQR 4-28 days).Type IV VSDs were associated with higher risk of postoperative LVSD compared to type I or type II VSDs.Postoperative LVSD was less common in patients with a genetic syndrome.Table III demonstrates comparisons between patients with and without a genetic syndrome.

Discussion
This study presents the largest cohort to date of infants who underwent surgical closure of isolated moderate or large VSDs, and there were four principal ndings.First, preoperative LV dilation is strongly associated with an increased risk of postoperative LVSD.Second, LVSD is common following VSD closure.Third, while common, LVSD almost uniformly resolves within 9 months of surgery.Finally, the presence of a genetic syndrome is associated with a decreased risk of postoperative LVSD.
Perhaps our most notable nding is that preoperative LV dilation can predict the development of postoperative LVSD.Speci cally, we have found that an LVIDd z-score > 3 is a fairly robust predictor.Studies describing the relationship between postoperative left ventricular volume status and ventricular performance are limited, [9,11,14,15] and no prior studies have identi ed a threshold of left ventricular dilation that predicts LVSD.Left ventricular dilation develops in response to the volume load on the LV due to left-to-right shunting across the VSD.Therefore, LVIDd is likely a surrogate of the hemodynamic impact of the VSD.As such, patients with increased LVIDd and LVIDd z-scores are considered to have more hemodynamically signi cant VSDs, which postoperatively is associated with increased rates of LV dysfunction.Objectively, preoperative LVIDd z-scores of > 3 were associated with both an increased incidence of postoperative dysfunction and also prolonged time to resolution of dysfunction.This important nding can guide the management of patients with VSDs.Speci cally, the medical and surgical teams may consider more precipitous VSD closure if the patient's z-score is approaching our threshold of 3. In so doing, our ndings suggest the likelihood of postoperative dysfunction would decrease, and the time to normalization of LV function would shorten.
The high incidence of postoperative LVSD is consistent with prior studies.[6,8,11,[15][16][17] In our study, approximately two-fths developed postoperative LVSD.Adamson et al. described a similar rate of postoperative LVSD in their cohort of 104 infants following surgical closure of isolated, large VSDs.[11] Similarly, Klitsie et al. and Baghdady et al. demonstrated LVSD after VSD closure using LV fractional shortening and myocardial performance index, respectively.[6,8] In contrast to these prior studies, we examined a signi cantly larger cohort of patients with hemodynamically signi cant VSDs.Our study also shows that although common, LVSD was rarely severe, with the majority of patients experiencing only mild LVSD.The pathophysiology of postoperative LVSD is postulated to be two-fold.First, acutely removing preload in a chronically volume overloaded LV is expected to cause some decrease in cardiac output.[3] Second, VSD closure increases the afterload for an LV habituated to pump against the lower afterload of the pulmonary vascular bed.[3] While common, LVSD uniformly resolved without intervention.The majority of patients in this study with LVSD had normalization by 3 months without medical intervention, and all had resolution of LVSD by 20 months.Multiple studies have corroborated these ndings and have demonstrated that LV function decreases immediately postoperatively and then gradually shows signs of improvement.[6,8,11,[15][16][17] Adamson et al. observed in their cohort of 39 infants with LVSD that all had normalization of LV function within 9 months of surgery.[11] Our results in concert with those of Adamson provide valuable insight into clinical management and follow-up.in patients with postoperative LVSD.Speci cally, our ndings suggest that medical providers can be reassured that intervention to address LVSD is highly unlikely.Additionally, should LVSD occur, families can be counseled that it is expected to resolve.
We have demonstrated that the presence of a genetic syndrome may be associated with both a decreased incidence of postoperative LVSD and shorter time to resolution.Patients with a genetic syndrome composed about one-third of our cohort, and the majority of those had Trisomy 21.Univariable analysis suggested the diagnosis of Trisomy 21 was associated with a decreased incidence of postoperative LVSD and decreased time to normalization should dysfunction occur.We postulate that this nding is related to the increased pulmonary vascular resistance that is often observed in patients with genetic syndromes, in particular Trisomy 21.[18-21]As a result of increased pulmonary vascular resistance, these patients have decreased left-to-right shunting preoperatively, and thus the acute shift in preload and afterload with VSD closure is less severe, resulting in lower rates of postoperative dysfunction.[18][19][20][21] This hypothesis is supported by our nding that patients with a genetic syndrome had both a decreased preoperative LVIDd and LVID z-score compared to patients without a genetic syndrome.Notably, while this observation was supported by our univariable analysis, on multivariable analysis, that association dissipated.This may be due to a relatively small number of patients with Trisomy 21 in our cohort and is an area in need of further study.
While we have identi ed several important ndings, there are limitations to our study.The study was retrospective and is subject to the limitations intrinsic to such a design.This is a single-center experience, which may limit its generalizability.Notably, authors from other institutions have also reported LVSD is common after surgical VSD closure. 11Because early in the study period our institution relied predominantly on LV fractional shortening to assess LV function, the echocardiographic images needed to estimate LV ejection fraction were not present for many of the earlier cases in the study.Therefore, we used fractional shortening throughout the study, as it was considered the standard in pediatric cardiology at the outset of our study period.Fractional shortening of left ventricular function has limitations, and, currently, estimates of left ventricular ejection fraction would be used and derived by Simpson's method or 5/6 area length.Notably, as with fractional shortening, those methods of estimating ejection fraction have inherent assumptions that limit their accuracy.Our congenital heart surgical program underwent multiple, signi cant personnel changes throughout the course of the study, which may have impacted variations in surgical techniques.We did not see signi cant era effect to suggest this had any impact on our study ndings.

Conclusion
In this study, we discovered that elevated LVIDd prior to surgery may be a signi cant predictor of postoperative LVSD.Postoperative LVSD following VSD closure is common.LVSD is expected to resolve in most patients within 9 months of surgery and uniformly by 20 months.The presence of a genetic syndrome appears to be protective against LVSD.

Tables
characterize the time to recovery of left ventricular systolic function among children experiencing postoperative LVSD, and (3) identify variables associated with postoperative LVSD.

Figure I presents
Figure I presents time to resolution of LVSD.The median time to resolution of LVSD was 76.5 days [12.5, 139].Complete recovery of left ventricular function occurred by the 9th postoperative month in 94%, with all 62 patients with postoperative LVSD eventually having recovery of normal function.Subjects with a preoperative LVIDd z-score > 3 and postoperative LVSD generally had longer time to recovery as compared to those with a preoperative LVIDd < 3. Subjects with a genetic syndrome generally had shorter time to resolution of LVSD.

Figure
Figure II demonstrates the receiver operative characteristic analysis of LVIDd z-scores, wherein a z-score of 3.1, with an area under the curve of 0.80, predicted postoperative LVSD with a sensitivity of 0.63 and speci city of 0.83.

Figure
Figure III presents the results of multivariable linear regression analysis wherein only LVIDd z-score was signi cantly associated with postoperative LVSD.

Figures
Figures

Figure 1 Time
Figure 1

Table II .
Comparison of patient characteristics in subjects with and without LVSD.