In this national cohort of incident hemodialysis patients we found no associations between elevated plasma levels of total p-cresol sulfate, MMA or DMA and all-cause or cardiovascular mortality. We did find an elevated risk of all-cause mortality with higher indoxyl sulfate levels but only in the fully-adjusted model. While this finding is similar to previous research detailed below, in the setting of multiple comparisons in the current analysis and the borderline p-value (0.043), we are cautious with the significance of this finding. The search for uremic toxins has been ongoing in the nephrology community for many years
. The use of urea measurement alone is a crude measure of uremia that does not reflect the multitude of solutes that accumulate during renal failure and contribute to disease. Discovery of toxic solutes that accumulate in renal failure and could be diminished by alterations in dialysis or by reduction in their production would allow for tailoring of dialysis therapy and ultimately improve patient outcomes. For example, protein-bound solutes such as p-cresol sulfate and indoxyl sulfate can be removed more effectively by increasing dialyzer surface area and dialysate flow
. Similarly, intracellularly sequestered compounds such as MMA would be expected to be removed best by longer dialysis sessions.
The solute p-cresol sulfate has had the most observational data to suggest toxicity. While our study showed a borderline significant association between elevated p-cresol sulfate and cardiovascular mortality, it disappeared with adjustment for other factors including residual renal function and baseline co-morbidities. In a cohort of 139 patients with different stages of CKD (ranging from stage 2 to stage 5 on dialysis), baseline free and total p-cresol sulfate showed an inverse relationship with renal function and a positive relationship with vascular calcification
. This study, which included only Caucasian prevalent CKD patients, measured free and total p-cresol sulfate
. They also found that free but not total p-cresol sulfate levels were associated with mortality
. This group, using the same Caucasian cohort, showed that elevated levels of indoxyl sulfate were also associated with overall and cardiovascular mortality
. Another study in 100 prevalent hemodialysis patients in Taiwan also showed an association between free and total p-cresol with cardiovascular events
. A study of 175 prevalent hemodialysis patients in Belgium showed an elevated risk of all-cause mortality for patients with high free p-cresol sulfate levels (HR 2.28; 95% CI: 1.12, 4.64)
. A later study from the same group in Belgium showed that in 499 patients with mild to moderate kidney disease, higher baseline free p-cresol levels were associated with cardiovascular events (HR 1.39, p = 0.04)
. Most of the previous studies were limited by small sample sizes with few events and, thus, an inability to fully adjust in statistical models. Ours is also the only US based study evaluating these associations with differing patient populations, medications, dialysis practice patterns and lastly outcomes (CV mortality versus CV events). These differences compared to previous studies may partially explain the differential results.
Our study measured total levels of the solutes. Assays of total levels of protein-bound solutes may not reflect free levels due to variations in albumin concentration and the extent of protein binding. Other studies have shown that free but not total levels show stronger associations with mortality
. However, stratification of our results by albumin status did not show differential effects. Our study also used specimens from 2 to 6 months after dialysis initiation. The ideal timing of specimen collection for this type of study is unknown. Collecting samples before the initiation of dialysis may lead to more confounding by residual renal function, whereas samples from later in the course of dialysis may lead to survival bias, as not all the participants who started on dialysis will have survived to specimen collection. The uremic toxins we studied are all poorly cleared by current dialysis therapies and by the dialysis therapies used at the initiation of the CHOICE study, therefore, one would not expect large changes in levels after dialysis initiation, but this has never been studied.
Among patients receiving conventional dialysis, we demonstrated that protein-bound uremic solutes were not independently associated with mortality in a well-dialyzed cohort. This finding was consistent among all the uremic solutes and did not differ by diabetic or nutritional status. Our finding are similar to other studies evaluating associations between potential uremic toxins and mortality which also had null results
. Recent trials comparing more frequent or nocturnal daily dialysis have shown considerable improvement in outcomes, which has been attributed to improved dialysis and clearance
[26, 27]. The quality indicators such as lower blood pressure, hospitalization rates, and mortality, however, may all be related to improved volume, electrolytes and blood pressure, rather than greater removal of uremic solutes. This hypothesis still needs to be tested in the recently completed trials with both cardiovascular and mortality end-points.
We have focused only on major clinical events, mortality and CVD mortality. However, other disabilities suffered by patients receiving nominally adequate dialysis, such as neurocognitive and sleep disorders, quality of life indicators, response to medication such as ESAs and malnutrition that may benefit from increased solute clearance but further work is needed to test the relation of these non-urea solutes to such outcomes. Identification of solutes or classes of solutes associated with mortality and morbidiy could also lead to strategies for reducing their production or enhancing their elimination.
There are several limitations to our study. The first is that we did not have data on the entire CHOICE hemodialysis cohort, which may have introduced an unmeasured selection bias, since only those who survived and had the correct sample in the specimen bank were included. Our final cohort had less severe comorbid illness and early mortality than the overall CHOICE sample, in part because some specimens were obtained later in patients’ course of dialysis therapy. Another limitation is the use of plasma from a single time point as the exposure. Unfortunately, most studies of uremic solutes have had this limitation. Finally, like all previous studies on these potential uremic toxins, our study is observational and therefore causality cannot be directly inferred. These limitations are balanced by the several strengths of our study, including prospective design with inclusion of only incident hemodialysis patients; detailed and precise information for demographic, clinical and treatment factors; and systematic adjudication of baseline comorbid conditions as well as outcomes. These comprehensive data allowed us to extensively adjust for potential confounders in our analysis.