Sleep apnea in end-stage renal disease patients: Impact on cardiovascular and neurological outcomes

ABSTRACT

Introduction:

Sleep apnea (SA) is an important comorbidity in end-stage renal disease (ESRD) patients. The association between SA and cardiac and neurological disease is known. This study investigates the relationship between SA and cardiovascular and cerebrovascular outcomes in the ESRD population.

Methods:

In a retrospective cohort study, the United States Renal Data System was queried to identify ESRD patients aged 18–100 years in whom hemodialysis had been initiated between 2005 and 2013. Diagnoses of SA and clinical comorbidities were determined from International Classification of Disease-9 codes. Demographic variables were obtained from Centers for Medicare and Medicaid Services Form-2728. Logistic regression was used to examine the association of SA with myocardial infarction (MI) or with stroke, controlling for demographic and clinical variables.

Results:

Of 858,131 subjects meeting the inclusion criteria, 587 had central SA, and 22,724 had obstructive SA. The SA cohort was younger, more likely to be male and Caucasian compared to the non-SA cohort. Patients with SA also had more tobacco and alcohol use, hypertension, heart failure, and diabetes. Central SA (aRR = 1.69, 95% CI = 1.28–2.23) and obstructive SA (aRR = 1.15, 95% CI = 1.09–1.21) were associated with an increased risk of stroke but not MI.

Conclusion:

In the ESRD population, a diagnosis of central SA or obstructive SA increased the risk of stroke, but not MI. Early identification and treatment of SA in the ESRD population may help reduce the risk of stroke in these patients.

Introduction

Sleep apnea (SA) is characterized by abnormal breathing during sleep.[1] Obstructive sleep apnea (OSA) occurs when episodes of complete or partial obstruction occur during sleep, and central sleep apnea (CSA) occurs when there is reduction or absence of central respiratory effort.[2]

SA increases the risk of stroke and MI in the general population.[3] In patients with chronic kidney disease (CKD) and ESRD, fluid overload and uremia affect the risk of cardiovascular events and potentially compound the risk of complex SA.[4] The role of SA as a risk factor for stroke and MI warrants investigation in the ESRD population.

Materials and Methods

A retrospective cohort study design was used to assess stroke or MI in ESRD patients with CSA or OSA compared to those with no apnea diagnosis. The data were collected from the United States Renal Data System (USRDS), which is a deidentified database that collects, analyzes, and distributes information about ESRD patients in the United States, who are enrolled in Medicare and entered into the database upon initiation of dialysis. This research was determined to be not human subjects research by the Augusta University Institutional Review Board on March 25, 2022.

Methods

All ESRD patients aged 18–100 who began hemodialysis between 2005 and 2013 and were without missing age, race, sex, ethnicity, or access type were considered for inclusion in the study sample. Additional exclusions included those with an SA diagnosis but no sleep study, less than six months of follow-up after an SA diagnosis or first date of dialysis, those without an SA diagnosis but having continuous positive airway pressure (CPAP) durable medical equipment claims, individuals with no claims, and individuals who died within the first six months of the SA diagnosis or their first date of dialysis. The total analysis sample size was 645,133 patients.

Main independent variable

The main independent variable for this study was a three-level variable based on a diagnosis of SA (no diagnosis, obstructive, or central). Those with a diagnosis of central or obstructive SA were determined by having at least one diagnosis of central (ICD-9 786.04, 327.26, 327.27, 327.29) or obstructive (ICD-9 327.23) apnea with a sleep study [Common Procedural Terminology (CPT) codes 95810, 95811, 95782, 95783, 95800, 95801, 95806, 95807, 95808] with or without a CPT code for CPAP (94660) within six months of the SA diagnosis. ICD-9 and CPT codes were extracted from hospital, physician supplier, or detailed claims.

Outcome variables

The two outcome variables were a stroke or MI and were determined using ICD-9 codes occurring at least six months following the SA diagnosis and/or at least six months following the start of dialysis. Stroke was defined by the presence of at least one stroke diagnosis ICD-9 code and the occurrence of a head CT procedure code within 30 days of the stroke diagnosis. An MI was defined by the presence of at least one MI diagnosis ICD-9 code and the presence of a troponin procedure code within 30 days. The person-years at risk for each outcome was defined as the number of years at risk from the first diagnosis of SA to the outcome, either stroke or MI, or from the start of dialysis for those without an SA diagnosis.

Risk factors

All risk factors were determined using demographic information from the patient file and from hospital claims data using ICD-9 codes combined with the CMS-Form 2728 where appropriate. Demographic risk factors included age at the start of dialysis, race, sex, ethnicity, access type [catheter, graft, arteriovenous (AV) fistula], and ESRD etiology (diabetes, hypertension, lupus, polycystic kidney disease, glomerulonephritis, or other etiology). Clinical diagnosis risk factors included alcohol or tobacco use, hypertension, diabetes, arrhythmias, heart failure with an echocardiogram within one year of diagnosis, and cerebrovascular disease.

Statistical analysis

All statistical analysis was performed using SAS 9.4, (SAS Institute Inc., Cary, NC USA) and statistical significance was assessed using an alpha level of 0.05. Descriptive statistics were determined within stroke or within MI diagnosis. Chi-square or t-tests were used to examine preliminary unadjusted differences between stroke or MI groups.

Logistic regression modeling was used to examine whether SA was a risk factor for stroke or for MI, controlling for various demographic and clinical diagnosis risk factors. All logistic regression models incorporated the person-years at risk by including an offset parameter of the natural log of the person-years at risk. Each risk factor was first examined in a simple logistic regression model, and the unadjusted relative risk (RR) and 95% confidence interval (CI) were estimated. All variables were then entered into a comprehensive full model, and a backward model building strategy was used to arrive at the final model. Variables with the highest, nonsignificant P-value in the full model were eliminated one at a time until the final model consisted of those variables that were statistically significant at the 0.05 alpha level or needed in the model based on model fit criteria. The model fit criterion, Akaike’s Information Criterion, and -2log (likelihood test) were examined after each nonsignificant variable was removed from the model. The final model consisted of all risk factors that were statistically significant or needed in the model to improve model fit to the data. The adjusted RR (aRR) and corresponding 95% confidence interval (CI) were estimated for each variable in the final model.

Results

Of the 2,462,344 ESRD patients, 858,131 met the criteria for inclusion. Tables 1A and 1B provide descriptive statistics including demographics and clinical diagnoses grouped by stroke and MI diagnoses. It should be noted that there were no subjects who had an MI and central SA, and thus, those without an MI who had central SA were excluded from all MI analyses [Table 1]. Both Table 2 and Figure 1 give the results of the simple and final logistic regression model, accounting for the person-years at risk for stroke. Tables 2, 3, and Figure 2 give the results of the simple and final logistic regression model, accounting for the person-years at risk for myocardial infarction (MI) [Table 3].

Table 1A.

Descriptive statistics [n (%)] by stroke or myocardial infarction (demographics)

Variable	Level	Stroke			Myocardial Infarction
		Diagnosis n=41,613 (6.5%)	No Diagnosis n=603,537 (93.5%)	P	Diagnosis n=481 (0.1%)	No Diagnosis n=644,563 (99.9%)	P
Outcome and Main Independent Variable
Sleep Apnea	Central	60 (0.1)	527 (0.1)	<0.0001
	Obstructive	1737 (4.2)	20987 (3.5)		24 (5.0)	22700 (3.5)	0.0815
	No Apnea	39816 (95.7)	582023 (96.4)		457 (95.0)	621382 (96.5)
Person-Years at Risk*		2.7 (1.9)	3.9 (2.5)	<0.0001	4.2 (2.5)	3.8 (2.5)	<0.0001
Demographics
Age*		65.1 (13.2)	62.2 (15.3)	<0.0001	65.0 (13.1)	62.4 (15.2)	<0.0001
Sex	Female	21942 (52.7)	259692 (43.0)	<0.0001	222 (46.2)	281210 (43.7)	0.2704
	Male	19671 (47.3)	343845 (57.0)		259 (53.9)	362872 (56.3)
Race	Black	15169 (36.5)	177674 (29.4)	<0.0001	78 (16.2)	192641 (29.9)	<0.0001
	Other Race	2035 (4.9)	37428 (6.2)		55 (11.4)	39390 (6.1)
	White	24409 (58.7)	388435 (64.4)		348 (72.4)	412051 (64.0)
Ethnicity	Hispanic	6130 (14.7)	96845 (16.1)	<0.0001	127 (26.4)	102773 (16.0)	<0.0001
	Non-Hispanic	35483 (85.3)	506692 (84)		354 (73.6)	541309 (84.0)
Access Type	Catheter	33464 (80.4)	480797 (79.7)	<0.0001	377 (78.4)	513435 (79.7)	0.7282
	Graft	1876 (4.5)	21759 (3.6)		20 (4.2)	23593 (3.7)
	AV Fistula	6273 (15.1)	100981 (16.7)		84 (17.5)	107054 (16.6)
ESRD Etiology	Diabetes	22815 (54.8)	273071 (45.3)	<0.0001	NR†	NR	<0.0001
	Glomerulonephritis	685 (1.7)	15154 (2.5)		NR	NR
	Hypertension	11640 (28.0)	170782 (28.3)		NR	NR
	Lupus	290 (0.7)	6315 (1.1)		NR	NR
	Polycystic Kidney Disease	427 (1.0)	11712 (1.9)		NR	NR
	Other Etiology	5756 (13.8)	126503 (21.0)		61 (12.7)	132096 (20.5)

*mean (SD), †NR=Not reportable due to some cells having frequencies <11, per the USRDS[13]

Table 1B.

Descriptive statistics [n (%)] by stroke or myocardial infarction (clinical factors)

Variable	Level	Stroke			Myocardial Infarction
		Diagnosis n=41,613 (6.5%)	No Diagnosis n=603,537 (93.5%)	P	Diagnosis n=481 (0.1%)	No Diagnosis n=644,563 (99.9%)	P
Clinical Diagnoses
Tobacco	Yes	19805 (47.6)	213472 (35.4)	<0.0001	224 (46.6)	232726 (36.1)	<0.0001
	No	21808 (52.4)	390065 (64.6)		257 (53.4)	411356 (63.9)
Alcohol	Yes	2966 (7.1)	30959 (5.1)	<0.0001	26 (5.4)	33861 (5.3)	0.8843
	No	38647 (92.9)	572578 (94.9)		455 (94.6)	610221 (94.7)
Heart Failure	Diagnosis	20049 (48.2)	140291 (23.2)	<0.0001	303 (63.0)	159773 (24.8)	<0.0001
	No Diagnosis	21564 (51.8)	463246 (76.8)		178 (37.0)	484309 (75.2)
Hypertension	Diagnosis	36748 (88.3)	410287 (68.0)	<0.0001	425 (88.4)	446097 (69.3)	<0.0001
	No Diagnosis	4865 (11.7)	193250 (32.0)		56 (11.6)	197985 (30.7)
Diabetes	Diagnosis	25804 (62.0)	251025 (41.6)	<0.0001	313 (65.1)	276104 (42.9)	<0.0001
	No Diagnosis	15809 (38.0)	352512 (58.4)		168 (34.9)	367978 (57.1)
Arrhythmias	Diagnosis	461 (1.1)	3579 (0.6)	<0.0001	NR†	NR	0.9959
	No Diagnosis	41152 (98.9)	599958 (99.4)		NR	NR
Cerebrovascular Disease	Diagnosis	7559 (18.2)	44121 (7.3)	<0.0001	74 (15.4)	51563 (8.0)	<0.0001
	No Diagnosis	34054 (81.8)	559416 (92.7)		407 (84.6)	592519 (92.0)

†NR=Not reportable due to some cells having frequencies <11, per the USRDS[13]

Table 2.

Logistic Regression Results on Stroke, *referent level

Variable	Level	Simple Unadjusted Models			Final Adjusted Model
		RR	95% CI	P	aRR	95% CI	P
Main Independent Variable
Sleep Apnea	Central versus No Apnea*	2.01	1.53–2.63	<0.0001	1.69	1.28–2.23	<0.0001
	Obstructive vs No Apnea*	1.38	1.31–1.45		1.15	1.09–1.21
Demographics
Age	1 yr increase	1.029	1.028–1.029	<0.0001	1.024	1.023–1.025	<0.0001
Sex	Female vs Male*	1.50	1.47–1.53	<0.0001	1.40	1.37–1.43	<0.0001
Race	Black vs White*	1.19	1.17–1.22	<0.0001	1.23	1.20–1.26	<0.0001
	Other race vs White*	0.73	0.70–0.77		0.79	0.76–0.84
Ethnicity	Hispanic vs Non-Hispanic*	0.79	0.77–0.82	<0.0001	0.94	0.91–0.97	0.0002
Access Type	Catheter vs AV Fistula*	1.25	1.22–1.29	<0.0001	1.28	1.24–1.32	<0.0001
	Graft vs AV Fistula*	1.47	1.40–1.56		1.20	1.13–1.27
ESRD Etiology	Diabetes vs Other Etiology*	2.06	2.00–2.12	<0.0001	1.72	1.66–1.78	<0.0001
	Glomerulonephritis vs Other Etiology*	0.86	0.80–0.94		0.93	0.86–1.01
	Hypertension vs Other Etiology*	1.63	1.58–1.69		1.23	1.19–1.27
	Lupus vs Other Etiology*	0.80	0.71–0.90		1.06	0.93–1.20
	Polycystic Kidney vs Other Etiology*	0.63	0.57–0.69		0.71	0.64–0.78
Clinical Diagnoses
Tobacco	Yes vs No*	1.56	1.53–1.60	<0.0001	1.16	1.14–1.19	<0.0001
Alcohol	Yes vs No*	1.33	1.27–1.38	<0.0001	1.27	1.22–1.33	<0.0001
Heart Failure	Diagnosis vs No Diagnosis*	2.76	2.70–2.82	<0.0001	1.97	1.94–2.01	<0.0001
Hypertension	Diagnosis vs No Diagnosis*	3.56	3.45–3.67	<0.0001	2.24	2.17–2.32	<0.0001
Diabetes	Diagnosis vs No Diagnosis*	2.05	2.01–2.10	<0.0001	0.94	0.92–0.97	<0.0001
Arrhythmias	Diagnosis vs No Diagnosis*	1.71	1.55–1.89	<0.0001
Cerebrovascular Disease	Diagnosis vs No Diagnosis*	3.56	3.46–3.66	<0.0001	2.51	2.44–2.58	<0.0001

*referent level

Figure 1.

Final logistic regression model of sleep apnea on stroke

Table 3.

Logistic Regression Results on Myocardial Infarction, *referent level

Variable	Level	Simple Unadjusted Models			Final Adjusted Model
		RR	95% CI	P	aRR	95% CI	P
Main Independent Variable
Sleep Apnea	Obstructive vs No Apnea*	1.64	1.09–2.47	0.0188	1.32	0.87–2.00	0.1854
Demographics
Age	1 yr increase	1.027	1.020–1.033	<0.0001	1.018	1.011–1.026	<0.0001
Sex	Female vs Male*	1.11	0.93–1.33	0.2380
Race	Black vs White*	0.42	0.33–0.54	<0.0001	0.49	0.38–0.63	<0.0001
	Other race vs White*	1.42	1.06–1.88		1.72	1.28–2.31
Ethnicity	Hispanic vs Non-Hispanic*	1.67	1.37–2.05	<0.0001	1.58	1.26–1.96	<0.0001
Access Type	Catheter vs AV Fistula*	1.04	0.82–1.32	0.8756
	Graft vs AV Fistula*	1.13	0.70–1.85
ESRD Etiology	Diabetes vs Other Etiology*	2.30	1.80–2.95	<0.0001	1.66	1.29–2.14	<0.0001
	Hypertension vs Other Etiology*	1.45	1.09–1.93		1.21	0.90–1.61
Clinical Diagnoses
Tobacco	Yes vs No*	1.44	1.20–1.72	<0.0001
Alcohol	Yes vs No*	0.96	0.64–1.42	0.8180
Heart Failure	Diagnosis vs No Diagnosis*	4.55	3.78–5.48	<0.0001	3.66	3.02–4.43	<0.0001
Hypertension	Diagnosis vs No Diagnosis*	3.29	2.49–4.35	<0.0001	1.88	1.41–2.52	<0.0001
Diabetes	Diagnosis vs No Diagnosis*	2.20	1.82–2.66	<0.0001
Arrhythmias	Diagnosis vs No Diagnosis*	0.90	0.29–2.80	0.8577
Cerebrovascular Disease	Diagnosis vs No Diagnosis*	2.52	1.97–3.23	<0.0001	1.86	1.45–2.39	<0.0001

*referent level

Figure 2.

Final logistic regression model of sleep apnea on myocardial infarction

Stroke

Those with a stroke had a significantly higher unadjusted mean age and had a significantly higher unadjusted percent of OSA, female sex, Black race, catheter and graft access types, diabetic ESRD etiology, tobacco use, alcohol use, heart failure, hypertension, diabetes, arrhythmias and cerebrovascular disease. Those with a stroke had a significantly lower unadjusted percentage of Hispanic ethnicity than those without. All variables were significantly associated with stroke in simple models, and the final logistic regression model accounting for person-years at risk contained all variables except arrhythmias. CSA (aRR = 1.69, 95% CI = 1.28–2.23) and OSA (aRR = 1.15, 95% CI = 1.09–1.21) were associated with a significantly increased risk of stroke. Age, female sex, Black race, catheter or graft versus AV fistula access types, diabetes and hypertensive ESRD etiology compared to other ESRD etiologies, tobacco use, alcohol use, heart failure, hypertension, and cerebrovascular disease were all associated with an increased risk of stroke. Other race, Hispanic ethnicity, polycystic kidney disease compared to other ESRD etiologies, and diabetes were associated with a decreased risk of stroke.

Myocardial infarction

Those with an MI had a significantly higher unadjusted mean age and had a significantly higher unadjusted percent of OSA, other race, Hispanic ethnicity, diabetic ESRD etiology, tobacco use, heart failure, hypertension, diabetes, and cerebrovascular disease. Those with an MI had a significantly lower unadjusted percentage of Black race and hypertensive ESRD etiologies versus other ESRD etiologies than those without an MI. Because of the low frequency of occurrence of glomerulonephritis, polycystic kidney disease, and lupus ESRD etiologies, these three etiologies were combined with the “other ESRD etiology” group for modeling purposes. While RR was determined for all variables in simple models, access type and arrhythmias were not used in the logistic regression model-building due to the low frequency of occurrence in specific categories. Obstructive SA (aRR = 1.32, 95% CI = 0.87–2.00) was associated with an MI neither in simple models nor in the final logistic regression model. Age, other race, Hispanic ethnicity, diabetic ESRD etiology compared to other etiologies, heart failure, hypertension, and cerebrovascular disease were all associated with an increased risk of an MI.

Discussion

At least 60% of ESRD patients also have SA based on other studies.[5] Different mechanisms such as uremic neuropathy and uremic myopathy in ESRD contribute to SA. Uremic neuropathy can cause impairment in the upper airway muscles’ ability to sense signals while uremic myopathy can cause the muscles to have decreased tone.[4] Furthermore, fluid overload in ESRD patients can cause fluid shifts from lower extremities to the neck and lungs, and this can contribute to upper airway collapsibility and the resultant sleep-disordered breathing.[4]

Sleep apnea has a known association and increased risk of adverse cardiovascular events including heart failure, atrial fibrillation, MI, and stroke.[4] Hypoxic episodes and inspiratory effort obstruction result in inadequate air flow and oxygenation. Pleural pressure decreases, and the pressure, stretch, and oxygen demand of the heart increases.[2] Detrimental cardiovascular remodeling downstream of OSA and other risk factors can occur.[6] Rapid oxygenation upon restoration of respiration leads to free radical creation, oxidative stress, and inflammation that can cause downstream damage.[1,2] Transcription factor-1 and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB) become activated, which results in an inflammatory process that leads to atherosclerosis.[6] This damage can be cardiovascular in nature and causes an event such as an MI, or it can impair cerebral perfusion and cause neurologic events such as cerebral hemorrhage or stroke.[7] Wang et al.[6] postulate that OSA is an independent risk factor contributing to cardiovascular outcomes. Other studies have found a bidirectional relationship between sleep-disordered breathing and stroke, with untreated OSA leading to a higher stroke risk in patients.[8,9]

This study investigated the association of SA in ESRD patients and the occurrence of cerebrovascular and cardiovascular events. The data demonstrated that ESRD patients with either OSA or CSA had an increased risk of stroke. People of older age, female sex, Black race, with hypertension, with cerebrovascular disease, with heart failure, and who used alcohol or tobacco also had increased risk of stroke. These findings highlight the importance of screening patients with ESRD for sleep-disordered breathing, and recognizing a patient’s risk factors and providing early prevention and treatment could help improve outcomes in the ESRD population.

This study did not look at the association of CSA in ESRD with MI due to finding no patients with CSA and MI in this population. The study did not support an increased risk of MI in this population (aRR = 1.32, 95% CI 0.87–2.00, P = 0.0815). This study did show that age, Hispanic ethnicity, hypertension, cerebrovascular disease, and heart failure all were associated with increased risk of MI within the ESRD population. There could have been a synergistic effect between demographics and MI risk within the ESRD population that is affecting this finding. Further research is needed to allow a better understanding of the association between SA and cardiovascular risks in patients with ESRD.

The data not supporting a higher risk of MI in ESRD patients with SA could be explained by the limitations in the methodology of the study. This analysis used the procedural code for CPAP as a criterion for the diagnosis of SA although it was not required. However, it is known that CPAP is the treatment for obstructive SA because it helps maintain the upper airway and normalizes breathing.[10,11] Studies have shown that CPAP treatment decreases cardiovascular morbidity and mortality by its effect on sympathetic nervous system overactivity, inflammation, endothelial dysfunction, and lipid metabolism.[10] Because all of the patients in this study could have been using CPAP therapy, this could mask the potential association between SA and MI risk in the ESRD population. The patients within this database also have Medicare, such that access to care and to CPAP treatment would not be an issue, potentially diluting the harmful effects of SA in this population. Additionally, the CPAP code does not provide any additional information about how compliant each patient is to the treatment. Use of a CPAP code provides no information on how often each patient is using their CPAP, the benefit they are receiving, and the potential reduction in their cardiovascular and cerebrovascular risk. The lack of detailed information on CPAP compliance and CPAP efficacy is a limitation of this data set.

There were far fewer patients in this study with CSA and ESRD versus OSA and ESRD. The pathophysiological mechanisms that cause the absence of respiratory drive in patients with CSA are varied, and as such, the treatments are varied as well.[12] Some of these treatments are still undergoing trials and further research. CPAP has been found to be beneficial in some patients with idiopathic central SA as well as in heart failure patients; however, there was a clinical trial that did not find decreases in mortality for CSA patients on CPAP.[12] CSA is already rare, and the requirement that patients with CSA needed a documented sleep study procedure code may have lessened the number of CSA patients in the study, and as a result, the risk of MI and stroke in CSA patients with ESRD may be biased.[12]

Additionally, the study relies on information within the USRDS database in terms of diagnostic and procedural codes, and there are limitations on the data provided. There is no way to verify the accuracy of these diagnoses on a clinical basis. Furthermore, the database includes information about patients on dialysis in the United States, and therefore, the findings may not be generalizable globally due to demographic differences.

Sleep apnea can go unrecognized and undiagnosed in populations with renal dysfunction. ESRD patients can have poor sleep quality and fatigue that is related to their kidney disease, and so these symptoms could make the diagnosis of SA more challenging in this population.[4] Additionally, many of the characteristics expected in patients with SA are absent in patients with chronic kidney disease, including increased body mass index, snoring, witnessed apneic episodes, and morning headaches.[4] Because this is a retrospective study with the discussed limitations, further studies systematically assessing SA in patients with chronic kidney disease and ESRD either retrospectively or prospectively, and using different thresholds of apnea/hypopnea indices to define SA, are needed. Including patients with different severities of chronic kidney disease and on different treatment modalities for SA such as surgery or oral appliances might help tease out the risk of these comorbidities on stroke and MI.

Data availability statement

The data analyzed in this article were supplied by the United States Renal Data System (USRDS) under a data use agreement. Data will be shared on request to the corresponding author with permission of the United States Renal Data System. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as official policy or an interpretation of the United States Government. The contents of this article do not represent the views of the Department of Veterans Affairs or the United States Government.

Financial support and sponsorship

The work was performed at the Medical College of Georgia in Augusta, GA and was supported by a grant award from Dialysis Clinics, Inc. to AM.

Conflicts of interest

There are no conflicts of interest.