Abstract
Background and objectives This study aimed to determine absolute and excess stroke risks in people with ESRD compared with the general population.
Design, setting, participants, & measurements This cohort study used data linkage between the Australia and New Zealand Dialysis and Transplant Registry and hospital and death records for 10,745 people with ESRD in New South Wales from 2000 to 2010. For the general population, Australian Institute of Health and Welfare hospital usage records and Australian Bureau of Statistics census data were used. Rates and standardized incidence rate ratios of hospitalization with a stroke were calculated.
Results People with ESRD had 640 hospitalizations with stroke in 49,472 person-years of follow-up (1294 per 100,000 person-years), and people in the general population had 338,392 hospitalizations with stroke (212 per 100,000 person-years), an incidence rate ratio of 3.32 (95% confidence interval, 3.31 to 3.33). Excess risk was greater for women (incidence rate ratio, 5.14; 95% confidence interval, 5.11 to 5.18) than men (incidence rate ratio, 2.52; 95% confidence interval, 2.51 to 2.54; P for interaction <0.001) and decreased with age. People ages 35–39 years old with ESRD had an 11 times increased risk of stroke (incidence rate ratio, 11.08; 95% confidence interval, 9.41 to 13.05), and risk in people ages ≥85 years old increased 2-fold (incidence rate ratio, 2.04; 95% confidence interval, 1.87 to 2.23; P for interaction <0.001). Excess risk was greater for intracerebral hemorrhage (incidence rate ratio, 4.18; 95% confidence interval, 4.11 to 4.26) than ischemic stroke (incidence rate ratio, 3.43; 95% confidence interval, 3.40 to 3.45; P for interaction <0.01).
Conclusions People with ESRD have a substantially higher risk of stroke, particularly women and young people, and hemorrhagic stroke. Future work could investigate effective and safe interventions for primary and secondary prevention of stroke in people with ESRD.
Introduction
Stroke affects one in every six people, is the biggest cause of severe acquired neurologic disability worldwide, and is the third biggest cause of death in the United States (1,2). People with CKD have a 40%–70% greater risk of having a stroke than people without CKD, but the epidemiology of stroke in ESRD is not well described. Studies investigating stroke risk in ESRD are all >10 years old, and their results are conflicting (3). A Japanese population-based cohort study reported a higher incidence of hemorrhagic stroke but lower incidence of ischemic stroke in people with ESRD (4), whereas an American study reported a 5- to 10-fold increased rate of both types of stroke compared with the general population (5).
Differences in stroke risk among people with ESRD compared with the general population might be explained by a higher prevalence of traditional risk factors for stroke, including hypertension, diabetes, and atrial fibrillation (6). Drugs that are used in primary and secondary prevention strategies to lower stroke risk associated with these risk factors in the general population (including aspirin, warfarin, and statins) may be ineffective or even increase the risk of stroke in people with ESRD (7–9). In addition, novel risk factors specific to having ESRD (accelerated calcific atherosclerosis, platelet dysfunction, anemia, and frequent anticoagulation for those on hemodialysis) might increase stroke risk but may affect the risks of ischemic and hemorrhagic strokes differently (10–12).
The aims of our study were to establish robust estimates of the absolute risk of stroke in people with ESRD, determine the magnitude of the excess risk of stroke in people with ESRD compared with the general population, and determine whether the excess stroke risk varied by sex, age, or stroke subtype (ischemic stroke, intracerebral, or intracranial or subarachnoid hemorrhage).
Materials and Methods
Setting
We performed a cohort study of all adults and children with ESRD who resided in New South Wales (NSW) between July 1, 2000, and June 30, 2010, and compared them with the entire Australian general population. NSW (population of 7.4 million people) is Australia’s most populous state and has a broadly multicultural population representative of the whole nation (current population of 23.6 million people); 2.5% of the NSW population is either Aboriginal or Torres Strait Islander people, and 30% of NSW residents were born overseas, of whom 21% were born in the United Kingdom, 9% were born in New Zealand, 6% were born in China, and 5.5% were born in India. Australia has a universal health care system that provides free care in public hospitals.
Study Populations
People with ESRD.
The Australia and New Zealand Dialysis and Transplant Registry (ANZDATA) collects demographic, clinical, and treatment data on all people whose native kidney function is not expected to recover and for whom indefinite treatment for ESRD is anticipated. Specifically, the ANZDATA records comorbidities, including diabetes, ischemic heart disease, peripheral artery disease, and cerebrovascular disease, along with smoking habit and cause of renal failure at the time of ESRD. The NSW Admitted Patient Data Collection is a mandatory state-wide census of all admitted patient services provided by hospitals. Clinical information (including diagnoses made during admission coded according to the International Classification of Diseases, 10th revision Australian modification [ICD-10-AM]) and health service use are recorded for every hospital discharge, death, or transfer. By statute, every death in NSW is certified by a doctor and notified to the registrar of the NSW Registry of Births, Deaths, and Marriages. We linked the ANZDATA records with the NSW Admitted Patient Data Collection and the NSW Registry of Births, Deaths, and Marriages to establish a cohort of people with ESRD and identify their hospitalizations with stroke and the date that they died. Data linkage was undertaken independently by the NSW Centre for Health Record Linkage using probabilistic matching and best practice privacy-preserving protocols, with ethical approval granted by the NSW Population and Health Services Research Ethics Committee (approval no. 2011/12/363).
General Population.
The Australian Institute of Health and Welfare collects data on every episode of care provided in hospitals, including the principal reason for admission (coded using the ICD-10-AM), and makes nationwide summary-level data available stratified by sex, age, and year, from which we identified stroke admissions. We obtained age- and sex-stratified distributions for the entire population from the Australian Bureau of Statistics and used these to calculate stroke admission rates.
Stroke Ascertainment.
Hospitalization with stroke was our primary outcome. Specifically, we identified ischemic strokes from ICD-10-AM diagnosis codes I63.0–I63.9, intracerebral hemorrhages from codes I61.0–I61.9, and subarachnoid hemorrhages from codes I60.0–I60.9. For all-cause strokes, we additionally included hospitalizations with unspecified strokes (I64) and intracranial hemorrhages (I62.0–I62.9).
Statistical Methods
For people with ESRD, time was measured from either July 1, 2000, or the date on which they were first treated for ESRD if it occurred after that day and continued until the date that they died or June 30, 2010, whichever came first. We had no data on hospitalizations with stroke for people with ESRD before June 1, 2000. To calculate rates of stroke (incident and recurrent) among people with ESRD, we divided their numbers of hospital admissions with a new stroke (identified from the NSW Admitted Patient Collection) by the total length of time that they were at risk of having a stroke. For the general population, to calculate stroke event rates, we used the Australian Institute of Health and Welfare hospital admissions data for the numerator and the Australian Bureau of Statistics population distribution data for the denominator. We then compared the rates of stroke in people with ESRD with those in the general population using indirect standardization by age, sex, and year and expressed them as stroke incidence rate ratios (IRRs) with 95% confidence intervals (95% CIs). For people who started treatment for ESRD during our study period, we then estimated the overall cumulative risk of having at least one stroke and compared stroke risk between men and women for different types of strokes using proportional hazards models. In these cumulative risk analyses, we only included data for the first stroke that people had after starting treatment for ESRD and specifically excluded people with prevalent ESRD to avoid the prevalent pool effect (overestimating stroke incidence by erroneous inclusion of prevalent patients) and immortal time bias (selection bias by including only people who survived long enough on dialysis to be included in analyses).
Results
Patient Characteristics
From July 1, 2000, to June 30, 2010, there were 10,745 people with ESRD in NSW, Australia (Table 1); 3700 people were already receiving treatment for ESRD at the start of our cohort study, with another 7045 people commencing treatment during follow-up. Of the total cohort, 6745 (62.8%) people were still alive, and 4000 (37.2%) had died by June 30, 2010. The median duration of follow-up was 3.8 years (interquartile range [IQR]=1.6–7.8 years), representing a total of 49,472 person-years of observation; 550 (5.1%) people had at least one stroke, and 90 (0.8%) people had two or more strokes (a total of 640 stroke events). There were 311 (48.6%) ischemic strokes, 112 (17.5%) intracerebral hemorrhages, 62 (9.7%) intracranial hemorrhages, 14 (2.2%) subarachnoid hemorrhages, and 141 (22.0%) strokes that were unclassified by pathologic subtype. During the same time period, we identified 338,392 stroke events in the general population (approximately 16 million people of a similar age to people with ESRD). These included 152,208 (45.0%) ischemic strokes, 46,006 (13.6%) intracerebral hemorrhages, 22,583 (6.7%) intracranial hemorrhages, 23,764 (7.0%) subarachnoid hemorrhages, and 93,831 (27.7%) strokes unclassified by subtype.
Characteristics of participants with ESRD
Absolute Incidence Rate of Stroke in People with ESRD and the General Population
Men and women with ESRD had higher rates of all types of stroke across all ages that were similar to those of people 20–30 years older in the general population (Figure 1). The overall crude rate of stroke in people with ESRD was 1294 per 100,000 person-years compared with 212 per 100,000 person-years in the Australian general population. A woman age 40–44 years old with ESRD had a higher risk of having a stroke than a woman age 65–69 years old in the general population (519 versus 304 per 100,000 person-years). Similarly, a man age 40–44 years old with ESRD had approximately the same risk of stroke as a man age 60–64 years old in the general population (461 versus 437 per 100,000 person-years). Although men in the general population had a higher overall rate of stroke (246 per 100,000 person-years) than women (185 per 100,000 person-years), women with ESRD had a higher overall risk of stroke (1466 per 100,000 person-years) than men with ESRD (1171 per 100,000 person-years).
Rates of hospitalization with stroke in men and women with ESRD and the general population. General population data are from the Australian Institute of Health and Welfare and the Australian Bureau of Statistics by age, sex, and year of diagnosis. All-cause stroke includes ischemic stroke, hemorrhagic stroke, unspecified strokes, and intracranial and subarachnoid hemorrhages.
Excess Risk of Stroke in People with ESRD Compared with the General Population
People with ESRD were more than three times more likely to have a stroke than people in the general population (IRR, 3.32; 95% CI, 3.31 to 3.33), but the size of this excess stroke risk varied by sex (P for interaction <0.001) and age (P for interaction <0.001) and differed between stroke subtypes (P for interaction <0.01) (Figure 2, Table 2). Women with ESRD had more than five times the risk of having a stroke compared with women in the general population (IRR, 5.14; 95% CI, 5.11 to 5.18). For men with ESRD, excess stroke risk was less than for women with ESRD, but it was still more than double that of men in the general population (IRR, 2.52; 95% CI, 2.51 to 2.54). Younger people with ESRD had a much greater excess risk of all types of stroke, which declined with age (Figure 2, Table 2). People ages 30–39 years old with ESRD had nearly a nine times excess risk of stroke (IRR, 8.74; 95% CI, 7.79 to 9.81), whereas for people >80 years old, stroke risk was 46% higher (IRR, 1.46; 95% CI, 1.43 to 1.50). Excess stroke risk also varied by stroke subtype and was greater for intracerebral hemorrhage (IRR, 4.48; 95% CI, 4.35 to 4.62) than for either ischemic stroke (IRR, 3.41; 95% CI, 3.39 to 3.43) or subarachnoid hemorrhage (IRR, 1.19; 95% CI, 1.03 to 1.39) (Figure 2, Table 2).
Incidence rate ratio for hospitalization with stroke for men and women with ESRD compared with the general population. General population data are from the Australian Institute of Health and Welfare and the Australian Bureau of Statistics by age, sex, and year of diagnosis. All-cause stroke includes ischemic stroke, hemorrhagic stroke, unspecified strokes, and intracranial and subarachnoid hemorrhages.
Risk of hospitalization for stroke subtypes in people with ESRD by age and sex
Proportion of People with ESRD Who Had at Least One Stroke during Follow-Up
Median time to first recorded stroke in people with incident ESRD was 1.5 years (IQR=0.7–3.5) after starting treatment: 1.4 years (IQR=0.6–3.2) for men and 1.6 years (IQR=0.7–3.8) for women. After 5 years of ESRD, 5.5% of men and 7.9% of women had experienced at least one stroke, rising to 8.8% of men and 11.8% of women by 10 years duration of ESRD. Overall, compared with men with ESRD, women with ESRD were 41% more likely to have an all-cause stroke (hazard ratio [HR], 1.41; 95% CI, 1.14 to 1.74), 45% more likely to have an ischemic stroke (HR, 1.45; 95% CI, 1.08 to 1.96), and 89% more likely to have an intracerebral hemorrhage (HR, 1.89; 95% CI, 1.10 to 3.26) (Figure 3).
Cumulative risk of hospitalization with stroke in men and women with ESRD. All-cause stroke includes ischemic stroke, hemorrhagic stroke, unspecified strokes, and intracranial and subarachnoid hemorrhages. 95% CI, 95% confidence interval; HR, hazard ratio.
Discussion
In this large population-based study that used data on 339,032 stroke events in 10,745 people with ESRD and the entire Australian population to obtain standardized comparator data, we have shown that people with ESRD were more than three times more likely to have a stroke than people in the general population. This excess stroke risk was not consistent and showed important and substantial differences depending on sex, age, and stroke subtype. Women with ESRD had a five times increased stroke risk, whereas men with ESRD had two times the risk of having a stroke. In the general population, men were more likely to have a stroke than women, but we observed that women with ESRD were >40% more likely than men with ESRD to have a stroke. We also found that younger people with ESRD had a far greater excess risk of having a stroke than older people with ESRD. People with ESRD aged 35–39 years old had an 11 times increased risk of stroke, and although excess stroke risk declined with age, people >65 years old still had greater than two times the risk of having a stroke compared with someone in the general population. Although the risk was increased across both types of stroke (ischemic and hemorrhagic), the excess risk of intracerebral hemorrhage was greater than the excess risk of ischemic stroke.
To our knowledge, this is the first study to report that women with ESRD have a greater excess risk of all types of stroke than men with ESRD (5). Our findings are consistent with the association of sex with another complication of ESRD—cancer. Women with ESRD have a 10%–40% greater excess risk of cancer and a greater excess risk of all-cause death, but little is known about the association of sex with cardiovascular events in ESRD (13–15). For stroke, these sex-dependent differences in incidence rates may be explained by age of onset and duration of ESRD, differences in severity and prevalence of risk factors for stroke, dialysis care, and access to kidney transplantation. Women are 70% more likely than men to start treatment for ESRD late (eGFR<5 ml/min per 1.73 m2), more likely to receive shorter dialysis (<20 hours in total per week), less likely to have an arteriovenous fistula, and more likely to have their dialysis adequacy overestimated (16,17). Some traditional stroke risk factors (including diabetes) have more severe effects in women than men, and starting treatment for ESRD late and having poor dialysis adequacy are both associated with an increased risk of cardiovascular disease. Preventive care may also be relatively less effective among women than men with ESRD because of the effect of estrogen on the absorption, bioavailability, and biologic effect of drugs, including aspirin and warfarin. Women with strokes are less likely to receive evidence-based care (including diagnostic and treatment procedures) and have worse functional outcomes after stroke than men (15,18). Whether such complacency extends to the prevention of vascular disease among women with ESRD is unknown.
Our second main finding was that the excess risk of stroke was much greater for younger people with ESRD than older people, which is consistent with the excess risks of all-cause cardiovascular death and cancer in people with ESRD (both decline with advancing age) (19). Although three times more ischemic strokes occurred in the ESRD group than hemorrhagic strokes, the excess risk was higher in the hemorrhagic group, suggesting that the standard interventions for stroke prevention (and the use of antiplatelet drugs, warfarin, and statins) and dialysis delivery (and the use of heparin), particularly in young women, may not be appropriate in the ESRD population. Drugs used to prevent strokes in the general population may have different effects on people with ESRD, in whom they may actually cause harm. The best evidence currently available to guide treatment in the setting of ESRD is inconclusive, particularly for people with ESRD who are young or have atrial fibrillation, highlighting a lack of good-quality randomized, controlled data. One meta-analysis found that antiplatelet agents had no effect on the risk of having a stroke, with uncertain effects on the risk of bleeding (including having a hemorrhagic stroke) in people with ESRD (20). Another meta-analysis found that statins had no effect on lowering the risk of stroke in people with ESRD, despite lowering cholesterol (21). The role for warfarin in people with ESRD and atrial fibrillation is similarly unclear, with poor-quality evidence mainly derived from cohort and cross-sectional studies that suggests a higher risk of hemorrhagic stroke with an unclear effect of treatment with warfarin on the risk of ischemic strokes (22). Finally, the sizes of the absolute and excess risks of stroke incidence that we saw were consistent with the only other study with a comparable racial mix of participants (23). In the general population, Asian and Hispanic people have a 2 times increased stroke risk and black people have a 2.5 times increased absolute risk of stroke compared with white people (24,25). We observed similar racial differences in absolute stroke risk between our study and cohort studies with mainly Asian (2429 per 100,000 person-years), Hispanic (2410 per 100,000 person-years), or black (4794 per 100,000 person-years) participants (5,26,27). Excess risk of stroke in people with ESRD was consistent among these same studies and similar in size to the 3-fold excess risk that we observed (5,26–30). This magnitude of excess stroke risk was also consistent with the 5–10 times increased risk of ischemic heart disease and all-cause cardiovascular death seen in other studies of people with ESRD (31–34).
The major strengths of our study are its size and long duration of follow-up. This is the first population-based cohort study of stroke in people with ESRD to be conducted in a broadly multicultural population, which means that our results provide generalizable estimates of stroke risk in people with ESRD. Selection and outcome ascertainment biases are likely to be minimized compared with other cohort study designs, because the ANZDATA collects data prospectively on everyone with treated ESRD, and the NSW Admitted Patient Data Collection and the NSW Registry of Births, Deaths, and Marriages are also universal in their coverage. There are some potential limitations to our study. First, we could not validate stroke hospitalizations, because data were deidentified, and there is no statutory stroke register in Australia. ICD-10 codes have been shown to have good sensitivity (between 74% and 90%) and even better specificity (between 95% and 97%) in identifying stroke hospitalizations in the general population (35). Using the first two discharge diagnoses in administrative datasets (rather than just the primary discharge diagnosis as we did) has been shown to increase the sensitivity of identifying ischemic strokes. However, the Australian Institute of Health and Welfare hospital admissions data only list the primary discharge diagnosis recorded in the admitted patient data collection for the general population, and therefore, we used just the primary discharge diagnosis to identify strokes among people with ESRD to avoid any differential misclassification bias. Although the joint American Heart Association/American Stroke Association definition of stroke excludes intracranial hemorrhages, including ICD codes for intracranial hemorrhages has been shown to increase the positive predictive value of identifying all-cause strokes (36). Furthermore, etiology of ischemic stroke includes large-artery disease, cardioembolism, and small-vessel disease, and we were unable to capture the heterogeneous nature of ischemic stroke using ICD codes. Second, between 15% and 46% of stroke events are never admitted to the hospital, and if people with ESRD are more likely to be admitted to the hospital after a stroke than the general population, differential bias in stroke ascertainment may exist. Third, we compared people with ESRD in NSW with general population data for the whole of Australia and cannot exclude geographic differences in stroke rates between NSW and the rest of Australia as an explanation for some of the excess risk that we have attributed to ESRD. Fourth, our estimates of excess risk are only adjusted for age, sex, and year. Data on other comorbidities and race among the general population were not available to us, and therefore, we could not adjust for the presence or severity of other disease processes and other factors that may confound the relationship between ESRD and stroke.
Our study highlights the gap between clinical trials and the reality of patients that clinicians encounter. Despite potentially having the most to gain, people with ESRD and women are among the least likely to be included in trials of primary and secondary stroke prevention measures (37). Altered drug absorption, metabolism, and excretion and interactions with medications prescribed for multiple other conditions may alter the balance between benefits and harms of preventive treatments in people with ESRD. To reconcile this gap, researchers must conduct trials that specifically recruit people with ESRD, or they must undertake additional population-based observational studies using linkage of existing electronic health and administrative records to establish robust estimates of treatment effects. Until definitive data exist, clinicians must be cautious when recommending preventive treatments for people with ESRD.
Disclosures
None.
Acknowledgments
We thank Dr. Sradha Kotwal, Martin Gallagher, and Allan Cass from the Renal and Metabolic Division, The George Institute for Global Health, with whom we collaborated on the design, methodology, and dataset preparation stages of this study. The data reported here have been supplied by the Australia and New Zealand Dialysis and Transplant Registry (ANZDATA). Finally, we thank the renal community of New South Wales who supplied original data to ANZDATA and without whom this study would not have been possible.
The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the ANZDATA. Statistical code is available from P.M.. Full datasets cannot be shared because of restrictions imposed by dataset custodians as a condition of their release to named researchers.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
- Received December 8, 2014.
- Accepted June 9, 2015.
- Copyright © 2015 by the American Society of Nephrology