The novel findings of our study are: (1) endothelial dysfunction, as measured by flow-mediated, endothelium-dependent vasodilation, was associated with intradialytic hypotension, and (2) arterial stiffness, as measured by carotid-femoral artery PWV, was associated with intradialytic hypertension. These associations were independent of ultrafiltration or predialysis SBP. Although causality cannot be proved by our cross-sectional study, these results suggest that intradialytic hypotension and hypertension may be governed by distinct vascular pathologies.
Higher PWV or lower FMD independently predict cardiovascular mortality in the dialysis population.12–14
Higher PWV is associated with traditional factors (in particular, age, diabetes, high blood pressure) as well as risk factors related to uremia (volume overload, abnormal calcium, and phosphate metabolism).18
The endothelial dysfunction of patients with CKD is associated with increased age, higher blood pressure, and inflammation,13
and may be related to higher levels of ADMA, an inhibitor of endothelial nitric oxide synthase that is elevated in subjects with reduced kidney function.19
In concordance with prior studies, our subjects had low FMD and high PWV. The median FMD in this cohort was 4.7%, compared with typical upper arm FMD > 10% for healthy subjects without kidney disease reported from the laboratory of one of the authors.16,17
Median PWV was 10.3 m/s, compared with 9.3 m/s reported for controls of similar age.20
The associations of increased age with higher PWV, diabetes with both PWV and FMD, and the high correlation between PWV and FMD in our cohort were also in agreement with prior studies.
The novel finding that lower FMD is associated with intradialytic hypotension may be explained by the effect of dialysis on several serum mediators of endothelial function. ADMA is an inhibitor of endothelial nitric oxide synthase that is elevated in dialysis patients.21,22
Several groups have found that ADMA levels decrease during dialysis,23–25
which could reverse endothelial dysfunction leading to vasodilation and hypotension, particularly in those with the highest levels of ADMA and the worst endothelial function at the start of the dialysis session. Endothelin-1 (ET-1), a vasoconstrictor, produced by the endothelium, is another candidate mediator of the association between endothelial dysfunction and intradialytic hypotension. Increased ET-1 activity tonically suppresses endothelium-dependent dilation. ET-1 is increased in dialysis patients, and intradialytic hypotension is associated with a decrease in ET-1 during dialysis.26
It is also possible that subjects with impaired endothelial function have dysregulation of coronary blood flow, limiting the ability of the myocardium to increase cardiac output during hypotensive episodes during dialysis and leading to ischemia and stunning as previously described.6
As for the association between arterial stiffness and intradialytic hypertension, an interaction with sympathetic activity is a possible mechanism. It is plausible that fluid removal causes sympathetic activation, and in those with severe arterial stiffness, this produces an exaggerated pressor response.
Our results highlight a potential role for peripheral vascular disease in the cardiovascular mortality of subjects with ESRD. Low FMD and intradialytic hypotension are known risk factors for cardiovascular mortality in this population,4,13
but whether brachial FMD is simply a surrogate for coronary endothelial dysfunction, and whether intradialytic hypotension is mainly cardiac in origin, is unknown. While our study does not answer these questions, it suggests a possible mechanistic pathway whereby peripheral vascular dysfunction contributes to intradialytic hemodynamic instability, which in turn causes chronic cardiac ischemia and fibrosis, predisposing to cardiac arrhythmias and death. Alternatively, chronic hemodynamic instability may worsen vascular dysfunction. The direction of this association should be clarified by longitudinal studies.
We found a strong association between higher serum phosphate and intradialytic hypotension. This phenomenon has been observed previously by Tisler and colleagues in a cohort of 958 dialysis patients. Compared with patients with occasional hypotension, those with frequent hypotension had higher phosphorus (1.99 vs. 1.79 mmol, P<0.005).27
Whether this represents patient noncompliance (leading to higher intradialytic weight gain and hypotension when this extra fluid is removed), an effect of phosphorus-related medications, or represents a real cardiovascular effect of serum phosphate levels is unknown. In our cohort, this association became stronger and remained significant (10mmHg drop in SBP per mg/dL higher phosphorus, P=0.014) even when the 9 subjects with phosphorus > 6 mg/dL were dropped from the analysis. Moreover, ultrafiltration was not associated with hypotension in our cohort. Therefore, we doubt that noncompliance is the driving force behind this association. Our study was not structured to study possible effects of type or dose of phosphorus binders and calcimimetics. Recent studies suggest that binders and calcimetics may improve vascular structure and function, 28,29
which could explain our findings. We feel it is possible that phosphorus itself has a detrimental vascular effect. While phosphorus has been shown to predict arterial stiffness in dialysis patients,30,31
this was not true in our cohort. However, subjects with higher phosphate showed a trend for worse endothelial dysfunction: after adjustment for age and diabetes, every 1 mg/dL increase in phosphorus was associated with 1.7% lower FMD (P=0.054). In animals and in healthy human subjects, phosphorus loading has been shown to induce endothelial dysfunction,32
but to our knowledge a potential association between serum phosphate levels and endothelial function has not been studied in dialysis patients. Larger studies with adequate power to measure associations of phosphorus, PTH, and doses of binders and calcimimetics with vascular dysfunction are warranted.
In designing this study, we considered what effect the dialysis procedure may have on endothelial function and arterial stiffness. Studies are inconsistent in this area. An improvement in endothelial function after hemodialysis, possibly due to removal of ADMA,33
is plausible; however, in other studies hemodialysis had a detrimental34,35
or negligible effect36,37
on endothelial function. Likewise, it is plausible that fluid removal during dialysis results in temporary improvement in arterial stiffness, as shown by Di Iorio et al.,38
although other studies show worsening of arterial stiffness with dialysis.18
We assumed that dialysis would have an effect on vascular function testing. Therefore, we standardized our measurements in relation to the dialysis procedure. In order to avoid the volume overload associated with the first dialysis session of the week, we measured endothelial function directly before the second or third dialysis session of the week, and arterial stiffness on the day two following the second or third session. All FMD’s and all PWV’s were always performed at the same time in relation to dialysis for consistency of within-group comparisons. While the timing of the measures in relation to dialysis is unlikely to affect the cross-sectional associations we report, it may affect the comparison of our cohort with other dialysis patients with respect to values obtained for FMD and PWV measurements.
Another challenge in the design of this type of study is the lack of evidence-based standards for assessing intradialytic hypotension and hypertension. KDOQI guidelines define intradialytic hypotension as drop in SBP of at least 20mmHg or a decrease in MAP of 10mmHg associated with symptoms such as muscle cramping,39
while the main outcome study reporting outcomes related to intradialytic hypotension used blood pressure only in the definition.4
Symptomatic hypotension occurs in about 20–30% of dialysis sessions.2
In our cohort the rate of asymptomatic hypotension was higher, as would be expected: 63% had a drop of SBP > 20mmHg. The average drop in SBP of our cohort was 31mmHg ( ± 18), which is comparable to the average decrease of 28mmHg ( ± 23) in the cohort of 1244 patients studied by Shoji et al.4
On the basis of this limited comparison, our cohort likely has a typical rate of asymptomatic hypotension.
In assessing the generalizability of our results one must consider demographic characteristics that are pertinent to vascular dysfunction. With respect to age and diabetes status, our cohort is fairly representative of the United States dialysis population. In fact, the high percentage of diabetes as a cause of ESRD is typical for the United States, Taiwan, Japan, and Mexico, where diabetes accounts for > 40% of incident cases of ESRD.40,41
In addition, our group was 43% African American, compared with the United States dialysis population, which is 33% African American.40
In non-ESRD populations, Black race is associated with higher PWV42–44
and lower FMD;45
and therefore it is possible that vascular dysfunction was more prevalent in our cohort than in comparable studies. The most notable difference between our cohort and other dialysis populations was the predominance of male gender. In the Framingham cohort (with no ESRD subjects), male gender is associated with lower % FMD.46
Two studies have shown that gender is not a significant correlate of PWV.47,48
Thus, it is difficult to surmise how the predominance of male gender affected our results. Overall, while our sample size is modest and drawn from a Veteran population primarily of male gender, we think it is likely that our results would generalize fairly well to other dialysis patients. In addition, although the composition of this cohort could affect the FMD and PWV results, we have no reason to think that the observed associations of FMD and PWV with hemodynamic changes during dialysis would not be generalizable.
Additional limitations of this study that merit consideration include the possibility of type 2 error given the modest number of subjects. Initial power calculations for this pilot study were based on PWV because it has lower variability and higher expected effect size than FMD, which may account for the lower P values obtained for FMD associations. In addition, this study was cross sectional and therefore does not provide evidence of causality. Nevertheless, despite these limitations, we have found significant and independent associations between endothelial dysfunction and intradialytic hypotension, and between arterial stiffness and intradialytic hypertension. Further studies to elucidate these potential mechanisms of intradialytic hypotension and hypertension are on going.