Abstract
Rationale & Objective:
The associations between ischemic stroke and time to dialysis initiation and/or death in adults with late-stage chronic kidney disease (CKD) have not been explored. We sought to measure the rate and factors associated with stroke in CKD stages 4 and 5 (CKD4–5) and assess the association of stroke with initiation of dialysis and death.
Study Design:
Retrospective cohort.
Setting & Participants:
Patients with CKD4–5 in Medicare 2007 to 2014.
Exposure or Predictor:
Ischemic stroke in CKD4–5.
Outcomes:
Initiation of maintenance dialysis or death.
Analytical Approach:
Cox proportional hazard modeling assessed factors associated with ischemic stroke. A matched analysis (stroke/no stroke) estimated the cumulative incidence of incident kidney failure and death, treated as competing events. Simulations using a state transition model determined differences in expected time to kidney failure or death and death alone for patients with and without stroke with CKD5.
Results:
123,251 patients with CKD4 and 22,054 with CKD5 were identified. Mean ages were 81.0 and 79.2 years, respectively. Female sex (HRs of 1.21 [95% CI, 1.12–1.31] and 1.39 [95% CI, 1.04–1.86] for CKD4 and CKD5, respectively) and black race (HRs of 1.25 [95% CI, 1.12–1.39] and 1.12 [95% CI, 0.80–1.58] for CKD4 and CKD5, respectively) were factors associated with ischemic stroke. Rates for 30-day mortality were 13.3% and 18.8%, and for 1-year mortality, 40.0% and 38.2%. For patients with CKD5, kidney failure or death occurred an average of 3.6 months sooner for patients with an ischemic stroke, and death (irrespective of kidney failure), a mean of 24.3 months sooner.
Limitations:
Study design cannot determine causality; lack of data for stroke severity.
Conclusions:
Female sex and black race were associated with increased risk for stroke in CKD4 and CKD5. In CKD5, stroke was associated with a shorter time to kidney failure or death by nearly 4 months, and to death, by more than 2 years.
Stroke, a common and potentially catastrophic health event,1 has profound implications for the affected individual and for society as a whole. As with most cardiovascular events, stroke occurs at a high rate in patients receiving maintenance dialysis. Reports have quantified its incidence and prevalence,2–7 geographic variation across the United States,8 and association with years of life lost.9
However, it is unclear whether catastrophic clinical events that occur in late-stage chronic kidney disease (CKD; ie, stages 4 and 5 [CKD4–5]) hasten CKD progression and the onset of dialysis. Regarding stroke, rates are known to increase abruptly before initiation of maintenance dialysis.10 This finding may suggest that stroke (or perhaps any major cardiovascular event) is associated with more rapid progression to kidney failure and initiation of maintenance dialysis. However, because Murray et al10 were able to study only patients who survived to initiate dialysis (ie, who were “immortal” during the predialysis period), they could not quantify the association of stroke with time to dialysis initiation or with death.
To our knowledge, the association between ischemic stroke and time to dialysis initiation and/or death in older adults with CKD4–5 has not been explored. To more closely examine stroke epidemiology and its potential implications in late-stage CKD, we calculated stroke rates in patients with CKD4 and CKD5, modeled factors associated with ischemic stroke, and estimated the association between ischemic stroke and time to dialysis initiation or death. To do so, we used a novel claims-based algorithm to apportion patient time to CKD4 or CKD5 more precisely than has been done previously.
Our overall goal was to provide clinical insights for nephrologists, who must often attempt to contextualize how risks for dialysis initiation and for death might be affected by major cardiovascular events that occur during late-stage CKD. We reasoned that better understanding of the potential implications of stroke in late-stage CKD would help nephrologists to better counsel older patients who experience an ischemic stroke and potentially to adjust dialysis planning in response.
Methods
Data Sources
The 20% Centers for Medicare & Medicaid Services (CMS) billing claims files for 2007 to 2014 were used to determine CKD stage (and the estimated duration of each stage), outcomes (ischemic stroke and initiation of maintenance dialysis), presence of comorbid conditions, and a putative marker of disability based in part on claims for the use of durable medical equipment, described in more detail later.
Study Design and Cohort Construction
This study used a retrospective cohort design. Prevalent patients with CKD with at least 2 International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) claims for CKD4 (585.4) and/or CKD5 (585.5) between January 1, 2008, and December 31, 2014, who were 65 years and older and younger than 110 years and had been insured by Medicare Parts A and B for at least 1 year were identified. Identification of CKD stage was based on claims. The algorithm we used, which was designed to overcome traditional limitations of other approaches, is described in detail in Item S1. Patients were censored if they required either maintenance hemodialysis or peritoneal dialysis or underwent kidney transplantation, but not if they underwent transient dialysis (such as for acute kidney injury). Analyses for patients with CKD4 and CKD5 were undertaken independently and in parallel; patients could contribute time to both stages if they transitioned from CKD4 to CKD5.
Demographic and Comorbid Condition Variables
Age, race, and sex were included for covariate control in the analysis, as were major comorbid conditions, ascertained over 1 year before the stage index date by the presence of 1 or more inpatient or 2 or more outpatient/Part B claims separated by at least 30 days. Missing data were rare: when race was missing (<1%), patients were categorized as other/unknown. No values were missing for sex. Census region included a “missing” category, but absolute numbers were low, requiring cell size suppression. Conditions assessed and the associated identification approach are shown in Item S1 and Table S1.
To create a claims-based score designed to reflect putative disability status, we used a variation of a previously developed claims-based algorithm11,12 to assign a proxy disability score, which we have previously found to be associated with outcomes. This approach is detailed in Item S1 and Table S2.
Outcomes
Ischemic stroke was determined using an algorithm originally used by Go et al13 and previously used by our group.8,9 Death was determined by Medicare date of death. Incident kidney failure was determined using claims for dialysis originating from dialysis units (“clinics”) or kidney transplant procedure codes. We then searched for inpatient claims with ICD-9-CM code 585.6 within the preceding 3 days to determine inpatient initiation of dialysis.
Statistical Analysis
Descriptive statistics (counts and percentages) were used to characterize patients with CKD. Ischemic stroke rates were for each CKD stage group ascertained overall and by strata of age, sex, race, disability proxy score, diabetes, census region, and prior stroke history. Cox proportional hazard modeling was performed to assess the factors associated with ischemic stroke in this population. Two sensitivity analyses were undertaken, designed to address the potential for collider bias. First, histories of ischemic and hemorrhagic stroke were eliminated from the models. Second, patients with a history of ischemic or hemorrhagic stroke were excluded from the models.
Patients with ischemic stroke were then matched (1:3) with individuals who did not experience stroke. Matching was performed exactly by CKD stage, presence of atrial fibrillation, prior ischemic stroke history, sex, disability proxy score, and Liu comorbidity score group.14 Matching also required age and CKD stage duration to be within 2.5 years. The success of the match was evaluated using a standardized difference < 0.1. Thus, the “index date” was the date of first ischemic stroke for patients with stroke and the corresponding number of days (years) post–stage-index for controls. We then used the 1-minus-Kaplan-Meier survival probability to estimate the cumulative incidence of kidney failure and death, treating these as competing events, and associated 95% confidence intervals (CIs) at 30, 90, and 180 days poststroke using bootstrapping (2,000 iterations).
A simulation was performed using a state transition model to determine the difference in expected time to: (1) kidney failure or death and (2) death irrespective of kidney failure for patients with CKD5 who did and did not experience ischemic stroke. The model assumed a Weibull distribution for time to death and separately for time to kidney failure or death. Parameters of the Weibull distribution were estimated from the matched cohorts. The fitted distributions, when compared with the observed, were highly similar. Note that because: (1) death rates, particularly in CKD5, were so much higher than rates of kidney failure, and (2) death represents informative rather than random censoring that is conditional on observed patient characteristics, the differences in time to kidney failure (alone) between patients with stroke and those without stroke cannot be compared.
Compliance and Protection of Human Research Participants
The research protocol was approved by the Institutional Review Board at Hennepin Healthcare, which provided a waiver of informed consent because all data were deidentified. Data Use Agreements between the Hennepin Healthcare Research Institute and the US Renal Data System and CMS were in place.
Results
Construction of the analytic cohort is shown in Figure 1. After relevant exclusions, 123,251 patients with CKD4 and 22,045 with CKD5 were identified. Characteristics are shown in Table 1.
Figure 1.
Construction of the analytic cohorts. 1Entry into a given chronic kidney disease (CKD) stage (eg, stage 5, stage 4) occurred on the date of the first claim for that stage, assuming there was 1 or more future claim for a less advanced stage. See Item S1 for a fuller description of how CKD stage was determined. Note that individuals were continually eligible for inclusion in the analytic cohort for their entire follow-up (ie, if an individual initially failed to meet criteria for inclusion into the stage 5 analytic cohort, the individual could be eligible at a later date, thereby contributing time to the appropriate stage). 2Age had to be between 65 and 110 years, inclusive. 3Represents individuals with less advanced CKD (eg, stages 1–4) who were potentially eligible for analysis in the stage 4 analytic cohort if they did not satisfy criteria for inclusion in the stage 5 analytic cohort. Thus, patients who did not formally meet stage 5 criteria were eligible for inclusion in the stage 4 analytic cohort (while remaining eligible to contribute time to the stage 5 analytic cohort if their CKD later advanced). Abbreviation: LVAD, left ventricular assist device.
Table 1.
Characteristics of Cohorts by Kidney Disease Stage
| CKD4 | CKD5 | |
|---|---|---|
| No. of patients | 123,251 | 22,054 |
| Years at stage | 1.18 ± 1.28 | 0.56 ± 0.88 |
| Age, y | 81.0 ± 8.2 | 79.2 ± 8.1 |
| Female sex | 64,988 (52.7%) | 11,255 (51.0%) |
| Race | ||
| White | 100,646 (81.7%) | 15,818 (71.7%) |
| Black | 15,327 (12.4%) | 4,238 (19.2%) |
| Other/unknown | 7,278 (5.9%) | 1,998 (9.1%) |
| Census region | ||
| Northeast | 23,253 (18.9%) | 4,224 (19.2%) |
| Midwest | 31,156 (25.3%) | 4,892 (22.2%) |
| South | 49,535 (40.2%) | 9,281 (42.1%) |
| West | 18,993 (15.4%) | 3,570 (16.2%) |
| Missing | 314 (0.3%) | 87 (0.4%) |
| Comorbid conditions | ||
| ASHD | 61,561 (50.0%) | 10,279 (46.6%) |
| CHF | 60,296 (48.9%) | 10,440 (473%) |
| PVD | 40,921 (33.2%) | 7,163 (32.5%) |
| Other cardiac | 38,471 (31.2%) | 6,502 (29.5%) |
| COPD | 34,450 (28.0%) | 5,339 (24.2%) |
| GI bleeding | 10,921 (8.9%) | 2,083 (9.4%) |
| Liver disease | 3,529 (2.9%) | 617 (2.8%) |
| Arrhythmias | 38,594 (31.3%) | 5,949 (27.0%) |
| Atrial fibrillation | 31,351 (25.4%) | 4,482 (20.3%) |
| Cancer | 20,615 (16.7%) | 3,510 (15.9%) |
| Diabetes | 64,976 (52.7%) | 12,476 (56.6%) |
| Hypertension | 115,928 (94.1%) | 21,061 (95.5%) |
| Anemia | 77,482 (62.9%) | 17,280 (78.4%) |
| Chronic infections | 1,360 (1.1%) | 316 (1.4%) |
| Prior hemorrhagic stroke | 184 (0.2%) | 41 (0.2%) |
| Prior ischemic stroke | 1,931 (1.6%) | 349 (1.6%) |
| Disability proxy score | 1.25 ± 2.1 | 1.26 ± 2.3 |
| Disability score category | ||
| ≤0 | 71,887 (58.3%) | 13,081 (59.3%) |
| 1–2 | 24,087 (19.5%) | 4,336 (19.7%) |
| 3–4 | 17,951 (14.6%) | 2,948 (13.4%) |
| 5–6 | 6,140 (5.0%) | 1,024 (4.6%) |
| ≥7 | 3,186 (2.6%) | 665 (3.0%) |
Note: Values for continuous variables given as mean ± standard deviation; for categorical variables, as count (percentage).
Abbreviations: ASHD, atherosclerotic heart disease; CHF, congestive heart failure; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; GI, gastrointestinal; PVD, peripheral vascular disease.
Stroke Rates and Factors Associated With Stroke
Observed rates of ischemic stroke are shown by CKD stage in Table 2, overall and by key characteristics. Overall ischemic stroke rates were 21.1 (95% CI, 20.3–21.8) and 17.3 (95% CI, 15.0–19.6) per 1,000 patient-years in CKD4 and CKD5, respectively. In both stages, observed rates were higher for women than for men, for older than for younger patients, for black than for white patients, and for patients with than without diabetes.
Table 2.
Stroke Rates per 1,000 Patient-Years by Kidney Disease Stage
| Rate (95% CI) | ||
|---|---|---|
| CKD4 | CKD5 | |
| All participants | 21.1 (20.3–21.8) | 17.3 (15.0–19.6) |
| Sex | ||
| Male | 18.7 (17.7–19.8) | 14.3 (11.2–17.4) |
| Female | 22.9 (21.9–24.0) | 19.9 (16.5–23.3) |
| Age category | ||
| <70 y | 15.4 (13.6–1 7.3) | 9.5 (5.2–13.8) |
| 70–80 y | 17.1 (16.0–18.2) | 19.3 (15.3–23.2) |
| ≥80 y | 24.9 (23.8–26.0) | 18.4 (14.9–22.0) |
| Race | ||
| White | 20.7 (19.9–21.5) | 17.9 (15.1–20.7) |
| Black | 25.03 (22.7–27.4) | 19.9 (14.2–25.6) |
| Other/unknown | 18.05 (15.2–20.9) | —a |
| Census region | ||
| Northeast | 21.3 (19.6–23.0) | 15.3 (10.4–20.2) |
| Midwest | 21.1 (19.7–22.6) | 22.6 (16.8–28.4) |
| South | 21.6 (20.4–22.8) | 16.7 (13.2–20.2) |
| West | 19.1 (17.3–20.9) | 14.7 (9.4–20.1) |
| Missing | —a | —a |
| Prior ischemic stroke | ||
| No | 20.3 (19.5–21.0) | 16.5 (14.2–18.9) |
| Yes | 81.2 (68.4–94.0) | 83.3 (36.2–130.5) |
| Diabetes | ||
| No | 20.1 (19.1–21.1) | 13.9 (11.0–16.9) |
| Yes | 22.0 (21.0–23.1) | 20.6 (1 7.0–24.2) |
| Disability proxy score | ||
| ≤0 | 19.0 (18.1–19.8) | 14.3 (11.7–16.9) |
| 1–2 | 24.4 (22.4–26.4) | 23.1 (16.7–29.5) |
| ≥3 | 26.4 (24.3–28.5) | 23.0 (16.4–29.7) |
Abbreviations: CI, confidence interval; CKD, chronic kidney disease.
Values suppressed if number less than 11.
Factors associated with ischemic stroke, by CKD stage, are shown in Table 3. In fully adjusted models, age of 80 years or older (vs <70 years), female sex, and black (compared with white) race were associated with higher risk for ischemic stroke for both CKD stages. For example, the hazard for ischemic stroke was about 20% to 40% higher for women than for men (hazard ratios [HRs] of 1.21 for CKD4 and 1.39 for CKD5) and 12% to 25% for black than for white patients (HRs of 1.25 for CKD4 and 1.12 for CKD5). The presence of atrial fibrillation was significantly associated with ischemic stroke.
Table 3.
Factors Associated With Ischemic Stroke by Kidney Disease Stage
| CKD4 | CKD5 | |||
|---|---|---|---|---|
| HR (95% CI) | P | HR (95% CI) | P | |
| Age category | ||||
| <70 y | 1.00 (reference) | 1.00 (reference) | ||
| 70–80 y | 1.13 (0.98–1.30) | 0.1 | 1.85 (1.12–3.05) | 0.02 |
| ≥80 y | 1.59 (1.39–1.83) | <0.001 | 1.74 (1.05–2.88) | 0.03 |
| Sex | ||||
| Male | 11.00 (reference) | 1.00 (reference) | ||
| Female | 1.21 (1.12–1.31) | <0.001 | 1.39 (1.04–1.86) | 0.03 |
| Race | ||||
| White | 1.00 (reference) | 1.00 (reference) | ||
| Black | 1.25 (1.12–1.39) | <0.001 | 1.12 (0.80–1.58) | 0.5 |
| Other/unknown | 0.95 (0.80–1.12) | 0.5 | 0.45 (0.23–0.90) | 0.03 |
| Census region | ||||
| Northeast | 1.00 (reference) | 1.00 (reference) | ||
| Midwest | 1.02 (0.91–1.13) | 0.8 | 1.37 (0.90–2.07) | 0.1 |
| South | 1.04 (0.94–1.15) | 0.4 | 1.05 (0.71–1.55) | 0.8 |
| West | 0.96 (0.84–1.10) | 0.6 | 1.10 (0.67–1.81) | 0.7 |
| Missing | 1.33 (0.68–2.57) | 0.4 | 0.00 (0.00–0.00) | 0.9 |
| Comorbid condition | ||||
| ASHD | 1.09 (1.00–1.18) | 0.05 | 1.05 (0.77–1.43) | 0.8 |
| CHF | 1.05 (0.96–1.15) | 0.3 | 0.83 (0.59–1.15) | 0.3 |
| PVD | 1.06 (0.98–1.15) | 0.2 | 1.40 (1.04–1.89) | 0.03 |
| Other cardiac | 1.06 (0.97–1.16) | 0.2 | 1.03 (0.73–1.44) | 0.9 |
| COPD | 1.02 (0.94–1.12) | 0.6 | 0.81 (0.56–1.17) | 0.3 |
| GI bleeding | 1.11 (0.97–1.27) | 0.2 | 1.11 (0.69–1.79) | 0.7 |
| Liver disease | 0.95 (0.72–1.25) | 0.7 | 0.68 (0.22–2.13) | 0.5 |
| Arrhythmias | 0.97 (0.89–1.06) | 0.5 | 0.89 (0.63–1.26) | 0.5 |
| Atrial fibrillation | 1.45 (1.33–1.58) | <0.001 | 1.53 (1.08–2.15) | 0.02 |
| Cancer | 0.84 (0.75–0.94) | 0.003 | 0.56 (0.35–0.92) | 0.02 |
| Diabetes | 1.09 (1.01–1.17) | 0.03 | 1.34 (1.00–1.79) | 0.05 |
| Hypertension | 1.28 (1.07–1.52) | 0.006 | 0.95 (0.49–1.83) | 0.9 |
| Anemia | 0.90 (0.83–0.97) | 0.009 | 1.09 (0.78–1.51) | 0.6 |
| Chronic infections | 1.42 (1.01–2.02) | 0.05 | 0.00 (0.00–0.00) | 0.9 |
| Prior hemorrhagic stroke | 1.06 (0.40–2.83) | 0.9 | 3.03 (0.42–21.88) | 0.3 |
| Prior ischemic stroke | 3.49 (2.95–4.14) | <0.001 | 3.28 (1.76–6.10) | <0.001 |
| Disability score category | ||||
| ≤0 | 1.00 (reference) | 1.00 (reference) | ||
| 1–2 | 1.00 (0.91–1.11) | 0.9 | 1.38 (0.98–1.96) | 0.07 |
| 3–4 | 1.00 (0.89–1.12) | 0.9 | 1.25 (0.80–1.95) | 0.3 |
| ≥5 | 0.93 (0.79–1.09) | 0.4 | 1.29 (0.75–2.20) | 0.4 |
Abbreviations: ASHD, atherosclerotic heart disease; CHF, congestive heart failure; CI, confidence interval; CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; GI, gastrointestinal; HR, hazard ratio; PVD, peripheral vascular disease.
For sensitivity analyses, several additional models were run. First, models were run in which histories of previous stroke (ischemic or hemorrhagic) were eliminated as terms in the model; results, as shown in Table S3 for patients with CKD4, were relatively unchanged. Second, models were run in which patients with a history of an ischemic or hemorrhagic stroke were excluded. These results (as in Table S3) were also virtually unchanged. Analogous results for patients with CKD5 are shown in Table S4.
Association of Stroke With Death and Kidney Failure
To determine the association of ischemic stroke with time to dialysis or death in patients with CKD4 or CKD5, matching was undertaken. Results of the stroke/no stroke matching process for patients with CKD 4 and, separately, CKD5 are shown in Table S5. The matching strategy generated highly similar groups, as intended. To compare and contrast patients who were matched with those who were unmatched, matched and unmatched patients by stroke/no stroke status and CKD stage are shown in Table S6.
Risk for death, kidney failure, and the combination of death or kidney failure (explicitly considering kidney failure and death as competing risks) are shown in Figure 2A (CKD5) and B (CKD4); results are shown such that the differences between the 2 outcome curves represent the death (alone) outcome. Corresponding cumulative incidence probabilities (and associated bootstrapped 95% CIs) at months 1, 3, and 6 and years 1 and 2 are shown in Table 4.
Figure 2.
Risk for death and the combination of death or kidney failure in (A) chronic kidney disease stage 5 (CKD5) and (B) CKD4. *For patients without stroke, days since stroke represents the number of days following a counterfactual stroke, that is, a day on which a matched patients without stroke could have had a stroke, but did not. Abbreviation; ESKD, end-stage kidney disease.
Table 4.
Bootstrapped Percentages for the Outcomes of Dialysis, Death, or the Combination of Dialysis Plus Death at Various Time Points by CKD Stage
| CKD4 | CKD5 | |||||
|---|---|---|---|---|---|---|
| Stroke | No Stroke | P | Stroke | No Stroke | P | |
| Kidney failurea | ||||||
| 30 d | 6.8% (5.8%-7.7%) | 3.3% (2.9%-3.6%) | <0.01 | 25.7% (19.7%-32.4%) | 11.7% (9.1%-14.5%) | <0.001 |
| 90 d | 11.3% (10.2%-12.5%) | 7.9% (7.3%-8.5%) | <0.01 | 35.9% (29.2%-43.2%) | 32.1% (27.9%-36.4%) | 0.3 |
| 6 mo | 14.6% (13.3%-16.0%) | 13.0% (12.2%-13.7%) | 0.03 | 45.7% (38.7%-53.2%) | 46.8% (42.4%-51.3%) | 0.8 |
| 1 y | 19.8% (18.2%-21.3%) | 20.9% (19.9%-21.9%) | 0.2 | 50.0% (43.2%-57.6%) | 60.1% (55.6%-64.4%) | 0.02 |
| 2 y | 26.5% (24.8%-28.3%) | 29.8% (28.6%-30.9%) | 0.001 | 54.2% (47.0%-62.0%) | 67.8% (63.4%-72.1%) | 0.002 |
| Deathb | ||||||
| 30 d | 22.4% (20.9%-24.0%) | 5.7% (5.3%-6.3%) | <0.001 | 21.2% (15.0%-27.5%) | 4.3% (2.7%-6.1%) | <0.001 |
| 90 d | 31.1% (29.4%-32.8%) | 12.2% (11.5%-12.9%) | <0.001 | 27.5% (20.7%-34.3%) | 7.5% (5.3%-9.8%) | <0.001 |
| 6 mo | 37.6% (35.7%-39.4%) | 18.2% (17.3%-19.1%) | <0.001 | 31.5% (24.6%-38.4%) | 10.1% (7.7%-12.8%) | <0.001 |
| 1 y | 45.7% (43.8%-47.8%) | 25.7% (24.7%-26.8%) | <0.001 | 34.5% (27.4%-41.3%) | 14.1% (11.2%-17.3%) | <0.001 |
| 2 y | 53.5% (51.5%-55.6%) | 34.3% (33.1%-35.6%) | <0.001 | 37.4% (30.0%-44.4%) | 16.6% (13.3%-20.0%) | <0.001 |
| Kidney failure or deathc | ||||||
| 30 d | 29.2% (27.4%-30.9%) | 9.0% (8.4%-9.6%) | <0.001 | 46.9% (40.0%-54.8%) | 16.0% (12.9%-19.3%) | <0.001 |
| 90 d | 42.4% (40.6%-44.2%) | 20.1% (19.2%-21.0%) | <0.001 | 63.3% (55.8%-70.3%) | 39.6% (35.3%-43.9%) | <0.001 |
| 6 mo | 52.2% (50.3%-54.1%) | 31.2% (30.1%-32.2%) | <0.001 | 77.2% (71.0%-83.6%) | 56.9% (52.6%-61.3%) | <0.001 |
| 1 y | 65.5% (63.6%-67.4%) | 46.6% (45.4%-47.8%) | <0.001 | 84.4% (78.8%-89.6%) | 74.2% (70.4%-78.1%) | 0.007 |
| 2 y | 80.0% (78.3%-81.9%) | 64.1% (62.9%-65.4%) | <0.001 | 91.4% (86.9%-95.6%) | 84.4% (81.0%-87.7%) | 0.01 |
Note: Values expressed as hazard ratio (95% confidence interval).
Abbreviation: CKD, chronic kidney disease.
Death considered as a competing risk.
Kidney failure considered as a competing risk.
Whichever outcome occurred first.
For patients with CKD5, stroke was associated with hastening of death and of kidney failure, although the association was more pronounced for the former. For example, the percentage of patients with stroke who died at 30 days (21.2%; 95% CI, 15.0%−27.5%) was higher than the percentage of patients without stroke who died at 2 years (16.6%; 95% CI, 13.3%−20.0%). Ischemic stroke was associated with substantially higher risk for kidney failure in the short but not the long term: at 30 days, the percentage of patients with stroke with kidney failure was 25.7% (95% CI, 19.7%−32.4%) versus 11.7% (95% CI, 9.1%−14.5%) of patients without stroke. The cumulative incidence probability curves for kidney failure from patients with and without stroke crossed at approximately day 140; by this time, approximately three-quarters of patients who experienced a stroke either developed kidney failure or died. In contrast, this was the case for less than half the patients without stroke. Patterns for patients with CKD4 were qualitatively similar, but percentages of all groups reaching the outcomes at any given time point were lower.
Results of the state transition model demonstrated that for patients with CKD5, the outcome of kidney failure or death occurred an average of 6.5 months sooner for patients who experienced an ischemic stroke than for those who did not. For patients who survived to develop kidney failure, it occurred an average of 3.6 months sooner, while death (irrespective of kidney failure) occurred a mean of 24.3 months sooner in patients who experienced an ischemic stroke.
Discussion
In this study, we sought to understand the potential implications of stroke in patients with late-stage CKD not requiring kidney replacement therapy. We found that ischemic stroke rates were extremely high in patients with CKD4 and CKD5 and that women and black patients were substantially more likely to experience an ischemic stroke than were men or patients of other races. Risk for kidney failure following an ischemic stroke more than doubled over the ensuing month and remained elevated for several months thereafter in patients with CKD4 or CKD5. In patients with CKD5, stroke was associated with shorter time to kidney failure or death by 6.5 months, and to death (irrespective of kidney failure) by more than 2 years. For patients who survived to develop kidney failure, it occurred about 3.5 months sooner. This large-scale observational study quantifies the association of ischemic stroke with kidney failure and death in patients with late-stage CKD not requiring kidney replacement therapy.
Our main finding is that ischemic stroke in older individuals with advanced CKD is associated with a period of substantially increased risk for progression to kidney failure or death in the ensuing months. We chose to study stroke for 2 reasons. First, CKD appears to be associated with increased risk for ischemic stroke, making stroke’s potential implications important to study. Although reported findings are heterogeneous, the balance of the evidence appears to suggest that ischemic stroke risk increases as estimated glomerular filtration decreases.15 Second, how stroke is associated with kidney failure is an important clinical question that arose from previous work10 that had shown that in patients who survived to initiate dialysis, stroke rates rapidly increased as dialysis initiation approached.
However, the association between stroke and dialysis initiation per se could not be determined using such a study design because it considered only patients with CKD who survived to (and who were judged appropriate candidates for) dialysis initiation. In contrast, the present design followed all patients forward from the onset of CKD4 or CKD5. By so doing, we found that an ischemic stroke was associated with a substantially higher likelihood of kidney failure over the ensuing 4 months among patients with CKD5, even after consideration of high short-term death rates. This suggests a substantially heightened period of danger not only for death (as would be expected), but also for progression to kidney failure after an ischemic stroke.
Notably, the association of ischemic stroke with kidney failure diminished over time, such that the risk was about the same for CKD5 patients with and without stroke at about 4.5 months. This was likely a result of a phenomenon known as “depletion of susceptibles,”16,17 in which patients who experience a catastrophic event (in this case, ischemic stroke) are most likely to experience the outcome of interest (kidney failure) soon after the inciting event. In contrast, patients who do not develop kidney failure soon after an ischemic stroke may represent a cohort of ischemic stroke “tolerators.”
We found that rates of ischemic stroke in advanced CKD were about 4 times higher than in a similarly aged Medicare population,18 which is not unexpected given the high rates of cardiovascular events in patients with kidney disease.19,20 The association of several key factors with outcomes is particularly noteworthy. Compared with white race, black race presented 1.3-fold (CKD4) and 1.1-fold (CKD5) increased risk for ischemic stroke (although only in CKD4 was the HR significantly different from unity). Although our study population and design differ from others, making direct comparisons imperfect, black compared with white race has consistently been shown to be a factor associated with risk for ischemic stroke in populations not selected on the basis of CKD.
For example, Koton et al,21 using data from the ARIC Study, showed that black race was associated with 1.2- to 1.7-fold increase in total stroke incidence (depending on the geographical location of the participants sampled), while Kissela et al,22 drawing on data from the Greater Cincinnati/Northern Kentucky Stroke Study, estimated the risk ratio of blacks compared with whites for first-ever stroke to be 1.8 in persons aged 65 to 74 years. Similarly, findings from REGARDS showed the stroke incidence rate ratio to be approximately 1.6 for black as compared with white persons aged 65 to 74 years.23 Our somewhat weaker HRs suggest the possibility that as CKD advances, black-white differences may become less pronounced.
Regarding diabetes, we also found a less pronounced association with ischemic risk than was found in the ARIC cohort,21 in which patients with diabetes 65 years and older had 1.5-fold increased risk for stroke compared with patients without diabetes, which suggests that “traditional” risk factors may become somewhat less important with very advanced CKD.
This might also apply to sex. Perhaps our most unexpected finding, that of increased risk among women, stands in contrast to most studies, which report that male sex is a risk factor for stroke.21,22,24 Reasons for this remain uncertain, although the precise magnitude of our HRs should be interpreted with caution given that our model-building approach was not designed to elicit precise HRs for any prespecified factor.
Late-stage CKD, a period of substantial health risks, is also the time for careful planning for a potential transition to maintenance dialysis. This transition represents a considerable challenge given that the duration of CKD4 or CKD5 is often short, highly variable, and beset with major clinical challenges. Major intercurrent events resulting in hospitalization contribute to these challenges by altering the relative immediacy of the many competing clinical priorities. Therefore, an understanding of how a major clinical event may affect death and incident kidney failure is important for patients and their physicians.
Unfortunately, our results demonstrate the considerable challenge of counseling older ischemic stroke survivors with advanced CKD. Merely accelerating the discussion about risks and benefits of dialysis, and if dialysis is considered appropriate, working rapidly to accomplish initiation-related milestones such as modality education and vascular access placement may not be the correct approach. Because death rates are also high following an ischemic stroke, much higher than rates of kidney failure, as would be expected,25 many older ischemic stroke survivors will die before they require dialysis. A sober discussion between nephrologists and patients about the risks for death and for progression to kidney failure is therefore required.
Future work should study how older ischemic stroke survivors (or survivors of any catastrophic event that occurs during late-stage CKD) fare following dialysis initiation and would likely contribute insights to dialysis planning; if older stroke survivors do especially poorly after dialysis initiation, this might suggest that conservative care is a relatively more attractive option than dialysis initiation.
We are uncertain as to the implications of findings for younger individuals with advanced CKD. In general, catastrophic events such as stroke might be expected to have fewer potential downstream consequences in younger as compared with older individuals. As such, time to kidney failure or death following an ischemic stroke in younger individuals might be longer than the estimate we provide. However, this is far from certain because individuals who have reached CKD4 or CKD5 at a relatively young age may be unexpectedly vulnerable to poor outcomes following a major cardiovascular event such as a stroke. Elucidation of this issue awaits study in a large database enriched in younger persons with advanced CKD.
Our study had several limitations. Because we studied only Medicare beneficiaries, the cohorts of individuals with CKD4 and CKD5 were relatively elderly, consistent with previous work.10 Thus, our findings cannot be safely extrapolated to other types of patients, particularly younger ones. Also, because ours was an observational study, we cannot determine causality, only temporal antecedence, and confounding, as with all observational studies, is a significant threat. As such, we cannot know whether ischemic stroke itself actually hastened kidney failure, whether an ischemic stroke is a marker of ill health that is collinear with advancing CKD, or whether dialysis initiation is viewed as helpful in the overall management of critically ill patients. Misclassification is also a substantial threat when using administrative claims data because some patients’ level of kidney function likely was not reflected accurately in the CKD claims; some patients without CKD5 were undoubtedly incorrectly assigned to this analytic cohort, while some true patients with CKD5 were almost certainly not identified. We also lacked patient-level data for estimated glomerular filtration, relying on medical billing claims to determine CKD stage. This is a significant limitation because estimated glomerular filtration can vary widely within a given stage. We also lacked access to degree of CKD-related abnormalities such as anemia, secondary hyperparathyroidism, uncontrolled hypertension, or volume overload, all of which factor into when a patient initiates dialysis and could be sources of residual confounding in our analysis. Finally, we lacked a measure of stroke severity, such as the National Institutes of Health stroke scale score, by which to adjust our models.
In conclusion, we found that ischemic stroke rates are high in older Medicare beneficiaries with late-stage CKD. Compared with white patients, black patients are more likely to experience an ischemic stroke in CKD4 or CKD5. Ischemic stroke is associated with higher risk for death and progression to dialysis-requiring kidney failure in the immediately ensuing months. Future work regarding survival following dialysis initiation in patients with CKD who experience an ischemic stroke or any other major clinical event may inform the nephrology community about the optimal approach during the potential transition from late-stage CKD to kidney failure for such patients.
Supplementary Material
Item S1: Supplemental methods.
Table S1: ICD-9-CM diagnosis and V codes used to define comorbid conditions.
Table S2: HCPCS and ICD-9-CM diagnosis codes used to define the disability proxy score.
Table S3: CKD4, original, without previous strokes as model terms, and excluding those with prior stroke.
Table S4: CKD5, original, without previous strokes as model terms, and excluding those with prior stroke.
Table S5: Ischemic stroke versus no ischemic stroke, matched cohorts, CKD4 and CKD5.
Table S6: Comparisons of unmatched and matched CKD4 no-stroke patients, unmatched and matched CKD4 stroke patients, and unmatched and matched CKD5 no-stroke patients.
PLAIN-LANGUAGE SUMMARY.
We studied whether an ischemic stroke “hastens” the time to death or incident kidney failure in patients with advanced chronic kidney disease. Using Medicare data, we matched patients who experienced an ischemic stroke to similar patients who did not. For patients with stage 5 chronic kidney disease who experienced a stroke, death or incident kidney failure occurred an average of 6.5 months sooner than for patients who did not experience a stroke; death, irrespective of kidney failure, occurred about 2 years sooner. Patients and their physicians should understand that a stroke likely puts them at increased risk for death or need for dialysis and should consider this when discussing plans to potentially transition to dialysis.
Acknowledgements:
The authors thank Chronic Disease Research Group colleagues Anne Shaw for manuscript preparation and Nan Booth, MSW, MPH, ELS, for manuscript editing.
Support: This study was funded by grant R56 HL132373 to Dr Wetmore. The funder had no role in study design; data collection, analysis, or reporting; or the decision to submit for publication.
Footnotes
Financial Disclosure: The authors declare that they have no relevant financial interests.
Contributor Information
James B. Wetmore, Chronic Disease Research Group, Hennepin Healthcare Research Institute Department of Medicine, Hennepin Healthcare, Minneapolis, MN.
Charles A. Herzog, Chronic Disease Research Group, Hennepin Healthcare Research Institute Department of Medicine, Hennepin Healthcare, Minneapolis, MN.
Anne Sexter, Chronic Disease Research Group, Hennepin Healthcare Research Institute.
David T. Gilbertson, Chronic Disease Research Group, Hennepin Healthcare Research Institute
Jiannong Liu, Chronic Disease Research Group, Hennepin Healthcare Research Institute.
Scott E. Kasner, Department of Neurology, University of Pennsylvania, Philadelphia, PA
References
- 1.Berry JD, Dyer A, Cai X, et al. Lifetime risks of cardiovascular disease. N Engl J Med. 2012;366(4):321–329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Saran R, Robinson B, Abbott KC, et al. US Renal Data System 2017 Annual Data Report: epidemiology of kidney disease in the United States. Am J Kidney Dis. 2018;71(3)(suppl 1):S1–S676. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Wiesholzer M, Harm F, Tomasec G, Barbieri G, Putz D, Balcke P. Incidence of stroke among chronic hemodialysis patients with nonrheumatic atrial fibrillation. Am J Nephrol. 2001;21(1):35–39. [DOI] [PubMed] [Google Scholar]
- 4.Seliger SL, Gillen DL, Longstreth WT Jr, Kestenbaum B, Stehman-Breen CO. Elevated risk of stroke among patients with end-stage renal disease. Kidney Int. 2003;64(2):603–609. [DOI] [PubMed] [Google Scholar]
- 5.Sozio SM, Armstrong PA, Coresh J, et al. Cerebrovascular disease incidence, characteristics, and outcomes in patients initiating dialysis: the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) study. Am J Kidney Dis. 2009;54(3):468–477. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Sanchez-Perales C, Vazquez E, Garcia-Cortes MJ, et al. Ischaemic stroke in incident dialysis patients. Nephrol Dial Transplant. 2010;25(10):3343–3348. [DOI] [PubMed] [Google Scholar]
- 7.Power A, Chan K, Singh SK, Taube D, Duncan N. Appraising stroke risk in maintenance hemodialysis patients: a large single-center cohort study. Am J Kidney Dis. 2012;59(2):249–257. [DOI] [PubMed] [Google Scholar]
- 8.Wetmore JB, Ellerbeck EF, Mahnken JD, et al. Stroke and the “stroke belt” in dialysis: contribution of patient characteristics to ischemic stroke rate and its geographic variation. J Am Soc Nephrol. 2013;24(12):2053–2061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Wetmore JB, Phadnis MA, Ellerbeck EF, Shireman TI, Rigler SK, Mahnken JD. Relationship between stroke and mortality in dialysis patients. Clin J Am Soc Nephrol. 2015;10(1):80–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Murray AM, Seliger S, Lakshminarayan K, Herzog CA, Solid CA. Incidence of stroke before and after dialysis initiation in older patients. J Am Soc Nephrol. 2013;24(7):1166–1173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Wetmore JB, Liu J, Wirtz HS, et al. Geovariation in fracture risk among patients receiving hemodialysis. Clin J Am Soc Nephrol. 2016;11(8):1413–1421. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Wetmore JB, Yan H, Hu Y, Gilbertson DT, Liu J. Factors associated with withdrawal from maintenance dialysis: a case-control analysis. Am J Kidney Dis. 2018;71(6):831–841. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Go AS, Hylek EM, Chang Y, et al. Anticoagulation therapy for stroke prevention in atrial fibrillation: how well do randomized trials translate into clinical practice? JAMA. 2003;290(20):2685–2692. [DOI] [PubMed] [Google Scholar]
- 14.Liu J, Huang Z, Gilbertson DT, Foley RN, Collins AJ. An improved comorbidity index for outcome analyses among dialysis patients. Kidney Int. 2010;77(2):141–151. [DOI] [PubMed] [Google Scholar]
- 15.Lee M, Saver JL, Chang KH, Liao HW, Chang SC, Ovbiagele B. Low glomerular filtration rate and risk of stroke: meta-analysis. BMJ. 2010;341:c4249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Miettinen OS, Caro JJ. Principles of nonexperimental assessment of excess risk, with special reference to adverse drug reactions. J Clin Epidemiol. 1989;42(4):325–331. [DOI] [PubMed] [Google Scholar]
- 17.Renoux C, Dell’Aniello S, Brenner B, Suissa S. Bias from depletion of susceptibles: the example of hormone replacement therapy and the risk of venous thromboembolism. Pharmacoepidemiol Drug Saf. 2017;26(5):554–560. [DOI] [PubMed] [Google Scholar]
- 18.Fang MC, Coca PM, Ghosh K, Cutler DM, Rosen AB. Trends in stroke rates, risk, and outcomes in the United States, 1988 to 2008. Am J Med. 2014;127(7):608–615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Go AS, Chertow GM, Fan D, McCulloch CE, Hsu CY. Chronic kidney disease and the risks of death, cardiovascular events, and hospitalization. N Engl J Med. 2004;351(13): 1296–1305. [DOI] [PubMed] [Google Scholar]
- 20.Matsushita K, van der Velde M, Astor BC, et al. Association of estimated glomerular filtration rate and albuminuria with all-cause and cardiovascular mortality in general population cohorts: a collaborative meta-analysis. Lancet. 2010;375(9731): 2073–2081. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Koton S, Schneider AL, Rosamond WD, et al. Stroke incidence and mortality trends in US communities, 1987 to 2011. JAMA. 2014;312(3):259–268. [DOI] [PubMed] [Google Scholar]
- 22.Kissela B, Schneider A, Kleindorfer D, et al. Stroke in a biracial population: the excess burden of stroke among blacks. Stroke. 2004;35(2):426–431. [DOI] [PubMed] [Google Scholar]
- 23.Howard VJ, Kleindorfer DO, Judd SE, et al. Disparities in stroke incidence contributing to disparities in stroke mortality. Ann Neurol. 2011;69(4):619–627. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Howard G, Cushman M, Kissela BM, et al. Traditional risk factors as the underlying cause of racial disparities in stroke: lessons from the half-full (empty?) glass. Stroke. 2011;42(12): 3369–3375. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.O’Hare AM, Choi AI, Bertenthal D, et al. Age affects outcomes in chronic kidney disease. J Am Soc Nephrol. 2007;18(10): 2758–2765. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Item S1: Supplemental methods.
Table S1: ICD-9-CM diagnosis and V codes used to define comorbid conditions.
Table S2: HCPCS and ICD-9-CM diagnosis codes used to define the disability proxy score.
Table S3: CKD4, original, without previous strokes as model terms, and excluding those with prior stroke.
Table S4: CKD5, original, without previous strokes as model terms, and excluding those with prior stroke.
Table S5: Ischemic stroke versus no ischemic stroke, matched cohorts, CKD4 and CKD5.
Table S6: Comparisons of unmatched and matched CKD4 no-stroke patients, unmatched and matched CKD4 stroke patients, and unmatched and matched CKD5 no-stroke patients.