HEMO Revisited: Why Kt/Vurea Only Tells Part of the Story.

In this issue of the Journal of the American Society of Nephrology, a posthoc analysis of the data and samples of the Hemodialysis (HEMO) study reports on the effect of raising Kt/Vurea on a broad range of low molecular weight uremic toxins.1 The HEMO study,2 one of the great pivotal studies in hemodialysis (HD), was itself grounded upon a previous, smaller—but equally important—study in patients with chronic dialysis. TheNationalCollaborativeDialysis Study (NCDS)3 randomized maintenance HD patients to two different (time-averaged) BUN concentrations and two different treatment times. Both the rate of withdrawal from the study for medical reasons and the hospitalization rate were higher in the group of patients with high BUN, whereas no effect of treatment time was seen. A simple and straightforward conclusion of the NCDS data wouldhavebeen thatBUNconcentrations are related topatient outcomes. The authors, however—as stated in the abstract— concluded that occurrence of morbid events is affected by the dialysis prescription.3 A secondary analysis of the same study data, based on kinetic modeling of BUN, popularized the concept of Kt/Vurea. Eventually, the paradigm of Kt/Vurea-driven dialysis prescription prevailed and became the universally used marker of dialysis adequacy. Ever since, Kt/Vurea, has been subject to criticisms, culminating in a recent debate of its pros and cons.5,6 If a common conclusion would be sought from both texts in spite of the opposite vision they defend, then it could be that Kt/Vurea remains a useful baseline metric of certain aspects of dialysis adequacy but that it misses a number of others. The HEMO study was designed to test the effects of increments in dialysis dose, expressed as Kt/Vurea, and the level of dialysis membrane flux on patient morbidity andmortality. In the primary analysis of the entire cohort, high-dose hemodialysis providing an average single-pool Kt/Vurea of 1.71 provided no benefit over a standard treatment providing a single-pool Kt/Vurea of 1.32, although subgroup analysis suggested that survival was increased for women in the high Kt/Vurea group. Also, comparison of chronic high-flux dialysis, as defined by higherb-2microglobulin (b(2)M) clearance, and low-flux dialysis did not significantly alter all-cause mortality. Of note, a secondary analysis showed that mean cumulative predialysis serum b(2)M levels but not dialyzer b(2) M clearance were associated with all-cause mortality.7 The study by Meyer et al. in this issue1 aims to dissect the effects of Kt/Vurea-driven dialysis prescription on serum concentrations of a wider panel of small uremic retention solutes. In a cohort of 1281 patients (about two-thirds of the original 1846 study participants), the authors measured predialysis serum levels of various small solutes in the first study year, at least three months after randomization, in an attempt to differentiate the response of highversus low-dose dialysis expressed as Kt/Vurea on individual uremic retention molecule concentrations. Whereas serum concentrations of some solutes were found to be similar between groups, i.e., p-cresyl sulfate and asymmetric dimethyl arginine (ADMA), others were found to be lower in the high Kt/Vurea arm. Unsurprisingly, predialysis serum BUN levels were lower, as were levels of trimethylamine oxide (TMAO), indoxyl sulfate, and methylguanidine but, overall, most decreases were not as impressive as would have been expected given the 33% Kt/Vurea increment. Furthermore, along the lines of the original NCDS study, Meyer et al. looked at the association between serum concentrations and outcomes.1 They could not find clear associations between serum levels of uremic retention solutes measured in predialysis samples in the first year of the HEMO study and overall mortality. It should, however, be noted that, for the protein-bound solutes, accurate measurements of free solute concentrations were not available for all patients and that only associations with total solute concentrations were reported. Especially with regards to the cresols, observational studies have shown associations with outcomes especially for the free—i.e., unbound—fraction.8 The assets of this study are that (1) samples were collected prospectively in the context of a randomized controlled trial; (2) although this is a posthoc analysis in a subset of the original study population, the number of samples is large and almost equal per group (643 and 638); and (3) the approach differs from previous analyses by comparing two independent arms rather than performing correlation analyses. All these factors help to minimize sources of external bias. Althoughmore or less unavoidable with a posthoc analysis, there are, however, also a number of drawbacks: (1) as the Published online ahead of print. Publication date available at www.jasn.org.