Prediction of spontaneous closure of ventricular septal defect and guidance for clinical follow‐up

Aim To predict the spontaneous closure of ventricular septal defect (VSD) and assist pediatrician to manage VSD children. Methods Between January 2008 and December 2016, 132 children diagnosed with isolated VSD by echocardiography were enrolled. All participating children were followed‐up by echocardiography yearly until the closure of VSD or 6 years old. The clinical indicators and echocardiographic parameters of patients were collected. Statistically significant factors were used to establish a Logistic Regression model for predicting spontaneous closure of VSD. Receiver operating characteristic (ROC) analysis was used to assess the specificity and sensitivity of Logistic Regression model. Results Spontaneous closure occurred in 60% of all patients; 57% in perimembranous VSD (p‐VSD) and 64% in muscular VSD (m‐VSD) patients. Initial diagnosis age, defect size, aneurysms tissue of the ventricular membranous septum (ATVMS), pulmonary hypertension (PH), and left ventricular diastolic dimension (LVDD) were statistically significant. Defect size, ATVMS and LVDD were determined by the Logistic Regression model as representative factor. P‐VSD and m‐VSD model had areas under the ROC curves 0.854 and 0.898, respectively. Conclusion We inferred that defect size, ATVMS and LVDD were characteristic and representative predictors for spontaneous closure of VSD. And we summarized the prognostic factors and recommended a follow‐up criteria to assist the pediatrician managing VSD children.


| INTRODUCTION
Ventricular septal defect (VSD) is the most common congenital heart disease (CHD), accounting for almost 50% of all CHD. 1,2 The estimated prevalence of VSD is 2 to 3.94 per 1000 live births diagnosed by echocardiography. 3,4 Two-dimensional (2D) combined color Doppler echocardiography makes it easy to diagnose VSD, and to measure size and position of defect, as well as to evaluate the anatomical relationship with the surrounding tissue. The spontaneous closure rate of VSD remains controversial, however, VSD generally close spontaneously in 12% to 84% cases. [5][6][7] This wide range may be attributed to age, size and position of VSD, diagnostic methods, and follow-up period. Recent studies focus on describing the incidence and natural history of VSD. 6 Others studies had a short follow-up time. 8 In this study, patients were followed-up from the infant period to 6 years old, and showed the spontaneous closure rate of perimembranous VSD (p-VSD) and muscular VSD (m-VSD). Our aim was to seek to the spontaneous closure predictive factors of VSD, and to recommend a follow-up criteria for assisting pediatrician in selecting the best management scheme for VSD children.

| Study population
Between January 2008 and December 2016, 132 children presented with isolated VSD at Shengjing Hospital of China Medical University by echocardiography. Out of 132, 127 patients, 66 males, and 61 females with average age of 7.99 ± 8.52 months, were diagnosed either with p-VSD or m-VSD. Five children were excluded from the study during the follow-up period due to surgical intervention (including 4 p-VSD and 1 m-VSD). This study did not include patients with inlet and outlet VSD. Because most patients were recommended for early surgical treatment, rather than follow-up observation. All participating children underwent echocardiography yearly till the closure of VSD or 6 years old. All parents and guardians were provided written informed consent for the minors to participate in this study. The school ethics committee approved the study.

| Echocardiographic parameters
All echocardiographic examinations were performed using Philips iE33 to select S5-1 (frequency 1.0-5.0 MHz) or S12-4 probe (frequency 4.0-12.0 MHz). Clinical indicators (sex, initial diagnosis age) were collected at the beginning of our study. Since most patients have a first diagnosis age of less than 6 months, they were divided into two groups: less than 6 months and more than 6 months. Then, children patients underwent conventional echocardiography. They were placed in the conventional left lateral position, once asleep or quiet, and parasternal long-axis, parasternal short-axis, and apical four-chamber views were obtained. The main purpose was to scan the ventricular septum completely, and combine with color Doppler echocardiography to determine the type and position of the defects. Defect size measured by 2D echocardiography. If the defect boundary was not clearly displayed, color flow mapping may be used. The diameter of defect was recorded as the maximal VSD diameter in multiple planes.
Defect size was divided into three groups 6 : (a) less than 3 mm was classified as small; (b) 3 to 5 mm was classified as medium; (c) more than 5 mm was classified as large. Number of defect were divided into two groups: single defect or multiple defects. Position of m-VSD was classified as (a) anterior; (b) midmuscular; (c) apical VSD as described in Ramaciotti et al. 9 According to whether the aneurysms tissue of the ventricular membranous septum (ATVMS) was formed, p-VSD was divided into two groups. Diagnostic criteria of pulmonary hypertension (PH) were defined as pulmonary artery systolic pressure > 36 mm Hg, or mean pulmonary artery pressure > 25 mm Hg. 10 We followed the recommended guidelines 11 to measure left ventricular end-diastolic dimension (LVDD), aortic root dimension (AOD), and pulmonary artery dimension (PAD). Spontaneous closure was confirmed by color Doppler mapping, and the heart murmur was undetectable.

| Establishment of logistic model
The model was generated with spontaneous closure as a dependent variable (spontaneous closure as 1, no closure as 0), defect size (X 1 ), ATVMS (X 2 ), and LVDD(X 3 ) as independent variables.

| Statistical analysis
Using SPSS 23.0 software, continuous variables were expressed as mean ± SD (X ± SD), and t-test was used for comparison between groups; dichotomous variables was expressed as percentage, and χ 2 tests was used for comparison between groups. Statistically significant clinical indicators and echocardiographic parameters were used to identify predictors by binary Logistic Regression analysis. Likelihood ratio test was used to evaluate the fit of the entire model. The predictability of the Logistic Regression model was evaluated by the ROC analyses. The area under the curve >0.5 would predict spontaneous closure, while the area under the curve ≤0.5 would predict open. P < 0.05 was considered statistically significant.
Males had a better closure rate (62%) compared with females (51%); By the number of defect, spontaneous closure rates were shown as follows: (a) single, 39 of 69 (57%); (b) multiple, 3 of 5 (60%). Patients less than 6 months of initial diagnosis age showed a higher rate (69%) than older than 6 months (45%). There was a negative association between defect size and rate of spontaneous closure; small defects had the highest rate (83%). ATVMS was a favorable factor for spontaneous closure. The spontaneous closure rate of p-VSD with ATVMS was 70%, while that without ATVMS was only 24%. PH was a negative factor for spontaneous closure. Although there were only five patients with PH in this study, none of them were closed. LVDD of spontaneous closure was smaller than no closure. Differences of results between patients with and without spontaneous closure of p-VSD are summarized in Table 1.

| M-ventricular septal defect
Spontaneous closure rate of 64% was observed among 53 patients with m-VSD. As in p-VSD, males showed a higher rate (66%) compared to females (63%) and 65% of single defects closed spontaneously compared to multiple defects (60%). Spontaneous closure was more frequent in age ≤ 6 months group (83%) vs than in age>6 months (39%) group. The small defects showed the highest frequency (80%) of spontaneous closure. Apical defects closed at a higher rate (67%) than mid-ventricular (66%) and apical (67%). Similar to p-VSD, PH was a negative factor for spontaneous closure. Three patients with PH were no closed. LVDD of spontaneous closure was smaller than no closure. Differences of results between patients with and without spontaneous closure of m-VSD are summarized in Table 2.
Binary Logistic Regression analysis showed that defect size

| DISCUSSION
With the improvement in sensitivity and clarity of 2D combined with color Doppler echocardiography in recent years, 11 the detection rate of VSD is increased. 3 However, this has advantages and disadvantages. Echocardiographic follow-up can improve the understanding of the natural history of VSD, and assess the severity of the disease.
Conversely, due to the increase of the detection rate, parents are anxious for VSD children, and pediatricians are also facing great pressure Show a process of ventricular septal defect (VSD) children spontaneously closed by echocardiography. A case of a 32-days girl baby who underwent echocardiography due to heart murmur. A, Color Doppler echocardiography shows parasternal long-axis viewed of a perimembranous defect, and the width of defect was 4.0 mm. B, When she was 90 days, color doppler echocardiography showed the defect of the parasternal short-axis had been reduced, and the width of defect was 3.1 mm. C, When she was 8 months, this defect became smaller to 2.0 mm. D, When she was 14 months, the defect by echocardiography has closed spontaneously, and heart murmur also disappeared. AO, aortic; LA, left atrium; LV, left ventricular; PA, pulmonary artery; RA, right atrium; RV, right ventricular; VSD, ventricular septal defect to manage patients with VSD. Our study, which included all the factors in common used and easily acquired might help to predict spontaneous closure of VSD and assist pediatrician in selecting the management scheme for VSD children.
In this study, VSD children were followed up to 6 years old with echocardiography. We thoroughly studied the natural history of p-VSD and m-VSD. On the one hand, we found the prevalence of p-VSD was more common (74 cases, 58%) than m-VSD (53 cases, 42%).
This confirmed the previous observation that the incidence of p-VSD was higher than m-VSD. 6 On the other hand, we demonstrated that the rate of spontaneous closure was 60% in VSD, and m-VSD are easier closure than p-VSD, which in similar to the conclusion from Eroglu et al 5 and Axt-Fliedner et al. 12 This difference may be attributed to different closed mechanisms. P-VSD closure was thought to occur as a result of damage caused by a high-velocity jet to damage the endothelium at the edge of the VSD or the septal leaflet of the tricuspid valve, both of which can lead to accumulation of platelets and production of fibrous tissue. 13 On the other hand, the spontaneous closure of m-VSD was due to muscular encroachment of the septal defect along with superimposed fibrosis or fibrous tissue formation around the margins leading to apposition of the edges of the defect. 4,14 Another reason may be that m-VSD size was slightly smaller than p-VSD. Our study showed the diameter of p-VSD was 3.43 ± 1.34 mm, and the diameter of m-VSD was 2.68 ± 1.08 mm.
Although high rate of spontaneous closure was satisfying, there was currently no simple and systematic way to predict spontaneous closure of VSD. 15 Therefore, it was essential to find out predictors of spontaneous closure of VSD. According to our analysis, defect size, ATVMS, and LVDD were characteristic and representative predictors.
Our study showed that defect size was one of the key factors affecting the spontaneous closure of VSD. For the same age of examination, the smaller the defect is, the easier it to close spontaneously. The spontaneous closure rate of small p-VSD was 83%, and small m-VSD was 80%. Our results are consistent with these studies, 2,16 the spontaneous closure probability of small defects was much higher than larger ones. ATVMS formation was also an important factor of the spontaneous closure of p-VSD. It was reported that the rate of ATVMS formation was 46% to 81%, 2,5,8  Whether the position of m-VSD affects the probability potential for spontaneous closure remains debatable. The results of this study indicated that the spontaneous closure of m-VSD is independent of the position of the defect, which is in line with published reports. 7,19,20 We observed that the incidence rate of mid-muscular VSD, which is the most common position for m-VSD, was 60% with a spontaneous closure rate of 66%. The spontaneous closure rate of apical defects was 67%, higher than the mid-muscular VSD and the apical VSD. This is contrary to previous reports which showed that apical VSD and anterior VSD persisted longer than mid-muscular VSD. 7,9 Possible reasons for this disparity are differences in follow-up time, diagnostic criteria, and ethnicity. 3,21 It is well known that spontaneous closure of VSD can occur at any age, but most commonly occurs in childhood. 22 According to the statis-

| LIMITATION
There were many limitations in this study. (a) The number of patients was not big enough, and the follow-up time was not long enough. The  (b) This study did not combine other clinical examinations and laboratory indicators. Therefore, our information can be used to predict the possibility of spontaneous closure and decide treatment options, but not to provide an accurate diagnosis.

| CONCLUTION
This study described the natural history and spontaneous closure rate of children with p-VSD and m-VSD. We inferred that defect size, ATVMS and LVDD were characteristic and representative predictors for spontaneous closure of VSD. A follow-up criteria was recommended by summarizing this study, which may assist pediatricians in selecting the appropriate treatment strategy to manage VSD children.

ACKNOWLEDGMENTS
This study was funded by National Natural Science Foundation of China (Grant No. 81571686).