An invited commentary on ‘Clinical outcomes of early closure versus conservative diagnosis with atrial septal defect: a nationwide population-based cohort study’

We read with interest the article by Min et al. [1] in which they reported the improved clinical outcomes after early closure versus conservative management of newly diagnosed atrial septal defects (ASDs) in a national population-based cohort study utilizing the Korean National Health Insurance Service (NHIS) database. In Korea, NHIS enrolment is mandatory with 97% of the population being covered. Citizens can only avoid doing so if they are covered by another medical program or if they are temporary or illegal residents.

The NHIS gives an excellent population-based overview of diagnosis and medical treatments delivered to a stable population with ‘all patients being followed without censor except for lost eligibility due to death or emigration’, and ‘the rate of emigration from Korea is negligible’.

The authors identified 15 311 newly diagnosed isolated ASDs between 2002 and 2006, in patients without a history of atrial arrhythmias, ischaemic stroke or Eisenmenger’s syndrome. This is a much larger cohort than similar studies [2, 3].

For clinical reasons not identifiable in the NHIS, 1685patients underwent early closure (within 6 months of diagnosis). The early closure patients tended to be older than 20 years (66.4%), while the majority of conservatively managed patients were <20 years (76%). While guidelines such as the American Heart Association guidelines recommend ASD intervention based on symptoms, echocardiography and cardiac catherization [4], this information is not available that introduces a large possibility of selection bias. Patients undergoing early ASD closure may have been more haemodynamically compromised at diagnosis, prompting urgent intervention. This obliges a review of the take-home message and the overall validity of dividing patients into groups based on a 6-month timeframe post-diagnosis, as patients in the conservative group may not have reached the criteria warranting intervention.

Following propensity matching, 1644 patients were followed in each group for a median of 12.8–12.9 years. Interestingly, the authors included 150 patients in the conservatively managed group that had ‘late’ ASD closure after a mean of 2.4 years for surgery and 7.1 years for device closure.

Despite this study design, the authors identified a significantly higher all-cause mortality rate in patients over 40 years who were managed conservatively (HR, 0.55 [95% CI, 0.43–0.77]).

All-cause mortality is considered a useful clinical end-point in population-based studies, but it can be flawed [5]. The cause of death is not available from the NHIS. Therefore, it is difficult to ascertain if death was ASD related. For example, some studies report an improvement of all-cause mortality with surgical ASD closure, despite most patients dying of non-cardiovascular-related causes [6, 7].

Interestingly, the authors also attempted to compare the incidence of atrial arrhythmias between groups. As the diagnosis of atrial arrhythmia is only captured in the NHIS if patients present to hospital, the study has no mechanism of capturing or evaluating the effect of ASD closure or conservative management on the development of atrial arrhythmia outside the hospital setting, as distinct from postoperative atrial arrhythmia.

Similarly, community-based transient ischaemic attacks may not be captured by the NHIS while presumably stroke patients will present for treatment and are hence recorded.

The variable prescribing practices concerning antiplatelet or anticoagulants between groups at best shows that patients are more likely to be started on antiplatelets or anticoagulants over time if they are managed conservatively and less likely to be stopped irrelevant of the ASD being open or closed.

The retrospective design utilizing the NHIS as the sole data source introduces further limitations. The propensity score matching and posthoc sensitivity analysis implemented can adjust for the impact of potential baseline cofounders. While statistically sound given the information available, the NHIS does not provide data regarding patient symptoms, echocardiography or cardiac catherization.

The impact of the selection bias could have been reduced using subgroup analysis of the 150 conservative strategy patients eventually undergoing ASD closure. They could not be subsequently added to the early ASD closure group without introducing immortal-time bias. However, it is likely that the intervention impacted the survival of these patients, thereby skewing the overall primary end-point of the conservative strategy. This effect on all-cause mortality cannot be deemed negligible, as almost 10% of conservative strategy patients required ASD closure.

CONCLUSION

The authors are to be complimented on a real-world study that confirms a survival benefit of isolated atrial septal defect closure in adult patients over the age of 40 years, without a history of atrial arrhythmias, ischaemic stroke or Eisenmenger’s syndrome. However, many unmeasured confounders weaken the overall conclusions drawn by the authors.

REFERENCES