Introduction

In 2006, the incidence of cardiac arrest in South Korea was 19,480, with a survival rate of 2.3%.1,2 The prognosis of this condition has gradually improved over time. In 2016, the survival rate increased to 7.6%, with 4.2% of patients showing good neurological outcomes. In 2018, the survival to discharge rate was 8.6% among 30,539 annual cardiac arrest cases, with 5.1% showing good neurological outcomes1,2. The overall survival rate of patients following cardiac arrest in other nations, including European countries, the United States, and Japan, was approximately 7–10%, which is similar to that observed in South Korea3,4,5. Owing to the increased number of successful cardiac resuscitations (SCRs), interest in patient outcomes has surged, triggering a significant increase in related research. Furthermore, owing to the limited number of patients who undergo SCR and the associated high mortality rate, many studies have focused on short-term prognosis, following up patients for 1 month or < 1 year, and reporting indices such as mortality, cerebral performance category (CPC) scale, and clinical characteristics6,7,8,9,10,11. Furthermore, studies in Thailand and Switzerland identified predictive factors for outcomes in patients who underwent SCR, with follow-up focusing on short-term prognosis assessments over a 1-year period12,13. Xiao et al.14 previously reported rehospitalization and mortality rates (38%) in 190 patients who underwent SCR and experienced severe functional neurological impairment (modified Rankin Scale [mRS] = 5) after 1 year. Using clinical features, No et al.15 evaluated the survival to discharge rates of patients admitted to the intensive care unit after cardiopulmonary resuscitation. Several studies have predicted the short-term prognosis of patients with SCR using various electrophysiological tests16,17,18,19,20,21, including neurological examinations, electroencephalography (EEG), somatosensory evoked potentials, neuron-specific enolase, and brain imaging to assess short-term outcomes (1 month to 1 year) using the CPC scale or mRS16,17,18,19,20,21. One study quantified and analyzed the EEG results of patients with ischemic brain injury and assessed their neurological prognosis after 1 month22. Moreover, one study evaluated the prognosis of patients who underwent SCR using brain biomarkers (neuron-specific enolase, S100 calcium-binding protein ꞵ, glial fibrillary acidic protein, tau, and neurofilament light) and reported an association between neurofilament light and poor neurological prognosis23. However, owing to the high mortality rate and challenges of long-term follow-up, few studies have examined the long-term prognosis (≥ 2 years) of patients who underwent SCR. In particular, data regarding the development of other brain disorders are lacking.

Ischemic hypoxia can cause up to 95% brain damage within the first 15 min of cardiac arrest24, to which the hippocampus, thalamus, basal ganglia, cerebellum, and cerebral cortex are particularly susceptible25,26. These brain structures are associated with brain disorders, such as stroke, epilepsy, Alzheimer’s disease, and Parkinson’s disease; thus, ischemic brain injury could potentially increase the incidence of these diseases.

In the present study, we performed a big data analysis based on the hypothesis that patients who undergo SCR would have a higher incidence of various brain disorders in the form of delayed neurological impairment than do controls. We aimed to estimate the incidence of brain disorders in this patient group and explore the associations between various confounding factors.

Results

Comparison of baseline clinical characteristics

Of the 4,014,168 patients in the study population, 572 and 1,144 were classified in the SCR and non-SCR groups, respectively (Fig. 1). Table 1 presents the baseline clinical characteristics of the patients. No significant between-group differences were observed in any demographic or comorbidity variables. All characteristics had small SMDs; therefore, the control group (non-SCR) was obtained through appropriate PS matching using these statistics.

Fig. 1

Study flowchart.NHIS, national health insurance service; PS, propensity score.

Brain disorders were present in 103 (18.0%) patients with a history of SCR and 119 (10.4%) patients without a history of SCR (Table 1). The cumulative incidence curve shows a qualitative comparison of the incidence of brain disorders over 36 months between the two groups (Fig. 2). The data showed that SCR was achieved in 21 of 1,000 patients within 1 year of onset, whereas it was achieved in 11.3 of 1000 patients in the control group without a history of SCR but with similar characteristics. Additionally, the adjusted HR value was significant at 1.82, showing that the risk of brain disorders was 1.82 times higher in the SCR group than in the non-SCR group (adjusted HR, 1.82; 95% CI, 1.40–2.37) (Supplementary Table 1; Fig. 2).

Fig. 2

Overall cumulative incidence curve of brain disorders.

Subgroup analysis

In the subgroup analysis, the risks in the SCR group were calculated for stroke, epilepsy, Alzheimer’s disease, and Parkinson’s disease (Fig. 3). The stroke incidence rate in the SCR group (7.27) was significantly higher than that in the control group (4.36) (adjusted HR, 1.68; 95% CI, 1.09–2.61) (Supplementary Table 2; Fig. 3A). Additionally, the epilepsy incidence rate in the SCR group (10.49) was significantly higher than that in the control group (4.34) (adjusted HR, 2.37; 95% CI, 1.58–3.53) (Supplementary Table 3; Fig. 3B). Furthermore, the incidence rate of Alzheimer’s disease in the SCR group (7.15) was significantly higher than that in the control group (4.49) (adjusted HR: 1.56; 95% CI, 1.02–2.40) (Supplementary Table 4; Fig. 3C). However, the incidence rate of Parkinson’s disease in the SCR group had an insignificant adjusted HR of 1.35 (95% CI, 0.48–3.78), owing to a relatively small number of patients with Parkinson’s disease (Supplementary Table 5; Fig. 3D).

Fig. 3

Cumulative incidence rates for each brain disorder. (a) Stroke. (b) Epilepsy. (c) Alzheimer’s disease. (d) Parkinson’s disease.

A forest plot revealed that the cumulative HRs were significant in patients aged ≥ 60 years and in males (Fig. 4). The HR for patients aged ≥ 60 years compared to those aged 40–59 years was 1.8 (95% CI, 1.35–2.40) and the HR for men compared to women was 0.59 (95% CI, 0.45–0.79). There were more females aged ≥ 60 years than males of the same age group (Supplementary Table 6). Among the comorbidity variables, the HR of underlying DM was 1.50, which was statistically significant (95% CI, 1.12–1.71) (Fig. 4).

Fig. 4

Forest plot of brain disorders in successful cardiac resuscitation.CKD, chronic kidney disease; DM, diabetes mellitus; HTN, hypertension.

Discussion

In this study, we examined the long-term (3 year) prognosis of patients with ischemic brain injury during temporary cessation of systemic circulation. The results revealed that the risk of developing brain disorders such as stroke, epilepsy, and Alzheimer’s disease was higher in the SCR group than in the control group. Among the brain structures, the hippocampus, cerebral cortex, cerebellum, thalamus, basal ganglia, and hypothalamus are particularly vulnerable to cerebral ischemia25,26 because of their high oxygen and energy demands. These structures are predominantly composed of grey matter, which has a greater requirement for oxygen than the white matter and approximately four times the blood flow (80–100 mL/100 g/min vs. 20–25 mL/100 g/min)27,28. In addition, the glutamate receptors on these structures are more intricately connected to neurons and more susceptible to glutamate-induced excitotoxic processes29,30. Thus, the neurons are damaged more easily during cerebral ischemia and more prone to necrosis than glial cells, astrocytes, and oligodendrocytes29,30. Furthermore, the low density of excitatory amino acid transporter proteins that remove glutamate from synapses can contribute to increased susceptibility to ischemic brain injury30.

In our study, the HR for stroke in the SCR group was 1.68 times (95% CI, 1.09–2.61) higher than that in the control group over the 3-year follow-up period. Reperfusion following cardiac resuscitation can lead to microcirculatory perturbation-associated neuronal dysfunction, which can cause cerebral hyperemia, ultimately resulting in hypoperfusion and no reflow31. Related mechanisms include impaired vasomotor regulation due to pericyte death or vasoconstriction owing to decreased nitric oxide production32. Endothelial function of the cerebral vasculature may be impaired during this process. The cerebrovascular endothelium plays an important role in maintaining the blood–brain barrier, regulating microcirculatory blood flow, and secreting auto-anticoagulants and mediators33. Dysfunction of endothelial auto-anticoagulants can lead to widespread microthrombi in the cerebral vasculature. Concomitant impaired vasodilation can increase the risk of stroke by increasing cerebrovascular resistance and decreasing cerebral blood flow33. In addition, reperfusion can trigger cerebrovascular injury by forming free radicals, producing glutamate, and driving intracellular calcium accumulation33.

Complications of ischemic brain damage can occur after several weeks or months and have been reported as delayed post-hypoxic leukoencephalopathy34. Although the exact mechanism is unknown, it is believed to be associated with demyelination throughout the brain35. Arylsulfatase A levels are reduced by 50% in affected individuals, suggesting a potential association with delayed post-hypoxic leukoencephalopathy35. to an animal study, selective neuronal damage and death over time is an alternative mechanism of delayed hypoxic brain injury because microglial activation begins with deoxyribonucleic acid degradation36.

The HR for epilepsy in the SCR group was 2.37 times (95% CI, 1.58–3.53) higher than that in the control group. Recurrent seizures and excessive glutamate secretion underlie the formation of the cerebral ischemia-induced epileptogenic zone37,38. Among various neurotransmitters, glutamate plays a crucial role in the initiation and propagation of epileptic seizures.

Abnormal amounts of glutamate secreted by astrocytes simultaneously activate numerous neurons, triggering seizures39,40. Furthermore, glutamate transport dysfunction contributes to elevated extracellular glutamate levels in the epileptogenic hippocampus. Repeated and intense seizures result in excessive glutamate secretion and overstimulation of glutamate receptors, which release large amounts of calcium into cells, resulting in neuronal cell death38. In the present study, epilepsy had the highest incidence (27 cases) among brain disorders within the first 3 months after SCR, indicating acute brain injury. In addition, extensive neuronal death can lead to abnormal reorganization of synaptic plasticity and neuronal circuity, changes in interneuron number and function, and dentate neurogenesis abnormalities38.

In the present study, the HR for Alzheimer’s disease in the SCR group was 1.56 times (95% CI: 1.02–2.40) higher than that in the control group. Although the mechanism underlying ischemic brain injury in neurodegenerative brain disorders remains unknown, it accelerates dementia onset by 10 years41. However, several hypotheses have proposed that ischemic brain injury may cause extensive neuronal death in the hippocampus and cerebral cortex, along with the accumulation of diffuse senile plaques and elevated tau protein levels42,43. In addition, ischemic brain injury can induce hyperphosphorylation, leading to the formation of paired helical filaments and neurofibrillary tangles44,45. Another mechanism involves impaired glutamate transport function, which causes cell death due to poor uptake. Animal studies have shown delayed (45–90 d) neuronal degeneration and death in rats following the prolonged administration of glutamate uptake inhibitors, which could be prevented by the administration of glutamic receptor antagonists46,47. Collapse of the blood–brain barrier due to ischemic brain injury may also play a role48,49,50, as it allows toxic substances such as amyloid and tau proteins to accumulate in the central nervous system, thus contributing to progressive neurodegenerative changes48,49,50. Blood–brain barrier dysfunction can cause inflammation in the central nervous system by enabling easy and abnormal passage of immune cells, facilitating the development of neurodegenerative diseases51. Furthermore, acetylcholine levels in the hippocampus decrease after ischemic brain injury52. Acetylcholine, a neurotransmitter, plays a crucial role in neuronal signaling and memory function; therefore, it is believed to contribute to the development of Alzheimer’s disease52. In addition, delayed neurological sequelae following ischemic brain injury may accelerate the progression of Alzheimer’s disease. Previous studies have reported cerebral cortex atrophy, hippocampal atrophy, and hydrocephalus in autopsies performed 1–2 years after ischemic brain injury53,54.

Glutamate plays a key role in Parkinson’s disease by acting as the primary excitatory transmitter found in the basal ganglia and glutamatergic projection areas, such as the striatum, subthalamic nucleus, and substantia nigra pars compacta in the brain55,56. The substantia nigra pars compacta performs important regulatory functions, as it receives glutamatergic innervation from the subthalamic nucleus, and it is connected to various nuclei in the basal ganglia circuit via dopaminergic projections56. Patients with Parkinson’s disease have less dopamine transmission to the striatum, and as a compensatory mechanism, neurons in the subthalamic nucleus are activated to stimulate dopamine release in the substantia nigra pars compacta. However, sustained glutamate release can cause excitotoxicity and neurodegeneration in dopaminergic neurons57. Given that glutamate is secreted in large amounts following ischemic brain injury, it may increase the long-term prevalence of Parkinson’s disease. In the present study, the HR for Parkinson’s disease in the SCR group was 1.35 times higher than that in the control group; however, the difference was not statistically significant (CI, 0.48–3.78). In addition, the small number of patients (n = 15) limited proper evaluation.

Hypoxic damage can be classified as hypoxemic hypoxia caused by a purely hypoxic event or ischemic hypoxia resulting from cardiovascular arrest. Hypoxemic hypoxia is characterized by low oxygen saturation and inadequate blood flow to the brain, and it is commonly caused by respiratory problems, such as pneumonia, asthma, or airway injury, rather than cardiac problems58. Hypoxemic hypoxia with adequate cerebrovascular circulation may not cause severe brain damage58. For example, primary respiratory arrest-associated hypoxemic hypoxia may cause temporary brain dysfunction but rarely results in permanent serious brain damage58. This is because when cerebral blood flow is maintained, even small amounts of oxygen are supplied and toxic substances from the brain are removed via blood circulation58. In the present study, the prevalence of brain disorders in the SCR group was higher than that in the control group, supporting the hypothesis that ischemic hypoxia-associated brain damage can result in serious outcomes.

This study had several limitations. First, although this study selected participants from a dataset of 4 million individuals, the final sample size was only 572 owing to the rare nature of SCR. Second, the follow-up period was limited to 3 years, and the study was retrospective in design. Third, the prevalence of the disease was determined using the codes and disease names of the Korean Standard Classification of Diseases. Third, we did not conduct latency-restricted or time-band analyses; thus, a portion of very-early post-index events may reflect peri-arrest brain injury rather than de novo chronic disease, despite our stringent 2-year washout to exclude prevalent disease. Therefore, this study may not have established causality and is limited in determining the exact relationship between brain disorders and patients with SCR.

Overall, this study provides insights into the potential brain disorders that may occur in patients with SCR, considering their long-term prognosis. Further studies with larger cohorts of patients with SCR are required to provide more robust evidence to predict the long-term prognosis and improve neurological outcomes in patients with SCR.

Methods

Data source

This study used data from the National Health Insurance Service (NHIS), a customized health information database. The National Health Insurance Corporation collects, manages, and maintains health information data available for policy and academic research purposes through this customized database. We randomly stratified and extracted data collected from 4,014,168 individuals between 2008 and 2020 from the 50 million patients in the NHIS database, considering information related to age, sex, economic status, and underlying diseases. Owing to the nature of the data, they could only be collected following approval from the Institutional Review Board; data analysis was performed in a room separately provided by the NHIS for security reasons. The NHIS serial number for this data is NHIS-2022–1–626. This study is a retrospective observational analysis using de-identified secondary data from the National Health Insurance Service (NHIS) database. There was no direct contact with the subjects, and no additional data collection was performed. As the data were already anonymized, the risk to participants was considered extremely low. Accordingly, the study was granted an exemption from ethical review and informed consent by the Institutional Review Board (IRB) of Jeonbuk National University Hospital. This exemption is documented in both the study protocol and the waiver request form, and the relevant information has been added to the Methods section of the manuscript. Data on the long-term prognoses of stroke, epilepsy, Alzheimer’s disease, and Parkinson’s disease in patients who received cardiac resuscitation and survived were compared with the individuals in the control group.

Patient definition

The patient group and target diseases must have operational definitions based on the characteristics of the health insurance data. The International Classification of Diseases, 10th revision (ICD-10) codes were mainly used for operational definitions. The definitions of the patient group, target diseases, and underlying diseases are presented below.

Brain disorders, the target diseases in this study, included stroke, epilepsy, Alzheimer’s disease, and Parkinson’s disease. For each individual, the occurrence of these four diseases within 3 years of follow-up was defined using the following ICD-10 codes: stroke (I61, I62, I63, I64), epilepsy (G40, G41), Alzheimer’s disease (F00, G30), and Parkinson’s disease (G20).

Patients who underwent SCR were defined as those with ICD-10 I460 or cardiac arrest ICD-10 I461 or I469. The following exclusion criteria were applied: 1) brain disorders during the 2-year washout period, 2) an additional history of SCR during the 2-year washout period, and 3) death within 3 years of follow-up after SCR. Patients who underwent SCR and were recruited between 2010 and 2017 with washout and follow-up periods of 2 and 3 years, respectively, were enrolled. Of the > 4 million people, 24,020 had a history of SCR. However, 22,537 patients were excluded based on the exclusion criteria; therefore, 572 patients were finally included in the SCR group (Fig. 1). To ensure incident outcomes, we applied a 2-year washout period prior to cohort entry and excluded anyone with pre-existing brain disorders during washout. Brain disorders during follow-up were defined a priori using ICD-10 codes within a fixed 3-year follow-up window.

For each patient, the five underlying diseases recorded during the washout period were defined as follows: hypertension (HTN) (I10, I15, and 65 antihypertensive drugs), diabetes mellitus (DM) (E10, E11, E12, E13, E14, and 116 diabetes medication), chronic kidney disease (CKD) (N03, N18, N19), cancer (any C code), and cardiovascular disease (I21, I22, I20, I470, I49, I43, and I50).

Patients in the control (non-SCR) group were matched with the study (SCR) group using propensity score (PS) matching considering eight variables (age, sex, economic status, HTN, DM, CKD, cancer, and cardiovascular disease), with a 1:2 greedy nearest neighbor algorithm. Differences between groups were computed on the logit of the propensity score and matches were selected without replacement, and caliper width (0.25) was applied. The appropriateness of PS matching was judged quantitatively using standardized mean differences (SMDs) and qualitatively using numerical values. An SMD < 0.1 indicated no difference between the two groups. Table 1 summarizes the PS matching results.

Hazard ratios (HRs) and 95% confidence intervals (CIs) for the four brain disorders were estimated using multivariate Cox proportional hazard models. After PS matching on 8 baseline variables, Cox models estimated hazard ratios with 95% CIs. Data preprocessing was performed using SAS, and all statistical analyses were performed using R.

Data availability

All data generated or analyzed during this study are included in this manuscript.

References

Yoon, H. et al. Main outcomes of the sudden cardiac arrest survey 2006 to 2016. Clin. Exp. Emerg. Med. 6, 183–188 (2019).

Article PubMed PubMed Central Google Scholar 1.

Korea Centers for Disease Control & Prevention. Cardiac arrest statistics. Act on the Prevention and Management of Cardio-Cerebrovascular Diseases. (Accessed 8 Nov 2023). 1.

Fink, E. L. et al. Unchanged pediatric out-of-hospital cardiac arrest incidence and survival rates with regional variation in North America. Resuscitation 107, 121–128 (2016).

Article PubMed PubMed Central Google Scholar 1.

Kitamura, T. et al. Nationwide improvements in survival from out-of-hospital cardiac arrest in Japan. Circulation 126, 2834–2843 (2012).

Article PubMed Google Scholar 1.

Straney, L. D. et al. Trends in PICU admission and survival rates in children in Australia and New Zealand following cardiac arrest. Pediatr. Crit. Care Med. 16, 613–620 (2015).

Article PubMed Google Scholar 1.

Thompson, L. E. et al. Long-term survival trends of Medicare patients after in-hospital cardiac arrest: Insights from Get With the Guidelines-Resuscitation®. Resuscitation 123, 58–64 (2018).

Article PubMed Google Scholar 1.

Widestedt, H., Giesecke, J., Karlsson, P. & Jakobsson, J. G. In-hospital cardiac arrest resuscitation performed by the hospital emergency team: A 6-S year retrospective register analysis at Danderyd University Hospital Sweden. F1000Res 7, 1013 (2018).

Article PubMed PubMed Central Google Scholar 1.

Moulaert, V. R. M. et al. Long-term outcome after survival of a cardiac arrest: A prospective longitudinal cohort study. Neurorehabil. Neural Repair 31, 530–539 (2017).

Article PubMed Google Scholar 1.

Peskine, A. et al. Long-term functional outcome of cardiac arrest survivors. Resuscitation 85, e65–e66 (2014).

Article CAS PubMed Google Scholar 1.

Chan, P. S. et al. Long-term outcomes in elderly survivors of in-hospital cardiac arrest. N. Engl. J. Med. 368, 1019–1026 (2013).

Article CAS PubMed PubMed Central Google Scholar 1.

Mateen, F. J. et al. Long-term cognitive outcomes following out-of-hospital cardiac arrest: A population-based study. Neurology 77, 1438–1445 (2011).

Article CAS PubMed Google Scholar 1.

Limpawattana, P. et al. Long-term outcomes and predictors of survival after cardiopulmonary resuscitation for in-hospital cardiac arrest in a tertiary care hospital in Thailand. Ther. Clin. Risk Manag. 14, 583–589 (2018).

Article PubMed PubMed Central Google Scholar 1.

Fuchs, A. et al. Survival and long-term outcomes following in-hospital cardiac arrest in a Swiss university hospital: A prospective observational study. Scand. J. Trauma Resusc. Emerg. Med. 29, 115 (2021).

Article PubMed PubMed Central [Google Scholar](http://scholar.google.com/scholar_lookup?&title=Survival%20and%20long-term%20outcomes%20following%20in-hospital%20cardiac%20arrest%20in%20a%20Swiss%20university%20hospital%3A%20A%20pros