Cardiovascular Outcomes in Patients on Home Hemodialysis and Peritoneal Dialysis

Visual Abstract


Introduction
Cardiovascular disease remains the leading cause of morbidity and mortality among people with kidney failure.Cardiovascular disease is present in .50% of patients undergoing dialysis, and the risk of cardiovascular death in the dialysis population is 20 times higher than in the general population. 1 This increased risk may be explained by the high prevalence of

Methods Study Population and Data Sources
The study cohort included 68,645 individuals 18 years and older from USRDS with linked claims for Medicare Part A and Part B or Medicare Primary Other as the primary payer, who had incident ESKD between January 1, 2005, and December 31, 2018, and who had initiated home dialysis within 6 months of the first ESKD service date.Patients with ,3 months of continuous home dialysis (excluding breaks of ,1 month) and those receiving care in a facility with .90% of patients residing in a skilled nursing facility were excluded.Figure 1 shows the cohort derivation.The University of Cincinnati Institutional Review Board Committee deemed the study exempt because the data were deidentified.

Covariates
The major predictor was type of home dialysis modality, defined as HHD or PD.Information on the covariates was obtained from the Centers for Medicare & Medicaid Services (CMS)-2728 form, USRDS patients file, USRDS treatment history file, and USRDS residence file. 15The USRDS patients file was used to obtain information on the date of the first ESKD service, age, sex, race/ethnicity, region, date of death, and cause of death.The USRDS treatment history file was used to identify dates and types of dialysis modalities.The USRDS residence file was used to obtain information on patients' zip codes of current residence at study entry.These zip codes were combined with zip code-level data from the United States Census Bureau American Community Survey 5-year estimates from 2007 to 2011 to determine neighborhood socioeconomic status, which we defined as the percentage of zip code residents living below the federal poverty level and grouped similarly to the United States Census Bureau literature into five categories: I (,13.8%),II (13.8%-19.9%),III (20.0%-39.9%),IV (40% or more), and unknown. 16The CMS form 2728, which is filed at the time of initiation of dialysis, was used to collect information on body mass index; dialysis modality; cause of kidney failure; comorbidities (cardiovascular disease, hypertension, diabetes mellitus, chronic obstructive pulmonary disease, cancer, and smoking); predialysis nephrology care; laboratory data; and poor functional status defined by inability to ambulate, inability to transfer, or need of assistance with daily activities.History of cardiovascular disease at baseline was defined by the presence of heart failure, atherosclerotic heart disease, peripheral vascular disease, amputation, or cerebrovascular accident as specified in the form CMS-2728. Patients with unavailable information on other covariates were categorized into a missing group for that covariate, as presented in Table 1.
Patients receiving care in a facility with .90% of patients residing in a skilled nursing facility were excluded.During the study era, an increasing but nonetheless small share of patients dialyzing at home, as indicated in USRDS data, resided in a skilled nursing facility, not a private residence.This trend has been more pronounced with HHD than with PD because of state regulations that have permitted the construction of dialysis dens.USRDS data do not clearly indicate whether a patient resides in a private residence or skilled nursing facility, and data also fail to indicate whether a Medicare-certified dialysis facility actually provides on-site hemodialysis to skilled nursing facilities.To address this issue in an objective manner that was applicable to both dialytic modalities, we identified first-time nursing home residency after home dialysis initiation by identifying the date of the first physician claim with place of service code 31 ("skilled nursing facility") or 32 ("nursing facility") or Current Procedural Terminology codes 99304-99318, indicating evaluation and management services in a nursing facility. 17We parameterized nursing home residency as a monotonic binary factor equal to 0 from home dialysis initiation until the date of the first qualifying claim and equal to 1 thereafter.This approach is supported by Chen et al., who showed that even a short stay in a nursing facility was associated with markedly higher risks of death and hospitalization during 1 year of follow-up. 18

Outcomes
The primary outcomes were the occurrence of a composite cardiovascular event, defined as hospitalization with a primary diagnosis of acute coronary syndrome (myocardial infarction or unstable angina), heart failure, or stroke; separate occurrence of acute coronary syndrome, heart failure, or stroke; cardiovascular death; and all-cause death. 191][22] Myocardial infarction, unstable angina, heart failure, and stroke were identified by ICD-9-CM codes as presented in Supplemental Material 1. [23][24][25] Death dates and causes of death as reported on CMS form 2746 were obtained from the USRDS patients file.Cardiovascular death indicated that the reported code for the cause of death was a cardiovascular one and included myocardial infarction; acute pericarditis, including cardiac tamponade; atherosclerotic heart disease; cardiomyopathy; cardiac arrhythmia; cardiac arrest cause unknown; valvular heart disease; congestive heart failure; cerebrovascular accident, including intracranial hemorrhage; and ischemic brain damage/ anoxic encephalopathy.

Statistical Analyses
Descriptive statistics were used to describe patients' characteristics at dialysis initiation.Categorical variables were summarized using frequencies and percentages, with chi-square or Fisher exact tests used to test for differences between HHD and PD.Continuous variables were summarized using means and SDs, with t tests used to test for differences in means.Rates were expressed as the number of events per thousand person-years (PTPY); exact Poisson confidence intervals (CIs) were calculated, and significance was tested using unadjusted Poisson models.
To test the association of type of home dialysis modality with outcomes, adjusted for our covariates, multivariable Cox proportional hazards time-to-event models were constructed.All observation times were censored at the end of an initial home dialysis period, which was continuous, except for breaks of less than one month, end of primary Medicare coverage, or end of the study period-December 31st, 2018; thus, death, recovery, and transplant were implicitly censored as well.Nursing home was used as a time-dependent covariate.The risk estimates were expressed as hazard ratios (HRs) and their 95% CIs.We adjusted directly for covariates in the Cox regression analyses rather than using propensity score matching using those covariates to see the separate effects of covariates on the outcomes.Because residual confounding does not go away with propensity score matching, some statistical experts have criticized matching because it discards information from some unexposed patients, resulting in less precise estimates of the association between exposure and outcome.All statistical analyses were conducted using SAS version 9.4, with two-sided hypothesis testing and a P value of ,0.05 as the criterion for statistical significance.

Baseline Characteristics
Patients' mean age at study entry was 64615 years, and 42.3% were women.Regarding race/ethnicity, 66.5% were non-Hispanic White, 18.1% were non-Hispanic Black, 9.8% were Hispanic, 3.8% were Asian, and 0.9% were Native American.Only 11.7% did not receive predialysis nephrology care.Most of the study cohort had a history of hypertension (87.6%) and 38.3% had a history of cardiovascular disease.Diabetes (44.5%) was the most common attributed cause of kidney failure among patients on dialysis, followed by hypertension (30.1%).Table 1 summarizes the baseline characteristics of the study cohort.
The mean age at study entry was higher for patients on HHD than for patients on PD (68613 versus 63615 years).The majority of both patients on HHD and PD were of White race/ethnicity.Patients on HHD were less likely to be Hispanic (5.8% versus 10.1%).Patients on HHD were more likely to have comorbidities of cardiovascular disease (50.3% versus 37.5%), diabetes (56.5% versus 51.9%), and poor functional status (29.1% versus 5.7%), Patients on HHD were more likely to have kidney failure due to hypertension/large vessel disease (34.0%versus 29.8%) and less likely to have kidney failure due to glomerulonephritis (4.7% versus 9.1%).More patients on HHD resided in the mid-Western United States region (45.0%versus 24.6%).More patients on HHD lived in zip codes of higher socioeconomic status (,13.8%poverty category) (69.5% versus 66.0%).Patients on HHD had more unfavorable laboratory parameters of albumin ,3.5 mg/dl (40.5% versus 32.9%) and hemoglobin ,11 g/dl (65.8% versus 59.9%).Table 1 summarizes the characteristics of patients by type of home dialysis modality.

Cardiovascular Events
The cardiovascular event rate was 95.1 PTPY (95% CI, 93.3 to 96.8), the most common of which was heart failure (44.4 PTPY, 95% CI, 43.2 to 45.6).The mean time of followup was 1.861.6 years overall, 1.461.6 years for patients on HHD, and 1.961.6 years for patients on PD.Overall, the cardiovascular event rate was higher in patients on HHD (127.8PTPY; 95% CI, 118.9 to 137.2), as compared with those on PD (93.3 PTPY; 95% CI, 91.5 to 95.1).Rates of heart failure (76.5 PTPY; 95% CI, 69.9 to 83.7 versus 42.6 PTPY; 95% CI, 41.4 to 43.8) were higher in patients on HHD than in those on PD.The stroke event rate was slightly lower in patients on HHD (19.7 PTPY; 95% CI,  2 provides the unadjusted cardiovascular rates by type of dialysis modality (events, events per 1000 person-years).
In the regression models adjusted for demographics, nursing home residency, socioeconomic characteristics, and other clinical characteristics, the risk of cardiovascular events was slightly lower in patients on HHD than in patients on PD (HR, 0.92; 95% CI, 0.85 to 0.997) (Figure 2).Patients on HHD experienced a lower risk of stroke (HR, 0.58; 95% CI, 0.48 to 0.71) and a lower risk of acute coronary syndrome (HR, 0.83; 95% CI, 0.72 to 0.95), as compared with those on PD.The risk of heart failure was similar in both modalities (HR, 1.05; 95% CI, 0.94 to 1.16) (Figure 3).

Discussion
Our study examined the cardiovascular outcomes among patients on incident PD and HHD from 2005 to 2018 in the United States.We found that patients on incident HHD had a lower adjusted risk of cardiovascular events, cardiovascular death, and all-cause death.Further classifying the cardiovascular events, the risk of heart failure was similar, although patients on HHD had a lower risk of strokes and a lower risk of acute coronary syndrome than patients on PD.
Our study shows a higher unadjusted risk but a slightly lower adjusted risk of cardiovascular events in HHD relative to PD; the differences between the two estimates are likely because of differences in baseline characteristics between the two home dialysis modality groups.For example, patients on HHD were older and had more diabetes than patients on PD, but after adjustment with these covariates, patients on home dialysis had lower hazards of cardiovascular events than those on PD.Most patients on HHD in the United States are prescribed at least four treatments per week, and a subgroup are prescribed nocturnal hemodialysis, which typically involves .20 treatment hours per week.Cardiovascular benefit of intensive hemodialysis as compared with thrice-weekly in-center dialysis has been reported because of its association with reductions in left ventricular mass and BP, decreased use of antihypertensive medications, and increased left ventricular ejection fraction.While no studies have specifically compared incidence of CV events in patients on incident HHD and PD, Weinhandl et al. showed a 11% lower risk of cardiovascular hospitalization and a shorter length of stay for cardiovascular hospitalization in matched patients on HHD as compared with patients on PD. 9 Similarly, Suri et al. reported a lower risk of cardiovascular hospitalizations in HHD relative to PD. 10 PD may negatively alter the risk of cardiovascular morbidity.Fluid overload is frequently seen in patients on PD because of poor adherence to fluid intake restriction, low effluent drain volume, high transporter status, infrequent prescription changes, mechanical complications of   PD catheters, or loss of residual kidney function. 26,27Patients on HHD have a lower adjusted risk of CV mortality as compared with patients on PD.Nocturnal HHD stabilizes coronary calcification burden and improves endothelial vasodilation, baroreflex sensitivity, and arterial compliance. 28,29CKD-mineral and bone disorder, including high levels of calcium, phosphorus, alkaline phosphatase, and parathyroid hormone, results in vascular calcification and remains an important contributor to cardiovascular mortality. 30,31nhanced solute clearance coupled with volume removal with more frequent or longer HHD results in lower cardiovascular mortality because of improved control of BP, improved control of parameters associated with bone and mineral metabolism, reductions in ventricular volumes, and regression of left ventricular mass. 5,32Oxidative stress from advanced glycation end products formed 28 during heat sterilization of dialysate has also emerged in the recent decades as a novel nontraditional, uremia-related risk factor of increased cardiovascular mortality and morbidity in patients on PD. 33 HHD was associated with slightly better adjusted survival as compared with PD.The Australia and New Zealand Dialysis and Transplant Registry study reported a much higher improvement in survival, with 53% reduction in mortality in 706 patients on patients compared with 10,710 patients on PD. 11 However, this study was limited by the sole inclusion of patients started very early on home dialysis (within 90 days), which, especially for HHD, might have introduced a selection bias.In addition, the greater association of HHD with lower mortality may also involve stricter selection criteria favoring healthier patients for HHD in these centers less experienced with HHD therapy.A Canadian study showed a 34% lower risk of mortality for patients on incident HHD compared with those on incident PD.The risks of mortality with HHD were more pronounced for older cohorts from 2005 to 2019 and were not significantly different from 2011 to 2013. 12These findings may be possibly related to the inclusion of higher acuity patients in HHD programs over time and simultaneous improvement in the PD technique and patient survival.By contrast, a study from the United States demonstrated that mortality rates in 4201 prevalent patients on HHD and matched patients on PD were not statistically significant when the analysis was restricted to patients who started home dialysis within 6 months of kidney replacement therapy initiation. 9While we also restricted analysis to patients who started home dialysis within 6 months of onset of kidney failure, these differences could be attributed to the variation in prescription patterns and practices (short/daily versus longer/nocturnal).Higher risk of mortality in patients on PD may also be explained by higher risk infection-related hospitalizations and morbidity. 34,35][11] Our study has some limitations.First, the observational design precludes determination of causality.Second, comorbidity reporting on the Medical Evidence Report form has not been validated, and prior studies have shown underreporting of comorbid conditions that may have resulted in some nondifferential misclassification.Third, the details on treatment sessions and prescription patterns are not available in Medicare.Fourth, there is variability in the quality and completeness of the data recorded on form 2728 of the USRDS data.Fifth, we did not have data for residual renal function and urine output for patients on home dialysis, which remains a limitation of the study.Our study has the following strengths.First, because we used the largest registry of patients on dialysis in the United States and restricted to patients with primary Medicare, it captures all patients with ESKD and all hospitalizations, therefore, accurately predicting the rates of cardiovascular events and cardiovascular deaths in both HHD and PD for this high-risk population.Second, it included 68,645 patients, with the largest sample size for the home dialysis population in the literature so far.
In conclusion, HHD relative to PD is associated with lower adjusted risks of cardiovascular events, including those of stroke and acute coronary syndrome; lower adjusted risks of cardiovascular mortality; and lower adjusted risk of all-cause mortality.

Figure 3 .
Figure 3. Time-to-event model showing adjusted hazard ratios for cardiovascular death.*The model predicts time to cardiovascular death, with patients censored at the time of end of continuous home dialysis (other than breaks of ,30 days), end of time with primary Medicare, or end of the study period; nursing home stay is a time-dependent covariate.

Table 1 .
Patient characteristics by type of home dialysis modality Figure 1.Study cohort selection flow diagram.

Table 1 .
(Continued)Continuous variables were summarized using means and standard SDs; t tests were used to test for differences in means.
a Reported in mean (SD).

Table 2 .
Rates of cardiovascular events, all-cause death, and cardiovascular death per thousand person-years by type of home dialysis modality Rates were expressed as the number of events per thousand person-years (PTPY); exact Poisson confidence intervals were calculated; significance was tested using unadjusted Poisson models.16.4 to 23.4), as compared with those on PD (23.6 PTPY; 95% CI, 22.7 to 24.5).Rates of acute coronary syndrome did not differ significantly between patients on HHD and PD(35.1 PTPY; 95% CI, 30.7 to 39.9 versus 33.4 PTPY; 95% CI, 32.4 to 34.5).Table Figure2.Time-to-event model showing adjusted hazard ratios for cardiovascular events.*The model predicts time to cardiovascular events, with patients censored at the time of end of continuous home dialysis (other than breaks of ,30 days), end of time with primary Medicare, or end of the study period; nursing home stay is a time-dependent covariate.
Figure 4. Time-to-event model hazard ratios for home hemodialysis versus peritoneal dialysis for outcomes of cardiovascular events, heart failure, stroke, acute coronary syndrome, cardiovascular death, and all-cause death.*The models predict time to cardiovascular events, heart failure, stroke, and acute coronary syndrome, cardiovascular death and all-cause death censored for end of continuous home dialysis (other than breaks of ,30 days), end of primary Medicare coverage, and end of the study period.Hazard ratios are adjusted for covariates of type of home dialysis, age, sex, race/ethnicity, body mass index, history of cardiovascular disease, hypertension, diabetes mellitus, chronic obstructive pulmonary disease, cancer, smoking, functional status, albumin, hemoglobin, cause of ESKD, predialysis nephrology care, neighborhood poverty, geographic region, nursing home stay, and time from start of ESKD to the start of home dialysis; nursing home stay is a time-dependent covariate.