Why do we need more efficient clinical trials?
Trial design is a multidimensional beast! The average cost of phase 1, 2, and 3 clinical trials is estimated to be around $4, 13, and 20 million respectively. From a bird’s eye view, it is hard to imagine the hundreds of components that influence trial costs; for example material for lab tests, packaging, storage, shipping drugs, infrastructure for tracking and allocation, quality control etc. Clinical trials need thousands of person-hours dedicated to recruitment, consent, follow up and handling of trial data. These costs restrict evidence generation for beneficial treatments with a low potential to recoup the expenses. To overcome these barriers we need novel ways to design and conduct trials more efficiently. This report (Treweek et al, Trials 2015) highlighted some of the problems with the current state of many clinical trials such as low participant recruitment and retention rates, lack of efficiency in data collection and “spin” bias (bias in the way outcomes are selected and reported). In 2015, the National Institute of Health Research (NIHR) put out a first call for efficient trials and formed a new initiative called TrialForge with the goal to focus on reducing complexity and inefficiency while maintaining an acceptable level of risk (i.e. risk based choices on the degree to which trials use things like monitoring and documentation to reduce the labor costs of RCTs).
What is a registry and why are registries useful?
A disease/patient registry is a database that routinely collects and organizes healthcare data, with an aim to evaluate and improve outcomes for a population defined by a particular condition, disease or exposure.
Some examples of most well known registries are the 9/11 registry or the Flint lead exposure registry created after the 2015 Flint water crisis. These registries have supported research for long term physical and mental health effects from these disasters as well as helped to improve health delivery to exposed individuals through community or federally funded programs. In nephrology we have several dialysis and transplant registries (eg USRDS, UK Renal Registry, CORR, UNOS, SRTR, and many more). Overall in nephrology, this review (Ng et al, Kidney Int 2022) identified 79 kidney failure registries spanning 77 countries and put out a call to action for inter-registry research to enhance capacity for global data collection. Apart from auditing practice patterns and evaluating service quality, kidney failure registries can play a huge role in supporting epidemiologic, health-outcomes and economic research with real world clinical trials. To support further establishment and development of renal registries worldwide, the ISN established a group of experts (the SharE-RR supporting group), working on the development of resources that kidney health advocates can use to establish or develop a renal registry in their countries.
What is a registry based RCT (RRCT) and how is it different from a traditional RCT?
A registry based trial is one in which the participants are either recruited from a registry; or the data related to the intervention and the outcomes are captured in the registry and are used in the RCT. The efficiencies from registry based trials may therefore be found in using this routinely collected data to reduce the cost of participant identification, screening and of follow-up and monitoring depending on the design.
A registry based trial is one in which the participants are either recruited from a registry; or the data related to the intervention and the outcomes are captured in the registry and are used in the RCT.
In a previous #Nephtrials we did a deep dive into the problem of the gap between efficacy and effectiveness in traditional RCTs (also called exploratory or explanatory trials). Explanatory trials are designed to demonstrate the efficacy of an intervention and are geared towards internal validity with strict inclusion-exclusion criteria in the optimal clinical setting and may be lacking generalizability. Without external validity, the findings of the trial may not be applicable to the world outside of the trial setting. Reliable evidence is generated through large trials, the biggest barriers to which are infrastructure and money. Registry based trials are a design that can reduce both the infrastructure and monetary requirements to do large trials. As with everything else in life, registry based trials do have their own strengths and limitations.
Strengths of the RRCT design
A registry based trial provides the ability to study patients under real world conditions with longitudinal follow up and helps to overcome these problems of generalizability.
Interventions and follow up are integrated in routine clinical care thereby reducing barriers and facilitating evidence generation.
Informed consent process is often incorporated into the clinical care setting.
Integration with routine care and an already existing patient database can help to reduce trial costs by up to 90% (although reducing costs per se isn't the end goal, it's improving health outcomes by evidence generation).
Figure modified from Yndigegn T, et al. Heart 2018
Limitations of the RRCT design
The quality of the data in RRCT is bound by the quality of data in the registry, hence a high quality registry is a precondition for a robust RRCT. RRCT may be lower cost but maintaining a quality registry in itself is generally not low cost.
The pragmatic nature of RRCT can cause hindrances in adherence and safety monitoring and are not suitable for most explanatory trials.
If clinical endpoints chosen for RRCT are not readily captured by the registry, it can cause abnormally low event rate leading to type II error and negative trial. Consider an endpoint like intradialytic hypotension which may not be captured well - but mortality or hospitalization is usually well documented.
Registries are often jurisdictional in nature and doing multi-jurisdictional work is often very difficult needing to be set up in multiple countries which often comes with its own challenges.
Enter the Hemodiafiltration Registry based trial
The wheels of dialysis therapy for patients with kidney failure started turning in 1943 after Dr. Kolff successfully dialyzed a young woman with uremia for the very first time. From those rotating drum cellophane dialyzers we have come a long way!
Since the early days of dialysis, several innovations in technology now allow nephrologists to offer dialysis to a large number of patients, however we still have ways to go in terms of improving outcomes and mortality in this group of patients with median survival rates as low as 50% in 5 years as well as high rates of hospitalizations and poor quality of life.
Hemodiafiltration (HDF) was first conceived to address the shortcomings of HD i.e. limited middle and large molecule clearance. Exploring the history of HDF takes us deep into the investigations of pioneers like Henderson and Leber who first coined the term HDF in 1977.
Solute removal in conventional high-flux hemodialysis occurs by diffusion with efficient clearance of small molecule solutes. Combining diffusion and convection in HDF provides both high diffusive clearance of small molecules as well as improved removal of larger molecules achieved with hemofiltration. The treatment of HDF has undergone several important and necessary modifications made possible by innovations in water treatment methods to allow online HD.
What is online HDF?
In the earlier days of HDF, commercially prepared replacement fluid in bags was used and would allow an average reinfusion rate of 9L per session. Technological advances now make it possible to produce high volumes of non-pyrogenic pure and safe water “on-line” and continuously in outpatient dialysis units. This can be used to freshly prepare ultrapure dialysate which is reinfused as replacement fluid. This allows HDF with high fluid turnovers up to 30-40L per session.
Figure from Canaud et al, CJASN 2018
What evidence do we have so far on HDF?
Four large RCTs comparing HDF with HD have attempted to assess superiority of HDF over HD with respect to relevant clinical endpoints of all cause mortality, CV mortality, and infection related morbidity or mortality. These are listed in the table below.
Figure from Canaud et al, CJASN 2018
Peters et al performed an individual participant data (IPD) analysis on the results of these 4 big RCTs. In this study of a median follow-up of 2.5 years, 769 out of 2793 participants died (292 cardiovascular deaths). Online HDF was associated with improved survival with a reduction of all-cause mortality by 14% and a reduction of cardiovascular mortality by 23% compared to HD. Survival benefit was greatest in the HDF group receiving the highest delivered convection volume (>23L per 1.73m2 body surface area per session). However this was not a randomized comparison; outcomes would be expected to be higher in this group. Potential confounding is introduced because some determinants of survival also are the determinants of received volume. Also, the dose results should be interpreted cautiously given that many of them are post hoc or per protocol (and not intention to treat) analysis. We have been fooled before with achieved adequacy in chronic dialysis (Greene et al, JASN 2005).
So is HDF superior to HD?
This is literally a million dollar question! Specifically the magnitude of convection volume that should be delivered could be the code breaker.
Middle to large uremic molecules have been proposed to be responsible for increased CV mortality and impaired immunity in ESKD. This could explain why the meta-analyses of existing RCTs have shown improved CV and infection related outcomes in high volume HDF. The IPD analysis shows that HDF was not associated with beneficial effects on fatal infections, fatal malignancies, treatment withdrawal or sudden cardiac death, however showed beneficial effects on cardiac causes. Although overall not significant, stratification of convection volume in tertiles showed a trend for sudden death: larger the convective volume, lower the mortality risk. Could this finding be real?
What are the mechanisms causing HDF to be potentially superior to conventional HD?
Effects of uremic toxins:
HDF improves urea clearance and KT/V however this has not been found to improve mortality in previous studies. Observational studies have indicated that use of convective therapies reduces incidence of carpal tunnel syndrome and other related manifestations of Beta-2 microglobulin amyloidosis.
Elimination of middle molecular weight substances like Beta-2 microglobulin, IL-6, TNF-alpha, is indeed enhanced by HDF however so far there is no convincing correlation to lower levels of mortality.
Removal of FGF23 was markedly higher during HDF and this has been proposed as one of the potential mechanisms of lower CVD mortality with HDF considering that FGF23 is associated with LVH and CVD events in CKD.
Middle and large molecular weight compounds such as p-cresol, indoxyl sulphate, indole acetic acid and advanced glycation endproducts (AGEs) have been attributed to oxidative stress and endothelial dysfunction and proposed to be contributing to CV mortality in patients with ESKD. Studies have shown improved clearance of these molecules with HDF, however it is still uncertain if it influences clinical outcomes in HDF.
Effects on hemodynamic factors:
Intradialytic hemodynamic stability is better preserved in HDF compared to high flux HD. IPD analysis of 4 large RCTs revealed that LV mass and EF tended to worsen over time in HD group but remained stable in HDF patients indicating favorable effect of HDF long term.
Why is a new trial needed on HDF?
Observational studies comparing HDF with low flux HD did find improved mortality in the HDF group. But in a study analyzing patients on dialysis >90 days from seven European countries in DOPPS Phases 4 and 5 (2009–15) there was no evidence of a survival benefit with HDF, even at replacement fluid volumes >20 L (Locatelli et al, NDT 2018). A potential selection bias cannot be ruled out in observational studies as delivering high convection is likely to be achieved in healthier dialysis patients who may have a lower mortality risk.
Wang et al did a meta-analysis comparing HDF with standard HD in adults with end-stage kidney disease (Wang et al, AJKD 2014). They identified 16 trials reported from 1996 to 2013 which found no significant benefit for clinical cardiovascular outcomes or all-cause mortality. The authors have highlighted the suboptimal quality and underpowered nature of many of these trials, with imbalance in some prognostic variables at baseline. Intention-to-treat analysis was not used in some trials.
RCTs aimed at studying superiority of HDF over HD have had significantly heterogeneous designs. The CONTRAST study used a low-flux membrane dialyzer for HD while the other 3 trials had use of high-flux HD. The convective volume delivered in all 4 trials is also widely variable. Post hoc analysis of CONTRAST, ESHOL and THDFS showed a significantly lower mortality on the group of patients treated with the highest convective volumes, however 10% of patients from the HDF arm were excluded after randomization in THFDS and ESHOL because their blood pump speeds were not adequate.
The ESHOL study achieved a median convection volume of 22.9-23.9L and showed a 30% lower risk of mortality in HDF as well as lower risk of hypotensive episodes, stroke and infection compared to high flux HD.
Similarly the FRENCHIE study which specifically performed HDF on patients over age 65 reported similar rate of treatment associated adverse events in HDF and HD groups but significantly lower intradialytic symptomatic hypotension and muscle cramps in HDF group.
Effect of phosphate clearance and benefit in ESA dose is still uncertain due to inadequately powered studies with heterogeneous interventions.
Though the cost has come down with online HDF - there still is an additional cost to consider if units not doing HDF decided to convert. We need solid data to make the case that this would be worthwhile.
Hence we need a well designed RCT to effectively prove (or disprove) the superiority of HDF compared to high flux HD and also to determine the optimal convective dose in different clinical settings.
H4RT
Informed by these prior findings, the high volume hemodifiltration vs high flux hemodialysis registry trial (H4RT)- a non-blinded, parallel group RCT of high dose HDF versus high flux HD. The trial was specifically designed to aim for a substitution volume of greater than or equal to 21L adjusted to BSA. This equates to a convection volume of greater than or equal to 23 L in a typical patient needing 2L UF to reach target weight. The primary objective is to determine the effectiveness of high volume HDF compared to high flux HD on non-cancer mortality and hospitalization due to CV events or infection.
Visual Abstract of the H4RT trial design from Denisse Arellano, NSMC intern class of 2020 and the ISN Social Media Team
One of the main secondary outcomes of interest is to compare the cost effectiveness of high volume HDF vs high flux HD. To study this, the trial has used an efficient design with linkage for outcomes data to Hospital Episode Statistics, Civil Registration, Public Health England, UK Renal Registry and equivalent registries in Scotland. Identifiable information, as agreed with partner organizations and set out in the participant information sheet, will be used to link this primary dataset with existing routine healthcare databases for follow up. The flow of data — identifiable and pseudonymized— is summarized in the figure below.
Dialysis units in England, Scotland and Wales are invited to participate in the trial with a goal of recruiting 1550 patients. The inclusion criteria for participant patients is quite broad- adult patients (over age 18 years) receiving in center HD or HDF, dialyzing at least 3 times a week and having the potential to achieve high volume HDF. Patients lacking capacity to consent, with prognosis of less than 3 months or expected to be transplanted in next 3 months or unsuitable for high volume HDF due to clinical reasons such as vascular access, will be excluded.
Anticipating the challenges based on previous trials’ experience, strategies to improve adherence like monthly compliance reports and ongoing training in achieving high volume HDF as well as tailored interventions to overcome recruitment have been implemented within the trial.
Follow-up will continue for a minimum of 32 months and a maximum of 91 months. It will be undertaken through a combination of 6-monthly patient questionnaires and by linkage to routine healthcare databases — Hospital Episode Statistics, Civil Registration, Public Health England (PHE) and UK Health Security Agency (UKHSA), and the UKRR in England, and the equivalent databases in Scotland and Wales, as necessary. Only data that are collected as part of routine care will be collected. As the primary outcome relies on data linkage with national routine healthcare databases, participants will only be lost to follow-up if they opt out of data linkage after being randomized. Patients who discontinue allocated treatment can choose to continue to receive questionnaires and allow data linkage. Analysis will be by ITT, with no patients withdrawn from the analysis because they are unable to achieve high volumes of HDF.
The efficient design of H4RT (relying on linkage to routinely collect healthcare data for outcomes) has kept the overall costs of this trial just under £3million and explores the opportunities to address clinically important questions.
So what’s the future for Hemodiafiltration?
The H4RT will report its findings in late 2025. Another RCT: CONVINCE was set to recruit 1800 ESKD patients to be randomized between high volume HDF and high-flux HD to address the benefits and harms of HDF compared to HD with focus on survival, patient perspectives and cost-effectiveness, though the enrolment stopped at 1360.
H4RT may transform the scene of dialysis therapy offered to our patients and take the pioneering work of Kolff and others to the next level - or it may demonstrate that HDF remains a cool and promising tool with little clinical benefit!
Manasi Bapat,
Nephrologist, California, USA
Reviewed by Mike Walsh, Fergus Caskey, Meg Jardine and Swapnil Hiremath