Blog entry by Arif Khwaja
Last year I wrote a blog outlining the results from the Dutch CONTRAST study which was a large RCT which failed to show any benefit of online-haemodiafitration (OL-HDF) when compared with haemodialysis in terms of mortality. Recently 2 further large RCTs have been published comparing OL-HDF to haemodialysis which add significantly to theevidence base about he relative merits of OL-HDF when compared to conventional HD.
Firstly, the Turkish Hemodialfiltration Study was recently published in NDT. In this RCT, 782 patients were randomised to receive either high-flux HD or OL-HDF. At 2 years there was no significant difference in survival between the two groups - 77.6% in OL-HDF versus 74.8% in the high-flux group, P = 0.28. Its worth pointing out here that here that the authors stated that 'statistical power for this analysis was lower than hypothesized during the design of the study' in part due to a lower than expected event rate in the control group. In a post-hoc analysis of those in the OL-HDF group who actually achieved substitution volumes of >17.4 litres per session there was a mind-boggling 46% reduction in overall mortality and 71% risk reduction for cardiovascular mortality when compared to high flux HD. However its worth pointing out that those in the OL-HDF group who achieved high convective volumes were less likely to be diabetic, had higher serum albumin and higher blood flow rates. Whilst this was controlled for in the analysis it is entirely possible that better survival rates in those who achieved convective volumes >17.4l were simply 'healthier' patients.
The second study is the ESHOL study sponsored by the Catalonian Society of Nephrology, which was a multicentre, RCT that randomised 906 patients to receive either OL-HDF or haemodialysis (92% of those randomised to the haemodialysis arm received high-flux HD) and has just been published in JASN. The headline figures are impressive and are in striking contrast to both the CONTRAST and the Turkish HDF studies which both failed to achieve their primary endpoints. Those assigned to OL-HDF had a 30% lower risk of all-cause mortality, a 33% lower risk of cardiovascular mortality, and a 55% lower risk of infection-related mortality. The reduction cardiovascular mortality was primarily driven by a reduction in the number of strokes. The estimated number needed to treat suggested that switching eight patients from hemodialysis to OL-HDF may prevent one annual death which suggests that OL-HDF was having an astonishing clinical impact.
So why the striking difference between the ESHOL and Dutch/Turkish studies? The answer is not obvious to me but it is worth considering the following:
i) In the ESHOL study no formal statistics are done on the baseline characteristics of each group but its worth noting that there are baseline differences between the two groups. 7.5% of the OL-HDF group dialysed via a line compared to 13.1% of the HD group. The mean age was 66.3 years in the HD group vs 64.5 years in the OL-HDF group and 22.8% of the OL-HDF group were diabetic compared to 27.1% of the HD group. The Charlson Comorbidity Index was 6 in the OL-HDF group and 7 in the HD group. What I genuinely dont understand is why there was no statistical analysis to see if these baseline differences were statistically significant. Instead these variables were included in multivariate analyses and then treatment risk estimates were calculated in all subgroups. I assume this is an accepted statistical approach but I am just left with this nagging doubt that the 55% reduction in infection related mortality in the OL-HDF group is in part related to the fact that were nearly 40% less lines in this group than in the HD group.
ii) the reduction in strokes with OL-HDF accounted for most of the reduction in cardiovascular risk and this may well have been related to the significant reduction in intradialytic hypotension that was observed in this group compared to the HD group.
iii) a consistent theme that emerges from all three studies is that that actual replacement volume delivered seems to matter. The median replacement volume in the ESHOL study was around 21 litres/session compared to 17 litres/session in the Turkish Study and around 20 litres/session in the CONTRAST study. Indeed post-hoc analyses of both the Turkish and CONTRAST studies showed higher convective volumes did associate with better survival. This was also seen in a post-hoc analysis in the ESHOL study where those with >25 litres/session of convective volume had a 45% reduction in mortality. However what isn't clear is why some patients who are randomised to OL-HDF are able to achieve high convection volumes and others aren't. In particular I wonder whether factors such as quality of access and cardiac function may somehow select out those who are able to achieve high convection volumes. Thus its plausible that those who can tolerate high volume OL-HDF are simply those with better cardiac function
iv) many will be surprised at the scale of the impact of OL-HDF in the ESHOL study. There have been so many negative RCTs in large dialysis populations it is surprising to see that simply switching 8 patients from high-flux HD to OL-HDF can prevent one death per year.. I am not sure if this is plausible particularly given the fact that most of the reduction in mortality is driven by a reduction in infections and if this is due to OL-HDF how is OL-HDF reducing the risk of infections? Furthermore I'm not sure if the scale of impact seen in ESHOL will be translated to having a similar impact in routine clinical practice. For example in the UK, the centre that has been using OL-HDF for the longest period of time is Stevenage and they published their rather impressive experience here in cJASN - yet UK Registry Data does not suggest that patients in Stevenage have better survival than other centres in the UK that don't routinely use OL-HDF
Therefore the evidence from these 3 studies is a bit mixed. ESHOL showing remarkable effects that some may think are 'too good to be true' whilst the CONTRAST and Turkish-HDF studies failing to meet their primary endpoint. If OL-HDF is to be used than the actual convective volume delivered seems to be critically important if its going to have an effect. The key question has to be whether OL-HDF is cost-effective (both economically and environmentally). If the cost-effectiveness analysis stacks up then clearly Ol-HDF should be standard therapy - as yet a quality of life analysis of CONTRAST has failed to show a positive impact of OL-HDF.