#NephJC Chat
Tuesday, July 11, 2023 at 9 pm Eastern (AEST = July 12, 11 am, CET= July 12, 2 am)
Wednesday, July 12, 2023, at 9 pm Indian Standard Time and GMT/BST, 4:30 pm CET
N Engl J Med. 2023 Jun 16. doi: 10.1056/NEJMoa2304820
Online ahead of print.
Effect of Hemodiafiltration or Hemodialysis on Mortality in Kidney Failure
Peter J Blankestijn, Robin W M Vernooij, Carinna Hockham, Giovanni F M Strippoli, Bernard Canaud, Jörgen Hegbrant, Claudia Barth, Adrian Covic, Krister Cromm, Andrea Cucui, Andrew Davenport, Matthias Rose, Marietta Török, Mark Woodward, Michiel L Bots; CONVINCE Scientific Committee Investigators
PMID: 37326323
Introduction
The survival of patients living with ESRD on dialysis is poor. Despite some marginal changes over the last several decades, the 5-year survival rate stands at about 40% from the USRDS data (2022 USRDS report). Needless to say - this survival is many years less than the general population, ranging from a whopping ~40 year shorter lifespan for someone in their 40s, to a not insignificant shortening by about 10 years for those in their 70s (Bello et al, Nat Rev Nephrol 2023). Even compared to the big C, dialysis fares poorly - survival is not better for dialysis compared with breast, colorectal, or prostate cancer, and only lung cancer or pancreatic cancer have shorter survival than dialysis (Naylor et al, AJKD 2019). This is not for a lack of trying - more dialysis (in terms of adequacy, time, frequency, cooler dialysate or high flux filters) have all failed to demonstrate an improvement in survival. But moving beyond just high flux filters, could hemodiafiltration be the magic bullet?
Under the pressure (sic! hydrostatic) of finding a superior dialysis alternative, convection can be added to old diffusion, and this combination is called hemodiafiltration (HDF). It offers a theoretical advantage of removing middle and large molecular weight substances, possibly better hemodynamic stability, and reduced requirement of ESAs. See the recent NephJC summary by Manasi Bapat reviewing the basics of online hemodiafiltration (HDF) in the context of its registry H4RT and the randomized trial CONVINCE.
Figure depicting diffusion and convection from Gaudry et al, N Engl J Med, 2022
Though there have been trials done in HDF, mostly in Europe where HDF use is widespread, prior trials have several limitations leaving the nephrology community unconvinced. Though survival has been reported only in those who have a delivered convection volume of >23 liters, this kind of an analysis is inherently biased and hasn't been performed with a rigorous intention-to-treat spirit (see Greene et al, JASN 2005 for an example of dose-targeting bias from a post hoc analysis of HEMO). Even in the ESHOL analysis (Maduell et al, JASN 2013) which did report a lower all cause mortality (but also infection related mortality interestingly), ~10% patients were withdrawn for inadequate blood flow, there was differential loss to follow up, different transplant rates between groups (dealt with by censoring rather than a competing risk analysis) thus not increasing the community’s confidence in the results.
Why haven't the benefits of high molecular weight toxin removal and high convection volume translated into a universal benefit for it to be adopted as a standard of care? What are the confounding factors? What are mechanisms of benefit or lack thereof? Or is it plainly a lack of evidence for patient-reported outcomes and cost effectiveness that’s holding us back? Hence (despite 4 large trials and a pooled analysis), the question remained - Is HDF superior to HF-HD in reducing mortality? CONVINCE is the RCT designed to resolve this very quandary.
The Study
Methods
Study Design
This pragmatic, open-label, eurocentric trial spanned across 61 centers from 8 countries, in which patients were randomized in a 1:1 ratio to receive either high dose HDF or HF-HD. The allocation of trial participants to different groups was through a centralized block randomization method, with stratification based on the participating centers. It included both academic and non-academic centers. The study was supervised by an independent academic institution, which made at least one visit to every included center, and more in centers with more than 31 patients included. Data were collected on each patient’s visit, and laboratory data were collected according to local protocols.
Study population
Being a pragmatic trial, eligibility was quite broad. The trial enrolled adult patients (>18 years) with ESKD who had been on high-flux hemodialysis for at least three months. Patients with severe non-adherence to the prescribed dialysis regimen or life expectancy <3 months or an anticipated live donor kidney transplant within six months were excluded. Also excluded were those already on HDF in the prior 90 days. To avoid needing to do post hoc analyses based on convection volume, only those patients likely to achieve a 23L HDF therapy were included.
Here are the inclusion and exclusion criteria:
Intervention
The intervention of high-dose hemodiafiltration with on-line production of substitution fluid and ultrapure bicarbonate-based dialysis fluid at a convection volume (UF+ substitution fluid) of ≥23 L per session in post-dilution mode was targeted to be delivered. In case the target convection volume was not achieved, measures were taken to gradually adjust the dialysis prescription over a period of 2–3 weeks in order to achieve the target. First, the treatment time was increased to at least 4 hours. Next, the blood flow rate was raised to a minimum of 350-400 mL/min, and the filtration fraction was gradually increased to 33%. If still not achieved, then the highest possible volume was used, and the reason for the same was recorded.
Appendix III. Protocol steps to obtain convection volume ≥23 L/session, supplemental material from Blankestijn PJ et al, BMJ Open, 2020
In addition, iron, ESAs, phosphate binders, vitamin D or its analogs, PTH antagonists, drugs for hyperkalemia and extracorporeal anticoagulants were prescribed as per routine care. Patient visits were scheduled monthly for up to 12 months then 3 monthly for 12-36 months.
Statistical Analysis
It was postulated that the enrollment of 1800 patients would provide the trial with 90% power to determine a risk reduction of 25% (~ HR 0.75) in the intervention group, with a two-sided type I error of 5% and a follow-up of 2.5 years. The underlying assumption was a mortality rate of about 40% (~16/100 person-years) over the period of follow-up. However, the supplement describes various other scenarios for sample size based on event rates which describe a benefit of more than 25%, and with event rates lower than the assumed 40%. It is not clear if this was a post hoc presentation, given the lower recruitment rate and lower event rate observed (see Results section).
Table 1 (Supplementary appendix): Sample size calculated from the nSurvival function in the gsDesign package of R- software. Peter J. Blankestijn et al, NEJM 2023
Cox proportional-hazards models were used to estimate hazard ratios (HRs) and corresponding 95% confidence intervals for the primary and key secondary outcomes. Competing risk analyses were performed for the primary outcome, considering kidney transplantation as a competing event. The trial site was considered a random effect throughout the analysis and assessed the models' assumptions. The data were censored on the date of the last trial visit. Sensitivity analyses to test the robustness of the primary outcome were performed by excluding the trial site as a random effect. Due to the Covid-19 pandemic, the authors stratified analyses related to infection-related hospitalization and death based on Covid-19 infection. Two interim analyses were planned. Predefined subgroup analyses were conducted according to age group (<50, 50 to 65, and >65 years), biological sex, cardiovascular disease history, diabetes, and residual urinary output, but they were not adjusted for multiple comparisons.
Outcomes
Primary outcome of the trial was all-cause mortality, and key secondary outcomes were cause-specific mortality, fatal and nonfatal cardiovascular events, kidney transplant, and recurrent hospitalizations. Event classifications were reported by treating physicians. There was no blinded adjudication committee. Importantly, not all adverse events were reported. Considering a stage V disease, patients did have a wide variety of frequent adverse events, and hence reporting was restricted to only serious adverse events (but not all of them). Patient-reported outcomes (PROs) were studied based on self-reporting on a questionnaire- PROMIS (see box 1 from Blankenstijn et al, BMJ Open for the complete list of PROMs) and cost-effectiveness assessments were made but will be reported in subsequent publications.
Funding
The trial was funded by the European Commission Research and Innovation, Horizon 2020. The data analysis, paper writing, and journal submission were done by the scientific committee independently, without any influence from the financial sponsors of the study.
Results
A total of 1360 patients were randomized (with 683 patients in the high-dose hemodiafiltration arm and 677 patients in the high-flux hemodialysis arm). The recruitment was lower than the anticipated 1800 sample size.
Figure 1. Patient Enrollment and Follow-up, from Blankestijn PJ et al, NEJM, 2023
Study population was largely homogenous as is typical for a European trial. Only about 30% had diabetes and 40% had underlying CV disease. Most patients (>80%) were dialysed with an AVF, and the median dialysis vintage was just under 3 years.
Table 1- Baseline characteristics of study population from Peter J. Blankestijn et al, NEJM 2023
The mean follow-up period was 30 months. Loss of follow-up occurred in 2.6% of patients in the HDF group and 1.8% in HD group. The target volume of at least 23±1L per session (mean convection volume 25.3L) for high-dose convection was achieved in 92% of delivered hemodiafiltration sessions. The median duration of dialysis sessions was 240 min in both arms.
Supplementary Figure 1: Convection volume across visits for the hemodiafiltration group from Peter J. Blankestijn et al, NEJM 2023
Single-pool Kt/V in the HDF arm was 1.74 (1.71-1.77) in the HDF arm and 1.65 (1.62-1.68) in the HF-HD arm. However, other parameters, such as hemoglobin and phosphate, were similar between groups.
Supplementary figure 2: Variation in Kt/V per session across visits, from Peter J. Blankestijn et al, NEJM 2023
During follow-up, the HDF group had a slightly higher Kt/V yet similar hemoglobin and phosphate levels.
Table S2: Mean characteristics of dialysis parameters and laboratory variables during follow-up from Peter J. Blankestijn et al, NEJM 2023
Primary outcome
The primary outcome of death from any cause occurred in 118 patients (17.3%) in the HDF arm and in 148 patients (21.9%) in the HF-HD arm (HR 0.77; 95%CI, 0.65 to 0.93), equivalent to 7.13 and 9.19 events per 100 patient-years in HDF and HF-HD arms, respectively. This event rate is notably lower than the assumed 40% for the sample size calculation.
Sensitivity analyses with the inclusion of competing risk factors for kidney transplantation and treatment switches did not differ from those of primary analyses.
Figure 2A: Overall survival from Peter J. Blankestijn et al, NEJM 2023
Subgroup analyses
There seemed to be some differences in the prespecified subgroup analyses. In patients with a history of cardiovascular disease, the risk of death was similar in both arms (HR 0.99; 0.76–1.28), but in the absence of a history of cardiovascular disease, the risk of death was lower in the HDF arm (HR 0.58; 0.42–0.79). Similarly, risk of death was found to be lower in the HDF arm in the absence of a history of diabetes mellitus (HR 0.65; 0.48-0.87). Interaction p values are not presented - and the analysis was not adjusted for the multiple comparisons being made.
Figure 2B: Subgroup analysis from Peter J. Blankestijn et al, NEJM 2023
Secondary outcomes
Risk of death from specific causes of cardiovascular disease and the composite outcome of fatal and non-fatal cardiovascular outcomes were similar in HDF and HF-HD treatment arms. The overall change in mortality noted in the primary outcome was driven by the lower risk of non-cardiovascular death in the HDF group (HR 0.76, 0.59–0.98).
Table 2: Primary and secondary outcomes from Peter J. Blankestijn et al, NEJM 2023
Digging deeper, the infection-related deaths were less in HDF arm when Covid-19 infections were included (HR 0.69; 0.49–0.96), and excluding COVID-related deaths (36 of total 92 infection-related deaths) pushed this outcome into non-significant territory. Sepsis-related (non-central venous catheter) deaths were also numerically less in the HDF arm (7 vs 14 deaths).
The risk of hospitalizations was similar in HDF and HF-HD arms, for both non-fatal hospitalizations and infection-related hospitalizations. Advantages seen in risk reduction of death in Covid -19 infections in the HDF arm were not seen in infection-related hospital admissions (HR 1.06; 0.86-1.30).
Discussion
In this largest trial of HDF, there was a significant 23% reduction in all-cause mortality, mostly driven by a decrease in infection-related mortality. Before we accept the results and roll HDF worldwide, let's examine the strengths and limitations.
Strengths
This was a well-designed trial with a large sample size - the largest HDF trial to date. There was an independent academic committee that reviewed the trial, which assured the collaboration between the public sector (trial funded by European Union), academic and private sector (big dialysis concerns- Diaverum, Braun, Fresenius). Other aspects to underline, CONVINCE was another pragmatic trial, this being reflected by the fact patients just followed their usual dialysis schedules, and laboratory data, calculated pool KT/V were done as per preexisting national protocols. Targeted convection volume, with a proven survival benefit of > 23L per session, was achieved throughout the study, and the authors created a comprehensive protocol to obtain convection volumes of > 23L in an admirable 92% of the total number of HDF sessions. A total of more than 56 000 sessions of HDF (more than 224 000 hours of HDF) were done during the trial with no extra serious adverse events. Despite the pandemic, the trial had a high recruitment rate (96.7% of patients who were approached agreed to participate) and a low dropout/losing from evidence rate (~ 5%).
Limitations
This was an open-label trial, which is inherently necessary given the nature of the intervention, which is difficult to blind. However, the events were reported by treating physicians (who knew the treatment assignment), and no independent blinded event adjudication committee existed. Secondly, despite a lower than planned enrollment, and a lower than planned event rate, which would reduce power, a significant beneficial effect was reported. The only reason for that was a greater than expected benefit, clocking in at 27% versus the 25% assumed. Most importantly, the benefit was not seen with CV mortality reduction (for which the event rates were almost exactly the same) but by an outsize reduction in infection (and withdrawal, plus other causes) related mortality. Though infection-related mortality was also lower (with much smaller numbers) in the ESHOL trial, most other trials also reported a lower CV mortality which was not seen here, puzzlingly enough. The labs (similar hemoglobin, phosphate) and lack of difference in hospitalization between groups also do not help shed light on this infection-related mortality difference. The outcomes not being adjudicated by a blinded committee comes up here as an important limitation again. Also, by electing not to report all adverse events, this study falls short of giving us a fair idea about tolerability of HDF compared to HF-HD. Lastly, with respect to generalizability of the population, the population is younger, with almost all having an AVF, and lower DM/underlying CV disease - which may be more typical for Europe than the typical dialysis patient in North America and elsewhere. Some of this might be related to the need for ensuring 23L convection is possible, which is an important aspect when mistakenly applying these results to all patients. The lack of data on race/ethnicity is in keeping with European studies, where these data are not routinely collected and may not be available for a pragmatic design trial.
For a review of how CONVINCE compares to previous HDF RCTs, see this infographic by Priti Meena.
So, where are we after discussing CONVINCE results? We do have higher quality data supporting the benefit in terms of mortality of HDF. On the other hand, this was driven by a decrease in infection-related mortality, coupled with the other limitations, giving us a pause before rolling out HDF in all patients. Since it is difficult to ignore a mortality benefit - could we implement this in a subgroup of patients? The benefits were more marked in older people, without DM and without pre-existing cardiovascular disease (though these are subgroup analyses not adjusted for multiple comparisons). The addition of the patient voice, and cost data would be helpful for making these decisions as well. HDF results in an extra approx 3900L/year (quoting professor Jan Kielstein back of envelope calculation) and more consumables (one more tube system for convection pump, more heparin etc). CONVINCE does also support the safety profile of HDF, given the prerequisites of hygienic and microbial standards are safeguarded.
Conclusion
We are still far away from emulating a blanket practice of HDF and remain to be convinced of unequivocal benefits of HDF, though promises are seen at least in selected groups. As suggested by the authors, updating the hemodiafiltration-pooling project with individual-participant data from the present trial and from other trials—and the results of the H4RT trial in particular—would allow a more precise exploration of treatment effects across all subgroups.
Summary by Brammah R Thangarajah
Senior Lecturer and Consultant Nephrologist
Department of Medicine, University of Jaffna, Sri Lanka
And
Saumya Vishnoi
Clinical Associate
Institute of Renal Sciences, Global Hospital
Mumbai, India
NSMC interns, class of 2023