, Introduction
Chronic kidney disease (CKD) is a well-known risk factor for cardiovascular morbidity and mortality [1]. Furthermore, kidney function often worsens progressively in patients with CKD, with a significant portion of patients eventually requiring kidney replacement therapy, such as hemodialysis (HD). The dialysis population has a higher prevalence of traditional risk factors for cardiovascular disease (CVD) than the general population [2]. In addition, the unique characteristics of patients on dialysis, including volume overload, disturbed calcium and phosphate metabolism, and CKD itself, further increase the risk of CVD [3,4]. Therefore, current clinical guidelines consider CKD a high or very high risk factor for CVD [5].
Regarding CVD occurrence, the period of transition to HD initiation can be particularly challenging, with significant pathophysiological consequences for patients with CKD [6,7]. The early period following HD initiation has been reported to have higher rates of mortality and morbidity than subsequent periods [8–11]. However, to date, there is a scarcity of reports on the risk of CVD after HD initiation. This self-controlled case series (SCCS) study investigated the changes in CVD risk following HD initiation in patients with CKD using a nationwide healthcare claims database.
Methods
This study was approved by the Institutional Review Board of Severance Hospital, Yonsei University Health System (No. 9-2020-0148). Informed consent was waived because of the retrospective nature of this study based on the anonymous health insurance claim database.
Data source
This study used population-based health claims data provided by the Health Insurance Review and Assessment Service (HIRA) in Korea. Korea’s health insurance system, referred to as the National Health Insurance System, is a public and single payer system that is provided to all citizens residing in Korea [12,13]. HIRA is a government agency under the Ministry of Health and Welfare and is responsible for claims review and quality assessment of health care services [12,13]. Under the universal health insurance system in Korea, all clinics and hospitals submit claims to the HIRA for inpatient and outpatient care, including data on diagnoses as determined by the 10th revision of the International Classification of Diseases (ICD-10), medical procedures, prescription records, and demographic information [12,13].
Study design and case identification
We conducted a SCCS study to investigate the risk of CVD after HD initiation in patients with CKD. The SCCS is a case-only design and an epidemiological study type that utilizes individuals as their own controls, comparing the incidence of events during different exposure periods for the same individual [14]. This approach is particularly useful for evaluating the association between a transient exposure, such as HD initiation, and an acute event, such as CVD [14].
In the current study, patients with CKD on HD between 2007 and 2019 were selected from the HIRA database. Patients with CKD were identified by the presence of the ICD-10 code ‘N18’ either in the outpatient clinic, or in hospitalization. Patients treated with HD were defined by procedure codes for HD (O7020-1, O9991, O7031-4, or O7051-4) [15,16]. To define patients with CKD who newly initiated HD, we screened only patients whose diagnostic code of CKD existed at least 6 months prior to the earliest claim for HD. The date of the earliest claim for procedural codes for HD was defined as the index date (date of HD initiation). Patients who underwent HD or had CVD in 2007 (the wash-out period) were excluded. Consequently, all study patients had at least an 1-year washout period free of HD and CVD. Patients who underwent kidney transplantation or extracorporeal membrane oxygenation before HD initiation, aged <20 years or those with insufficient data for analysis were excluded. Patients who had both HD initiation and CVD on the same day were excluded, as we could not determine whether HD occurred before or after CVD [17]. This study included 74,584 patients with underlying CKD who were newly initiated HD between 2008 and 2019. As the SCCS design is a case-only method, 12,875 patients who had their incident CVD occurrence between 2008 and 2020 and fulfilled our criteria were included in this SCCS study (Supplementary Fig. 1, available online). We assessed the presence of comorbidities such as hypertension, diabetes mellitus, atrial fibrillation, heart failure, and malignancy. The relevant definitions are detailed in the Supplementary Methods (available online).
Definition of event outcome
The primary composite outcome, CVD, was defined as the composite development of acute stroke or myocardial infarction (MI) between 2008 and 2020. Secondary outcomes were defined as individual outcomes of CVD, including ischemic stroke, hemorrhagic stroke, and MI. The development of stroke was determined by admission with an accordant primary diagnostic code (ischemic stroke [I63] or hemorrhagic stroke [I60–62]) and an imaging study of brain computed tomography or magnetic resonance imaging during admission [18]. The development of acute MI was determined by admission with a primary diagnosis I21 code. The accuracy of diagnosing stroke or MI based on health claims data in Korea has been reported to be high [19,20]. Considering that the SCCS design assumes that each outcome event should be independent of the others and that the development of CVD could trigger subsequent events [21], we only identified the first CVD during the observation period per patient in the analysis.
Statistical analyses
We examined the changes in CVD risk during the transition period of HD initiation among patients with incident CVD occurrence between 2008 and 2020. Specifically, we calculated the proportion of patients with primary and secondary outcomes within the 12 months before and after HD initiation.
To assess the risk of CVD following HD initiation, we calculated the age-adjusted incidence rate ratio (IRR) for CVD during different time periods over the HD initiation period using the SCCS method. We operationally defined the risk periods based on the index date (day 0) of HD initiation. The pre-dialysis periods were defined as –60 to –31 days, –30 to –11 days, and –10 to –1 days, whereas the post-dialysis periods were defined as +1 to +10 days, +11 to +30 days, +31 and +60 days, and +61 to +150 days. The remaining period was treated as the baseline period (<–60 days, or >+150 days). The risk and control periods were divided based on the observed increased risk for CVD in Supplementary Table 1 (available online) and a previous study [22]. We used a conditional Poisson regression model to calculate the age-adjusted IRRs for CVD during each risk period and compare them to the baseline period. Fig. 1 illustrates the SCCS design and risk periods according to the HD initiation date. We also analyzed the age-adjusted IRRs for the secondary outcomes and conducted subgroup analyses according to age, sex, diabetes, and heart failure.
We conducted several sensitivity analyses to check the validity of the study findings, as well as the assumptions of the SCCS design, including the assumption that the observation periods are independent of events. Instead of using a time period of 30 days, we performed an SCCS analysis using a shorter time period of 15 days. Considering that CVD events may be fatal and reduce observation periods, we performed an additional analysis using a modified SCCS model for the event-dependent observation period [23,24]. Another key assumption of the SCCS design is that events do not affect exposure [14]. In the acute phase of CVD, it is possible that kidney function may deteriorate further and require HD. To assess if our results violated the mentioned assumption, we fitted the modified SCCS model assuming the presence of event-dependent exposure [23,25]. We further conducted a sensitivity analysis using another type of case-only method, a case-crossover design, which does not require the assumption that events do not affect exposure, for investigating the possible associations between transient exposure during a risk period and the development of an acute event [26]. In the case-crossover design, we selected case patients in whom HD was initiated within 180 days prior to the development of CVD. We compared the HD initiation immediately before the occurrence of CVD (hazard period, 1–30 days before CVD) to that at a different time (control period, 31–180 days before CVD) in each participant and estimated the odds ratio using conditional logistic regression. All data processing and statistical analyses were performed using SAS software (version 9.4.2; SAS Institute) and R software (version 3.5.1; The R Foundation for Statistical Computing) [23]. Two-sided p-values of <0.05 were considered statistically significant.
Results
Cardiovascular events during the 12 months before and after hemodialysis
Of the 253,894 patients with CKD who underwent HD between 2007 and 2019, 74,584 were newly initiated on HD between 2008 and 2019. Among these, 12,875 patients experienced CVD events, including stroke or MI, between 2008 and 2020 (Supplementary Fig. 1, available online). Patients with CVD were older, more likely to be male, and had hypertension, diabetes, atrial fibrillation, and heart failure (Supplementary Table 2, available online).
Regarding the variations in CVD risk per month during the 12 months before and after HD initiation, CVD occurred more frequently around the start of HD (Fig. 2A; Supplementary Table 1, available online). All secondary outcomes demonstrated similar trends, with MI being the most frequent event (Fig. 2B).
Self-controlled case series for the primary outcome
A total of 12,875 patients were finally included in the SCCS analysis: 6,367 patients with ischemic stroke, 2,396 patients with hemorrhagic stroke, and 4,112 patients with MI (Supplementary Fig. 1, available online). The patients enrolled had a mean age of 66.4 ± 12.0 years, and 8,224 of them (63.9%) were male (Table 1).
We compared the relative incidence of the primary outcomes in the seven risk periods with that in the control period. When focusing only on the post-dialysis period, we found that the risk of CVD was highest (approximately three times higher) during the 10 days following HD initiation than in the control period (age-adjusted IRR, 2.95; 95% CI, 2.44–3.56; p < 0.001) (Table 2) and that the risk gradually decreased over time.
The risk of CVD after HD initiation consistently increased across all subgroups, regardless of age (≥65 years), sex, or history of diabetes or heart failure (Supplementary Table 3, available online). The risk remained elevated throughout the entire post-dialysis risk period, with the highest value observed immediately after exposure and gradually decreasing over time.
Secondary outcomes and sensitivity analyses
Regarding secondary outcomes and focusing only on the post-dialysis period, the risk of MI was found to be highest during the first 10 days after HD initiation (age-adjusted IRR, 6.36; 95% CI, 5.00–7.91; p < 0.001) (Fig. 3). The risks of MI and ischemic stroke increased until 2 and 1 months after HD initiation, respectively, whereas the risk of hemorrhagic stroke remained relatively stable for up to 5 months following HD initiation (Fig. 3).
We conducted several sensitivity analyses to verify the robustness of the study findings and address the potential bias resulting from violations of the assumptions of the SCCS model. These included an SCCS model using a time period of 15 days (Supplementary Table 4, available online) and modified SCCS models with event-dependent observation periods and event-dependent exposure (Supplementary Table 5, available online). All analyses consistently demonstrated an increased CVD risk during the post-dialysis risk period (Supplementary Tables 4, 5; available online). Additionally, we performed a case-crossover analysis with 873 selected patients in whom HD was initiated within 6 months before the occurrence of CVD to confirm the temporal relationship. There was a higher likelihood of CVD in the week after HD initiation than during the control period (odds ratio, 1.96; 95% CI, 1.69–2.27; p < 0.001).
Discussion
This study demonstrated that patients with CKD on incident dialysis are at an increased risk of CVD around the time of HD initiation, with the composite outcome of CVD (acute stroke or MI) being the highest (up to three times) within 10 days after exposure compared to the control period and decreasing over time. MI was the most frequent CVD occurrence. Although the risk of ischemic stroke and MI decreased over time, it remained significantly higher for 1 to 2 months after HD initiation. An increased risk of hemorrhagic stroke was observed for up to 5 months. The subgroup and sensitivity analyses confirmed that these findings were robust.
Previously, several studies have shown increased mortality after dialysis initiation in patients with CKD [10,11], particularly in the first several months, which have been identified as a high-risk period for CVD [8,9]. Of 6,308 patients in a European cohort of patients initiating HD, the risk of CVD events, including cardiovascular, cerebrovascular, and peripheral vascular events, was approximately five times higher during the first month after HD initiation compared to that in the second year, continuing to increase until the fifth month after starting HD [8]. Based on the Korean HIRA database, 37% of patients with HD in Korea experienced CVD during a median follow-up of 22 months, with the incidence of these events being the highest during the first year after initiation of HD and decreasing significantly in the following years [9]. Regarding stroke in incident dialysis patients, the incidence of stroke has been reported to be 2.4 to 20.9 per 100 patient-years during the first 2 to 3 years [27,28]. Our study confirmed that the risk of CVD was highest in the period following HD initiation, decreasing over the next 5 months after HD initiation. Considering the increased risk of CVD during the peri-HD period, clinicians should carefully observe their patients for CVD until 5 months after HD initiation.
In patients with CKD on HD, the risk of CVD including stroke and MI known to be increased [29–31]. This increased risk of CVD in patients with CKD can be attributed to various factors, including traditional risk factors, such as hypertension and diabetes [2], as well as nontraditional uremia-related risk factors, including left ventricular hypertrophy, myocardial fibrosis, cardiac arrhythmia, and disturbed calcium phosphate metabolism [3,4]. With CKD stage progression, the risk of CVD may increase [32,33]. In our analysis, the incidence of CVD was higher several weeks before HD initiation, suggesting that patients who were about to initiate HD were at the highest risk of CVD. Considering the detrimental effects of terminal-stage CKD on the cardiovascular system [32,33], HD should, in theory and practice, improve cardiovascular outcomes by correcting fluid overload and uremic accumulation [34]. During the natural course of CKD, the risk of CVD increased over time [32,33]; and then decreased to baseline 1 year after HD initiation, which may imply a protective effect of HD. However, HD may cause a transient increase in the risk of CVD, especially during the early initiation period (in this study, up to 5 months after HD initiation). The HD procedure itself may also represent a risk for CVD [35]. HD has been reported to induce myocardial stunning, ventricular dysfunction, vascular changes, and endothelial dysfunction [3,35–37]. Furthermore, the use of heparin during HD has been linked to an increased risk of hemorrhagic stroke and endothelial dysfunction, further contributing to the development of CVD [37]. After this early adaptation period, HD may have a long-term protective effect for patients with CKD as mentioned above; therefore, it is recommended to carefully manage patients during the risky period of the first 5 months after HD initiation.
Our study also showed that different CVD events had different trends after HD initiation. The most common CVD event in the study population, MI, had an increased risk for up to 2 months after HD initiation, returning to baseline thereafter. In line with our findings, a European cohort was found to have a risk of coronary events significantly higher than that of cerebrovascular events, and this increased risk was maintained for up to 2 months after HD initiation [8]. The risk of ischemic stroke showed a similar trend. A study conducted using the United States Renal Data System database with 20,969 patients undergoing incident HD further supports our results, showing that stroke rates peaked at 0.7% to 1.5% patients per month during the 30 days after HD initiation and rapidly declined by 1 to 2 months after initiation [22]. Several reasons might be behind the stabilization of the risk for MI and ischemic stroke. For example, volume overload at the beginning of dialysis could be responsible for early CVD, whereas a decrease in the risk of CVD could be explained by the partial restoration or adaptation of the cardiovascular system in response to volume or electrolyte shifts after several months of dialysis. The rapid stabilization of the risk of stroke may derive from the partial recovery of cerebrovascular autoregulation resulting from acute volume and electrolyte shifts after the first month of dialysis [22].
In this study, the risk of hemorrhagic stroke remained consistently elevated throughout the 5-month risk period following HD initiation. A nationwide cohort study in Taiwan showed higher rates of hospitalization for both ischemic and hemorrhagic stroke in dialysis patients when compared to the general population [38]. End-stage kidney disease is associated with several conditions associated with an increased risk of hemorrhagic stroke, such as uncontrolled hypertension, bleeding diathesis, and blood vessel wall fragility [35,39]. Additionally, CKD was found to be associated with cerebral small vessel diseases, including cerebral microbleeds, which is an independent risk factor for hemorrhagic stroke [40]. Considering that the CKD stage is linked to the burden of cerebral microbleeds [40] and that heparin is commonly used during HD [37], the prolonged risk of hemorrhagic stroke after HD initiation may be attributed to these factors.
Our study has several strengths. First, we comprehensively evaluated the risk of CVD in patients undergoing incident HD using a large population derived from nationwide health claims-based data, which could minimize potential selection bias and provide information from a real-world setting. Additionally, the primary composite outcomes of this study included stroke and MI, with stroke divided into two subtypes (hemorrhagic and ischemic stroke). Our secondary outcome analysis revealed the differential risks of MI, ischemic stroke, and hemorrhagic stroke after incident HD. Finally, several sensitivity analyses were conducted to validate the robustness of our results, consistently demonstrating an increased CVD risk during the post-dialysis period.
This study had some inherent limitations. First, the study population may include patients undergoing transient HD who have recovered from acute kidney injury. To reduce such cases, we have only included patients with a diagnostic code of CKD at least 6 months before the first HD claim. However, there is still an inherent limitation in operationally distinguishing such cases based on electronic records. Second, the findings may not be generalizable because they are solely based on a Korean cohort. Third, the HIRA database does not provide laboratory values. This precludes determining the stroke etiology, such as large-artery atherosclerosis, cardioembolism, or small-vessel occlusion. Despite these limitations, our study uncovered the temporal changes in CVD risk after HD initiation using a nationwide, population-based cohort that included nearly the entire population of patients undergoing incident HD.
This study showed that patients with CKD in whom HD is initiated are at an elevated risk of developing CVD. This risk significantly increases shortly after HD initiation and gradually decreases over time. Therefore, it is essential to conduct thorough risk assessments and implement effective CVD prevention strategies for patients with CKD who are to begin HD. Given the increased risks of ischemic events and hemorrhagic stroke during the first 2 and 5 months, respectively, careful monitoring during this period is necessary.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement 1
Print
