BP lowering with medication has been demonstrated to reduce all-cause and CV mortality in people on dialysis (134). The pooled reduction in BP was -4.5/-2.3 mmHg and relative risk (RR) of CV events, CV mortality and all-cause mortality were 0.71, 0.71 and 0.8 respectively. Eight RCTs were included in this meta-analysis of which 7 were in haemodialysis populations and only 4 explicitly included participants with elevated BP. Therefore, there is little direct evidence of effects of BP lowering medication on outcomes in people on peritoneal dialysis. BP-lowering medications used in the intervention arm of included trials was heterogenous: ARBs in three, ACEi in two, βBs in two and CCBs in one. A further meta-analysis published at the same time was not included in our evidence synthesis since it included the same original publications, did not provide a pooled estimate of BP reduction, but it did reach very similar conclusions regarding event rates (135).
We identified three RCTs of BP lowering medication in dialysis patients with a primary or secondary outcome relevant to our search criteria that have been published since this 2009 meta-analyses (136-138). These three trials were not included in the network meta-analysis described below as these were all ‘treat to target’ trials. One of these trials compared ACEi vs. non-renin-angiotensin system inhibition (RASi) and a second compared ARB vs. non-RAS. Both trials achieved similar BPs in intervention and comparator arms as per their respective designs, but found no difference in their respective primary composites end-point of CV mortality, non-fatal stroke or myocardial infarction (plus heart failure admission in the first trial) (137, 138). A third trial compared a βB to ACEi in predominantly African-American haemodialysis patients (136). Despite a ‘treat to target’ design with a goal home BP of ≤140/90 mmHg, in post-hoc analysis there was a slightly lower home BP (p=0.037) in the atenolol arm; on 44-hr ABPM, there was numerically lower BP in the atenolol arm (-3.6/3 mmHg). This trial was terminated early due to an excess of CV events in the ACEi group, with no difference found in primary outcome of change in left ventricular mass index.
We were unable to draw firm conclusions regarding choice of antihypertensive drug class in dialysis patients due to lack of consistent evidence of reduction in all-cause, CV mortality or other relevant endpoints favouring any particular class. We have therefore based several of our suggestions on the findings of a recent network meta-analysis that assessed comparative BP lowering efficacy of antihypertensive medications in people on haemodialysis (139). Mineralocorticoid receptor antagonists (MRAs) and βBs were most effective at reducing systolic BP (SBP), compared to both placebo (MRAs: -10.8 mmHg; βBs: -8.7 mmHg) and other classes (e.g. MRAs -6.4 mmHg vs. ACEi and βBs -4.4 mmHg vs. ACEi). CCBs and ACEi lowered BP compared to placebo by -4.6 mmHg and -4.3 mmHg respectively. Additionally, the βB vs. placebo comparison provided a high confidence rating for effect estimate whereas other comparisons in this meta-analysis varied from moderate (e.g. CCB and ACEi vs. placebo; CCB and βB vs. ACEi) to low or very low confidence ratings.
Hypotension and discontinuation due to adverse effects were more common with ACEi (RR for hypotension and discontinuation were 6.62 and 1.77 vs. control). For these reasons, and in the absence of a clear cardioprotective class effect of RAS blockade in dialysis patients, we have suggested use of ACEi as third line in those on haemodialysis. In those on peritoneal dialysis there is evidence from small RCTs that ACEi or ARBs preserve residual renal function and urine output compared to control groups, despite similar reductions in BP over 12 to 24 months (140, 141) with a mean difference in GFR compared to controls of +0.93 mL/min/1.73 m2 (95% CI 0.11-1.75) for ACEi and +1.11 mL/min/1.73 m2 (95% CI 0.38-1.83) for ARB in pooled analyses (142, 143), leading us to suggest these classes as first line antihypertensives in people on peritoneal dialysis.
Although MRAs are most efficacious at lowering SBP, higher rates of discontinuation due to adverse effects (RR 3.35) were observed for this class, with a numerically increased risk for hyperkalaemia (RR 1.63, 95% CI 0.75 to 3.57). Additionally, the confidence rating for the effect estimate of MRA vs. comparators, including placebo, was either low or very low. Finally, although three meta-analyses (144-146) confirm BP-lowering efficacy of MRAs and impressive reductions in pooled estimates of CV and all-cause mortality (RR around 0.4 for all-cause mortality), concerns exist around the quality of some of the original trials of MRAs. Two large global RCTs of MRAs in haemodialysis populations (ACHIEVE and ALCHEMIST) are expected to complete recruitment shortly, and these are anticipated to define the efficacy and safety of MRAs with greater confidence. For these reasons, and until publication of ACHIEVE and ALCHEMIST studies, we recommend that MRAs are considered for use only in those dialysis patients with more problematic hypertension.
Alpha-blockers (αBs), ARBs and renin inhibitors were not found to lower BP more than placebo, although confidence intervals for the former two classes were wide (-6.7 mmHg, 95% CI -14.1 to 0.7 for αB vs. placebo; -3.0 mmHg, 95% CI -8.7 to 2.6 for ARB vs placebo) and only indirect comparison between αBs with placebo was available. Discontinuation due to adverse effects was also higher for ARBs (RR 1.57).
There are significant differences between antihypertensives in the extent to which they are removed by dialysis (see Appendix D). Theoretically, choice of antihypertensives based on dialysability may affect BP lowering efficacy, enhance BP variability and, in the case of drugs that reduce cardiac events by other mechanisms e.g. antiarrhythmic effects of βBs, dialysability may confer more a immediate effect on risk of CV events. There exists some lower quality evidence to support this hypothesis: a large propensity-matched retrospective cohort study from a Canadian haemodialysis population (n=6588) found increased risk of all-cause (RR 1.4) and CV (RR 1.2) mortality in those receiving high (atenolol, acebutolol, metoprolol) vs. low dialysability (bisoprolol, propanolol) βBs (147). Conversely, in non-adherent patients, it may be helpful to administer antihypertensives under direct supervision at the end of a dialysis session, particularly if using highly dialysed antihypertensives (e.g. atenolol, lisinopril, enalapril) that have prolonged half-lives in people on haemodialysis. Given the continuous nature of peritoneal dialysis, and paucity of relevant evidence in this population, we do not make any recommendations about in-class choice of BP lowering medication in those on peritoneal dialysis.
Anecdotal evidence suggests that dialysis patients may omit prescribed antihypertensives prior to haemodialysis sessions, either on advice of healthcare professionals or by their own decision (148). We were unable to find evidence to support this practice. However, it is reasonable to assume that, where a person takes one or more antihypertensive medications early in the morning on 4 days of the week and 6 or more hours later than this on the remaining 3 days, this may increase BP variability. Observational studies have established increased short- and long-term BP variability as independent predictors of all-cause and CV mortality and major adverse CV events in people on haemodialysis (149). Considering that the TIME study found that evening dosing of antihypertensive medication did not differ from morning dosing in terms of major CV outcomes in the general population (150), we suggest that clinicians discourage omission of antihypertensives prior to dialysis sessions and instead encourage consistent night-time dosing of antihypertensive medication in those patients for whom such medication is thought to be contributory to IDH. We acknowledge that the efficacy of this approach should be tested in a clinical trial.
We found limited evidence to support pharmacological approaches to management of IDH. Evidence to support the use of midodrine is inconclusive. A meta-analysis, published in 2004, included 10 studies with 117 participants and reported nadir BP 13.3/5.9 mmHg higher in those receiving midodrine, but no consistent benefit in terms of symptom reduction (151). All included studies had significant methodological flaws (e.g. none were of parallel group design) and were subjective to substantial risk of bias. A retrospective cohort study of 3083 patients, albeit not necessarily with confirmed IDH and subject to confounding by indication, found an adjusted incidence rate ratio of 1.37, 1.31 and 1.41 for all-cause mortality, all-cause hospitalization and hospitalization for CV causes respectively for those prescribed midodrine vs. controls (30).
Evidence to support use of L-carnitine supplementation is similarly inconclusive. A recent meta-analysis (8 studies, 224 participants) of 6 to 24 weeks duration using either cross-over or parallel group design (152). Of the included studies, only two were judged to be at low overall risk of bias. Compared to controls, participants allocated to L-carnitine supplementation had a pooled odds ratio for incidence of IDH of 0.26 (95% CI 0.1-0.72). Subgroup analysis suggested that only oral, as opposed to intravenous, supplementation was effective and that a minimum weekly dose of 4,200 mg is required. Conversely, a recent Cochrane review (3 RCTs, 128 participants) found insufficient evidence that L-carnitine prevented IDH (RR 0.76, 95% CI 0.34-1.69; low certainty evidence) (153).