Clinical Factors for Sudden Unexpected Death in Patients Undergoing Peritoneal Dialysis

Background: Sudden unexpected death (SUD) accounts for a signicant proportion of overall mortality in patients undergoing peritoneal dialysis (PD).This study aimed to investigate the SUD clinical proles in patients undergoing PD. Methods: Medical records from January 2009 to December 2018 were retrospectively reviewed in a hospital-facilitated PD center in Taiwan. Demographic data, laboratory parameters, comorbidities, drug history, physical performance status, cardiac function parameters, and peritoneal transport category were abstracted. Cox proportional hazard regression was used to determine hazard ratio(HR) in SUD clinical proles in patients undergoing PD. Results: Based on 28 patients undergoing PD with SUD, 60 controls were matched for date of death on a 2:1 ratio for comparison. The incidence of diabetes and the prevalence of low physical performance status, analgesic drug use, concordant comorbidity number, mental health and chronic pain were higher in patients undergoing PD with SUD than in controls. In the cardiac function analysis, the QTc interval on the electrocardiogram was longer (467 vs. 453 ms, P=0.010) in patients undergoing PD with SUD than in controls. At 3 months before death, patients who experienced SUD demonstrated progressive lower serum potassium concentrations. Cox proportional hazard regression analysis revealed that diabetes [adjusted HR, 4.73; 95% condence interval (CI), 2.75–8.14; P<0.001)] and mental health and chronic pain (adjusted HR, 2.07; 95% CI, 1.06–4.05; P=0.033) remain signicant predictors for SUD in patients undergoing PD. Conclusions: Diabetes mellitus, mental health and chronic pain, incidence of hypokalemia before death are signicant clinical factors for SUD in patients undergoing PD. Conceptualization, W.C.L. and J.B.C.; methodology, J.B.C.; validation, J.B.C., formal analysis, W.Y.Z., data curation, W.Y.Z. and T.W.H.; writing-original draft preparation, W.Y.Z.; writing-review and W.X.C. and

Recently, a few studies have performed survival analyses between HD and PD [1,[16][17][18][19][20][21]. Most of them focused on comparing the survival bene ts in a de ned period between HD and PD, with results presenting variable differences for survival bene ts. Recently, a multicenter retrospective cohort study showed that differences in the incidence of sudden death were not signi cant between patients undergoing HD and PD [1]. A case-control study in one center showed that recent blood transfusion, male sex, and diabetes mellitus increased the odds of sudden death in patients undergoing PD [22]. Moreover, investigators identi ed that other predictors were older age, ischemic heart disease, and decreased left ventricular ejection fraction (≤ 35%) [1]. However, whether additional clinical factors, e.g., demographic pro le, trajectory electrolyte, and cardiac function changes, could be considered as potential factors for sudden death in patients undergoing PD remains unclear.
Thus, this study aimed to examine the relationship between clinical factors and SUD in patients undergoing PD. The identi cation of potential risk factors for SUD in patients undergoing PD was via a wide-range clinical data collection.

Subjects
This retrospective study was conducted to collect clinical information on prevalence in patients undergoing PD in Kaohsiung Chang Gung Memorial Hospital in Taiwan using data from January 2009 to December 2018. Patients undergoing maintenance PD for >3 months and experienced SUD during the study period were enrolled in this study. For the control selection, two controls were extracted from the dataset matched for date of one death during the study period by computer-generated block randomization method. SUD De nition SUD was de ned as spontaneous death preceded by a sudden loss of consciousness within 1 h after the onset of acute symptoms, even in the presence of pre-existing heart disease, but with unexplained causes and unexpected timing and mode [23]. Con rmation of SUD was approved by medical professionals either in the hospital or the patients' residence.

Informative data collection
Various clinical data were collected, including demographics, laboratory parameters, comorbidities, cardiac function measurements, drug history, and physical functional performance. Laboratory parameters were retrospectively collected continuously three months before SUD, including parameters of hemogram, biochemistry, and intact parathyroid hormone. Cardiac function measurements consisted of electrocardiogram (ECG) and echocardiography performed at the nearest time of SUD occurrence. PR, QRS, and QT interval values in ECG were collected. QTc interval was measured using an ECG machine with Bazett's formula correction (QTc = QT/(RR) 1/2 ). Cardiac function was evaluated using Pulsed Doppler echocardiography. On the other hand, physical functional performance status was measured using the Karnofsky Performance Status scoring system [24].
Peritoneal equilibration test (PET) Standard PETs were performed at 6-month intervals. The last values of PET-related parameters were collected before SUD. Glucose load was calculated using the peritoneal glucose load index (PGLI), referred to as the net glucose content in PD solutions in the daily PD dwell divided by the body weight (kg).

Comorbidities
Previously validated algorithms were used for comorbidities based on claims data [25]. Brie y, comorbidities were categorized into concordant, discordant, and mental health/chronic pain. Comorbidity status was determined for an individual patient within each scal year, using data in the scal year before the index data as baseline.

Statistical analysis
Baseline characteristics, cardiac function parameters, PET, and laboratory measurements of patients and controls were summarized as frequencies (percentage), means (standard deviation), or medians (interquartile range). For categorical variables, differences between patients and controls were estimated using chi-squared or Fisher's exact test. For continuous variables, differences were estimated using the independent two-sample t-test. On the other hand, the difference of repeated laboratory measurements in each group was estimated using a one-way repeated analysis of variance (ANOVA) test, while a two-way repeated ANOVA was used for estimation of the difference of repeated laboratory measurements between patients and controls. Cox proportional hazard regression was used to determine the association between sudden death mortality and each included variable.

Cardiac function measurements of the entire cohort
In the ECG analysis, patients with SUD had longer QTc interval at baseline compared with the controls (467 ms vs 453 ms, P = 0.01). There were no signi cant differences in the left ventricular ejection fraction between SUD and controls ( Table 2). PET parameters PET parameters, including peritoneal transport category, adequacy indices, residual creatinine clearance, and PGLI, were not signi cantly different between patients with SUD and controls (Table 3). Longitudinal laboratory parameter changes at 3 months before SUD Patients with SUD demonstrated signi cantly lower serum K levels at three months before death (4.1 to 4.0 to 3.63 mEq/L; P < 0.001); however, these trend changes were not observed in controls. Considering between-group interactions and time, the difference in serum K trend changes was signi cant between patients with SUD and controls (P < 0.001) ( Table 4). After an adjusted analysis, diabetes (adjusted HR, 4.73; 95% CI, 2.75-8.14; P < 0.001) and mental health and chronic pain (adjusted HR, 2.07; 95% CI, 1.06-4.05; P = 0.033) remain as predictors for SUD in patients undergoing PD (Table 5).

Discussion
This study attempted to examine the clinical factors associated with SUD in patients undergoing PD. Using a long-term observational cohort, diabetes, K trajectory changes, and mental health and chronic pain were identi ed as risk factors for these patients. Given that majority of previous studies have focused on few parameters on sudden death in patients undergoing dialysis and rarely report on patients undergoing PD, more demographic data, clinical variables, and comorbidity discrimination were included in this study to examine their associations with SUD in patients undergoing PD. Our ndings suggest that a multifaceted approach is potentially bene cial for SUD prevention in patients undergoing PD.
When examining the associations of SUD with demographics and comorbidities, patients with and without SUD were found to have similarities in age, sex, PD modalities, and antihypertensive drug use.
Remarkably, patients with SUD had more concordant comorbidities at baseline compared with controls. These include some important cardiac comorbidities, e.g., hypertension, chronic heart failure and atrial brillation, diabetes, peripheral vascular disease, and stroke or transient ischemic attack [25]. All these comorbidities were observed in patients undergoing PD with SUD. In addition, majority of these patients demonstrated a prolonged QTc interval in this study. Given that previous studies reported that sudden cardiac death represents a major cause of mortality in patients with end-stage renal disease [1,9,11,26], our results implicated that a planned schedule for cardiac function examination was essential for patients undergoing long-term dialysis.
To date, no studies have reported the association between mental health and chronic pain comorbidities and SUD in patients undergoing PD. We found that patients with SUD had more mental health and chronic pain comorbidities and were more likely to use analgesic drugs, mainly for chronic osteoarticular complaints. Furthermore, by hazard analysis, mental health and chronic pain comorbidities were identi ed as signi cant risk factors for SUD in patients undergoing PD. In our cohort, the proportion of patients undergoing PD with SUD who used analgesics was higher than that of controls, i.e., 67.9% and 35.0%, respectively. Analgesic categories included nonsteroidal anti-in ammatory drugs (NSAIDs) and a combination of NSAIDs with weak opioids. NSAIDs have been known to be associated with several adverse effects, including gastrointestinal and cardiac toxicity, blood pressure impairment, and renal toxicity [27]. These adverse effects may be related to fatal consequences [28]. Owing to unwitnessed SUD in some of our subjects, de nitive causes of death could not be determined in our cohort. Nevertheless, an accumulating evidence has delivered a warning signaling the association between NSAIDs and fatal events in chronic kidney disease. Thus, we believe that an informative evaluation on the indication for NSAIDs in patients undergoing PD is mandatory to avoid SUD.
Diabetes has also been previously recognized as a risk factor for sudden death in the general population and patients undergoing dialysis especially in Asian population [6,21,[29][30][31]. Proposed pathophysiology mechanisms for sudden death in diabetes include microvascular and macrovascular diseases. Among the causes of sudden death, sudden cardiac arrest is known as the primary mechanism [6,29]. Diabetes triggers various untoward biological effects in the body. Consequently, pathologic changes appear in the organs and account for any observed functional impairment, especially in the heart. In this study, diabetes was a factor for SUD in patients undergoing PD. Our ndings are consistent with these studies that showed diabetes led to a higher mortality rate in PD [21]. However, whether diabetic complication control could reduce the incidence of sudden death in diabetic patients undergoing dialysis remains to be elucidated.
Sudden cardiac arrest has been reported as the primary cause of sudden death in patients undergoing dialysis [4,31]. Uremic milieu predisposes individuals with chronic kidney disease (CKD) to have cardiomyopathy and other vascular diseases [11]. These adverse effects consequently lead to arrhythmias, conduction abnormalities, and sudden cardiac arrest. One study investigated longitudinal changes in cardiac function and structure by echocardiography examinations in patients undergoing PD.
The results showed the prevalence of altered cardiac structure and function upon PD initiation [32]. Patients undergoing dialysis are prone to sudden death through prolonged QTc interval or through increased arrhythmogenesis [11]. In this study, we found patients who were undergoing PD and experienced SUD showed longer QTc intervals compared to controls. Thus, patients undergoing dialysis are predisposed to cardiac structure and function abnormalities during the period of renal replacement therapy. Primary and secondary preventions of cardiac arrest could thus reduce SUD in patients undergoing dialysis.
Serum K pattern and its contribution to high cardiac risk has been reported in a large population of patients undergoing PD [33], wherein a U-shaped relationship between time-averaged serum K concentrations and cardiovascular mortality and all-cause mortality was found. The risk for all-cause mortality in cut-off serum K concentrations was < 3.5 mEq/L and ≥ 5.5 mEq/L, respectively [33]. Another smaller study showed that patients undergoing PD with lower time-averaged serum K concentrations and higher standard deviation had higher all-cause and cardiovascular mortality [34]. We found that patients who were undergoing PD and experienced SUD revealed progressively declining serum K concentrations at 3 months before SUD. Serum albumin concentrations also exhibited similar trends. However, synchronized data of ECG and plasma K concentrations were not obtained in our study. In addition, the de nitive causes of hypokalemia in patients who were undergoing PD and experienced SUD also could not be obtained in this study. Considering serum K abnormalities, several spectra of cardiovascular diseases could be involved, including conduction defects, heart attack, and sudden cardiac death [35]. We propose that serum albumin and potassium trends in the last months could be a warning signal for SUD in patients undergoing PD.
The present study has notable limitations. First, information on the predictive performance for comorbidity on sudden death in patients undergoing dialysis is lacking. We used concordant and disconcordant variables for death prediction according to previous reports [25]. However, the tool needs to be validated further for consistent discrimination performance. Second, cardiac function and calci cation scores were not comprehensively evaluated in this study. Therefore, the true in uence of cardiac function on SUD in patients undergoing PD cannot be completely demonstrated in our study. Third, the sample size of SUD is relatively small, and to date, no universally accepted de nition of SUD exists. Finally, this study was conducted retrospectively; therefore, anything reported in this study could be a hypothesis and prospective studies need to be conducted in the future.

Conclusions
In this study, we found that diabetes mellitus and mental health and chronic pain are signi cant predictors for SUD in patients undergoing PD. In addition, a trend for hypokalemia before death is signi cant in these patients. Given these results, a comprehensive framework has been attempted to design a therapeutic strategy to prevent SUD in prevalent patients undergoing PD. Abbreviations PD, peritoneal dialysis; HD, hemodialysis; SUD, sudden unexpected death; ECG, electrocardiogram; DM, diabetes mellitus; CCB, calcium channel blocker; ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blockade; iPTH, intact parathyroid hormone; LV, Left ventricle; LVEF Left ventricular ejection fraction; PET, peritoneal equilibration test; RRF, residual renal function; WCC, weekly creatinine clearance; PGLI, peritoneal glucose loading index; nPCR, normalized protein catabolic rate; Alb, albumin; Hb, hemoglobin; Na, sodium; K, potassium; Ca, calcium; P, phosphate;

Declarations
The authors declared no con ict of interest.

Ethics approval and consent to participate
The Committee on Human Research at the Kaohsiung Chang Gung Memorial Hospital approved the data review protocol for this study (document no: 201800595B0). The requirement for patient consent was waived for medical chart review. The study was conducted in accordance with the principles of the Declaration of Helsinki.

Consent for Publication
Not applicable.

Availability of data and materials
All data supporting the study is presented in the manuscript or available upon request from the corresponding author of this manuscript, Jin-Bor Chen.
Funding None Authors' contributions