Article Text

Effects of haemodiafiltration or haemofiltration compared with haemodialysis on prognosis in patients with end-stage renal disease: protocol an updated systematic review and meta-analysis of randomised trials with trial sequential analysis
  1. Fan Zhang1,
  2. Jing Liao1,
  3. Yan Bai2,
  4. Zixuan Zhang1,
  5. Liuyan Huang1,
  6. Yifei Zhong1
  1. 1Department of Nephrology A, Longhua Hospital Shanghai University of Traditional Chinese Medicine, Shanghai, China
  2. 2Department of Cardiology, Longhua Hospital Shanghai University of Traditional Chinese Medicine, Shanghai, China
  1. Correspondence to Dr Yifei Zhong; sh_zhongyifei{at}


Introduction Haemodialysis is the most common treatment option for patients with life-sustaining end-stage kidney disease (ESKD). In recent years, haemodiafiltration or haemofiltration has been widely used in patients with ESKD, and there are still conflicting findings as to whether both are superior to traditional haemodialysis. This systematic review and meta-analysis were designed to determine whether haemodiafiltration or haemofiltration is more effective than haemodialysis in reducing all-cause mortality risk in patients with ESKD.

Methods and analysis We will perform a systematic PubMed, Embase, Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library and Scopus search, including studies published before September 2023. Randomised controlled trials will be included exploring the effects of haemodiafiltration or haemofiltration compared with haemodialysis on prognosis in patients with ESKD. Outcomes of interest include all-cause mortality, cardiovascular events, dialysis adequacy and adverse effects. The Cochrane Collaboration tools (ROB-2) will assess the bias risk. Available data will be used to calculate effect sizes. Heterogeneity between studies will be evaluated with I2. The trial sequential analysis will be used to eliminate false-positive results. The certainty of the evidence will be assessed using Grading of Recommendations, Assessment, Development and Evaluation criteria.

Ethics and dissemination This systematic review and meta-analysis was deemed exempt from ethics review. Results will be disseminated through publication in peer-reviewed journals and research conferences.

PROSPERO registration number CRD42023464509.

  • Mortality
  • Systematic Review
  • Meta-Analysis
  • End stage renal failure

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • This study will conduct a trial sequential analysis to assess reliability and conclusiveness of existing evidence in meta-analysis.

  • Populations from different ethnic groups may cause high heterogeneity.

  • Some small sample studies may make the results subject to publication bias.


Chronic kidney disease (CKD) has become a severe public health problem because an estimated 840 million people worldwide are reported to be affected by CKD.1 End-stage kidney disease (ESKD) caused by CKD is one of the leading causes of increased cardiovascular morbidity and mortality.2

To sustain life, patients with ESKD require renal replacement therapies such as peritoneal dialysis, haemodialysis or kidney transplantation.3 Among them, haemodialysis is the most common treatment worldwide, and the population is still growing; it is also used as a transitional treatment for kidney transplantation.4 5 With the development and improvement of dialysis technology, haemofiltration and haemodiafiltration have been gradually applied in clinical practice.6–8 The former is based on convection to remove excess water and toxins from the body, while the latter combines haemofiltration and haemodialysis.6

Four meta-analyses have been published to determine the impact of haemodiafiltration or haemofiltration compared with haemodialysis in patients with ESKD, but all reported no significant effect on reducing mortality.9–12 Recently, the New England Journal of Medicine published the results of a comparison of high-dose haemodiafiltration with high-flux haemodialysis (CONVINCE) trial, a large-sample, high-quality randomised controlled trial (RCT) reporting that haemodiafiltration can reduce an all-cause mortality risk in patients with ESKD compared with haemodialysis.13 14 This contradictory result necessitates an update of the existing systematic review.

Considering this, we will systematically review the evidence on haemodiafiltration, or haemofiltration, compared with haemodialysis for patients with ESKD to answer the following specific research questions and conduct a trial sequential analysis (TSA) of the primary outcomes to reduce the risk of drawing false negative conclusions and to help clarify the need for additional trials.

  1. Does haemodiafiltration and haemofiltration, compared with haemodialysis, reduce all-cause mortality in patients with ESKD?

  2. Does haemodiafiltration and haemofiltration, compared with haemodialysis, reduce the incidence of cardiovascular events in patients with ESKD?

  3. Does haemodiafiltration and haemofiltration improve dialysis adequacy compared with haemodialysis in patients with ESKD?

  4. Does haemodiafiltration or haemofiltration cause adverse events?


This systematic review and meta-analysis will be conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocol (PRISMA-P) guidelines15 (online supplemental table S1). The protocol has been registered in the International Prospective Register of Systematic Reviews (PROSPERO) with CRD42023464509. The systematic review is scheduled to commence in February 2024, and completion is scheduled for June 2024.

Patient and public involvement

Patients and/or the public were not involved in this research’s design, conduct, reporting or dissemination plans.

Data sources and search strategies

We will systematically search four electronic databases (Embase, PubMed, Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library and Web of Science) using a combination of medical subject terms (MeSH) and keywords related to haemodiafiltration, haemofiltration, and ESKD and search for data from registered but unpublished studies. The timeline was from the inception to September 2023. The search strategies for each database are shown in table 1 and online supplemental table S2–S4. In addition, we will also manually search for records in the reference lists of previous systematic reviews.16 17 Literature downloaded from the databases will be imported into EndNote V.20 software for management. An updated search will be conducted 2 months before the publication of the review or before the final analysis to ensure that any relevant studies are included to the extent possible.

Table 1

Search details for PubMed

Eligibility criteria

Eligibility criteria will be based on PICOS (Population, Intervention, Comparator, Outcome, Study design) elements developed by potential RCTs should include (1) adults with ESKD received dialysis treatment, (2) participants in the intervention group receiving haemodiafiltration or haemofiltration, (3) participants in the control group received haemodialysis and (4) primary outcomes of interest are all-cause mortality, and secondary outcomes including cardiovascular outcomes, which was defined a composite of cardiovascular mortality, myocardial infarction, and stroke or defined by study author, and dialysis adequacy (assessed by Kt/V), and adverse events.

Selection process

Two authors (ZZ and YB) will independently conduct the selection. The titles and abstracts will first be screened for relevance based on eligibility criteria, and then potential full-text articles will be reviewed. Disagreements will be adjudicated by a third author (FZ). Reasons for full-text exclusion will be documented (online supplemental table S5) and summarised in a PRISMA flow chart (figure 1).

Figure 1

Flow chart shows literature search and selection of trials.

Data extraction

Two independent reviewers will extract data using an established form, including literature information, demographic characteristics of participants, dialysis model and outcomes. We will attempt to contact the appropriate author for any incomplete or missing data. If no response is received within 2 weeks, it will be considered no response and included in the qualitative analysis. Any discrepancies will be resolved through discussion. We will calculate the kappa coefficient as a measure of consistency in study selection and data extraction.

Baseline and final follow-up outcomes will be extracted when outcomes are measured more than twice. If we encounter a study that presents results in the figure, we will use GetData software to extract the mean and SD.

To ensure correct and complete data extraction, we conducted a pilot extraction for five eligible studies (table 2), and the table will be subsequently revised as needed. All included studies will be listed chronologically by year of publication.

Table 2

The characteristics of included studies

Risk of bias assessment

Two independent authors (ZZ and JL) will assess the risk of bias for each included RCT according to the Cochrane Collaboration’s risk of bias tool 2 (RoB-2) as having a low risk of bias, some concerns or a high risk of bias: bias arising from the randomisation process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measuring the outcome and bias in the selection of the reported result.18 The Excel macro tool provided on the RoB-2 official website ( will be used to generate the risk of bias summary table. Any discrepancies will be resolved through a third author (FZ).

Handling missing data

For missing data that cannot be obtained from the text, we will contact the corresponding author, and if we do not hear back within 1 month, we will consider the contact a failure. Without a response, we will calculate effect estimates using the available data described in the Cochrane Handbook, where possible. In the case of studies reporting data format of (1) median, first and third quartiles or minimum and maximum; (2) mean, 95% CI; (3) median with IQR or median±SD and (4) mean, SE, we will use the following corresponding reference equations to estimate the mean and SD:

Mean±SD will be estimated using the method of Luo et al.19 Calculation results are available on the web


The formula A and B recommended by the Cochrane Handbook20 will be used to calculate the mean±SD.

Formula A (sample size of each group ≥100): SD= Embedded Image/3.92.

Formula B (sample size of each group <60): SD= Embedded Image/(2*t).

Mean±SD will be estimated by the method of Hozo et al.21 For studies with samples >25 per group, the mean was equal to the median, and the SD was calculated as IQR/4.

4) SD=SE error* Embedded Image .

Data analysis

Data synthesis

We will use the meta22 and metafor23 packages in R software to include studies with continuous or dichotomous data for outcomes in the meta-analysis. A z-statistic with a p value of 0.05 assessed the overall effect. The primary analyses will use as-treated effect measures. The summary statistics for the primary outcome (eg, all-cause mortality) were HR and 95% CI. Continuous data analysis will select the standardised mean difference and use Hedge’s g to explain the effect size. The restricted maximum likelihood method will be used as the estimator for the meta-analysis because the results of this method are more robust. In addition, we will calculate 95% prediction intervals to predict the range of actual effects.

Assessment of heterogeneity

Statistical heterogeneity between studies will be assessed with I2; a value greater than 50% will be considered to indicate substantial heterogeneity,24 in which case a random effects model will be used, and otherwise, a fixed effects model.

Subgroup analyses

If the data allow, we will perform several subgroup analyses to test interactions according to treatment (haemodiafiltration and haemofiltration); age (<65 and ≥65 years); male (≤50% and >50%); follow-up (<24 and ≥24 months); dialysis vintage (≤36 and >36 months); sample size (≤100 and >100) and publication time (before 2015 and after 2015).


If ≥10 studies are included, a random-effects meta-regression will be performed using the Hartung-Knapp-Sidik-Jonkman model to explore the magnitude of the effect of covariates on the outcome. The meta-regression model includes prespecified study levels, including mean age, proportion of males and sample size. A p<0.05 indicates a significant moderating effect of the variable.

Publication bias

Contourenhanced funnel plots will be used to assess publication bias for outcomes with ≥10 studies.25 Further exploratory analyses will be performed using Egger regression tests to assess asymmetry statistically.

Sensitivity analysis

We will use the ‘leave-one-out’ method, that is, deleting one study each time and repeating the analysis, to assess the robustness of the results.

Trial sequential analysis

We will assess the risk of false positives or false negatives in a meta-analysis by TSA. The Lan-DeMets method will be used to construct the O'Brien-Fleming monitoring boundary and the optimal amount of information, which is set to an alpha of 0.05, a two-sided beta of 0.80, and a relative risk reduction of 20%.26 27 The expected intervention effect may reach the level of sufficient evidence when the cumulative Z-curve enters the null zone or crosses the trial sequential monitoring boundary. The evidence could not draw a conclusion if the Z-curve did not cross any boundary or reach the required information size. TSA will be performed using TSA software (V. Beta).

Qualitative analysis

We will perform a narrative analysis to present the results for articles that meet the eligibility criteria set by this systematic review and meta-analysis but for which meta-analysis is not possible (eg, incomplete data).

Grading of evidence assessment

We will assess the quality of evidence using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) methodology to determine the certainty of the evidence for the following five domains28:

  1. Risk of bias: We will reduce the certainty of evidence if a sensitivity analysis shows significant differences between studies with low, medium or high bias.

  2. Indirectness: If the research questions in the included studies are consistent with the PICO questions, the evidence level will not be downgraded.

  3. Inconsistency: Unexplained heterogeneity will be a reason for downgrading, that is, I2 greater than 50%.

  4. Imprecision: Whether the CI corresponding to the effect estimate is narrow enough.

  5. Publication bias: If publication bias is found by funnel plot or Egger’s test, we will downgrade the evidence level of the domain.

The certainty of evidence will be rated as high, medium, low and very low by the GRADE tool. High means that further research is unlikely to change confidence in effect estimates; medium implies that further research is likely to have a significant impact on confidence in effect estimates; low means that further research is very likely to have an impact on confidence in effect estimates; and very low means that estimates of effects are very uncertain. The evidence summary table is provided in online supplemental table S6.


This study will integrate a recent large RCT based on a prior systematic review and meta-analysis to confirm the effect of haemodiafiltration, or haemofiltration compared with haemodialysis on all-cause mortality risk in ESKD. Furthermore, we should recognise that skill set and prescription for HDF are not entirely identical and lack of proper physician hand-on knowledge, represent a limitation for centres transitioning to HDF.29 Predictable limitations include (1) populations from different ethnicities may cause higher heterogeneity; (2) long follow-up studies are bound to have high dropout rates, which may cause bias in the results and (3) some small-sample studies may cause publication bias and lower evidence levels.

Ethics and dissemination

Due to the nature of the protocol, ethical approval is not required. We will follow the PRISMA guidelines to disseminate the study results through publication and conference presentations.

Ethics statements

Patient consent for publication


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • FZ and JL are joint first authors.

  • Contributors YZ conceptualised and designed the study and were major contributors in writing the protocol. FZ, ZZ and YB contributed to the development of the selection criteria, the risk of bias assessment and data extraction criteria. FZ, LH and JL developed the search strategy. All authors read, revised and approved the final manuscript.

  • Funding The work was supported by Shanghai Hospital Development Center (SHDC2022CRD003).

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.