Numerous observational studies have described a ‘U’ or ‘J’ shaped relationship between pre-dialysis systolic BP and outcomes in people on haemodialysis where only low and, usually, very high BPs are associated with excess all-cause or CV mortality. Such studies have typically utilised ‘usual’ non-standardised BP measurements and there is poor agreement between ‘usual’ BP readings and those taken in standardised conditions (20). Besides a likelihood that ascertainment of ‘usual’ BP readings doesn’t adhere to best practice (21) other factors account for the ‘U’ or ‘J’ shaped relationship which contrasts with the linear relationship between BP and adverse outcomes seen in the general hypertensive population and in haemodialysis patients where other measurement methods (standardised, HBPM, or ABPM) are utilised. For example, in the CRIC study standardised BP (mean of 3 seated BP measurements) taken away from the dialysis unit is linearly related to mortality, whereas pre-dialysis measurements retain a U-shaped association within the same cohort of patients (22). There are likely to be several explanations for these discrepant relationships, not least that pre-dialysis BP reflects an individual’s physiological ability to tolerate volume loading. Importantly, people on haemodialysis differ from the general hypertensive population in another important regard: a tendency to suffer IDH which affects ~30% of dialysis sessions. Accumulating evidence suggests that IDH and ultrafiltration volumes are independently predictive of myocardial stunning which, in turn, is predictive of left ventricular systolic dysfunction and myocardial fibrosis (23), vascular access thrombosis (24) and mortality (25).
A single feasibility and safety study randomised 126 hypertensive haemodialysis patients to a “standard” BP target range of 155-165 mmHg or an “intensive” range of 110 to 140 mmHg, measured in a standardised manner in the immediate pre-dialysis period (26). Of seven pre-specified feasibility objectives (27), two were achieved (mean separation of BP between arms by >10 mmHg; 75% participants providing minimum required number of standardised unit BP measurements), two were not achieved (IDH in intensive arm not >20% higher than standard arm; ≥66% of required HBPM/ABPM measurements) and three were not reported. Additionally, 55% of the 281 participants that consented did not progress to randomization, predominantly due to not achieving SBP 155 mmHg despite back-titration of medications (n=65, 23% of those consented) or participant-initiated withdrawal (45, 17%). Nonetheless the authors judged that a full scale RCT would be feasible, presumably with modifications. Importantly, and despite not being powered for definitive conclusions, a number of safety signals emerged around intensive BP lowering: those in the intensive arm experienced a non-significant three-fold increase in vascular access thrombosis (incidence rate ratio, IRR 3.09) and an IRR of 1.61 for hospitalisation.
Besides this study, the remaining data to guide BP targets is predominantly derived from observational cohort studies utilising non-standardised pre- and post-dialysis BP measurements described above. These studies fairly consistently show worse outcomes where pre-haemodialysis BP is below ~140 mmHg. A facility-level analysis of DOPPS data designed to minimise effects of unmeasured patient-level confounding found lowest mortality in those with a pre-haemodialysis BP of 130 to 159 mmHg (28). Other studies have found lowest risk of CV events with a pre-haemodialysis systolic BP range of 140 to 170 mmHg (5); lowest risk for all-cause mortality at 165 mmHg and for CV mortality at 157 mmHg (29) and lowest risk for all-cause mortality at 152 mmHg and CV events at 143 mmHg (30). Considering the totality of data available, we suggest for haemodialysis units or individual clinicians wishing to base BP targets on pre-haemodialysis measurements, that a systolic BP range of approximately 140 to 165 mmHg appears to be associated with the fewest short- and long-term adverse outcomes.
Data on pre-dialysis diastolic and post-dialysis BP are more limited. In the Robinson study, facility-level pre-dialysis DBP between 60 and 99 mmHg and post-dialysis SBP between 120 and 139 mmHg were associated with lowest mortality (31). The same analysis identified post-dialysis DBP <70 mmHg to be associated with increased mortality with no corresponding upper limit. Hannedouche did not find a pre-dialysis DBP level that was indicative of minimal risk for all-cause mortality. For CV mortality, a DBP of 90 mmHg was found to be associated with lowest CV mortality, although 95% confidence intervals for hazard ratio were wide, and depart from 1.0 at a DBP of approximately 70 mmHg (29). Post-dialysis BP readings were not analysed.
Limited observational data suggest that demographic characteristics such as age (32), co-morbidities including diabetes (33), atrial fibrillation and heart failure (34) and biomarkers, e.g. troponin I and NT-proBNP (35), influence the nature of the relationship between BP and outcomes, such that in younger patients without co-morbidity and normal biomarkers only higher SBP tends to be associated with adverse outcomes, leading us to suggest that aiming for lower BP targets may be acceptable in this population.
Studies using HBPM, ABPM and standardised clinic BP readings generally demonstrate a positive linear relationship between BP and outcomes (4, 5, 11). For example, a study of 326 predominantly African American haemodialysis patients dialysing in Indiana, US found an average home SBP of 120-130 mmHg to be associated with lowest all-cause mortality (4). Similarly the CRIC study investigators found in 377 haemodialysis patients that, compared to those with a standardised SBP <128 mmHg, adjusted HR for CV events was 2.14 (95% CI 1.17 to 3.9) and 2.9 (95% CI 1.55 to 5.42) for those with SBP 128 to 145 and >145 mmHg respectively (5). A similar relationship was demonstrated in a recent Brazilian study in 2,672 haemodialysis patients, where BP was measured in the inter-dialytic period albeit probably not in a standardized manner. Compared to a reference SBP of ≥ 171 mmHg, incidence of CV events was reduced in those with SBP 101-110 (HR 0.65, 95% CI 0.46-0.90), 111-120 (HR 0.66, CI 0.49-0.89), 121-130 (HR 0.75, CI 0.57-0.98), and 131-140 mmHg (HR 0.76, CI 0.6-0.97) (36).
However, it is important to recognise important limitations of these studies. Firstly, they were conducted in small, often single-centre, populations that are unrepresentative of the wider UK dialysis population; and, secondly, some excluded participants with important CV co-morbidities such as atrial fibrillation (4, 11). In a European study using ABPM in 344 haemodialysis patients, a U-shaped relationship between ambulatory SBP and CV and all-cause mortality was observed in the whole study cohort, whereas when ~30% with either atrial fibrillation or heart failure were excluded a positive linear relationship between BP and outcomes was observed (34). These contradictory findings emphasize the inherent difficulties in using observational data to guide clinical practice even when using ‘gold-standard’ BP measurement methodologies, and the need for well-designed interventional trials to define appropriate BP targets in those on dialysis.
In the peritoneal dialysis population there are few data to guide BP targets with the ISPD recommended target of <140/90 mmHg being extrapolated from the general and CKD population (37). One large prospective cohort study undertaken in China found a U-shaped relationship between usual clinic SBP – presumably non-standardized – and all-cause and CV mortality, with an SBP range of 119-141 being associated with lowest hazard ratio for adverse outcomes (38). An analysis of UK Renal Registry data in peritoneal dialysis suggested that higher BP is associated with reduced mortality in the first 12 months following commencement of renal replacement therapy (RRT), except in the subgroup listed for transplantation within six months of starting RRT (39). Assuming, as the authors suggest, that early transplant listing is a proxy for minimal comorbidity, it is suggested that we should aim for lower BP targets in less comorbid peritoneal dialysis patients although no specific BP targets are identified by this study.