Drug Selection for Treating Hypertension in Dialysis Patients

More to Consider than BP-Lowering Potency

Aggressively treating BP, to much lower levels than were previously considered acceptable, leads to large reductions in cardiovascular events and death in the elderly general population (¹). The evidence supporting BP lowering is so strong that the US Food and Drug Administration considers BP a valid, mechanistically disease–agnostic, surrogate end point for preventing cardiovascular disease (²). Hypertension likely also contributes to cardiovascular disease and mortality in patients treated with dialysis. However, BP lowering also carries risks, including intradialytic hypotension and other potentially associated morbidity (e.g. myocardial stunning, reduced cerebral perfusion), accelerated loss of residual kidney function, and vascular access thrombosis, among others. Lowering BP in some patients may cause greater harm than the elevated BP. The level of BP at which the benefits of BP lowering outweigh the risks is unknown and will be different for different types of patients. In clinical practice, treatment decisions are made on the basis of pre- and postdialysis BP readings, despite data suggesting “out-of-dialysis-unit” BPs may provide greater prognostic information (³); however, home BP monitors are not covered by insurance, and whether patients will adhere with home readings is unknown. The one thing we do know is how to lower BP, and that is with more frequent or prolonged hemodialysis treatments. Both have been associated with reductions in BP and left ventricular mass, and (in the case of frequent dialysis) mortality (⁴,⁵). Despite this, few patients are willing to deviate from thrice weekly in-center treatments of 3.5–4 hours duration. Volume is a key mediator of hypertension for most, albeit not all, patients (⁶). In the absence of more accurate tools, we continue to use dry weight challenge and our physical exam to guide volume management. From a practical standpoint, the achievement of euvolemia is limited by the amount of fluid that can be safely removed during each treatment. Such tools would not help in managing patients with large interdialytic weight gains who are unable to reduce fluid intake, or patients who are intolerant of ultrafiltration but are unwilling to attend extra treatments. When we have exhausted efforts at lowering dry weight—the intensity at which this is pursued is highly variable by clinic—we turn to antihypertensive agents. In a nutshell, the issues surrounding BP management in patients on dialysis are numerous and vastly more complex than in the general population.

With this context in mind, how does the choice of antihypertensive medications and their BP-lowering efficacy relate to the management of hypertension in patients on dialysis? There are several considerations. First, in anuric patients, none of the antihypertensive medications address volume overload. Second, in choosing an antihypertensive agent, we should consider dialyzability, mechanism of action, and pleiotropic effects, rather than just the magnitude of BP lowering. Third, high BP-lowering potency may not be a desirable characteristic because of the risks of intradialytic complications.

In this edition of CJASN, Shaman et al. (⁷) report the results of a network meta-analysis in which 40 randomized, clinical trials were combined to assess the BP-lowering efficacy of antihypertensive medications in patients on dialysis. The standard meta-analysis technique is to combine studies assessing similar interventions (e.g., β-blockers) with similar comparators (e.g., placebo) in similar populations (e.g., anuric hypertensive patients on hemodialysis) and settings (in-center, thrice weekly) with similar outcomes (e.g., ambulatory BP), from which a pooled estimate of treatment effects is derived. Meta-analyses are only as good as the sum of their parts, i.e., the quality of the individual studies and pooled estimates are only useful if the contributing studies are comparable. Network meta-analyses extend the meta-analysis technique by performing indirect comparisons of interventions that were not compared head-to-head (e.g., β-blockers versus aldosterone antagonists) but had a common comparator (e.g., placebo). Additional assumptions for network meta-analyses include transitivity (interventions are jointly randomizable), congruency (effects on underlying populations and subpopulations are similar), and consistency (findings across studies are similar) (⁸).

The meta-analysis of Shaman et al. finds a greater systolic BP-lowering effect of aldosterone antagonists and β-blockers as compared with angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, calcium channel blockers, and renin inhibitors, and that aldosterone antagonists were more often discontinued (versus β-blockers) for side effects. Considering the fundamental requirements of a meta-analysis, we identify several problems that may undermine the validity of these results. First, the study populations do not appear to be comparable. Some studies included patients receiving peritoneal dialysis, and at least one involved patients with heart failure with reduced ejection fraction. The magnitude of BP lowering depends on the severity of the hypertension. Studies with population mean systolic BP >170 mm Hg were combined with those with mean systolic BP of 130–140 mm Hg, and, in at least two studies, systolic BP was <125 mm Hg, making it doubtful that the antihypertensive was being used to lower BP. It was unclear whether the drug of interest was a first- or fourth-line agent. It was also uncertain whether factors that affect BP independently of antihypertensive agents (e.g., dialysate sodium concentration, dietary sodium intake, dialysis session length, and intensity of volume management) were uniform across studies. Even the primary outcome, BP lowering, was not uniformly measured, with different studies defining it by ambulatory, home, office, predialysis, and postdialysis measurements. These BP measurements are not equivalent but were treated as such. We know that ambulatory and postdialysis BPs are at least 10 mm Hg lower than a predialysis BP (³). Of the included trials, 14 (35%) were considered at low risk of bias, and only 14% of the direct evidence for the primary outcome came from trials without any concern for quality. The differences in populations, the potential for variability in other aspects of care that affect BP, the lack of uniformity in outcome measurements, and the low quality of many of the studies could readily obscure the pooled results. Regarding consistency, it is puzzling that aldosterone antagonists were found to be so potent in lowering BP, when six (albeit small) prior trials have reported no significant BP-lowering effect (⁹).

What are the implications of this meta-analysis for clinical practice and future research? We caution against applying these results to clinical practice, given our concerns. Shaman et al. conclude that “β-blockers should be evaluated in future research as the first class of agents considered for the treatment of hypertension.” We note that β-blockers are not recommended as a first-line antihypertensive or as an appropriate choice for monotherapy in the general population because of their lower efficacy in reducing risk of stroke, despite similar BP-lowering efficacy (¹⁰). Whether this is true in the dialysis population is unknown, and it may be that β-blockers are the preferred first-line agent because of sympathetic nervous system overactivity or via another mechanism. But it is also conceivable that β-blockers may contribute to bradyarrhythmic sudden death (¹¹) or (by impairing counterregulatory responses to shock) worsen intradialytic hypotension.

Antihypertensive agents may confer benefits (e.g., improving endothelial function) or harm (per the above), independently of BP lowering. We will not determine this unless we conduct randomized, clinical trials with meaningful clinical outcomes. Although epidemiologic studies find an association of elevated BP with adverse outcomes, this is not evidence that lowering BP improves outcomes, and therefore, BP-lowering effect in itself cannot be considered a surrogate for hard clinical outcomes. To assess whether intensive BP lowering is of benefit, we need a study that randomizes patients to two different BP levels and (given the primacy of volume control underlying hypertension) intensively manages volume overload in the “low” arm before adding antihypertensive agents. This is a challenging study to conduct, requiring frequent evaluation of volume and harm (e.g., intradialytic hypotension), constant reinforcement about reducing salt and water intake, the participation of patients who will agree to extra treatments when needed, and dialysis units that can accommodate all of this. Identifying the most potent antihypertensive agent seems premature until we know which BP level we should be treating to, and whether there is a benefit of a particular antihypertensive drug independent of BP lowering. However, when we do get to that point, the results of this meta-analysis may provide hypothesis-generating evidence that could be tested.

We fully recognize that the dearth of clinical trial data in patients with kidney failure makes it challenging to choose evidence-based therapies for our patients. There are many unanswered questions. Should we put our effort and resources into volume control by reducing dietary salt and water intake or in overcoming barriers to the use of more frequent or prolonged dialysis or into measures that preserve residual kidney function? Should we compare second-line therapies (antihypertensives) head-to-head, knowing that in reality, if the volume is not optimized, most patients will require several antihypertensives, and BP lowering via this means may not be as beneficial as by controlling volume? Should we consider multifactorial interventions, including antihypertensives, statins, and lifestyle modifications? Should we select these interventions for any patient on dialysis or a priori identify populations most likely to benefit? The dearth of evidence in this area can be addressed only by carefully designed, randomized, clinical trials that are conducted in a well defined population, with meticulous control for other factors that affect BP and volume, so that the effect of the intervention(s) can be measured and that it is measured by its effects on important clinical outcomes.

Disclosures

D. Miskulin receives salary support from Dialysis Clinic Inc. T. Shafi reports receiving personal fees from Hershey Medical Center, National Institutes of Health Center for Scientific Review, University of Mississippi Medical Center, and University of Tennessee; consulting fees and personal fees from Siemens; and nonfinancial support from American Society of Nephrology and National Kidney Foundation, outside the submitted work.

Funding

None.