Moderate Kidney Dysfunction Independently Increases Sudden Cardiac Arrest Risk: A Community-Based Study

Data availability statement

The data in this manuscript are from an ongoing study, and there is currently no IRB-approved mechanism by which this data will be deposited in a public repository. All analytical methods are included in this published article. De-identified participant data will be made available after publication upon reasonable request to the corresponding author, following approval of a proposal and a signed data use agreement.

Study population

Study participants were selected from the Oregon Sudden Unexpected Death Study (ORSUDS), an ongoing prospective, population-based study that enrolls individuals experiencing out-of-hospital SCA within the Portland, Oregon metropolitan area (catchment population approximately 1 million). Cases were identified through multiple sources, including first responders (Portland Fire Department and local ambulance services), the county medical examiner’s office, and emergency departments of participating hospitals. The study methods have been detailed in prior publications.¹⁶ Briefly, after receiving the SCA subject’s information, a comprehensive review of each case was conducted, encompassing circumstances of the event, clinical history, and, when available, autopsy findings. Cases of SCA with a likely cardiac etiology were determined through an adjudication process involving three physicians. SCA was defined as a sudden, pulseless collapse due to a likely cardiac etiology, occurring rapidly after symptom onset when witnessed or if unwitnessed, within 24 h of the subject being last seen in their usual state of health.¹⁷ Cases with non-cardiac etiologies, such as trauma, drowning, overdose/substance abuse, terminal illness, malignancies not in remission, or extracardiac causes (e.g., pulmonary embolism), were excluded.

Controls for ORSUDS were prospectively recruited from the same geographic region and time frame as the cases, ensuring a population with a medium-high risk profile, marked by a significant prevalence of coronary artery disease but no prior history of ventricular arrhythmias or sudden cardiac arrest. Participants were enrolled from several sources, including individuals transported by EMS for symptoms of acute coronary ischemia, patients undergoing angiography or attending outpatient cardiology clinics at a participating hospital, and members of a regional health maintenance organization.

Study design

For this study, we performed a case-control study in which all cases and controls from February 1, 2002, to December 31, 2020 were selected if they were aged 40 to 75 with detailed medical records available and serum creatinine levels measured (prior to arrest for cases). If subjects had multiple serum creatinine measures, the closest to the cardiac arrest event was selected for SCA cases. For controls, the most recently measured creatinine prior to study entry was selected.

The institutional review boards of Cedars-Sinai Medical Center, Oregon Health & Science University, and all relevant hospitals/health systems have approved the study protocol.

CKD was classified based on eGFR calculated using the CKD-EPI 2021 formula and the following categories: stage 1 CKD (chart history of kidney disease, with eGFR ≥ 90 mL/min/1.73m²), stage 2 CKD (eGFR < 90 and ≥ 60 mL/min/1.73m²), stage 3a CKD (eGFR < 60 and ≥ 45 mL/min/1.73m²), stage 3b CKD (eGFR < 45 and ≥ 30 mL/min/1.73m²), stage 4 CKD (eGFR < 30 and ≥ 15 mL/min/1.73m²), and stage 5 CKD or end-stage renal disease (ESRD) (eGFR < 15 mL/min/1.73m²).⁴ Moderate CKD was defined as including both stage 3a CKD and stage 3b CKD.⁴ Demographic information and comorbidities were obtained from medical records from collaborating hospitals.

External validation cohort selection

To validate our findings, we replicated the analysis using data from February 1, 2015, to February 1, 2023, from the ongoing Ventura PRESTO (Prediction of Sudden Death in Multi- Ethnic Communities) study, which covers a catchment population of ∼850,000. SCA cases in Ventura PRESTO were identified using methods similar to those employed in ORSUDS, including case ascertainment, adjudication, inclusion criteria, data retrieval, and definitions. The control group for Ventura PRESTO were patients from a regional health system (Cedars-Sinai Medical Center Medical Network, Los Angeles, CA) (n=4251) designed to represent members of the general population who obtain regular medical care. For the validation cohort, both cases and controls were selected using the criteria: individuals aged 40 to 75 years with detailed medical records and documented serum creatinine levels.

Statistical analysis

Data were presented as mean and standard deviation for continuous variables and as counts with percentages for categorical variables. Tests for differences between continuous variables was evaluated using the independent samples t-test, while categorical variables were analyzed using Pearson’s chi-square test. For continuous variables with a non-normal distribution, the Mann- Whitney U test was applied to assess association. Multivariable logistic regression analysis was conducted to calculate adjusted odds ratios (ORs) for moderate kidney dysfunction as a predictor of SCA. Covariates included in the regression model were demographic factors (age, race/ethnicity, and sex) and relevant risk factors or comorbidities associated with cardiac arrest and CKD: body mass index (BMI), diabetes, hypertension, chronic obstructive pulmonary disease (COPD), liver disease, congestive heart failure (CHF), coronary artery disease (CAD), and atrial fibrillation. We also calculated the adjusted ORs for each 1 mL/min/1.73m² and 10 mL/min/1.73m² decline in individuals with eGFR less or equal than 90 mL/min/1.73m² in discovery, validation, and pooled cohort, where the pooled cohort included all individuals from both the discovery and validation groups. Furthermore, the interaction between CKD severity and sex, age (≥ 65 vs <65), history of hypertension, diabetes, CAD, CHF in predicting SCA was also tested using the logistic regression model.

Subgroup analysis was conducted among SCA cases to highlight the relationship between CKD severity and resuscitation outcomes using multivariable regression model with covariates including demographic factors (age, race/ethnicity, and sex), history of hypertension, diabetes, CAD, CHF, witnesses, cardiopulmonary resuscitation by bystander, and initial rhythm recorded.

Initial rhythm was divided into non-shockable rhythms (pulseless electrical activity (PEA) and asystole) and shockable rhythms pulseless ventricular tachycardia and ventricular fibrillation (VT/VF). Particularly, because eGFR may vary with time, we performed a sensitivity analysis in the discovery cohort, limiting the analysis to individuals with creatinine levels measured within 90 days of the cardiac arrest event for SCA cases and within 90 days of the last physician visit for controls.

Finally, we conducted a subgroup analysis to highlight potential differences in the mechanisms linking varying severities of CKD and SCA. The cohort was stratified by CKD severity into mild/normal (eGFR ≥60 mL/min/1.73 m²), moderate (eGFR 30–59 mL/min/1.73 m²), and severe (eGFR <30 mL/min/1.73 m²) categories. We then compared clinical profiles and ECG characteristics between SCA cases and controls within each subgroup. All analyses were repeated for the PRESTO cohort to validate the findings. Statistical significance was determined at a two-sided p-value threshold of <0.05. All analyses were performed using IBM SPSS Statistics, version 24.

Study overview and baseline characteristics of individuals with SCA

From February 1, 2002, to December 31, 2020, in the ORSUDS study, 6,576 SCA cases were enrolled, of which 2,068 met inclusion criteria for this analysis. Additionally, 852 controls meeting the criteria were also included (Figure 1). The mean ages of SCA cases and controls were 61.4 and 62.7 years, respectively. SCA cases exhibited a higher prevalence of risk factors and comorbidities, except for CAD, compared to controls. These included COPD, hypertension, diabetes, CHF, and atrial fibrillation (Table 1).

• Download figure
• Open in new tab

Figure 1. Flow chart of discovery and validation study population selection.

The discovery population comprised 2,068 sudden cardiac arrest (SCA) cases and 852 non-SCA controls, aged 40 to 75 years, with detailed medical records and pre-SCA serum creatinine measurements from the Oregon Sudden Unexpected Death Study (ORSUDS). The validation population included 817 SCA cases and 3,249 non-SCA controls, aged 40 to 75 years, with detailed medical records and pre-SCA serum creatinine measurements from the Prediction of Sudden Death in Multi-Ethnic Communities (PRESTO) study in Ventura, California.

Association of moderate CKD and SCA in general population

A higher proportion of SCA cases had moderate and severe CKD (stage 4-5) compared to controls (17.7% and 15.4% vs. 14.7% and 2.6%, respectively; p < 0.001). Additionally, SCA cases exhibited a lower median eGFR than controls (74.7 vs. 80.9 mL/min/1.73m², p < 0.001) (Figure 2). Using individuals with normal kidney function (stage 1) or stage 2 CKD (eGFR ≥ 60 mL/min/1.73m²) as the reference group, the unadjusted odds ratios for SCA were 1.49 (95% CI: 1.19–1.87) for moderate CKD and 7.39 (95% CI: 4.76–11.50) for severe CKD (stages 4–5). In models adjusted for demographic factors and comorbidities, CKD remained independently associated with risk of SCA with ORs of 1.33 (95% CI: 1.03–1.72) and 5.68 (95% CI: 3.54– 9.14) for moderate and severe CKD (Table 2), respectively. The adjusted ORs for SCA associated with other comorbidities are also presented in Table 2. Furthermore, in study participants with an eGFR ≤ 90 mL/min/1.73m², each decline 10 mL/min/1.73m² was associated with a 24% (OR: 1.24, 95% CI: 1.18–1.31) increase in the odds of SCA, after adjusting for demographic factors and comorbidities (Figure 3). No interaction was found between CKD severity and sex, age (≥65 years vs. <65), or history of hypertension, diabetes, CAD, or CHF in predicting the risk of SCA (Table S1). Finally, a sensitivity analysis was performed on participants with creatinine measured within 90 days of the cardiac arrest event for SCA cases and within 90 days of the last physician visit for controls. This analysis included 681 SCA cases and 278 controls. Moderate CKD remained independently associated with SCA, with a crude OR of 1.68 (95% CI: 1.18–2.38) and an adjusted OR of 1.54 (95% CI: 1.03–2.29).

• Download figure
• Open in new tab

Figure 2. Estimated glomerular filtration rate of sudden cardiac arrest cases and controls in discovery and validation cohorts.

In the discovery cohort, SCA cases had a lower median estimated glomerular filtration rate (eGFR) compared to their non-SCA control counterparts (74.7 [48.9 – 94.4] vs. 80.9 [66.9 – 94.1], p<0.001). A similar finding was observed in the validation cohort, where SCA cases exhibited significantly lower eGFR than controls (64.5 [33.0 – 90.3] vs. 78.3 [65.9 – 90.7], p<0.001).

• Download figure
• Open in new tab

Figure 3. Odds ratio for risk of sudden cardiac arrest by estimated glomerular filtration rate (with ≥90 mL/min/1.73 m² as reference) in discovery, validation, and pooled population.

The risk of sudden cardiac arrest (SCA) increases as the estimated glomerular filtration rate (eGFR) decreases in both the discovery and validation cohorts, as well as in the pooled population.

Subgroup analysis

Among SCA cases with initial rhythm data available (n=2026), individuals with normal or mild CKD were less likely to present with an initial rhythm of PEA/Asystole compared to those with moderate and severe CKD (57.4% vs. 68.7% and 69.2%, p < 0.001). Moreover, they also demonstrated a trend towards higher survival-to-hospital-discharge proportion (14.8% vs. 11.9% and 10.4%, p = 0.06) (Table 3). After adjusting for demographic factors (age, race/ethnicity, and sex), history of CAD and CHF, whether the event was witnessed, bystander-provided cardiopulmonary resuscitation, and the initial recorded rhythm, moderate CKD appeared as an independent predictor of lower survival to hospital discharge, with an adjusted OR of 0.60 (95% CI: 0.39–0.92, p = 0.02) (Table S2).

Finally, while analyzing the differences in clinical profiles of SCA cases and controls across varying CKD severity groups, we found that COPD and BMI emerged as significant predictors of SCA in moderate CKD and normal/mild CKD but not severe CKD. The odds ratios for COPD were 3.62 (95% CI:2.61–5.02), 2.24 (95% CI: 1.19–4.24) vs. 0.93 (95% CI: 0.30–2.86), and for BMI, they were 1.014 (95% CI:1.001–1.026), 1.035 (95% CI:1.009–1.061) vs. 0.992 (95% CI:0.944–1.043) (Table S3) (Figure S1).

Regarding pre-arrest ECG parameters, which were available for 72.3% of individuals in the discovery cohort, SCA cases in both the normal/mild CKD and moderate CKD groups had significantly higher resting heart rates compared to their controls (80.3 ± 19.2 vs. 68.8 ± 14.1, p < 0.001, and 81.4 ± 19.2 vs. 69.8 ± 15.4, p < 0.001, respectively). No difference was observed in the severe CKD group (78.7 ± 17.4 vs. 78.9 ± 20.5, p = 0.96). Even after adjusting for demographics and comorbidities, resting heart rate remained a significant predictor of SCA in the normal/mild and moderate CKD groups (Table S3). Similar findings were confirmed in the validation cohort, PRESTO.

Validation analysis

There 817 SCA cases and 3,249 controls from the PRESTO study who met the inclusion criteria, with mean ages of 63.2 and 62.1 years, respectively. As in ORSUDS, SCA cases in PRESTO had a higher prevalence of most comorbidities evaluated compared to controls. Additionally, a greater proportion of SCA cases had moderate CKD compared to controls (22.6% vs. 14.9%, p < 0.001). Using eGFR ≥ 60 mL/min/1.73m² as the reference group, the crude ORs for SCA were 2.34 (95% CI: 1.92–2.85) for moderate CKD and 25.22 (95% CI: 17.94–35.45) for severe CKD (Table S4). After adjusting for demographic variables and comorbidities the adjusted ORs were 1.51 (95% CI: 1.17–1.97) for moderate CKD and 9.79 (95% CI: 6.33–15.12) for severe CKD (Table 2). Moreover, a decline in eGFR below 90 mL/min/1.73m² by each 1 mL/min/1.73m² and 10 mL/min/1.73m² was associated to an increased adjusted OR for SCA of 1.04 (95% CI: 1.03– 1.04) and 1.41 (95% CI: 1.31–1.50), respectively.

Among SCA cases, individuals with normal or mild CKD were less likely to present with an initial rhythm of PEA/Asystole compared to those with moderate and severe CKD (76.7% vs. 78.9% and 84.4%, p=0.09). Additionally, SCA cases with normal/mild CKD demonstrated a higher proportion of ROSC and survival to hospital discharge (STHD) than their moderate and severe CKD counterparts (Table 3). After adjusting for demographic factors (age, race/ethnicity, and sex), history of CAD, CHF, witnessed status, bystander-performed cardiopulmonary resuscitation, and the initial rhythm recorded, moderate CKD appeared as an independent predictor of failure to survive until hospital discharge, with an adjusted OR of 3.19 (95% CI: 1.13–9.09, p=0.03) (Table S2).

Strengths and Limitations

Our study possesses several strengths worth noting. Firstly, it is a community-based study encompassing two distinct populations: the ORSUDS cohort in Oregon, characterized by younger participants, a higher proportion of Caucasians, and a more prominent cardiac disease burden, and the PRESTO cohort in Ventura County, which included older participants, a higher representation of Hispanic individuals, and lower cardiac disease prevalence. Despite these demographic and clinical differences, most of our findings were consistent across cohorts.

Additionally, compared to other community-based studies, our work had a broader age range and racial/ethnic diversity, enhancing the generalizability of our findings.^7,14 Secondly, our study design integrated data from the EMS system and a strict and systematic adjudication of SCA, providing a more comprehensive assessment and robust evaluation of the burden of SCA among CKD patients in the community compared to prior studies relying solely on ICD codes for SCA definition. Finally, our analyses incorporated adjustments for various confounders previously demonstrated to be associated with CKD and SCA, strengthening the evidence for an independent association between CKD and SCA.

However, certain limitations should be acknowledged when interpreting our findings. First, the research is based on two large community-based observational studies. While we utilized multivariable regression models to reduce confounding bias, some residual confounding may persist. Secondly, the lack of alternative methods to estimate renal function, such as eGFR calculated using cystatin C or the urine albumin-to-creatinine ratio, limits our ability to explore other perspectives of renal function and albuminuria in relation to SCA. Finally, given the nature of out-of-hospital SCA, the initial rhythm data could be influenced by delays in EMS response, potentially underestimating the proportion of shockable rhythms and limiting our ability to characterize the true underlying rhythm at the time of arrest.