Outcomes of continuous kidney replacement therapy versus intermittent hemodialysis in dialysis-dependent patients with acute intracerebral hemorrhage

Hideaki Watanabe; Akira Okada; Keita Shibahashi; Hiroki Matsui; Kiyohide Fushimi; Hideo Yasunaga

doi:10.23876/j.krcp.24.297

Kidney Res Clin Pract > Epub ahead of print

Watanabe, Okada, Shibahashi, Matsui, Fushimi, and Yasunaga: Outcomes of continuous kidney replacement therapy versus intermittent hemodialysis in dialysis-dependent patients with acute intracerebral hemorrhage

Original Article

Kidney Research and Clinical Practice 2025;j.krcp.24.297.

Published online: September 9, 2025

DOI: https://doi.org/10.23876/j.krcp.24.297

Outcomes of continuous kidney replacement therapy versus intermittent hemodialysis in dialysis-dependent patients with acute intracerebral hemorrhage

Hideaki Watanabe¹

, Akira Okada²

, Keita Shibahashi³

, Hiroki Matsui⁵

, Kiyohide Fushimi⁴

, Hideo Yasunaga⁵

¹Department of Eat-loss Medicine, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan

²Department of Prevention of Diabetes and Lifestyle-Related Diseases, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan

³Tertiary Emergency Medical Center, Tokyo Metropolitan Bokutoh Hospital, Tokyo, Japan

⁴Department of Health Policy and Informatics, Tokyo Medical and Dental University, Tokyo, Japan

⁵Department of Clinical Epidemiology and Health Economics, The University of Tokyo, Tokyo, Japan

Correspondence: Hideaki Watanabe Department of Eat-loss Medicine, Graduate School of Medicine, The University of Tokyo, 7-3-1 Hongo, Bunkyo City, Tokyo 1130033, Japan. E-mail: hwatana@m.u-tokyo.ac.jp

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial and No Derivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/) which permits unrestricted non-commercial use, distribution of the material without any modifications, and reproduction in any medium, provided the original works properly cited.

Abstract

Background

This study was performed to investigate whether continuous kidney replacement therapy is more beneficial than intermittent hemodialysis in patients with acute intracerebral hemorrhage undergoing chronic dialysis, while adjusting for more clinically important variables.

Methods

Dialysis-dependent patients diagnosed with acute intracerebral hemorrhage were identified using the Japanese Diagnosis Procedure Combination database. We compared the in-hospital mortality rates and improvement in modified Rankin Scale scores before intracerebral hemorrhage development to discharge among patients who received continuous kidney replacement therapy or intermittent hemodialysis within 2 days of hospitalization. Overlap weighting based on propensity scores was performed to adjust for potential confounders.

Results

Among 922 eligible patients, 204 received continuous kidney replacement therapy and 718 received intermittent hemodialysis within 2 days of hospitalization. Propensity score overlap weighting analyses showed no significant difference between the continuous kidney replacement therapy and the intermittent hemodialysis groups in terms of in-hospital mortality (45.8% vs. 46.1%; risk difference, –0.3%; 95% confidence interval, –8.8 to 8.2%; p = 0.94) or decline in modified Rankin Scale (–3.1 vs. –3.2; difference 0.1; 95% confidence interval, –0.3 to 0.5; p = 0.64).

Conclusion

In this nationwide retrospective cohort study, continuous kidney replacement therapy and intermittent hemodialysis initiated within 2 days of admission showed no significant difference in in-hospital mortality or changes in modified Rankin Scale scores among dialysis-dependent patients with acute intracerebral hemorrhage.

Keywords: Cerebral hemorrhage, Continuous renal replacement therapy, Epidemiology, Renal dialysis, Hospital mortality

Introduction

Dialysis-dependent patients face an increased risk of stroke, with studies indicating an 8- to 10-fold higher likelihood compared to the general population [1,2]. Studies have shown that patients with stroke with chronic kidney disease have higher mortality and neurological deterioration rates than those without [3–5]. This association is reportedly more apparent among patients undergoing dialysis [6,7].

The use of intermittent hemodialysis (IHD) or continuous kidney replacement therapy (CKRT) has been discussed for patients with kidney failure complicated by the acute phase of stroke. Compared with IHD, CKRT is theoretically expected to achieve greater stability in cerebral perfusion pressure and less intracranial pressure elevation and is considered superior to IHD in treating patients with acute stroke [1,8]. The 2012 KDIGO (Kidney Disease: Improving Global Outcomes) guidelines recommend using CKRT rather than IHD for patients with acute kidney injury who develop acute brain injury or other causes of increased intracranial pressure or generalized brain edema [9]. However, this statement is mainly based on case reports and single-armed studies [10–13].

The comparative effectiveness of CKRT and IHD in patients with dialysis-dependent stroke remains unclear. A recent small-scale study revealed a superior consciousness level at 1 week after acute stroke in the CKRT-treated group than in the IHD-treated group, of which approximately half of the patients presented with intracerebral hemorrhage [14]. Conversely, a large-scale cohort study using the United States Renal Data System reported higher mortality rates in the CKRT-treated group than in the IHD-treated group [2]. However, these previous studies failed to adjust for important confounders, such as the Glasgow Coma Scale, modified Rankin Scale, Barthel Index, and type of stroke on admission.

Type-specific analyses are warranted because stroke type affects prognosis and pathophysiology [15]. Intracerebral hemorrhage is the second most common cause of stroke after ischemic stroke and accounts for 9% to 27% of all stroke cases [16,17]. Therefore, the present study focused on intracerebral hemorrhage and aimed to compare the mortality and morbidity rates between CKRT and IHD in patients with dialysis-dependent stroke with acute intracerebral hemorrhage. We used a Japanese national inpatient database, which includes information on the type of stroke, anatomical location of the hemorrhage, and disease severity, to overcome the abovementioned limitations of previous studies.

Methods

Data source

We used data from the Diagnosis Procedure Combination (DPC) database, a national inpatient database of administrative claims data, and discharge abstracts from Japan. The details of the database are described elsewhere [18]. A validation study of the diagnostic procedure code database suggested high specificity and sensitivity of procedure records, whereas most diagnoses had high specificity but moderate sensitivity [18].

This study was approved by the Institutional Review Board of the Graduate School of Medicine, University of Tokyo (No. 3501-[5]). The requirement for informed consent was waived due to data anonymity.

Study design and population

We extracted data from patients aged ≥18 years who were admitted between July 1, 2010, and March 31, 2020. The inclusion criteria were as follows: a recorded diagnosis of intracerebral hemorrhage was recorded as the primary diagnosis with the International Classification of Diseases and Related Health Problems, 10th revision (ICD-10) code I61, a brain computed tomography or magnetic resonance imaging examination performed within 2 days of hospitalization, and kidney replacement therapy administered within 2 days of hospitalization. We only included patients with a Glasgow Coma Scale score of ≤8 points on admission because we assumed that a clearer difference in effectiveness would be observed among severely affected patients.

Exclusions were made for patients transferred from different medical facilities, those diagnosed with acute kidney injury on admission, and those discharged within 2 days of hospitalization. Additionally, cases where ≥3 days had passed since the onset of intracerebral hemorrhage were excluded because the main research interest was the association between acute-phase intracranial pressure management and outcomes. In instances of multiple eligible hospitalizations for the same patient, we selected the initial admission for analysis.

Study outcomes and variables

The primary outcome assessed was in-hospital mortality, with the secondary outcome being the difference in the modified Rankin Scale between the periods “before intracerebral hemorrhage development” and “on hospital discharge.”

Data collected from the database encompassed various patient characteristics, including age, body mass index, sex, consciousness level on admission, smoking status, activities of daily living (ADLs), comorbidities, locus of stroke, and the use of drugs and procedures within 2 days of hospitalization. Additionally, information was gathered regarding whether the institution was an academic hospital, whether the patient was transported by ambulance, and the fiscal year of admission.

Upon admission, the level of consciousness was evaluated using the Japan Coma Scale, which was subsequently converted to the Glasgow Coma Scale, given its strong correlation [19]. The Charlson Comorbidity Index was calculated as previously described [20]. The modified Rankin Scale before intracerebral hemorrhage is determined based on the patient’s history and present illness at admission. This assessment reflects the patient’s ADLs approximately 1 week before intracerebral hemorrhage onset [21]. Diagnoses of atrial fibrillation, excluding concomitant mitral valve stenosis or status after mitral valve replacement, were identified using Japanese text, while diagnoses of diabetes mellitus, hypertension, and dyslipidemia were detected using the ICD-10 codes of E10–E14; I10, I11.0, I11.9, I12.0, I12.9, I13.9, I15.0–I15.2, and I15.9; and E78.0–E78.5, respectively.

Hemorrhages in specific brain regions, such as the putamen and thalamus, were identified using Japanese text, while subcortical, cerebellum, brainstem, intraventricular, multiple hemorrhages, and other intracerebral hemorrhages, including nontraumatic intracerebral hemorrhage in the hemisphere, were detected using ICD-10 codes I610, I614, I613, I615, I616, I611, I612, I618, and I619, respectively. ADLs were recorded using the Barthel Index as described previously [22]. The database records 10 components of the Barthel Index: feeding, bathing, grooming, dressing, bowels, bladder, toilet, transfers, mobility, and stairs. The total obtainable Barthel Index score was 100 points; however, considering the possibility of anuria in patients undergoing hemodialysis, the urination component was excluded, yielding a total score of 90 points. Patients were categorized as physically dependent or independent using a cut-off value of 90% of the total score, as previously employed [23].

Information regarding the following medications administered within 2 days of hospitalization was collected: vitamin K reversal; protamine; prothrombin complex concentrate; mannitol, glycerol; vasopressors (epinephrine, norepinephrine, dopamine, dobutamine, isoprenaline, ivabradine, etilefrine, olprinone, colforsin, denopamine, vasopressin, pimobendane, phenylephrine, bucladesine, vesnarinone, milrinone, ubidecarenone, or atropine); anti-arrhythmic (Class IA, Class IB, Class IC, amiodarone, digitalis, sotalol, nifekalant, bepridil, verapamil); anti-hypertensive drugs (propranolol, diltiazem, nicardipine, verapamil); anti-peptic ulcer disease drug; insulin; red blood cell transfusion; tranexamic acid; albumin; and fresh frozen plasma. Additionally, information was gathered on the use of the following devices or facilities within 2 days of hospitalization: cardiopulmonary resuscitation, external ventricular drainage, craniotomy, stereotactic aspiration of intracerebral hematoma, endoscopic hematoma evacuation, rehabilitation, intensive care unit, ventilator, and deep vein thrombosis prophylaxis.

Although the exact cause of death is not available in the DPC database, for patients who died during hospitalization, the database includes a binary variable indicating whether the patient died of the disease that consumed the most medical resources during hospitalization. Based on this, we summarized the cause of death for patients where this information was available.

Statistical analyses

First, patient demographics were summarized based on whether they received CKRT or IHD within 2 days of hospitalization. After adjusting for potential confounders, we performed weighted generalized linear model analyses with a binomial distribution and log-link function to determine risk differences for the primary outcome and modified Rankin Scale score differences for the secondary outcome.

We used overlap weighting based on the propensity score to balance background characteristics between the two groups [24,25]. Instead of propensity score matching, propensity score-based overlap weighting, which has distinct advantages, was applied: it retains all samples (unlike propensity score matching, which excludes some) and ensures better covariate balance by reducing standardized mean differences to zero [25]. Overlap weighting has been shown to outperform the inverse probability of treatment weighting in terms of bias reduction [26]. The propensity score was estimated using logistic regression analysis, with CKRT being the dependent variable. The calculation of propensity score necessitates the inclusion of all available variables that are clinically relevant and may influence the outcome of interest, the following independent variables: patient characteristics, drugs administered (e.g., vasopressors, transfusion), procedures performed (e.g., craniotomy, cardiopulmonary resuscitation, or endoscopic hematoma evacuation), and locus of intracerebral hemorrhage (putamen, thalamus, subcortical, cerebellum, brainstem, intraventricular, multiple hemorrhages, and other intracerebral hemorrhages, including nontraumatic intracerebral hemorrhage in the hemisphere). We calculated propensity score C-statistics to assess the ability to differentiate between the two groups. Then, the estimated propensity score was used to derive overlap weights, defined as 1 – the propensity score for patients who underwent CKRT and the propensity score for those who underwent IHD. We compared patient backgrounds between the two groups before and after weighting. Standardized mean differences were used to evaluate the differences between the groups, and an absolute standardized mean difference of >10% indicated an imbalance [27].

We also performed an instrumental variable (IV) analysis to adjust for unmeasured confounders. We defined IV as the hospital’s preference for CKRT because the use of CKRT for intracerebral hemorrhage may depend on the hospital’s preference and expertise. We used the two-stage residual inclusion method for IV analysis [28]. In the first-stage model, we examined the association between IV and treatment assignment while adjusting for covariates. We determined the raw residual for each patient by calculating the difference between the model-predicted probability of CKRT administration and the actual CKRT administration. In the second-stage model, we included the residuals as additional covariates and analyzed the associations between treatment and the primary outcome. The IV analysis utilized robust standard error estimation. To verify the strength of the IV, we used a partial F test with a value >10 indicating that the instrument is not weak [29].

All statistical analyses were performed using the Stata/SE version 17.0 software (StataCorp.). A two-tailed significance level of p < 0.05.

Results

Study population

We identified 978 eligible patients undergoing chronic dialysis who developed intracerebral hemorrhage. After excluding 56 patients, we included 922 eligible patients from 423 hospitals: 204 received CKRT and 718 received IHD within 2 days of hospitalization. A flow chart of the patient selection process is presented in Fig. 1.

Among the 204 patients who received CKRT, 92 died during hospitalization. Of these, 81 patients (88.0%) were recorded as having died due to the disease that required the most medical resources. The breakdown of causes was as follows: intracerebral hemorrhage (n = 76), sepsis (n = 1), disseminated intravascular coagulation (n = 1), and end-stage kidney disease (n = 3). Among the 718 patients who received IHD, 329 died. Of these, 295 (89.7%) were reported to have died from the disease requiring the most medical resources, including intracerebral hemorrhage (n = 294) and sepsis (n = 1).

Table 1 presents the patients’ background characteristics, stratified according to whether they underwent CKRT or IHD. Overall, he mean age was 64 years, and the overall in-hospital mortality rate was 45.7%. Before implementing overlap weights, patients who received CKRT were more likely to be younger, had lower Glasgow Coma Scale scores on admission, higher modified Rankin Scale scores before intracerebral hemorrhage development, higher prevalence of ambulance use, underwent more craniotomies, and received deep vein thrombosis prophylaxis. After implementing the overlapping weights, the distribution of the patient characteristics was well-balanced. The C-statistic of the propensity score was 0.79 points, and the distribution of propensity scores in the unweighted and weighted models is shown in Supplementary Fig. 1 (available online). The kernel density plot of the propensity scores (Supplementary Fig. 1B, available online) may still show differences in the shape of the distributions between groups. This is because overlap weighting balances the covariates, not the entire distribution of the propensity scores.

Outcomes

Primary analysis

Table 2 presents the results of the generalized linear regression analysis with and without overlap weighting for the outcomes of the primary analysis. Patients who received CKRT had lower in-hospital mortality rates than those who did not; the unweighted analysis yielded a risk difference of –0.7% (95% confidence interval [CI], –8.5% to 7.0%; p = 0.86), while the weighted analysis yielded a risk difference of –0.3% (95% CI, –8.8% to 8.2%; p = 0.94). We observed comparable proportions of modified Rankin Scale differences between before intracerebral hemorrhage development and on hospital discharge in the CKRT and IHD groups; the unweighted analysis yielded a scale difference of –0.2 (95% CI, –0.6 to 0.1; p = 0.22), while the weighted analysis yielded a scale difference of 0.1 (95% CI, –0.3 to 0.5; p = 0.64).

The IV analysis, where the F statistic was calculated as 3,075.80 (>10), did not show a significant difference in the in-hospital mortality risk between those groups (risk difference, –2.7%; 95% CI, –16.7% to 11.3%; p = 0.71).

Discussion

In this observational study using real-world data from a national inpatient database, we examined the use of CKRT within 2 days of hospitalization, in comparison to IHD, among patients with chronic kidney disease on dialysis for acute intracerebral hemorrhage. Our findings revealed that CKRT administration was not associated with decreased in-hospital mortality or improvements in modified Rankin Scale scores from before intracerebral hemorrhage development to hospital discharge.

Our findings highlight the limited effectiveness of CKRT in improving the prognosis of this specific patient population, for which there exists insufficient evidence from randomized controlled trials. Despite our efforts to adjust for disease severity by restricting the analysis to cases with a Glasgow Coma Scale score ≤8 on admission and analyzing a larger sample size (n = 922 vs. n = 35) compared to a prior study [14], no significant differences were observed. This could be attributed to Japanese physicians restricting the IHD blood flow rate in accordance with the Japanese guidelines, which recommend this practice if CKRT is unavailable during the acute period of intracerebral hemorrhage [30]. Theoretically, this could lead to a similar blood flow rate as that achieved with CKRT, supporting the previous article’s perspective [31] that IHD can be administered safely if the IHD prescription is adjusted carefully.

Our study showed a similar mortality rate of 45.7% and a mean hospitalization duration of 46.7 days, consistent with findings from a previous study [15], which reported a relatively stable mortality rate of approximately –40% at 1 month and 54% at 1 year after intracerebral hemorrhage, despite advances in neurocritical care over the recent decades. Although we included patients with similar severity, our results with multivariate adjustment yielded different results from those of a previous study [2], such as comorbidities and locus of stroke, use of drugs, and procedures within 2 days of hospitalization.

Our study had some limitations. First, although we employed the overlap weighting method based on propensity scores to mitigate differences in baseline characteristics that may have influenced the decision to administer CKRT, biases caused by unmeasured confounders could still be present. Detailed information on intracranial pressure, net total in-hospital volume of input and output, size of intracerebral hemorrhage, differences in supportive care in each facility, and dialysis prescriptions are examples of such unmeasured confounders. However, this would be negligible because we used the IV method in the sensitivity analysis. Even though overlap weighting enables each absolute standardized mean difference to be <10%, kernel density plots may not completely overlap (Supplementary Fig. 1, available online), as the distribution of the adjusted covariates is not expected to match exactly [32,33]. Second, patients who presented to the hospital at ≥3 days after intracerebral hemorrhage development were excluded, as our research’s main focus was on the acute phase of intracerebral hemorrhage. Despite these limitations, our 10-year dataset from a comprehensive database covering 90% of Japanese tertiary care hospitals offers valuable insights into the use of CKRT in patients with chronic kidney disease undergoing dialysis for acute intracerebral hemorrhage, particularly in the context of the challenges associated with conducting large-scale randomized controlled trials in this population. Third, although patients with acute kidney injury were excluded using the ICD-10 code N17, some patients who experienced acute kidney injury but did not receive this code may have been misclassified. Fourth, the DPC database does not contain direct information on the exact cause of death. However, it includes a binary variable indicating whether the patient died of the disease that required the most medical resources during hospitalization. Using this information, it was observed that most deaths in both CKRT and IHD groups were attributed to intracerebral hemorrhage. Fifth, detailed information regarding dialysis (median dialysis vintage, cause of end-stage kidney disease, mode of dialysis [peritoneal dialysis or hemodialysis]) was unknown in this study population, though over 95% of the Japanese population is receiving outpatient hemodialysis [34].

In conclusion, although a previous study [2] suggested the inferiority of CKRT compared to IHD, this retrospective cohort study using a national database with more extensive covariate adjustment showed no significant difference in in-hospital mortality between CKRT and IHD initiated within 2 days of admission among dialysis-dependent patients with acute intracerebral hemorrhage. Given the possibility of residual confounding due to unmeasured severity factors such as hemodynamic instability and intracranial pressure, further research using detailed clinical data is warranted to determine whether CKRT may offer potential benefits in more critically ill patients.

Supplementary Materials

Supplementary data are available at Kidney Research and Clinical Practice online (https://doi.org/10.23876/j.krcp.24.297).

j-krcp-24-297-Supplementary-Fig-1.pdf

Notes

Conflicts of interest

Akira Okada is a member of the Department of Prevention of Diabetes and Lifestyle-Related Diseases, a cooperative program between The University of Tokyo and the Asahi Mutual Life Insurance Company. The other authors have no other conflicts of interest to declare.

Funding

This work was supported by grants from the Ministry of Health, Labor and Welfare, Japan (23AA2003 and 22AA2003).

Data sharing statement

The dataset analyzed in the current study is not publicly available because of contracts with hospitals that provide data to the database.

Authors’ contributions

Conceptualization: HW, AO, KS, HY

Data curation: KF

Formal analysis: HW

Investigation: HW, HM

Supervision: AO, HY

Validation: AO, KS, HY

Writing–original draft: HW.

Writing–review & editing: AO, KS, HM, KF, HY

All authors read and approved the final manuscript.

Figure 1.

Selection of new-onset intracerebral hemorrhage with ESKD on patients undergoing dialysis from the Diagnosis Procedure Combination (DPC) database.

CKRT, continuous kidney replacement therapy; CT, computed tomography; ESKD, end-stage kidney disease; ICD-10, International Statistical Classification of Diseases and Related Health Problems, 10th Revision; IHD, intermittent hemodialysis; MRI, magnetic resonance imaging.

Table 1.

Characteristics of eligible patients with intracerebral hemorrhage before and after using overlap weights

Characteristic	Before using overlap weights			After using overlap weights
Characteristic	CKRT	IHD	SMD	CKRT	IHD	SMD
Age (yr), mean	59.5	64.3	–39.6	60.9	60.9	0.0
Male sex	65.0	66.0	–3.0	66.0	66.0	0.0
Smoking status
Nonsmoker	57.4	56.8	1.2	55.2	55.2	0.0
Current/past smoker	20.8	23.7	–7.0	22.3	22.3	0.0
Missing	21.8	19.5	5.7	22.5	22.5	0.0
Glasgow Coma Scale
3	26.2	19.9	15.1	24.1	24.1	0.0
6	40.6	44.9	–8.6	41.7	41.7	0.0
7	33.2	35.3	–4.4	34.2	34.2	0.0
Modified Rankin Scale
0	45.5	38.8	13.7	44.2	44.2	0.0
1	13.4	12.7	2.0	12.7	12.7	0.0
2	8.4	8.0	1.4	8.2	8.2	0.0
3	2.5	4.4	–10.4	2.7	2.7	0.0
4	3.0	6.9	–18.3	3.2	3.2	0.0
5	27.2	29.2	–4.4	29.0	29.0	0.0
Charlson Comorbidity index
1	68.8	64.9	8.4	68.1	68.1	0.0
2	16.8	16.1	2.0	17.3	17.3	0.0
≥3	14.4	19.0	–12.6	14.6	14.6	0.0
Diabetes mellitus	28.0	34.0	–12.8	31.0	31.0	0.0
Hypertension	55.0	54.0	2.9	55.0	55.0	0.0
Dyslipidemia	3.0	3.0	0.8	3.0	3.0	0.0
Body mass index (kg/m²)
<18.50	16.8	21.7	–12.4	17.2	17.2	0.0
18.50–24.9	62.9	54.0	18.0	60.4	60.4	0.0
25.0–29.9	9.9	12.4	–8.0	11.2	11.2	0.0
≥30.0	3.0	2.4	3.5	2.9	2.9	0.0
Missing	7.4	9.4	–7.3	8.2	8.2	0.0
Barthel Index score
<80	82.7	80.8	4.8	82.2	82.2	0.0
≥80	2.0	3.4	–8.7	2.3	2.3	0.0
Missing	15.3	15.8	–1.2	15.5	15.5	0.0
Academic hospital	15.0	11.0	14.2	15.0	15.0	0.0
Cardiopulmonary resuscitation^a	1.0	1.0	5.9	1.0	1.0	0.0
Atrial fibrillation	2.0	3.0	–7.1	2.0	2.0	0.0
Use of ambulance	95.0	90.0	19.3	94.0	94.0	0.0
Fiscal year
2010	9.4	7.9	5.4	9.9	9.9	0.0
2011	9.9	10.0	–0.4	10.1	10.1	0.0
2012	11.9	11.7	0.5	10.8	10.8	0.0
2013	11.9	10.6	4.1	11.7	11.7	0.0
2014	7.4	7.8	–1.3	6.5	6.5	0.0
2015	7.4	7.3	0.3	7.2	7.2	0.0
2016	7.9	8.6	–2.5	9.2	9.2	0.0
2017	7.9	10.6	–9.2	8.9	8.9	0.0
2018	8.4	8.5	–0.2	8.2	8.2	0.0
2019	10.4	7.6	9.7	9.3	9.3	0.0
2020	7.4	9.4	–7.3	8.1	8.1	0.0
Use of vitamin K^a	6.0	6.0	1.3	6.0	6.0	0.0
Use of mannitol^a	23.0	13.0	25.7	20.0	20.0	0.0
Use of glyceol^a	34.0	27.0	15.6	32.0	32.0	0.0
External ventricular drainage^a	18.0	16.0	5.4	18.0	18.0	0.0
Craniotomy^a	50.0	29.0	43.9	45.0	45.0	0.0
Hematoma evacuation^a	1.0	1.0	7.5	1.0	1.0	0.0
Endoscopic hematoma evacuation^a	5.0	5.0	1.4	5.0	5.0	0.0
Rehabilitation^a	17.0	28.0	–26.1	19.0	19.0	0.0
Use of vasopressors^a	51.0	28.0	50.2	45.0	45.0	0.0
Use of anti-arrythmics^a	7.0	4.0	13.2	6.0	6.0	0.0
Use of anti-hypertensives^a	96.0	95.0	2.7	97.0	97.0	0.0
ICU admission^a	37.0	31.0	13.2	36.0	36.0	0.0
Use of ventilator^a	77.0	51.0	56.3	72.0	72.0	0.0
DVT prophylaxis^a	63.0	48.0	31.1	61.0	61.0	0.0
Use of anti-peptic ulcer disease medications^a	89.0	84.0	13.6	88.0	88.0	0.0
Use of insulin^a	40.0	32.0	16.9	39.0	39.0	0.0
pRBC transfusion^a	23.0	12.0	28.7	20.0	20.0	0.0
Use of tranexamic acid^a	44.0	48.0	–7.5	45.0	45.0	0.0
Use of albumin^a	5.0	2.0	18.4	4.0	4.0	0.0
Use of FFP^a	15.0	6.0	28.4	12.0	12.0	0.0
Platelet transfusion^a	8.0	4.0	18.2	7.0	7.0	0.0
Anatomical hemorrhage location
Putamen	39.0	31.0	15.6	36.0	36.0	0.0
Thalamus	22.0	26.0	–8.3	24.0	24.0	0.0
Subcortical	11.0	17.0	–18.2	13.0	13.0	0.0
Brainstem	6.0	9.0	–11.7	7.0	7.0	0.0
Cerebellum	9.0	6.0	9.7	8.0	8.0	0.0
Intraventricular	5.0	5.0	2.1	5.0	5.0	0.0
Multiple hemorrhage	0.0	0.0	–9.2	0.0	0.0	–7.3
Others	8.0	6.0	9.1	8.0	8.0	0.0

Data are expressed as percent unless otherwise indicated.

CKRT, continuous kidney replacement therapy; DVT, deep vein thrombosis; FFP, freshly frozen plasma; ICU, intensive care unit; IHD, intermittent hemodialysis; pRBC, pure red blood cell; SMD, standardized mean difference.

^aWithin 2 days of hospitalization.

Table 2.

Results of generalized linear regression analysis comparing IHD and CKRT in patients with intracerebral hemorrhage

Outcome	Model	CKRT group (n = 204)	IHD group (n = 718)	Risk difference	95% CI	p-value
Primary outcome
In-hospital mortality (%)	Unweighted	45.1	45.8	–0.7	–8.5 to 7.0	0.86
In-hospital mortality (%)	Weighted	45.8	46.1	–0.3	–8.8 to 8.2	0.94
Sensitivity analysis: instrumental variable method^a
In-hospital mortality (%)	Weighted	44.8	47.4	–2.7	–16.7 to 11.3	0.71
Secondary outcome
Modified Rankin Scale difference^b	Unweighted	–3.1	–2.9	–0.2	–0.6 to 0.1	0.22
Modified Rankin Scale difference^b	Weighted	–3.1	–3.2	0.1	–0.3 to 0.5	0.64

CI, confidence interval; CKRT, continuous kidney replacement therapy; IHD, intermittent hemodialysis.

^aHospital preference for CKRT, the ratio of CKRT performed in all renal replacement therapies, was used as the instrumental variable.

^bThis was calculated by subtracting before intracerebral hemorrhage development from the discharge modified Rankin Scale score.

References

1. Davenport A. Continuous renal replacement therapies in patients with acute neurological injury. Semin Dial 2009;22:165–168.

2. Morgan MC, Waller JL, Bollag WB, et al. Association of intermittent versus continuous hemodialysis modalities with mortality in the setting of acute stroke among patients with end-stage renal disease. J Investig Med 2022;70:1513–1519.

3. Henry RM, Kostense PJ, Bos G, et al. Mild renal insufficiency is associated with increased cardiovascular mortality: the Hoorn Study. Kidney Int 2002;62:1402–1407.

4. Kon S, Konta T, Ichikawa K, et al. Association between renal function and cardiovascular and all-cause mortality in the community-based elderly population: results from the Specific Health Check and Guidance Program in Japan. Clin Exp Nephrol 2018;22:346–352.

5. Kumai Y, Kamouchi M, Hata J, et al. Proteinuria and clinical outcomes after ischemic stroke. Neurology 2012;78:1909–1915.

6. Lee JG, Lee KB, Jang IM, et al. Low glomerular filtration rate increases hemorrhagic transformation in acute ischemic stroke. Cerebrovasc Dis 2013;35:53–59.

7. Cherng YG, Lin CS, Shih CC, et al. Stroke risk and outcomes in patients with chronic kidney disease or end-stage renal disease: two nationwide studies. PLoS One 2018;13:e0191155.

8. Osgood M, Muehlschlegel S. POINT: should continuous venovenous hemofiltration always be the preferred mode of renal replacement therapy for the patient with acute brain injury? Yes. Chest 2017;152:1109–1111.

9. Khwaja A. KDIGO clinical practice guidelines for acute kidney injury. Nephron Clin Pract 2012;120:c179–c184.

10. Bagshaw SM, Peets AD, Hameed M, Boiteau PJ, Laupland KB, Doig CJ. Dialysis disequilibrium syndrome: brain death following hemodialysis for metabolic acidosis and acute renal failure: a case report. BMC Nephrol 2004;5:9.

11. Lin CM, Lin JW, Tsai JT, et al. Intracranial pressure fluctuation during hemodialysis in renal failure patients with intracranial hemorrhage. Acta Neurochir Suppl 2008;101:141–144.

12. Ronco C, Bellomo R, Brendolan A, Pinna V, La Greca G. Brain density changes during renal replacement in critically ill patients with acute renal failure: continuous hemofiltration versus intermittent hemodialysis. J Nephrol 1999;12:173–178.

13. Ghoshal S, Parikh A. Dialysis-associated neurovascular injury (DANI) in acute brain injury: practical considerations for intermittent dialysis in the neuro-ICU. Clin J Am Soc Nephrol 2021;16:1110–1112.

14. Matsui T, Nakagawa T, Kikuchi H, Horio H, Hashimura K. The effect of continuous renal replacement therapy with the AN69ST membrane on inflammatory markers and the level of consciousness of hemodialysis patients with stroke: comparison with hemodialysis with low blood flow rate. Pril (Makedon Akad Nauk Umet Odd Med Nauki) 2018;39:29–35.

15. Montaño A, Hanley DF, Hemphill JC 3rd. Hemorrhagic stroke. Handb Clin Neurol 2021;176:229–248.

16. Steiner T, Al-Shahi Salman R, Beer R, et al. European Stroke Organisation (ESO) guidelines for the management of spontaneous intracerebral hemorrhage. Int J Stroke 2014;9:840–855.

17. Feigin VL, Lawes CM, Bennett DA, Barker-Collo SL, Parag V. Worldwide stroke incidence and early case fatality reported in 56 population-based studies: a systematic review. Lancet Neurol 2009;8:355–369.

18. Yamana H, Moriwaki M, Horiguchi H, Kodan M, Fushimi K, Yasunaga H. Validity of diagnoses, procedures, and laboratory data in Japanese administrative data. J Epidemiol 2017;27:476–482.

19. Nakajima M, Okada Y, Sonoo T, Goto T. Development and validation of a novel method for converting the Japan Coma Scale to Glasgow Coma Scale. J Epidemiol 2023;33:531–535.

20. Quan H, Li B, Couris CM, et al. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am J Epidemiol 2011;173:676–682.

21. Ministry of Health Labour and Welfare. Explanation materials for the survey regarding the impact assessment of the introduction of the Diagnosis Procedure Combination (DPC) [Internet]. [cited 2024 Dec 02]. Available from: https://www.mhlw.go.jp/content/12404000/000923138.pdf

22. Wada T, Yasunaga H, Yamana H, Matsui H, Fushimi K, Morimura N. Development and validation of an ICD-10-based disability predictive index for patients admitted to hospitals with trauma. Injury 2018;49:556–563.

23. Okada A, Yamana H, Morita K, et al. Potassium concentration in initial fluid therapy and in-hospital mortality of patients with diabetic ketoacidosis. J Clin Endocrinol Metab 2021;106:e2162–e2175.

24. D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998;17:2265–2281.

25. Desai RJ, Franklin JM. Alternative approaches for confounding adjustment in observational studies using weighting based on the propensity score: a primer for practitioners. BMJ 2019;367:l5657.

26. Zhou Y, Matsouaka RA, Thomas L. Propensity score weighting under limited overlap and model misspecification. Stat Methods Med Res 2020;29:3721–3756.

27. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009;28:3083–3107.

28. Terza JV, Bradford WD, Dismuke CE. The use of linear instrumental variables methods in health services research and health economics: a cautionary note. Health Serv Res 2008;43:1102–1120.

29. Staiger D, Stock JH. Instrumental variables regression with weak instruments. Econometrica 1997;65:557–586.

30. Hirakata H, Nitta K, Inaba M, et al. Japanese Society for Dialysis Therapy guidelines for management of cardiovascular diseases in patients on chronic hemodialysis. Ther Apher Dial 2012;16:387–435.

31. Davenport A. Practical guidance for dialyzing a hemodialysis patient following acute brain injury. Hemodial Int 2008;12:307–312.

32. Li F, Morgan KL, Zaslavsky AM. Balancing covariates via propensity score weighting. J Am Stat Assoc 2018;113:390–400.

33. Thomas LE, Li F, Pencina MJ. Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial. JAMA 2020;323:2417–2418.

34. Japanese Society for Dialysis Therapy Renal Data Registry. 2022 Annual Dialysis Data Report, JSDT Renal Data Registry. Nihon Toseki Igakkai Zasshi 2023;56:473–536.