Analysis of 2-Year Survival Outcomes of Japanese Older Populations on Hemodiafiltration: A Propensity Score-Matched Study Based on Insurance Claims Data

Abstract

Despite the lack of evidence that suggests hemodiafiltration (HDF) offers a better survival outcome than standard hemodialysis (HD), the number of patients initiating HDF in Japan continues to rise. This study examined the temporal change in the number of HDF incidents, evaluated factors associated with all-cause mortality, and compared the mortality risk and survival time of patients on HDF with patients receiving standard HD in three sets of 2-year cohorts. The primary analyses included the insurance claims data of 460 HDF patients and propensity score-matched 903 standard HD patients who initiated dialysis therapy between 1 April 2012 and 31 March 2018. Patient follow-up was censored at the time of death or the end of the 2-year study period. The number of patients who initiated HDF and the proportion of all-cause mortality cases were evaluated. Additionally, the survival outcomes between propensity score-matched HDF and standard HD patient groups were compared throughout cohorts. The number of HDF patients increased throughout cohorts, but the proportions of mortality cases across cohorts slowly decreased. Adjusting for all study covariates, we observed that HDF patients had a lower mortality risk and longer survival time than patients on standard HD. This study supports the notion that HDF lowers all-cause mortality compared with standard HD in an incident dialysis population in Fukuoka Prefecture, Japan.

1. Introduction

In Japan, hemodiafiltration (HDF) and hemodialysis (HD) are the two most commonly used treatment modalities for end-stage renal disease (ESRD). Recent statistics showed that 55.1% of the dialysis population in Japan received HDF as the primary treatment, while 41.5% of the dialysis population in Japan received HD on a regular basis [1]. Although HD care has been available to the public as a treatment for ESRD through public insurance schemes since the 1960s, HDF was only officially recognized as a treatment for ESRD patients and reimbursed by the Japanese health insurance system in April 2012 [2,3]. Since then, the number of HDF patients has increased considerably.

Notable randomized clinical trials (RCTs), which have compared the outcomes of HDF with conventional HD, have been conducted in several European countries. However, the RCT results have been inconclusive, and only two studies have shown the benefit of HDF over conventional HD in terms of survival [4,5]. In the ESHOL trial examining the effectiveness of online HDF, the data revealed that HDF patients experienced a 30% lower risk of all-cause mortality, a 33% lower risk of cardiovascular mortality, and a 55% lower risk of infection-related mortality compared to those receiving standard HD care [4]. These encouraging findings were further supported by the more recent CONVINCE trial, wherein the HDF patient group had a significantly lower all-cause mortality rate than the high-flux HD patient group [5].

Nonetheless, all published RCTs contain a potential risk of bias, leading to overestimation or underestimation of the actual effect [6]. On the other hand, results from retrospective cohort studies showed favorable results for HDF, with statistically and significantly lower all-cause mortality or/and cardiovascular mortality among HDF patients [7,8,9].

Notwithstanding, the results of studies which were conducted in European countries could not be generalized to the Japanese population due to significant variations in practice [10]. Unlike the HD treatment for patients who treated in European countries, the use of low-flux HD in Japan is uncommon, as most patients are treated with high or super-high dialyzers [1,2,3]. Studies also showed that the average blood flow for Japanese patients was between 200 and 220 mL/min [3]. With such low average blood flow, a predilution treatment for HDF is commonly used due to the difficulty of substituting an adequate volume during treatment time (average of 4 h). Additionally, the volume of substitution fluid is also different between Europe and Japan. For on-line HDF patients in Japan, the mean volumes were 9.2 and 40.6 L for the post-and-predilution treatment, respectively [3,11]. The mean volume of 9.2 L for post-dilution on-line HDF is lower than that of European countries and the level reported to be necessary for a better outcome for on-line HDF patients. Likewise, favorable findings for on-line HDF reported in the ESHOL and CONVINCE trials were based on post-dilution mode. Due to a low average blood flow, this dilution mode is uncommon in Japan; it was estimated that less than 10% of HDF patients were treated with post-dilution on-line HDF [11].

There are not many comparative studies of HD and HDF survival among Asian populations that use population data over an extended period. A study that compares the secular trend in mortality between HD and HDF is currently unavailable. Therefore, this study was conducted to investigate the temporal change in mortality of patients in three sets of 2-year cohorts (1 April 2012–31 March 2014, 1 April 2014–31 March 2016, and 1 April 2016–31 March 2018). The researchers measured the hazard ratio and mean restricted survival time for patients who were treated with HD and HDF for mortality within 2 years of follow-up and compared the trends between these three cohorts. The differences in survival over time between HD and HDF treatments were also compared.

2. Methods

2.1. Study Design

This study was designed as a retrospective cohort study. By using the electronic database, all ESRD patients that began on-line HDF or HD treatment between 1 April 2012 and 31 March 2017 were identified. Based on the treatment initiation date, the patients were divided into three cohorts: Cohort 1 (1 April 2012–31 March 2014), Cohort 2 (1 April 2014–31 March 2016), and Cohort 3 (1 April 2016–1 April 2018). Patients in each cohort were followed for two years, starting from the day they received treatment until the day they died, emigrated to other prefectures, or by the end of 2 year of follow-up or whichever came first. The primary endpoint of the analysis was death.

2.2. Study Location

This research was carried out in Fukuoka Prefecture, which is located on Kyushu Island in the southwestern region of Japan. Dialysis care is provided by a total of 199 institutions that are located all over the prefecture [1]. These establishments are operated as ambulatory hemodialysis units which are attached to a hospital or as standalone hemodialysis centers or clinics.

2.3. Data Source

The data were mainly based on claims that were submitted to the Fukuoka Prefecture’s Insurance Association for the Elderly. In Japan, citizens aged ≥ 75 are entitled to medical coverage under the Latter-stage Health Insurance Scheme. Insurance coverage extends to patients aged 65 to 74 with a particular disability such as ESRD who require ongoing dialysis care. Insurance claim records were stored electronically. The data were extracted from the insurance database using SQL Server 2014 (Microsoft Corporation, Redmond, WA, USA).

2.4. Participant Identification

The population of this study was ESRD patients with active insurance status who began on-line HDF or HD treatment from 1 April 2012 to 31 March 2018. The International Classification of Disease, 10th revision, code N18.0 was used to identify chronic kidney disease diagnosis, and specific insurance codes were utilized to confirm patients’ HD or HDF maintenance status. This inclusion was limited to patients who received on-line HDF or HD exclusively as the primary treatment for ESRD. Thus, patients who primarily received peritoneal dialysis but also needed intermittent on-line HDF or HD care were excluded. Moreover, patients who changed treatment modality during the study period were excluded as well.

2.5. Matching

In this study, the baseline characteristics of HDF and HD patients were analyzed, and the propensity scores for each patient were calculated based on sex, age and Charlson’s comorbidity index score. Then, each HDF patient was matched with two patients on HD by applying a greedy matching algorithm with a caliper set at 0.2 [12,13].

2.6. Final Sample

A total of 460 HDF patients were identified and found to meet the inclusion criteria. These consisted of patients who began HDF treatment between 1 April 2012 and 31 March 2014 (Cohort 1, n = 48), 1 April 2014 and 31 March 2016 (Cohort 2, n = 110), and 1 April 2016 and 31 March 2018 (Cohort 3, n = 303). For a comparative analysis, 907 propensity score-matched patients who began HD treatment between 1 April 2012 and 31 March 2014 (Cohort 1, n = 96), 1 April 2014 and 31 March 2016 (Cohort 2, n = 220), and 1 April 2016 and 31 March 2018 (Cohort 3, n = 587) were identified. For Cohort 3, only 1 propensity score-matched HD case was found for 19 HDF cases. The claim data of all patients (N = 1367) were retrospectively reviewed, and the reported all-cause mortality cases were analyzed. Figure 1 shows the participant selection process.

2.7. Definition of Variables

The patients were categorized by sex and two age categories: 65–74 years old and ≥75 years old. The International Classification of Disease, 10th revision (ICD-10) codes were used to identify the status of heart failure (I11.0; I13.0; I42.0–I43; I50.0–I50.9), diabetes (E10, E11, E13, E14), cerebral stroke (I60.0–I66.9), malignancy (C00.0–C43.9; C45.0–C75.9; C76.0–D03.9; D05.00–D09.9), and dementia (F00, F01, F02, F03). By using unique patient identification, the data from the Long-term Care Insurance database were linked to identify the status of elderly care needs. In Japan, long-term care services are provided to the insured when people aged ≥65 and above require long-term care or support for daily activities. Available statuses include Support Level 1 (SL1) to Support Level 2 (SL2), and Care Level 1 (CL1) to Care Level 5 (CL5) [14]. A higher care level indicates a higher degree of physical disability and the need for support. In this study, the status was categorized into three groups: low-support (SL1–SL2 and CL1), moderate-support (CL2–CL3), and high-support (CL4–CL5).

The Charlson Comorbidity Index (CCI) was adapted and used to assess the presence and severity of comorbid conditions [15,16]. This validated instrument assigns weights (1, 3, or 6) to each of 19 major comorbid conditions that would likely influence treatment prognosis and survival outcomes. This study modified the weight (m-CCI) to exclude chronic kidney disease. Subsequently, three categories were created based on the calculated weights to represent the degree of severity: mild (m-CCI ≤ 1), moderate (m-CCI = 2–6), and severe (m-CCI ≥ 7). The use of the CCI to measure the severity of comorbid conditions based on insurance claims data in Japan has been widely documented [17,18].

2.8. Definition of Outcome

The primary outcome was all-cause mortality during the follow-up period. The insurance claim code 202 was used to determine mortality status. In the insurance data, such a code signifies loss of insurance eligibility due to death. Several epidemiological studies in Japan have used a claim-based definition of death. Past studies that validated such a mortality status indicated high specificity and positive predictive values, suggesting a low likelihood of outcome misclassification [19,20].

2.9. Data Analysis

In this study, data were analyzed by cohort (Cohort 1, Cohort 2, Cohort 3). Data in each cohort were also combined and analyzed separately. The descriptive statistics for categorical variables were expressed in numerical values and percentages. The possible trends of HDF enrollment and all-cause mortality cases across cohorts were evaluated using the Cochran–Armitage chi-squared test for trends and the Jonckheere–Terpstra test [21,22,23].

Cox regression analyses were performed to compare the mortality risks between HD and HDF patients, and restricted mean survival time (RMST) was calculated to measure the average survival time over the follow-up period. All reported p-values were two-tailed, and the significance level was set at p < 0.05. Stata Statistical Software: Release 17 (StataCorp LP, College Station, TX, USA) was used to analyze the data.

2.10. Ethical Considerations

This study used anonymized claim insurance data. Thus, the requirement to obtain informed consent was waived in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan. This study was also approved by the Institutional Review Board of Kyushu University (Clinical Bioethics Committee of the Graduate School of Medical Sciences, Kyushu University), approval Code: 2020-9.

3. Results

3.1. Baseline Characteristics of HDF Patients

The baseline characteristics of HDF patients were analyzed by cohorts. The number of patients that began HDF had increased dramatically across cohorts. The proportion of patients that were grouped by age and sex was almost balanced in all cohorts. Among patients who initiated HDF in Cohort 1, a significant number had a history of heart failure and cerebral stroke. A similar trend was also observed among those who started HDF in Cohort 2, where patients with heart failure and a history of cerebral stroke was prevalent. A high number of HDF patients with a history of heart failure was also observed in Cohort 3, but the proportion of patients with a history of cerebral stroke was slightly reduced when compared with those without such status. In all cohorts, most HDF patients had no records of support care needs. This indicated that these patients were most likely physically independent and did not require daily living assistance. We failed to observe a significant trend in the proportion of HDF patients when they were analyzed against baseline characteristics except for diabetes status.

Analyses of mortality cases showed a decreasing proportion of HDF patients who died across the cohort, despite the number of reported cases in each cohort continuing to rise. A high number of cases were reported among older patients as well as those with heart failure, malignancy, and dementia. In terms of the status of care level, a high number of mortality cases observed among patients who began HDF in Cohort 1 and Cohort 2 were those who did not require daily living assistance or support. However, in Cohort 3, those who require a moderate level of care showed the highest proportion of mortality cases.

Table 1. Descriptive statistics of HDF initiation and mortality cases within a 2-year follow-up.

Cohort	Cohort 1 (1 April 2012–31 March 2014)				Cohort 2 (1 April 2014–31 March 2016)				Cohort 3 (1 April 2016–31 March 2018)				All (1 April 2012–31 March 2018)
	HDF (N = 48)		Died (N = 16, 33%)		HDF (N = 110)		Died (N = 26, 24%)		HDF (N = 305)		Died (N = 57, 19%)		HDF (N = 463)		Died (N = 99, 21%)
	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	* p	§ p
Age
<75	24	50.0	8	33.3	58	52.7	7	12.1	151	49.5	14	9.27	233	50.3	29	12.5	0.74	0.34
≥75	24	50.0	8	33.3	52	47.3	19	36.5	154	50.5	43	27.9	230	49.7	70	30.4
Sex
Male	24	50.0	8	33.3	50	45.5	12	24.0	169	55.4	37	21.9	243	52.5	57	23.5	0.16	0.34
Female	24	50.0	8	33.3	60	54.6	14	23.3	136	44.6	20	14.7	220	47.5	42	19.1
Heart Failure
No	23	47.9	5	21.8	44	40.0	8	18.2	146	47.9	24	16.4	213	46.0	37	17.4	0.50	0.30
Yes	25	52.1	11	44.0	66	60.0	18	27.2	159	52.1	33	20.8	250	54.0	62	24.8
Diabetes
No	39	81.3	15	38.5	79	71.8	20	25.3	204	66.9	36	17.7	322	69.6	71	22.1	0.03	0.88
Yes	9	18.8	1	11.1	31	28.2	6	19.4	101	33.1	21	20.8	141	30.5	28	19.9
Malignancy
No	37	77.1	12	32.4	98	89.1	23	23.5	272	89.2	48	17.7	407	87.9	83	20.4	0.57	0.55
Yes	11	22.9	4	36.4	12	10.9	3	25.0	33	10.8	9	27.3	56	12.1	16	28.6
Stroke
No	22	45.8	7	31.9	52	47.3	12	23.1	167	54.8	34	20.4	241	52.1	53	21.9	0.12	0.64
Yes	26	54.2	9	34.6	58	52.7	14	24.1	138	45.3	23	16.7	222	47.9	46	20.7
Dementia
No	40	83.3	10	25.0	98	89.1	23	23.5	256	83.9	36	14.1	394	85.1	69	17.5	0.60	0.45
Yes	8	16.7	6	75.0	12	10.9	3	25.0	49	16.1	21	42.9	69	14.9	30	43.5
SCL
NA	33	68.6	15	93.8	79	71.8	16	61.5	206	67.5	32	18.4	318	68.7	63	24.7	0.29	0.01
Low	9	18.8	0	0.00	13	11.8	7	26.9	31	10.2	4	14.8	53	11.5	11	26.2
Moderate	5	10.5	0	0.00	12	10.9	0	0.00	44	14.4	14	46.7	61	13.2	14	23.4
High	1	2.08	1	6.25	6	5.45	3	11.5	24	7.87	7	41.1	31	6.70	11	55.0
m-CCI
Mild	19	39.6	6	39.6	31	28.2	3	9.68	108	35.4	20	18.5	158	35.1	29	18.4	0.67	0.81
Moderate	14	29.8	3	29.2	42	38.2	15	35.8	100	32.8	13	13.0	156	33.7	31	19.9
Severe	15	31.3	7	31.3	37	33.6	8	21.6	97	31.8	24	24.7	149	32.2	39	26.2

* p for trend for incident HDF proportions (variables SCL and m-CCI were tested using the Jonckheere-Terpstra test, and others were tested using Cochran-Armitage Test for trends). ^§ p for trend for the proportion of HDF patients who died (variables SCL and m-CCI were tested using the JonckheereTerpstra test, and others were tested using the Cochran-Armitage Test for trends). Abbreviations: HDF = hemodiafiltration; SCL = support and care level; NA = not applicable, m-CCI = modified Charlson Comorbidity Index.

summarizes the results of descriptive statistics of patients on HDF and the survival outcome within 2 years of follow-up according to the HDF initiation cohort.

3.2. Evaluation of Treatment Modality

The hazard ratios for patients on HDF were calculated by using patients on standard HD as a reference group. The analyses indicated a lower mortality risk among HDF patients than HD patients. However, the results were only statistically significant among patients who began HDF care in Cohort 2 and Cohort 3. By combining all propensity score-matched samples from three different cohorts, a lower mortality risk was also statistically evident. The calculated RMSTs for both groups also showed that patients with HDF treatment had longer survival time compared to those with HD treatment. There were also some improvements in terms of survival time based on cohorts. Although both modalities showed an improved survival outcome over time, HDF is still superior in terms of survival benefit as it offers a longer survival time.

Table 2. The estimates of 2-year mortality risk and survival time of older patients according to treatment initiation cohorts, treatment modalities, and statistical models.

	Cox Model			Restricted Mean Survival Time (RMST)
	HR	95% CI	p	M	95% CI	Diff	95% CI	p	Diff a	95% CI	p
Cohort 1
HD	Ref.			1.13	1.03–1.24
HDF	0.62	0.34–1.13	0.117	1.26	1.12–1.40	0.13	0.30–0.15	0.154	-	-	-
Cohort 2
HD	Ref.			1.15	1.06–1.24
HDF	0.33	0.22–0.51	<0.000	1.65	1.55–1.75	0.50	0.37–0.63	<0.000	0.45	0.31–0.59	<0.000
Cohort 3
HD	Ref.			1.19	1.14–1.28
HDF	0.27	0.20–0.36	<0.000	1.67	1.62–1.72	0.48	0.40–0.55	<0.000	0.45	0.37–0.52	<0.000
All
HD	Ref.			1.23	1.18–1.27
HDF	0.32	0.26–0.40	<0.001	1.70	1.65–1.75	0.47	0.41–0.54	<0.000	0.45	0.37–0.52	<0.000

HD = hemodialysis, HDF = hemodiafiltration, HR = hazard ratio, M = mean for survival time, Diff = difference. Sample was based on propensity score matching (1:2) using a Greedy algorithm with a caliper set at 0.20, and the estimates were adjusted for all study covariates. ^a Difference was adjusted for all covariates.

summarizes the results of the analyses.

Kaplan-Meier plots were constructed to compare the survival function by plotting the cumulative incidence proportion (1–survival). Log-rank chi-squared tests indicated a statistically significant difference in cumulative incidence proportion between HDF and HD groups for Cohorts 2 and 3. Figure 2 presents the Kaplan-Meier plots of cumulative incidence proportion (1–survival) for HDF and HD patients by cohorts.

4. Discussion

In this epidemiological cohort study, the researchers examined the incidence of ESRD patients on on-line HDF and evaluated the performance of HDF and standard HD in lowering all-cause mortality.

By evaluating patients’ survival on HDF and conventional HD, the findings suggested that HDF may confer a survival advantage compared to conventional HD among Japanese patients. The analysis results for the propensity score-matched model (HR 0.32, 95% CI 0.26–0.40, p < 0.001) showed a statistically significant lower mortality risk among HDF patients compared to patients with standard HD. Evaluation of survival estimates based on cohort also showed that the survival of HDF patients improved over time. These findings were not surprising, given that several studies had previously documented a superior performance of HDF over standard HD. A propensity score-matched cohort study that was conducted in Spain observed a significant 24% reduction in all-cause mortality among HDF patients (HR 0.76, 95% CI 0.62–0.94) compared with standard HD patients [9]. Likewise, a cohort study that used data from Australia and New Zealand also found that HDF patients were associated with a significantly lower risk of all-cause mortality (Australia: HR 0.79, 95% CI 0.72–0.87; New Zealand: HR 0.88, 95% CI 0.78–1.00) [7]. Comparatively, estimates of hazard ratios of this study were smaller, implying a significantly lower all-cause mortality risk. These results might be related to variations in HDF practice such as the prevalent use of predilution mode among on-line HDF. A previous study that was conducted in Japan also found the superiority of the predilution mode of on-line HDF compared to the post-dilution mode in lowering all-cause mortality [24]. Improving clinical practice in HDF care might also explain the improving survival risk across cohorts [25].

The estimates of survival benefits were based on the Cox regression hazard ratio. However, several researchers who examined the utility of such estimates posited that they are not intuitive to interpret and have restrictive assumptions of proportional hazards, and that their statistical power depends heavily on the number of events [26,27,28]. Although the median survival time difference provides an intuitive interpretation, the researchers also argued that such an estimate is insensitive to outliers and is often less precise [29]. In this study, the event rate was low (<50%); therefore, the median survival time and the difference could not be estimated. By considering data and methodological limitations, the analyses were supplemented with the calculation of RMST. As the statistical power depends on the exposed follow-up time, the RMST could provide a more precise estimate in case of a low event rate. The applicability of RMST in evaluating treatment effects has also been documented in many medical studies. Nonetheless, analyses based on RMST have also indicated that patients with HDF treatment had longer survival time in propensity score-matched analyses. Like Cox regression results, the survival benefits of HDF increased over time across all cohorts. With that, additional statistical evidence was provided to the effect that HDF provides superior patient survival.

The improvement in survival attributed to on-line HDF might also be linked to the high convection volumes, as indicated by previous studies. In the CONTRAST trial, no significant differences in overall mortality were found between the HD and on-line HDF groups [30]. However, further analysis revealed that on-line HDF patients with high convection volume (>21.85 L) experienced better survival outcomes [30]. In the ESHOL trial, which focused on high-efficiency post-dilution HDF patients with median convection volumes ranging from 22.9 to 23.9 L, a 30% decrease in overall mortality rate was reported compared to HD patients [4]. Despite these promising findings, it is crucial to consider the potential impact of selection bias. The intervention and control groups had some differences in their baseline characteristics. Upon examining the characteristics of patients in ESCHOL trial, it is clear that HDF patients were slightly younger, fewer were diabetic, and the median Charlson comorbidity index was lower. Additionally, the number of patients with central venous catheters for vascular access was statistically and significantly lower in the HDF group compared to the conventional HD group. Even though these characteristics were accounted for in the statistical analysis, it is still plausible that these group differences could have influenced the survival outcomes [31]. A similar positive outcome was also seen in the CONVINCE trial that restricted patients who received high-dose on-line HDF with a convection volume ≥23 L in post-dilution mode per session [5]. In this trial, a lower all-cause mortality risk was observed among patients with HDF than among those with conventional high-flux HD. While the results of these studies strongly imply that a high convection volume is needed for improved survival outcomes among post-dilution HDF patients, the positive effect of high-volume HDF could also be seen in pre-dilution HDF, as evidenced in a Japanese study that analyzed the survival outcomes between conventional HD and pre-dilution HDF patient groups with a substitution volume of ≥40 L/session, where an optimal substitution volume of 50.5 L was determined to improve survival outcome [32]. Perhaps, with this preponderance of evidence, it is safe to say that the most critical factor is not performing HDF but instead supplying enough convection volume during HDF [33].

This study has several limitations. As it had an observational design, HDF treatment was not assigned randomly. Therefore, causality could not be assessed. The absence of random assignment also increases the risk of selection and confounding biases, potentially leading to skewed inferences and erroneous conclusions. Despite the bias being statistically controlled using propensity score matching (PSM), it is essential to note that no method can completely eliminate selection bias, as the success of PSM is contingent upon the availability and accurate measurement of all relevant confounders. While PSM significantly reduces selection bias and improves the reliability of inferences in observational studies, it cannot fully account for all confounding variables, particularly unmeasured ones. Additionally, such a technique could not be used as a substitute for randomization because residual confounding or reverse causality phenomena could not be ruled out. In our study, the issue is further complicated as the data used in the analyses were primarily obtained from the claim records provided by the insurance association (company). As the records were purposely submitted by the health care providers for reimbursement, they rarely contain crucial clinical information such as patients’ prescriptions, type of vascular access, dialysis vintage, blood flow rate, serum albumin level, blood flow rate, and fluid volume replacement. Additionally, the mode of on-line HDF, whether pre-dilution or post-dilution, could not be determined from the claims record. Nonetheless, the available data suggest that over 90% of on-line HDF in Japan is pre-dilution HDF [11]. The standard characteristics of pre-dilution HDF are as follows: blood flow rate: 200–250 mL/min; substitution fluid volume: 40–60 L/session; dialysate flow: 500–600 mL/min; dialysis time: 4–5 h; target removal rate of αMG 35–40% and βMG 80%; a protein leakage membrane hemodiafilter; and a central dialysate delivery system [11].

Despite its limitations, to the best of our knowledge, this study is the first to analyze the all-cause mortality of incident hemodiafiltration (HDF) patients in Japan. While there are several dialysis-related studies focusing on the older population in Japan, there is currently a lack of population-based studies evaluating HDF performance. Therefore, the results of this study may still have clinical relevance and can assist clinicians in selecting dialysis modalities for patients, especially older ones. Additionally, patients who changed their dialysis modality during the analysis were excluded, which helped eliminate any bias that may have been introduced by modality crossovers.

Analyzing healthcare insurance data is pivotal in identifying trends and patterns in healthcare utilization, offering insights that can drive policy decisions, improve patient outcomes, and enhance the efficiency of healthcare systems. In Japan, the number of patients with ESRD requiring dialysis care continues to increase, as the country recorded 347,474 ESRD cases at the end of 2022, with a prevalence rate of 2781 cases per million people [1]. A significant change in the trend of dialysis modality selection has also been observed in recent years, where the number of ESRD patients treated with HDF has far surpassed the number of patients treated with HD and other modalities. As the number of ESRD patients initiating or switching treatment modality to HDF is expected to increase further, our findings underscore the importance of HDF studies in Japan, not only in providing information regarding improved patient survival but also in advancing scientific understanding of dialysis treatment, expanding therapeutic applications, and assisting policymakers in making evidence-based decisions.

In conclusion, this study supports the notion that HDF lowers all-cause mortality compared with conventional HD among the dialysis population. Therefore, this study adds to the observational evidence that suggests the superiority of HDF among the dialysis population of Fukuoka Prefecture in Japan.