Long-term mortality in patients with atrial septal defect: a nationwide cohort-study.

Aims: In this nationwide cohort of atrial septal defect (ASD) patients, the largest to date, we report the longest follow-up time with and without closure in childhood and adulthood compared with a general population cohort. Methods and results: Using population-based registries, we included Danish individuals born before 1994 who received an ASD diagnosis between 1959 and 2013. All diagnoses were subsequently validated (n = 2277). Using the Kaplan-Meier estimates and Cox proportional hazards regression adjusted for sex, birth year, and a modified Charlson Comorbidity Index, we compared the mortality of ASD patients with that of a birth year and sex matched general population cohort. The median follow-up from ASD diagnosis was 18.1 years (range 1-53 years). Patients with ASD had a higher mortality [adjusted hazard ratio (HR): 1.7; 95% confidence interval (CI): 1.5-1.9] compared with the general population cohort. The adjusted HR 30 days after closure was 1.4 (95% CI: 1.2-1.7), and it was 2.4 (95% CI: 2.0-2.9) for patients without closure.
Conclusion: Overall, ASD patients had a higher long-term mortality than a general population cohort matched on birth year and gender. Our data indicate a lower relative mortality of those ASD patients undergoing closure than the ASD patients not undergoing closure.

Introduction

Half a century ago, the life expectancy of ASD patients was reported to be less than 50–60 years; however, later studies have demonstrated higher survival rates. Murphy et al. reported that closure before the age of 24 years normalized the mortality risk compared with the expected survival of the background population, whereas patients with closure after the age of 24 had an increased mortality risk. Cuypers et al. investigated 139 patients with surgical closure before the age of 15 and found low morbidity and mortality rates 35 years after closure. Similar results were found by Roos-Hesselink et al. in a study including 135 patients with up to 26 years of follow-up. However, none of these studies were directly compared with a matched background population. In adulthood, it is unclear whether ASD closure changes the natural course. Closure has been reported to reduce the 10-year mortality rate in some studies and had no impact in others., John Sutton et al. found that patients with ASD closure had a survival rate comparable with those with no ASD after 5 and 10 years but with no follow-up beyond this time. Longer follow-up has shown reduced survival of ASD patients, regardless of closure, compared with matched controls, but these data were collected more than 25 years ago and contemporary studies of large population with a long-term follow-up are lacking. Hence, evidence is needed to clarify whether a small ASD impact on survival and whether ASD patients normalize mortality risk after closure in either childhood or adulthood. We therefore aimed to estimate the long-term mortality of adult ASD patients with and without closure compared with a general population cohort in a nationwide cohort covering up to 53 years of follow-up.

Materials and methods

Study cohort and design

This nationwide cohort study was based on national medical registries in Denmark. Part of the study cohort has previously been used in investigations of the impact of an ASD on pneumonia, atrial fibrillation, and stroke., The unified Danish health care system is publicly funded with equal access for all Danes. Hospital data are gathered in nationwide public registries linked to a unique personal identification number (the CPR number) provided for all inhabitants since 1968. Information on the date of birth, date of immigration, and date of death was identified in the Danish Civil Registration System.

Two data sources were used to identify ASD patients for this study. We used the Danish National Patient Registry (DNPR) established in 1977 to identify patients born before 1994 and diagnosed with an ASD from 1977 to 2013. The DNPR contains information on all inpatient and outpatient hospital contacts in Denmark, dates of admission and discharge, surgical procedures, and discharge diagnoses coded according to the International Classification of Disease (ICD). The 8th edition was used until 1993 after which the 10th version has been used. In Denmark, it is mandatory by law for all healthcare centres and hospitals to report to the register, and it is considered to be nearly complete. Two independent doctors reviewed all hospital records to validate the diagnosis registered in the DNPR. The ASD closing procedure was likewise registered in the DNPR, and the date was confirmed in the review of hospital records. During this medical record review, we identified a small number of patients with ASD diagnosis dates and/or ASD closure dates recorded before the initiation of the DNRP. Thus, the dates of diagnoses of the included subjects ranged from 1959 to 1 January 2013. The second data source included patients diagnosed with ASD before the age of 15 years between 1963 and 1974. These patients are a subgroup of a cohort used in a previous study of patients with congenital heart defects. The patients were identified by an experienced physician from 1970 to 1974 by review of inpatient and outpatient hospital records in all Danish medical and paediatric departments. The CPR number of the patients alive in 1968 was manually identified, and their diagnoses were translated into ICD-10 codes. Patients who died before 1968 (n = 5), and therefore did not receive a CPR number, were excluded. We only included patients with an isolated ASD except those few (n = 3) who had a concomitant patent arterial duct, thought to be of no relevance in adulthood. Patients were excluded if the defect was described as a patent foramen ovale. Thus, only patients with an ASD diagnosis confirmed by medical record review were included in this study.

For every validated ASD patient, 10 people from the general population were matched on sex and birth year using the Danish Civil Registration System.

Baseline comorbidity was identified using the ICD diagnoses in the DNPR covering both inpatients and outpatient diagnoses before or at the time of the ASD diagnosis. Only comorbidities registered after 1977 were included in the baseline comorbidity due to the lack of information before this time. Comorbidity was indexed in the Charlson Comorbidity Index, which includes diagnoses of heart-, lung-, cerebral-, renal-, and liver disease, as well as cancer, Human Immune Deficiency Virus, and diabetes, for a total of 17 diagnoses. Patients were categorized into four groups depending on their score as follows: no coexisting comorbidity, mild, moderate, and severe comorbidity. We modified the index by excluding diagnoses that could be caused by the ASD in the intermediate step in the causal pathway between the ASD and mortality. Hence, we did not include comorbidities such as cerebrovascular disease (except neurological diseases), pulmonary heart disease and those of the hemiplegia/paraplegia diagnoses caused by stroke as adjustment in our analyses.,

Mortality

Death was identified using data from the Civil Registration System. The causes of death were obtained from the Danish Register of Causes of Death. Mortalities in the period between 1959 and 1977 were identified in the hospital records and added separately. Patients without a CPR number in 1968 (n = 5) were unaccounted for in terms of death or emigration.

Statistical analysis

Follow-up of ASD patients and the comparison cohort started on the date of diagnosis or closure (the index date for the matched comparison cohort members). Follow-up continued until death, emigration, or the end of Follow-up (1 January 2013), whichever came first. For patients with closure, 30 days of mortality after the procedure was computed both in total and in terms of closure before or after 1990 when catheter closure was introduced in Denmark. The Kaplan–Meier curves were computed with either age or time since diagnosis/closure as a time-scale.

We used Cox proportional hazards regression to compute hazard ratios (HR) of mortality for ASD patients, beginning at the time of diagnosis for those without closure and 30 days after closure for those with closure, and compared them with the general population cohort. Age was the underlying time scale. The patients were stratified by age at time of the closure, and the HR was adjusted for gender, Downs Syndrome, and the Charlson Comorbidity Index.,

In a separate analysis, we entered closure as a time varying exposure to compare mortality of ASD patients with and without closure. This analysis was done by adjustment for pulmonary arterial hypertension and Eisenmenger syndrome. We graphically verified the assumption of proportional hazards with log minus log plots. Analyses were performed using Stata 13 (StataCorp LP, College Station, TX, USA). This study was approved by The Danish Data Protection Agency (j.nr. 2010-41-4649) and by the National Board of Health (j.nr. 7-604-04-2/193/KWH).

Results

The hospital records of all patients eligible for validation (n = 4408) diagnosed in one of 80 hospitals in Denmark since 1959 were reviewed. The majority of the patients (74%) were diagnosed in one of the four university hospitals in Denmark. Surgical closure was performed in five public hospitals and transcatheter closure in two public hospitals all with highly specialized teams. No ASD closures were performed in private institutions. Patients were excluded from the validation process if the ASD diagnosis was inadequate or missing in the hospital record or if a concomitant congenital heart disease was present. A total of 2277 patients met the inclusion criteria after validation.

Baseline characteristics

Most of the ASD patients (n = 1554, 68%) had closure of their defect during follow-up. Median follow-up was 18.1 years (range 1–53.3 years) after diagnosis. The ASD patients, especially those without closure, had more baseline comorbidities than the comparison cohort at the time of diagnosis (Table ).

Table 1

 Baseline characteristics at time of diagnosis

	ASD total (n = 2277)	ASD closed (n = 1554)	ASD unclosed (n = 723)	Comparison cohort (n = 22756)
Female, n (%)	1384 (61)	979 (63)	405 (56)	13 834 (61)
Mean age, end of FU (years)	45.4 ± 0.4	46.6 ± 0.5	43.0 ± 0.9	46.0 ± 0.1
Chronic lung disease, n (%)	181 (8)	113 (7)	68 (9)	1024 (5)
Diabetes, n (%)	116 (5)	74 (5)	42 (6)	790 (3)
Pulmonary heart disease, n (%)	123 (5)	60 (4)	63 (9)	52 (0.02)
Hypertension, n (%)	262 (12)	188 (12)	74 (10)	1661 (7)
Ischaemic heart disease, n (%)	325 (14)	239 (15)	86 (12)	1212 (5)
Cerebrovascular events, n (%)	180 (8)	132 (8)	48 (7)	613 (3)
Arrhythmia, n (%)	514 (23)	396 (25)	118 (16)	715 (3)

ASD, atrial septal defect.

The 30 days of mortality after closure was 0.9% (n = 8), of which five were after surgery, one was after transcatheter treatment, and two were after unknown procedures. Of the eight patients, six were adults (three were more than 65 years old), and two were children.

The ASD patients had an overall increased mortality risk (Figure ) compared with the comparison cohort with an adjusted HR of 1.7 (95% CI: 1.5–1.9) (Table ). The HR was higher [adjusted HR: 2.4 (95% CI: 2.0–2.9)] for those with no closure compared with those who had closure during follow-up [adjusted HR: 1.4 (95% CI: 1.2–1.7)]. Comparison of mortality for patients with ASD closure vs. those without closure showed an adjusted HR of 1.6 (95% CI: 1.3–2.1).

Figure 1

(A) Survival in years in atrial septal defect (ASD) patients without closure compared with the comparison cohort with age as an underlying time scale. (B) Survival in ASD patients after closure compared with the comparison cohort with years after closure as an underlying time scale. (C) Survival in patients with closure before the age of 18 compared with the comparison cohorts with years after closure as an underlying timescale. (D) Survival in patients with closure after the age of 18 compared with the comparison cohorts with years after closure as an underlying timescale.

Table 2

 HR for mortality in patients with closed or unclosed atrial septal defect compared with a general population comparison cohort in different age groups and within two time intervals after diagnosis or closure

Hazard ratio (95% CI)
	ASD closeda	ASD unclosed	ASD total
ASD all	1.6 (1.4–1.9)	3.1 (2.6–3.7)	2.1 (1.8–2.3)
	1.4 (1.2–1.7)b	2.5 (2.1–3.0)b	1.7 (1.5–1.9)b
	1.4 (1.2–1.7)c	2.4 (2.0–2.9)c	1.7 (1.5–1.9)c
ASD closed <18 years	1.7 (1.03–2.9)
	1.7 (1.02–2.8)b
	1.6 (0.9–2.6)c
ASD closed 18–50 years	1.9 (1.4–2.5)
	1.7 (1.3–2.3)b
	1.7 (1.2–2.2)c
ASD closed >50 years	1.5 (1.2–1.9)
	1.3 (1.02–1.6)b
	1.3 (1.02–1.6)c

Adjusted mortality risk for ASD patient with closure, no closure during the follow-up period and the total cohort of ASD patients compared with the comparison cohort.

ASD, atrial septal defect.

Mortality risk starting 30 days after closure.

Adjusted for the modified Charlson Comorbidity Index and gender.

Adjusted for the modified Charlson Comorbidity Index, gender, and Downs.

For patients with closure there was an increased mortality risk if closure was performed after the age of 18 years. The adjusted HR comparing the ASD patients with the general population was 1.6 (95% CI: 1.2–2.2) in the group of ASD patients who had closure performed between 18 and 50 years of age and the adjusted HR was 1.3 (95% CI: 1.02–1.6) in patients with closure after the age of 50 years. For those ASD patients with closure performed before the age of 18 years, the long-term mortality remained increased when adjusted for the modified Charlson comorbidity index, with a HR of 1.7 (95% CI: 1.03–2.9). When adjusted also for Down syndrome, the relative risk was the same, but no longer statistically significant [HR: 1.6 (95% CI: 0.9–2.7)]. In a direct comparison, there was a higher, but not statistically significant, mortality risk for males [adjusted HR: 1.2 (95% CI: 0.95–1.5)] compared to females.

Causes of death

Patients with closure had significantly fewer cardiac deaths (P < 0.01) than those without closure where almost half the patients died from cardiac disease (Table ). Heart failure in particular was a more frequent cause of death in the group of patients who had no ASD closure. In both groups, 14% and 15%, respectively, ‘ASD’ was reported as being the cause of death. It remains unknown to which extent this covers heart failure and pulmonary heart disease. There was a higher number of cancer-related deaths in the closure group but this was not statistically significant (P = 0.064). Patients who had closure by surgery did not have a higher risk of long-term mortality compared with those who had closure performed by catheter (HR 0.6, 95% CI: 0.2–1.8; P = 0.39).

Table 3

Causes of death for atrial septal defect patients

	No closure, n = 723	After closure, n = 1554
	n (%)	n (%)
Total deaths	156	159
Cause of death
Cardiac, non-ischaemic	49 (31)	31 (19)
Heart failure	19	4
Atrial fibrillation	2	4
Mitral valve pathology	4	2
Othera	24	21
Cardiac, ischaemic	24 (15)	23 (14)
Ischaemic heart disease	11	16
AMI	13	7
Cardiac, total	73 (47)	54 (34)
Infection	10 (6)	10 (6)
Pneumonia	8 (5)	7 (4)
Stroke	9 (6)	11 (7)
Cancer	22 (14)	38 (24)
Diabetes/kidney disease	8 (5)	10 (6.3)
Chronic lunge disease	7 (4)	8 (5)
Downs	9 (6)	2 (1.3)
Other	15 (10)	16 (10)
Unknown	3 (2)	10 (6.3)

Causes of deaths divided into categories. Percentages in each category is of the total number of deaths for patients either with or without closure.

ASD, atrial septal defect.

Other cardiac disease includes those listed only as death caused by ASD, cardiomyopathy, and heart disease of no further specification.

Discussion

This is the largest nationwide cohort study with the longest follow-up time including closed as well as unclosed ASD and including all ASDs closed by surgery as well as those closed by catheter. It shows, that mortality was higher for ASD patients than for the background population.

The 30 days of mortality found in this study is in concordance with most recent studies emphasizing the low risk of the closure procedures in ASD patients. The long-term survival has; however, been a subject of debate. Recent studies primarily focus on the short- to medium-term outcome of single-centre study populations, often with good results within 5–8 years after closure. Kutty et al. found that only 3% of ASD patients with closure had died at the end of follow-up (range 5–20 years), but only half of the patients were older than 20 years at time of closure, and with a 20% loss to follow-up the mortality could be underestimated. In contrast, Jeong et al. found much higher mortalities of 21% after closure and 27% with no closure with a mean follow-up of 6.8 years. John Sutton et al. found that adult patients with ASD closure did not have altered survival compared with a matched control group 5–10 years after closure. The very low long-term survival of 40% in patients with closure found by Murphy et al. was not replicated in this study. We found that more than 80% were alive 30 years after closure and that the survival rate for the comparison cohort was similarly higher than the one found by Murphy et al. This difference could be explained by general health improvements and increased life expectancy in the population over the last 50 years. The general life expectancy has improved by nearly 5 years since 1990 when the study by Murphy et al. was published and much more since the 1950 s when the patients in the study by Murphy et al. had surgery. Furthermore, the expanded knowledge in medical treatment, intensive care facilities, and improved treatment options for the patients could improve the overall life expectancy of the patients in this study.

The mortality risk for patients without closure in our study was increased compared to those who had closure during follow-up. However, the patients without closure were far from comparable with those who had closure during follow-up. The patients without closure were older at the time of diagnosis and had more comorbidity. Six patients were not eligible for closure due to high pulmonary blood pressure or Eisenmenger syndrome. We excluded those patients in the direct comparison between treatment with or without closure well aware that these complications may possibly be a consequence of the ASD. Some of the differences in baseline comorbidity could be caused by the ASD, such as atrial fibrillation, stroke, and pulmonary heart disease. We excluded these diagnoses from the Charlson Comorbidity Index to avoid adjustment for intermediate steps between ASD and mortality that are potentially influenced by ASD closure. Hence, the present outcomes could result from a difference in ASD-related comorbidity, which was consequently not adjusted for, that is atrial fibrillation, stroke, and pulmonary infections, as these are intermediate steps between ASD and mortality. In Denmark, patients with no symptoms and no right ventricular enlargement or a Qp/Qs ratio smaller than 1.5 are typically not offered closure. Those patients without closure are anticipated to have a completely normal life expectancy compared with the comparison cohort. However, with time the magnitude and importance of a shunt may increase due to age-related changes in the circulation and we may need to revisit the indication for closure in those patients as they age. This is supported by the fact that cardiac-related mortality is very high in this group. We do not know the number of patients with large shunts who have not been offered closure due to comorbidity or who declined further treatment. Further knowledge regarding survival and changes in indication of closure with increasing age of those with small and haemodynamically insignificant ASDs is warranted. Adult ASD patients without closure could possibly benefit from a medical re-examination including symptoms, shunt, and right ventricle dimensions. In patients without closure, it is also important to remember that there is a proportion of patients with a small defect, that are not yet diagnosed. These patients might not have had reason to consult the health care system, and their morbidity and mortality could possibly be lower than in patients diagnosed with a small ASD.

Closure of an ASD before the age of 18 years does not seem to normalize the mortality risk compared to the background population, although the increased risk does not reach significance. We know that the risk of arrhythmia, stroke, and pneumonia starts to increase after the 4th or 5th decade., Most of the patients in our study and other similar research with closure in childhood have not reached this age yet. We can therefore only speculate about the possible increase in mortality risk in these patients in the years to come.

Limitations

Certain limitations should be kept in mind while interpreting the registry-based data in this study. Information on the type of ASD and the shunt dimensions is not available in these registries, and the results of this study must be interpreted for the entire ASD population. Misclassification is possible although all hospital records were validated for the correct diagnosis, status, and date of closure. Losses to follow-up could lead to selection bias if those lost to follow-up differ in outcome to those with full follow-up. The five patients with no CPR number are likely to have died before receiving a CPR number and thereby possibly caused an underestimation of the study results. However, the number of patients with no CPR number was small, and with the almost complete follow-up of the DNPR and Central Person Registry, bias due to loss of follow-up is not relevant in this study. The registration of the death dates in the Civil Registration System is considered to be practically without error. Selection bias could be of importance in patients with an unclosed ASD. The majority of patients with an unclosed ASD have no symptoms and the diagnosis is therefore often given in relation to other hospital admittances with possibly significant disease. However, we have minimized the influence of comorbidities not related to the ASD by adjusting for the Charlson Comorbidity Index. Baseline comorbidity is not included in patients diagnosed before 1977 and their comparison cohort. This could underestimate the difference in baseline comorbidity between these two groups. Later occurrence of comorbidity used in the Charlson Comorbidity Index is; however, complete for all groups.

The causes of death registered in The Danish Register of Causes of Death are only as precise as what the doctor interpreting and entering the cause of death writes on the death certificate. In cases where no autopsy was performed, the cause can be registered as most likely, but ailments such as ASD, Downs, etc. are obviously not valid death causes and are likely to cover unknown causes of death.

Conclusions

Patients with an ASD have increased mortality compared with the general population regardless of closure in childhood or adulthood. Closure—especially in childhood seems to improve survival in ASD patients. Our data indicate the importance of long-term medical follow-up of ASD patients, even with simple small defects. Adult patients with an unclosed ASD may need a re-evaluation 5–10 years after diagnosis, since symptoms, shunting, impact on cardiac function, and indication for closure may change over time.