Nationwide Increase in Complex Congenital Heart Diseases After the Fukushima Nuclear Accident

Background After the Chernobyl nuclear accident in 1986, an increase in the incidence of congenital heart disease (CHDs) in the neighboring countries was reported. In 2011, Japan experienced the Great East Japan Earthquake and the nuclear accidents at Fukushima. However, a nationwide study of their effects has not been conducted yet. Methods and Results We used data covering the period between 2007 and 2014 from the annual surveys conducted by the Japanese Association for Thoracic Surgery, which included almost all of the operations pertaining to 46 types of CHDs in Japan. CHDs were divided into 2 groups based on complexity, the time of occurrence during heart development, and age at operation. We estimated the change in the number of the operations per 100 000 live births between pre‐ and postdisaster using a negative binomial generalized linear mixed model. Overall, a significant 14.2% (95% CI, 9.3–19.4) increase in the number of operations for complex CHDs in neonates and infants per 100 000 live births was found, whereas those performed for patients of 1 to 17 years old showed no significant change during the study period. Conclusions The number of operations for complex CHDs in neonates and infants in Japan significantly increased after the massive disaster, and its level was maintained thereafter. The number of operations for complex CHD was not equal but closely correlated to the live birth prevalence of complex CHDs. Therefore, some meaningful increase in the live birth prevalence can be assumed; however, the precise cause of the increase is unknown.

O n March 11, 2011, the Great East Japan Earthquake occurred followed by a great tsunami and the nuclear accidents at the Fukushima Daiichi Nuclear Power Plant. It is known that various health problems are associated with massive disasters. Reproductive outcomes, such as stillbirth, preterm birth, low birth weight, and congenital anomalies, are also affected by such disasters. 1 The Great East Japan Earthquake was accompanied by the nuclear accidents. After the Chernobyl accident, which was also a nuclear disaster, congenital anomalies apparently increased in the countries near the disaster site, but their direct relevance to radiation exposure was considered to be minor. 2 A paper describing the conclusion by EUROCAT (European Registry of Congenital Anomalies and Twins) showed that no detectable increase in the incidence of congenital anomalies due to ionizing radiation was observed in Western Europe, relatively far from the incident site. 3 In terms of congenital heart anomalies, which are one of the most common congenital anomalies, 4 a paper reported an increase in congenital heart diseases (CHDs) after the Chernobyl accident. 5 However, the study did not receive a positive reception because of the lack of methodological details. 6,7 What was the situation in Japan after the massive disaster in 2011? The Fukushima Health Management Survey launched by the Fukushima prefecture reported no changes in reproductive outcome compared with the recent averages elsewhere in Japan. 8 However, the nationwide trend, including the adjacent prefectures, was not studied. The survey may have underestimated the impact of the disaster because the evacuees' were not fully investigated. Furthermore, patient discharge rates after surgery for cryptorchidism, which is a congenital abnormality, had increased nationwide (13.4%; 95% CI, 4.7-23.0%) after the Fukushima nuclear accidents in 2011. 9 As a result, we wondered whether there was a similar tendency for CHDs. We focused our attention on the data collected by the Japanese Association for Thoracic Surgery. The data can be considered as an entire patient population enabling comparison of the preand postdisaster periods because the association has been recording almost all operations for CHD in Japan since before the disaster in 2011. 10 Because Japan provides universal healthcare coverage, 11 the number of surgeries in Japan is independent of the patients' socioeconomic level. This would help minimize the bias in the data. Another advantageous point is that in contrast to the Chernobyl study, the number of specific CHD operations was recorded. This enables us to examine the association between the disaster and the type of CHD. When defects occur in the early stages of heart development, the CHDs tend to be complex and severe, and require difficult surgical treatment; consequently, the incidence trends of such complex CHDs are an area of concern. Thus, in this paper, we examined the change in the number of operations performed for CHDs, dividing the CHDs into complex and noncomplex.

Methods
The data and analytic methods that support the findings of this study are available from the corresponding author upon reasonable request.

Study Population and Data Sources
We used the data from the annual surveys conducted by the Japanese Association for Thoracic Surgery since 1986. 10 The surveys cover almost all of the surgeries pertaining to CHD in Japan. The association sends out questionnaires to almost all institutions, receives the answers and compiles the annual report. In this study, we considered the 4 years before and after the Great East Japan Earthquake on March 11, 2011March 11, (ie, 2007March 11, -2014 These data contained the number of operations pertaining to CHD, whether cardiopulmonary bypass (CPB) was used, age (neonate, infant, 1-17 years, ≥18 years), and surgical classification (50 types, including 4 types of "redo"; these types were characterized according to the complexity of the operative procedure and the need for additional operations). We analyzed 46 surgical classifications, excluding the redo operations and those in patients under 1 year old (neonate and infant), regardless of the use of CPB, in order to infer the change in the incidence (more accurately, live birth prevalence) of CHD. The numbers were aggregated according to the surgical classification and year.

Congenital Heart Diseases
Different types of cardiovascular malformation occur according to the developmental stage of their responsible anomaly. 12 Generally, the morphological aberration tends to be more serious when occurring at an earlier developmental stage. In this study, we classified the surgical classifications according to pre-and postorganogenesis period, except for "coronary disease" and "others." The surgical classifications were also classified according to the complexity of the malformation (complex or noncomplex malformation) except for "aneurysms of the sinus of Valsalva," "aortic valve lesion," "mitral valve lesion," "coronary disease," and "others." The classification is shown in Table 1.

Yearly Change in the Number of Operations
To analyze the yearly change in the total number of CHD operations, we calculated the total number of operations per 100 000 live births each year. The 95% CIs were also calculated assuming the numbers followed a Poisson distribution. Moreover, we separated the 46 CHDs into "early" and "nonearly" categories according to their developmental stage, and the early CHDs were also divided into "complex" and "noncomplex" as described in Table 1. We also calculated the number of the operations per 100 000 live births and 95% CIs for each division in the same way. The number of live births was excerpted from the Population Survey Report published by the Ministry of Health, Labour, and Welfare. 13 Additionally, to review the yearly changes in the 46 CHDs, we calculated

Clinical Perspective
What Is New?
• This study investigated the influence of a huge earthquake followed by a nuclear accident on complex congenital heart diseases in terms of the number of the operations in 1-yearold patients. • The nationwide impact was assessed by comparing the preand postaccident periods in Japan under the conditions of the existing well-established medical care system and medical statistics.
What Are the Clinical Implications?
• A disaster associated with a nuclear accident might increase the risk of a wide range of complex congenital heart diseases. • Establishment of a medical care system and medical statistics is required for assessing the influence of unexpected disasters.
the number of operations per 100 000 live births for each surgical classification and accumulated them in ascending order by the 8-year total numbers of respective surgical classifications. We also calculated the number of complex and nonearly operations in patients 1 to 17 years old and examined the change and deviation in trend in order to determine the relation between the number of the operations for ages 0 to 1 years and that for 1 to 17 years.

Statistical Analysis
We performed a generalized linear mixed model (GLMM) analysis to examine the change in the number of operations between the predisaster period (2007-2010) and postdisaster period (2011-2014). The analysis employed a "log" link function and contained a random intercept and a random change rate in the numbers between the pre-and postdisaster period for each CHD. We performed Poisson GLMM and a negative-binomial GLMM analysis to determine the appropriate distribution for the data. We employed the "aods3" package of R to assess the goodness of fit of the models. Subsequently, we also performed a Bayesian analysis using the appropriate distribution as a sensitivity analysis (Data S1). Moreover, we performed a Cochran-Armitage trend test to examine whether a trend existed since the predisaster period and examined the deviation if such a significant trend were detected. When a significant deviation was detected, multiple comparisons according to the methods of Benjamini and   15,16 The data we used in this study were from the Society of Thoracic Surgery in Japan, and they were classified according to the type of CHD surgery. Some types of CHDs classified according to surgery type fit the descriptions of multiple types of CHDs classified according to the time of development. The corresponding relationships are shown in Tables 2 and 3. Interruption of the aortic arc was classified as type A or type B according to the location of interruption (type A and type B correspond to left ventricular outflow tract obstruction and conotruncal, respectively); however, this classification was not used for our data. We divided the number of interruption of aortic arc surgeries into 70% for left ventricular outflow tract obstruction and 30% for conotruncal because the interruption of aortic arc type ratio is 7:3 (type A:type B) in Japan.
We obtained study approval from the Institutional Review Board of Nagoya City University. The board waived the requirement for informed consent due to the anonymous nature of the data.

Results
The yearly change in the number of operations for CHDs per 100 000 live births is shown in Figure 1. The data used are shown in Table S1. The early group showed a remarkable increase in 2011 ( Figure 1A), while the nonearly group showed a monotonic increase from the predisaster period with no significant deviation ( Figure 1B, Cochran-Armitage trend test: P<0.001; deviations from the linear trend: P=0.161). The majority of the early CHDs were complex, and the yearly change in early CHDs was largely accounted for by the complex CHDs ( Figure 1C), whereas noncomplex CHDs seemed to have no specific tendency ( Figure 1D). The complex CHDs showed a nonmonotonic increase ( Figure 1C; Cochran-Armitage trend test: P<0.001; deviations from the linear trend: P<0.05). Combined with the result of the multiple comparisons, a large increase between 2010 and 2011 was demonstrated and the increased level was maintained.
The large increase consisted of small increases in several CHDs rather than a large increase in a few particular CHDs ( Figure 2C) whereas in the case of the nonearly CHDs, a few major CHDs, ventricular septal defect (VSD; perimembranous/muscular) and patent ductus arteriosus, comprised a majority of the increase ( Figure 2B).
When we compared the number of operations for complex CHDs between the ages of 0 and 1 (Figures 1C and 3A) and 1 to 17 years ( Figure 3B), the former showed a maintained rate of change after 2011 while the latter showed no changes ( Figure 3B; Cochran-Armitage trend test: P=0.456; deviations from the linear trend: P=0.119). In contrast, in the case of nonearly CHDs, operations at the age of 0 to 1 year showed a monotonic increase ( Figure 3C; Cochran-Armitage trend test: P<0.001; deviations from the linear trend: P=0.161) and those at 1 to 17 years showed a monotonic decrease ( Figure 3D; Cochran-Armitage trend test: P<0.001; deviations from the linear trend: P=0.555).
The goodness-of-fit test indicated that the negative binomial GLMM fit the data better (v 2 =236.5, df=227, P=0.32) in comparison to the Poisson GLMM (v 2 =286.7, df=228, P<0.01). Therefore, the negative binomial GLMM, was considered to be more appropriate for analyzing the data. The result of the negative binomial GLMM analysis of the change in the number of operations is shown in Figure 4. Overall, a 14.2% (95% CI, 9.3-19.4) significant increase was estimated, and 9 CHDs (coarctation+single ventricle, atrioventricular  Figure S1) overall increase. The results of GLMM based on the time of development classification are shown in Table 4. The ventricle, ventricular septum, and conotruncal groups showed a significant increase according to the classification by Yamagishi et al. 15 However, the conotruncal, left ventricular outflow tract obstruction, septal, AVSD, and complex groups showed a significant increase according to the classification by Botto et al. 16

Discussion
The number of operations for CHD in neonates and infants per 100 000 live births in Japan, primarily for complex CHDs, increased by %14% in 2011, the year of the Great East Japan Earthquake, and that level was maintained over the following 3 years. Both congenital malformations, CHDs and cryptorchidism, 9 showed significant increases after the Fukushima nuclear accident. The complex CHDs that showed significant increases were those known to occur during various developmental stages of the heart (eg, single ventricle or hypoplastic left heart syndrome: very early; tetralogy of Fallot: relatively  early), 12 and no particular tendency was found. The analyses based on the time of development showed that a broad range of disease groups was affected, which suggested that whole heart development would be impaired. In the case of operations for complex CHD for patients 1 to 17 years old, no significant changes were detected during the study period. Thus, it would be reasonable to assume that there is no association between the number of operations performed in each age group. The ratio of the number of abortions to the number of live births during the study period in Japan decreased slowly and monotonically ( Figure 5). No new guidelines were created or changed during our study period. Therefore, because there were no major changes in the screening, referral, indication for surgery, or treatment of complex CHDs during the study period in Japan, and because the institutions performing these surgeries had answered the questionnaire before 2011, the incidence of complex CHDs would have to have increased since 2011.
In contrast, in the case of the nonearly CHDs, there appeared to be an association between the monotonic increase in patients 0 to 1 year old ( Figure 3C) and the monotonic decrease in those 1 to 17 years old ( Figure 3D). Operation for VSD, one of the most common nonearly CHDs, is currently being performed earlier. This might explain the increase and decrease.
What caused the increase in the incidence of complex CHDs since the disaster? Given the data, we may hypothesize that the damage to gametes or the developing embryo/fetus from radionuclides emitted from the Fukushima nuclear accident was a direct factor that caused the increase. However, we have no other choice but to state that the association between radionuclides and complex CHDs is unknown at this moment because no data on individual exposure levels is available.
As an indirect effect, maternal stress (job loss, divorce, separation, or death of a close relative or friend) 17,18 is known to be one of the risk factors of conotruncal CHDs (ie, some complex CHDs). Because such stressful events were likely to have occurred during the course of the disaster, maternal stress is considered to be one of the possible causes of this increase. Another possible risk factor is diabetes mellitus. Several types of CHD have been associated with maternal diabetes mellitus, 19 which may explain the increased incidence of single ventricle or hypoplastic left heart syndrome as well as conotruncal CHDs. Some studies have reported that diabetes mellitus has increased in the evacuees over the age of 40, 20,21 but another showed no significant change in  pregnant women. 22 Thus, it is currently unknown whether pregestational or gestational diabetes mellitus has increased in the disaster-affected or adjacent areas. There may also be other possible factors, and further studies will be needed.

Strengths and Limitations
Because of the universal health insurance coverage in Japan, 11 infants with CHDs are treated properly regardless of their family's socioeconomic level. This is a unique characteristic of our data in comparison to other countries'. Moreover, almost all of the institutions performing surgeries for CHD had answered the questionnaire before 2011. Thus, the data we used in this study were almost that of the entire Japanese patient population undergoing surgeries pertaining to CHDs in Japan and are considered to have little bias. Bias attributable to the compositional change of the participating institutions would also be unlikely.
Our study does have some limitations. First, the significant increase was estimated using the number of operations, not the incidence. We should note that the increase in the number of operations is higher than that of the incidence of complex CHDs because complex CHDs often require multiple operations within the first year of life. Second, the data we used in this study lack regional information and data on the exposure level of individuals. More specific patient data such as time, location, and amount of maternal exposure are required to determine the influence of radionuclides and regional difference. If the exposure were increased in a particular region, the number of operations would be greater in the region in The "group" column indicates the disease group defined by Yamagishi (2007)   question. Similarly, if the date of the operation or the weeks of pregnancy at the delivery day are available, we may be able to specify the event responsible for the increase in complex CHDs. Third, we were not able to identify the starting point of the increase because the data were compiled annually. If the cause of the increase had occurred in March 2011, and the fetus was in the organogenesis period, then the infants at risk would be born at the beginning of October in full term or September in moderate or late preterm. If the increase had begun before September 2011, we would not be able to attribute the increase to the disaster alone.

Conclusions
The number of operations for complex CHDs in neonates and infants per 100 000 live births in Japan increased by %14% after the Great East Japan Earthquake. The increase in the incidence of complex CHDs would be lower than that of the number of operations because complex CHDs often require multiple surgical interventions within the first year of life, but some meaningful increase in the incidence may still be assumed. The cause of the increase is unknown, but we should consider the influence of the radionuclides emitted from the Fukushima nuclear power plant. More specific patient data such as time, location, and amount of maternal exposure would be required to determine the cause.