Blood Pressure and Mortality in Long-Term Hemodialysis—Time to Move Forward

Abstract

Contrary to the direct, graded, and causal relationship of hypertension with cardiovascular outcomes in the general population, among dialysis patients, blood pressure (BP) recorded either predialysis or postdialysis displays a U-shaped curve with mortality. This paradoxical phenomenon of lower BP or a decline in BP over time being associated with increased mortality and higher BP being associated with a lower mortality is described as “reverse” epidemiology of hypertension, raising substantial controversy on whether BP lowering causes harms or benefits among dialysis patients. Unlike the inverse relationship of peridialytic BP with mortality, elevated BP recorded outside of dialysis is directly associated with poor long-term outcomes. Apart from the timing and technique of BP measurement, the U-shaped association of BP with mortality is also modified when accounting for factors related to patient’s clinical characteristics and level of illness, dialysis practices, and patterns as well as factors related to the methodology of survival analysis. Most importantly, deliberate BP lowering with antihypertensive drugs is associated with reduced cardiovascular morbidity and mortality. In this review, we explore the complex association of peridialytic, intradialytic, and interdialytic BP with outcomes among dialysis patients. We conclude with recommendations for a wider use of out-of-dialysis BP monitoring as a tool to better evaluate the cardiovascular risk and optimize the management of hypertension in this high-risk population. Rather than more cohort studies, we call for randomized trials to test the level of BP in dialysis patients that is optimal for cardiovascular outcomes.

Hypertension in end-stage renal disease patients on chronic dialysis is common, difficult to diagnose and often poorly controlled.^1,2 Unlike the direct and linear association of blood pressure (BP) with cardiovascular outcomes and mortality in the general population, cohort studies in dialysis patients showed an inverse relationship of BP recorded before or after dialysis with mortality.^3,4 The “paradoxical” phenomenon of lower BP being associated with poor long-term outcomes and higher BP being associated with a better outcome has been labeled as “reverse” epidemiology of hypertension, raising substantial uncertainty on whether hypertension is harmful and should be corrected in this particular population.^2–4 Contrary to the reverse prognostic association of conventional peridialytic BP recordings, elevated BP recorded outside of the dialysis unit either with self-measured home BP recordings or with 44-hour interdialytic ambulatory BP monitoring (ABPM) is directly associated with excess mortality risk.^5–7 In sharp contrast to the observational data, meta-analyses of randomized trials provided evidence that the risk of cardiovascular morbidity and mortality can be cut by a third when BP is lowered with the use of antihypertensive medications among dialysis patients,^8,9 particularly when they are hypertensive.⁸ Thus, it appears that spontaneous changes in systolic BP carry with them a prognosis that is fundamentally different from lowering BP in a deliberate manner using antihypertensive drugs.

In this review, we explore the “reverse” epidemiology of hypertension among patients on dialysis, discussing the role of confounding factors that magnify the inverse relationship of BP with mortality and oppose the ability of hypertension to predict risk. We compare this inverse relationship with the direct relationship of systolic BP lowering with an improvement in cardiovascular prognosis. We conclude with clinical practice recommendations for the necessity to widely apply the use of out-of-dialysis BP monitoring as an approach to better assess cardiovascular risk imposed by hypertension. We call for randomized trials, rather than more observational studies, to improve cardiovascular risk and optimize the management of hypertension among patients on dialysis.

FACTORS THAT MAY EXPLAIN THE REVERSE RELATIONSHIP OF BP WITH MORTALITY

The prognostic associations of BP recordings obtained before or after dialysis with cardiovascular outcomes and survival in numerous cohorts of incident and prevalent dialysis patients are summarized in Tables 1 and 2, respectively. The data suggest that lower predialysis BP is inversely associated with all-cause mortality.^10–14 On the other hand, high-peridialytic BP recordings were either not associated or minimally associated with increased mortality risk.^10–14 This “reverse” or “U-shaped” association may be due to (i) the BP measurement setting and technique used to assess the hypertension status of study participants,^5–7,15 (ii) patient clinical characteristics and level of illness,^16–19 (iii) the methodology used to the assess the value of hypertension as risk predictor in Cox-proportional hazard models.^11,14 These associations are discussed further.

Table 1.

Cohort studies evaluating the association of predialysis/postdialysis BP with mortality and clinical outcomes among incident dialysis patients

Author	Year	N	Region	Follow-up (years)	Age (years)	Baseline BP (mm Hg)		(%) of patients on antihypertensive therapy	Association of BP with outcome
						Pre-HD	Post-HD		CVE/CVM	ACM
Tomita54	1995	195	Japan	3	54	170/91	NA	NA	NA	Reduction in SBP <160mm Hg after the initiation of dialysis was associated with better 3-year survival
Kimura55	1996	195	Japan	3	54	170/91	NA	NA	NA	No association between BP and ACM among patients with pre-existing cardiovascular disease
Lynn32	2002	168	New Zealand	4.2	48.2	147.3/81.5	NA	76% before dialysis initiation; only 4.5% continued antihypertensive therapy while on dialysis	NA	U-shaped association between MAP at commencement of dialysis and ACM
Lucas56	2003	184	Spain	5.4	52	NA	NA	At the time of dialysis initiation, 36% had controlled hypertension and 37% uncontrolled hypertension	Uncontrolled hypertension before initiation of HD was associated with increased risk of CVM (RR: 2.93; 95% CI: 1.63–5.12)	History of uncontrolled hypertension before the initiation of dialysis was an independent predictor of ACM (RR: 1.79; 95% CI: 1.15–2.80)
Stidley19	2006	16,959	USA	34,705 person- years	NA	151.8/78.9	144.8/76.2	60.1	NA	Low pre-HD SBP (<120mm Hg) was associated with increased risk of ACM at years 1 and 2 of dialysis; among dialysis patients who survived for >3 years, high pre-HD SBP (>150mm Hg) was an independent predictor of ACM
Lib11	2006	56,338	USA	3	62.2	147.4/75.5	146.2/74.4	NA	NA	Compared with the referent category 160–180mm Hg, pre-HD SBP <120 was associated with 3.19-fold higher risk of ACM (HR: 3.19; 95% CI: 2.96–3.44)
Myers17	2010	16,283	USA	1.5	NA	150/78	NA	NA	NA	U-shaped association between pre-HD SBP and ACM. This association was more pronounced among older patients and diabetics
Shafi18	2014	446	USA	3.1	58	SBP: 152.5	NA	NA	Among patients at low- risk, pre-HD SBP was directly associated with increased risk of CVE (HR: 1.04; 95% CI: 0.96–1.13) and CVM (HR: 1.10; 95% CI: 0.96–1.25)	The level of illness modified the association between pre-HD SBP and ACM. In the low-risk group, a linear association between pre-HD SBP and ACM was noted (HR: 1.07; 95% CI: 1.01–1.14)
Duranton31	2016	357	France	2.2	74.9	150.8/71	NA	25.5	NA	High-baseline pre-HD SBP was associated with 8% reduced risk of ACM (HR: 0.92; 95% CI: 0.85–0.99 per 10mm Hg increase)

Abbreviations: ACM, all-cause mortality; BP, blood pressure; CI, confidence interval; CVE, cardiovascular events; CVM, cardiovascular mortality; DBP, diastolic blood pressure; HD, hemodialysis; HR, hazard ratio; MAP, mean arterial pressure; NA, not available; OR, odds ratio; RR, risk ratio; SBP, systolic blood pressure; SD, standard deviation.

Table 2.

Cohort studies evaluating the association of predialysis/postdialysis BP with mortality and clinical outcomes among prevalent dialysis patients

Author	Year	N	Region	Follow-up (years)	Age (years)	Baseline BP (mm Hg)		(%) of patients on antihypertensive therapy	Association of BP with outcome
						Pre-HD	Post-HD		CVE/CVM	ACM
Charra57	1992	445	France	20	48.4	NA	NA	1.6	Pre-HD MAP ≥99mm Hg was associated with increased risk of CVM	NA
Salem58	1996	649	USA	1	56	MAP: 110	NA	81.5	NA	No association between pre-HD MAP and 1-year risk of ACM (RR: 0.99; 95% CI: 0.97–1.01)
Duranti59	1999	370	Italy	NA	53	NA	NA	NA	NA	Hypertension had no effect on ACM
Iseki60	1997	1,243	Japan	5	52.4	151.5/80.5	NA	51.7	NA	Inverse association between DBP and ACM (OR: 0.84; 95% CI: 0.71–0.99)
Zager14	1998	5,433	USA	2.6	58.9	150.9/79.8	139.5/79.8	64.6	U-shaped association of BP with CVM; either low <110 (RR:2.04; 95% CI: 1.70–2.45) or high post-HD SBP >180mm Hg (RR: 1.73; 95% CI: 1.19–2.50) were associated with increased CVM	NA.
Salem61	1999	649	USA	2	56	MAP: 110	NA	81.5	NA	Patients classified as hypertensives had a trend towards lower risk of ACM (RR: 0.73, P = 0.06)
Blacher42	1999	241	France	6	51	157/85	NA	NA	No significant association between DBP and CVM (HR: 0.90; 95% CI: 0.60–1.50)	Inverse association between DBP and ACM (HR: 0.60; 95% CI: 0.3–1.0)
Port12	1999	4,499	USA	2.5	57.8	NA	NA	NA	NA	Pre-HD SBP <110mm Hg was associated with increased adjusted risk of ACM (RR: 1.86; 95% CI: 1.46–2.38)
Mazzuchi16	2000	405	Uruguay	2.74	62.1	142/81	NA	45.2	NA	Low pre-HD DBP <74.5 mmHg was associated with increased risk for early ACM (RR; 2.52, P = 0.004). High pre-HD SBP >160mm Hg was associated with late ACM (RR: 2.19; P = 0.04)
Tozawa36	2002	1,243	Japan	6.3	52.3	SBP: 151.5; DBP: 80.9; PP: 70.6	NA	52	Higher PP was associated with increased risk of CVE (HR: 1.138; 95% CI: 1.010–1.280) in nondiabetic but not in diabetic HD patients.	Higher PP was independent predictor of ACM (HR: 1.072; 95% CI: 1.010–1.149, P < 0.001) and cardiovascular events
Klassen38	2002	37,069	USA	1	60.1	154.3/79.3	139.6/72.7	NA	NA	Inverse association between post-HD SBP and 1-year risk of ACM (HR: 0.87; 95% CI: 0.84– 0.90). Direct linear association between post-HD PP and ACM (HR: 1.12; 95% CI: 1.06–1.18)
Foley37	2002	11,142	USA	3.8	59.7	151.8/79.7	137/74.1	50.4	NA	Increasing post-HD SBP was associated with increased ACM risk (HR: 3.00, P = 0.03) and increasing post-HD DBP with reduced risk of ACM (HR: 0.86, P = 0.006),
Takeda62	2005	164	Japan	3.8	59	157/82	NA	52	Elevated pre-HD SBP was independently associated with increased risk of CVE (RR: 1.23; 95% CI: 1.07–1.43).and CVM (RR: 1.25; 95% CI: 0.99–1.59)	No association between pre-HD SBP and ACM.
Kalantar- Zadeh10	2005	40,933	USA	1.3	59.8	153/79	NA	NA	Inverse association between pre-HD SBP and CVM	Lower pre-HD SBP (<110mm Hg) was associated with 60% higher risk of ACM (HR: 1.60; 95% CI: 1.42–1.81)
Li11	2006	69,590	USA	3 years	60.9	153.1/79.2	138.0/71.5	NA	NA	A reverse J-shaped association of pre-HD SBP with 1-year and 3-year ACM risk. Low pre-HD SBP (<120mm Hg) was associated with the highest risk for ACM (HR: 2.634; 95% CI: 2.443–2.840)
Ishimitsu63	2008	553	Japan	5 years	57	155/81	NA	79.7	A linear association of pre-HD PP with CVM (HR: 1.023; 95% CI: 1.008–1.037)	A linear association of pre-HD PP with ACM (HR: 1.015; 95% CI: 1.005–1.025)
Molnar64	2010	1,579 patients with PKD	USA	5 years	58	NA	NA	NA	NA	Among patients with PKD, low pre-HD SBP <120mm Hg (HR: 1.30; 95% CI: 1.06–1.92) was associated with increased ACM risk. A similar J-shaped association was evident in 67,085 HD patients without PKD
Chang65	2011	1,846 patients participating in the HEMO study	USA	2.5	57.6	NA	NA	79.1	NA	Low pre-HD SBP (<120mm Hg) was associated with 84% higher risk for ACM (HR: 1.84; 95% CI: 1.45–2.30, P < 0.001)
Robinson13	2012	24,525 patients participating in the DOPPS study	USA	47,124 person- years	NA	150/78	NA	NA	NA	Low pre-HD SBP 110–119 mmHg was associated with increased risk of ACM (HR: 1.14; 95% CI: 1.01–1.28). Higher pre-HD SBP 150–159mm Hg was associated with reduced risk of ACM (HR: 0.90, 95% CI: 0.81–0.99)
			Europe
			New Zealand
			Japan
Losito66	2013	3,674	Italy	2.2	67.2	137.1/73.4	131.0/72.0	56.5	NA	No association between BP and ACM in the overall study cohort. An inverse association between post-HD SBP and ACM among nontreated (HR: 0.74; 95% CI: 0.57–0.96)

Abbreviations: ACM, all-cause mortality; BP, blood pressure; CI, confidence interval; CVE, cardiovascular events; CVM, cardiovascular mortality; DBP, diastolic blood pressure; DOPPS, Dialysis Outcomes and Practice Patterns Study; HD, hemodialysis; HR, hazard ratio; MAP, mean arterial pressure; NA, not available; OR, odds ratio; PKD, polycystic kidney disease; RR, risk ratio; SBP, systolic blood pressure.

BP recordings obtained in-dialysis vs. out-of-dialysis provide different hazards with all-cause mortality

Conventional BP measurements in the dialysis unit are inaccurate estimates of the true BP. Inaccurate determination of hypertension status with the use of conventional BP recordings obtained either predialysis or postdialysis leads to “regression dilution bias,”³ limiting the ability of these measurements to provide clear and direct prognostic associations, even when such an association truly exists. Compared with nondialysis populations, BP among patients on dialysis is highly variable from predialysis to postdialysis as well as from 1 day to the next.²⁰ This high BP variability imposes particular difficulties in detecting hypertension with the use of peridialytic BP recordings; these measurements exhibit poor short-term reproducibility²¹ and provide imprecise reflection of the actual BP burden recorded with the “gold-standard” method of 44-hour interdialytic ABPM.²² Apart from their poor diagnostic accuracy in detecting interdialytic hypertension, another reason that could explain the “reverse” prognostic association of peridialytic BP with mortality is the fact that these measurements maybe influenced by the white-coat effect (i.e., elevated BP only in the dialysis setting), which is less pronounced with self-measured home BP recordings and eliminated with the use of interdialytic ABPM.^23,24 Another explanation could be that in-dialysis unit BP readings are unable to detect the masked hypertension effect (i.e., high BP at home and normal BP within dialysis), which may be also of prognostic significance.^24,25 Finally, BP recordings obtained shortly before or after dialysis maybe less representative of the patient’s typical BP as compared with a pool of BP measurements sampled from a broader spectrum of situations and activities during the out-of-dialysis period.²⁶

The notion that better precision in determination of hypertension status would provide direct and clear prognostic associations between BP and mortality is supported by a number of cohort studies using out-of-dialysis BP recordings as risk predictors (Table 3). In the first study to explore this hypothesis, Amar et al.⁷ prospectively followed 57 French hypertensive dialysis patients for a mean period of 2.8 years. Each 10mm Hg higher 24-hour pulse pressure (PP) recorded during the dialysis-off day was associated with 85% higher risk of cardiovascular mortality (risk ratio (RR): 1.85; 95% confidence interval (CI): 1.28–2.65); nocturnal systolic hypertension was also an independent cardiovascular risk predictor (RR: 1.41; 95% CI: 1.08–1.84 for each 10mm Hg higher night-time systolic BP).⁷ In a subsequent cohort of 168 Italian, nondiabetic dialysis patients without overt cardiovascular disease at baseline, an inverse night/day systolic BP ratio was associated with 2.52-fold higher risk of all-cause mortality (hazard ratio (HR): 2.52; 95% CI: 1.03–6.15) and 4.33-fold higher risk of cardiovascular mortality (HR: 4.33; 95% CI: 0.96–19.61) over 3.2 years of follow-up.²⁷ Among 73 Slovenian dialysis patients, Ekart et al.²⁸ reported that 16 patients died due to cardiovascular causes during a 10-year-long observational period. After adjustment for several conventional and nonconventional risk factors, mean 44-hour ambulatory BP was independently associated with increased risk of cardiovascular death.²⁸ Moriya et al.²⁹ prospectively evaluated 96 prevalent Japanese dialysis patients for a mean follow-up of 35 months to assess the endpoints of all-cause and cardiovascular mortality. Mean weekly PP (defined as the average of 14 BP measurements obtained by the patients themselves at home and 6 conventional peridialytic BP recordings) was calculated and used as fixed predictor in Kaplan–Myer survival analysis. After multivariate adjustment, weekly averaged PP >70mm Hg was independently associated with all-cause and cardiovascular mortality; in contrast, predialysis PP recorded in a single time-point had no association with either of the 2 outcomes.²⁹

Table 3.

Cohort studies evaluating the association of out-of-dialysis BP recordings with mortality and clinical outcomes among dialysis patients

Author	Year	N	Region	Follow-up (years)	Age (years)	(%) of patients on antihypertensive therapy	BP measurement technique	Association of BP with outcome
Amar7	2000	57	France	2.8	56.9	100	ABPM	Higher 24-hour PP was associated with increased risk of CVM (RR: 1.85; 95% CI: 1.28–2.65) and higher night- time SBP with increased risk of CVM (RR: 1.41; 95% CI: 1.08–1.84)
Tripepi27	2005	168	Italy	3.2	57.2	54	ABPM	Patients with night/day SBP ratio >1.01 had increased risk of ACM (HR: 2.52; 95% CI: 1.03–6.15) and CVM (HR: 4.33; 95% CI: 0.96–19.61)
Alborzi6	2007	150	USA	2	55.8	83	HBPM	Each 1 SD higher self-measured home SBP increased by 35% the risk of ACM (HR: 1.35; 95% CI: 0.99–18.4) and by 47% the risk of cardiovascular death (HR: 1.47; 95% CI: 0.98–2.19)
							ABPM
Agarwal30	2008	133	USA	2	56.8	83	HBPM	Mean 44-hour interdialytic ambulatory SBP (HR: 1.22; 95% CI: 1.07–1.32) had the same prognostic significance for ACM with that of 6 randomly selected SBP recordings from the same ABPM (HR: 1.20; 95% CI: 1.07–1.34)
							ABPM
Moriya29	2008	96	Japan	2.9	66	65.6	HBPM	Weekly averaged PP was an independent predictor of ACM
Agarwal5	2010	326	USA	2.7	54.9	74	HBPM	Increasing quartile of 44-hour ambulatory and home SBP were associated increased risk of ACM (adjusted HRs for increasing quartiles of ambulatory SBP: 2.51, 3.43, 2.62; and for home SBP: 2.15, 1.7, 1.44)
							ABPM
Ekart28	2012	73	Slovenia	10	54.2	64	ABPM	44-hour ambulatory SBP and DBP independent predictors of CVM
Bansal15	2015	326 incident HD patients participating in the CRIC study		2.83	NA	NA	Out-of-dialysis BP measurement in a single visit	U-shaped association between dialysis-unit SBP and ACM. A direct linear association between out-of-dialysis SBP and ACM (HR: 1.26; 95% CI: 1.14–1.40).

Abbreviations: ABPM, ambulatory blood pressure monitoring; ACM, all-cause mortality; BP, blood pressure; CI, confidence interval; CRIC, Chronic Renal Insufficiency Cohort study; CVM, cardiovascular mortality; DBP, diastolic blood pressure; HBPM, home blood pressure monitoring; HD, hemodialysis; HR, hazard ratio; NA, not available; SBP, systolic blood pressure.

The notion that the technique of measurement is likely to contribute to the variation in the association of BP with long-term outcomes is supported by 2 larger cohort studies, in which BP was assessed with self-monitored home BP recordings for 1 week, ABPM over a 44-hour interdialytic interval as well as predialysis and postdialysis for 6 consecutive dialysis sessions.^5,6 In the first, Alborzi et al.⁶ showed that among 150 prevalent dialysis patients, each 1 standard deviation higher home systolic BP was associated with 35% increased risk of all-cause death (HR: 1.35; 95% CI: 0.99–18.4) and 47% increased risk of cardiovascular death (HR: 1.47; 95% CI: 0.98–2.19).⁶ A similar relationship with mortality was detected when systolic BP was assessed by performing interdialytic ABPM (Figure 1). In the second, 326 prevalent dialysis patients were prospectively followed for a longer period of 32 months; 102 mortal events were recorded, yielding a crude mortality rate of 118/1000 patient-years. After multivariate adjustment, increasing quartiles of 44-hour ambulatory and self-measured home systolic BP were associated with increased risk of all-cause mortality (adjusted HRs for increasing quartiles of ambulatory systolic BP: 2.51, 3.43, 2.62; and for home systolic BP: 2.15, 1.7, 1.44).⁵ Once again, BP recorded within the dialysis unit either with routine or with standardized measurements was of no prognostic significance.⁵

Figure 1.

Out-of-dialysis BP recordings are of greater significance in predicting all-cause mortality among patients on dialysis. This figure depicts hazard ratios (HRs) for all-cause mortality for increasing quartiles of systolic BP. Increasing quartiles of home BP and ambulatory BP were associated with increased risk of mortality; in contrast, in-dialysis unit BP recordings did not achieve statistical significance in terms of prognostic value. Home systolic BP of 125–145mm Hg and of 115–125mm Hg by ambulatory BP was associated with the best prognosis in hemodialysis patients. Abbreviations: BP, blood pressure; HD, hemodialysis. (Reprinted with permission from Alborzi et al. Home blood pressures are of greater prognostic value than hemodialysis unit recordings. Clin J Am Soc Nephrol 2007; 2:1228–34.)

Of importance, out-of-dialysis BP recordings retain their strong prognostic significance even when fewer randomly selected measurements are used to assess the BP burden. To test this hypothesis, Agarwal et al.³⁰ compared the associated mortality risk of mean 44-hour ambulatory BP with that of 6 BP recordings randomly selected from the same 44-hour ABPM in a cohort study of 133 dialysis patients. Over a mean follow-up of 24 months, the HRs for all-cause mortality were 1.22/1.18 per 10/5mm Hg increase in mean 44-hour ambulatory BP and 1.20/1.15 when 6 BP recordings from the same ABPM were randomly selected and averaged.³⁰ This notion is also supported by a recent subanalysis of 326 incident dialysis patients participating in the Chronic Renal Insufficiency Cohort (CRIC) study.¹⁵ In this study, in-dialysis unit systolic BP showed a U-shaped curve with mortality risk, in line with several previous cohorts of incident dialysis patients.^11,31,32 In contrast, when systolic BP was recorded during the dialysis-off day in a prespecified follow-up visit (in the office setting), a direct and linear association with all-cause mortality was evident (HR: 1.26; 95% CI: 1.14–1.40 per 10mm Hg higher systolic BP).¹⁵ Notably, neither home BP monitoring nor ABPM was necessary to detect this diverse prognostic association of out-of-dialysis BP readings with mortality.

Level of illness and severity of comorbidities modifies the relationship of BP and mortality

The level of illness and the severity of comorbid conditions with opposing effect on BP may confound the role of hypertension as predictor of risk among dialysis patients.^3,4

In a longitudinal analysis of 4,499 prevalent dialysis patients participating in the Case-Mix Adequacy Study of the US Renal Data System (USRDS), Port et al.¹² showed that a low predialysis systolic BP of <110mm Hg was associated with 86% higher risk of cardiovascular mortality (RR: 1.86; 95% CI: 1.46–2.38, P < 0.001). When the analysis was stratified according to the history of pre-existing heart disease, the risk of mortality associated with low predialysis systolic BP was 2.5-fold higher among those with congestive heart failure.¹² However, when neither coronary artery disease nor congestive heart failure was present, the risk for all-cause mortality was 1.5 and not statistically significant. Accordingly, this landmark cohort study suggests that the underlying history of congestive heart failure may modify the risk for all-cause mortality and systolic BP.

To investigate the influence of age, race, and diabetes on the association of BP with mortality, Myers et al.¹⁷ studied 16,283 dialysis patients initiating renal replacement therapy at facilities operated by a large not-for-profit dialysis care provider in the United States during 2000–2006. They explored the association of age, diabetes, and race with outcomes. Age and systolic BP was modeled using cubic splines that allow the detection of nonlinear relationships of these continuous variables with all-cause mortality. An older age and the presence of diabetes magnified the association of low predialysis systolic BP with excess mortality. Among younger participants, a higher predialysis systolic BP was associated with increased mortality regardless of the diabetic status or race; in contrast, among older patients, high predialysis systolic BP was not predictive of risk.¹⁷

In the Choices for Healthy Outcomes in Caring for ESRD (CHOICE) Study,¹⁸ 446 incident US dialysis patients were classified into a high- vs. a low-biomarker group according to the levels of cardiac troponin I (cTnI ≥ 0.1ng/ml) and N-terminal fragment of prohormone B-type natriuretic peptide (NT-proBNP ≥ 9,252 pg/ml). As expected, patients in the high-biomarker group were older, had more frequently history of pre-existing heart disease and had higher adjusted mortality risk relative to those classified in the low-biomarker group (HR: 1.75; 95% CI: 1.37–2.24).¹⁸ The presence of subclinical cardiovascular disease modified the prognostic significance of predialysis BP (P for interaction 0.01 for all-cause mortality, 0.2 for cardiovascular mortality, and 0.07 for first cardiovascular event). Increasing systolic BP was associated with greater risk of all-cause mortality (HR: 1.07; 95% CI: 1.01–1.14, P = 0.03) only among patients classified in the low-biomarker group.¹⁸ In contrast, predialysis systolic BP was not associated with all-cause mortality in the high-biomarker group (HR: 0.99; 95% CI: 0.92–1.06, P = 0.70).

Dialysis vintage confounds the relationship of BP and mortality

Another source of confounding in the complex association of BP with mortality appears to be the dialysis vintage, i.e., time elapsed from onset of dialysis therapy. In a cohort of 405 patients who had survived at least 2 years on dialysis before the study enrollment, Mazzuchi et al.¹⁶ explored the association of baseline BP with the risk of death during the first 2 years of follow-up (years 3 and 4 of dialysis, early mortality) and the risk of death after the second year of follow-up (>5 years on dialysis, late mortality). When early mortality was analyzed, patients with lower systolic BP of <110mm Hg had 3.9-fold higher risk of death (RR: 3.91; 95% CI: 1.32–11.55). In contrast, when late mortality was analyzed, systolic BP >160mm Hg was the only independent predictor of mortality (RR: 2.42, P = 0.04).¹⁶ This differentiation is possibly explained by the fact that the most frequent causes of early deaths were withdrawal of dialysis therapy and malignancies, whereas late mortality was predominantly due to cardiac diseases.¹⁶ In a subsequent analysis of 16,959 incident dialysis patients in the United States,¹⁹ low predialysis systolic BP of <120mm Hg was associated with increased mortality risk in years 1 and 2 after dialysis initiation. In contrast, among patients who survived at least 3 years on dialysis, high-baseline systolic BP ≥150mm Hg was associated with excess mortality.¹⁹ It appears, therefore, that the assumption of proportionality inherent in the Cox models may be violated; early deaths among those with low systolic BP and late deaths among those with a higher systolic BP may point to distinct pathophysiological pathways.

Modeling BP as a time-varying covariate

The reverse association of low-peridialytic BP with increased mortality risk was shown to be more pronounced in cohort studies treating BP as a time-varying covariate in Cox-proportional hazard models.^3,4 In a cohort of 5,433 prevalent US dialysis patients prospectively followed over a mean follow-up of 2.6 years, Zager et al.¹⁴ showed a U-shaped association of predialysis and postdialysis BP with mortality. When BP was modelled as fixed predictor, the RR for all-cause mortality attributable to low predialysis systolic BP of <110mm Hg was 1.64 (95% CI: 1.22–2.20). When BP was treated as a time-varying covariate, predialysis systolic BP of <110mm Hg was associated with 3.07-fold increased risk of death (RR: 3.07; 95% CI: 2.50–3.76).¹⁴

In a larger cohort of 56,338 patients initiating dialysis therapy during 1997–2001 in a large dialysis chain in the United States,¹¹ predialysis systolic BP showed again a reverse J-shaped curve with 1-year mortality risk. The mortality risk for predialysis systolic BP of <120mm Hg was 3.19 times higher when treated as fixed covariate and 6.27 times higher when treated as time-varying covariate.¹¹

It has to be noted, however, that the magnified mortality risk arising in time-varying Cox-regression models does not prove causality between the risk factor (hypertension) and the outcome (mortality). For example, if a hypertensive patient developed heart failure or another condition with opposing effect on BP, the effect of pre-existing hypertension on mortality would be diluted.^3,4 In contrast, interventional trials where BP is deliberately lowered with the use of antihypertensive agents consistently show cardiovascular protection (vide infra).

Pulsatile BP component as predictor of risk

The association of BP with mortality among dialysis patients is modified depending on the BP component used as predictor of risk in hazards models.^3,4 Analyzing systolic and diastolic BP in isolation may ignore the effect of the pulsatile BP component, which is a powerful predictor of risk both in the general population³³ and among people with chronic kidney disease.^34,35

The relationship between different BP components and all-cause mortality was evaluated in a Japanese cohort of 1,243 prevalent dialysis patients.³⁶ When BP components were considered individually, a positive association with all-cause mortality was observed for PP (HR: 1.080, P = 0.002) and systolic BP (HR: 1.040, P = 0.04), but not for diastolic BP. When systolic and diastolic BP were considered conjointly, systolic BP was associated positively (HR: 1.083, P = 0.001) and diastolic BP inversely with all-cause mortality (HR: 0.086, P = 0.01), implying that widened PP is a strong and direct predictor of risk.³⁶

In a larger cohort of 11,142 prevalent dialysis patients participating in the USRDS Dialysis Morbidity and Mortality Waves 3 and 4 Study, Foley et al.³⁷ explored the prognostic significance of systolic and diastolic BP individually and conjointly in a Cox-proportional hazard model. A reverse association between BP and mortality was noted when systolic and diastolic BP components were considered in isolation. When considered conjointly, a direct association of systolic BP and an inverse association of diastolic BP with all-cause mortality emerged either predialysis or postdialysis. The HRs for all-cause mortality were 1.06/0.79 for predialysis BP and 3.0/0.87 for postdialysis BP (per 10mm Hg increase in BP).³⁷

The association of PP with 1-year mortality risk was assessed in another cohort of 37,069 prevalent dialysis patients from a large dialysis chain in the United States.³⁸ After adjusting for level of systolic BP, a direct and linear association between increasing PP and increasing mortality risk was evident. Each 10mm Hg increase in postdialysis PP increased by 12% the risk of death (HR: 1.12; 95% CI: 1.06–1.18).³⁸

Taken together, the above associations of PP with poor long-term outcomes may reflect a state of accelerated arterial ageing, which is a potent cardiovascular risk predictor among people on chronic dialysis.^39–42 Although PP is informative as a prognostic marker, its value as a clinical measure to provide guidance with respect to clinical decision making is limited. Nonetheless, it should be recognized that for any given systolic BP in a patient on long-term dialysis a lower diastolic BP magnifies the risk for future mortality.

Randomized evidence vs. observational data

The reverse association of BP with all-cause and cardiovascular mortality seen in observational studies is contrasted by the results of 2 meta-analyses of randomized controlled trials showing that BP lowering with antihypertensive drug use confers cardiovascular risk reduction in dialysis patients.^8,9

The first meta-analysis included 8 randomized placebo-controlled trials incorporating data from 1,697 dialysis patients and 495 cardiovascular events. The weighted mean difference in the change of BP between the active-treatment and placebo groups during follow-up was −4.5mm Hg for systolic and −2.3mm Hg for diastolic BP.⁹ This BP-lowering effect was associated with 29% reduction in the risk of all-cause mortality (pooled RR: 0.71; 95% CI: 0.55–0.92) and 29% reduction in the risk of cardiovascular mortality (pooled RR: 0.71; 95% CI: 0.50–0.99).⁹

In the second, Agarwal et al.⁸ included 5 randomized trials with 1,202 study participants. Compared with placebo or control therapy, the overall cardiovascular benefit of BP lowering with antihypertensive drug therapy was a 31% reduced risk of future cardiovascular events (pooled HR: 0.69; 95% CIs: 0.56–0.84). The cardiovascular protection offered by BP lowering was lesser when normotensive patients were included in the analysis (pooled HR: 0.86; 95% CI: 0.67–1.12) (Figure 2).⁸ These interventional data are at variance with the cohort studies but add to the growing body of evidence that hypertension among people on dialysis is harmful and should be controlled.

Figure 2.

Effect of deliberate blood pressure (BP) lowering with antihypertensive drug therapy on cardiovascular events among patients on dialysis. The figure depicts the hazard ratios (HRs) of BP lowering with antihypertensive medications on cardiovascular events. The analysis was separated on the basis of the hypertension status of patients included in individual studies. Deliberate BP lowering was shown to confer cardiovascular protection among dialysis patients, particularly when they are hypertensive. (Reprinted with permission from Agarwal R, Sinha AD. Cardiovascular protection with antihypertensive drugs in dialysis patients: systematic review and meta-analysis. Hypertension 2009; 53:860–6.)

INTRADIALYTIC HYPERTENSION AND MORTALITY

Contrary to the typical decline in BP from predialysis to postdialysis, in approximately 10–15% of dialysis patients, BP exhibits a “paradoxical” intradialytic elevation.^43,44 The exact mechanisms responsible for this complex dialysis-related complication are not yet fully clear. In a post-hoc analysis of the Dry-weight Reduction In Hemodialysis Patients (DRIP) trial, it was shown that probing of dry-weight in those patients with intradialytic hypertension at study initiation resulted in significant reductions of both intradialytic and interdialytic BP, suggesting that intradialytic hypertension is a marker of volume excess and is associated with systemic background hypertension.⁴⁵ This notion is also supported by studies showing that patients experiencing intradialytic hypertension have significantly higher 44-hour ambulatory systolic and diastolic BP levels as compared with carefully matched patients without intradialytic hypertension.^46,47 This phenomenon has gained increased attention in light of recent cohort studies associating intradialytic hypertension with higher mortality risk (Table 4).

Table 4.

Cohort studies evaluating the association of intradialytic change in BP with mortality and clinical outcomes among dialysis patients

Author	Year	N	Region	Follow-up	Predictor	Outcome	Risk
Inrig48	2007	438	USA	6 months	ΔSBP from pre- to post-HD ≥10mm Hg	Combined end-point of nondialysis- related hospitalization or ACM	OR: 2.17; 95% CI: 1.13–4.15 per 10mm Hg increase in SBP during HD
Inrig49	2009	1,748	USA	2 years	ΔSBP from pre- to post-HD	ACM	HR: 1.12; 95% CI: 1.05–1.21 per 10mm Hg increase in SBP during HD
Yang50	2012	115	Taiwan	4 years	ΔSBP >5mm Hg from pre- to post-HD	ACM	HR: 3.925; 95% CI: 1.410–10.846
Park52	2013	113,255	USA	2.2 years	ΔSBP from pre- to post-HD	ACM	U-shaped association of ΔSBP during dialysis with ACM and CVM in Cox-regression analysis with restricted cubic splines
						CVM
Lertdumrongluk51	2015	98,577	USA	NA	Time-varying ΔPP from pre- to post-HD	ACM	U-Shaped association between ΔPP during HD and ACM. Compared to the referent group of ΔPP <−5 to 5mm Hg, those with ΔPP <−25mm Hg had 21% higher risk of ACM (HR: 1.21; 95% CI: 1.14–1.29) and those with ΔPP ≥25mm Hg had 27% elevated risk of ACM (HR: 1.27; 95% CI: 1.18–1.35)

Abbreviations: ACM, all-cause mortality; CI, confidence interval; CVM, cardiovascular mortality; HD, hemodialysis; HR, hazard ratio; NA, not available; OR, odds ratio; ΔPP, change in pulse pressure; ΔSBP, change in systolic blood pressure.

In a post-hoc analysis of the Critic-Line Intradialytic Monitoring Benefit (CLIMB) trial, intradialytic hypertension (defined as a rise is systolic BP of >10mm Hg from predialysis to postdialysis) was detected in 13.2% of study participants.⁴⁸ Compared with patients whose systolic BP fell with dialysis, those with intradialytic hypertension had 2.17-fold elevated risk of reaching the combined end-point of all-cause death and non-access-related hospitalization over 6 months of follow-up (odds ratio (OR): 2.17; 95% CI: 1.13–4.15).⁴⁸

The association of intradialytic hypertension with 2-year survival was evaluated in a longitudinal analysis of 1,748 patients participating in the USRDS Dialysis Morbidity and Mortality Wave II study.⁴⁹ After controlling for several risk factors and postdialysis systolic BP levels, each 10mm Hg increase in systolic BP from predialysis to postdialysis was associated with 12% excess risk of death (HR: 1.12; 95% CI: 1.05–1.21). When the analysis was stratified depending on the levels of predialysis BP, intradialytic hypertension remained prognostically significant only in the subgroup of patients with predialysis systolic BP of 120mm Hg or lower (HR: 1.12; 95% CI: 1.05–1.21).⁴⁹ Another cohort study of 115 dialysis patients showed that a mean rise in systolic BP of >5mm Hg during dialysis over 25 consecutive dialysis sessions was associated with 3.93-fold increased risk of all-cause mortality (HR: 3.93; 95% CI: 1.42–10.85).⁵⁰

In 2 recent large studies using Cox-regression hazard models with restricted cubic splines, it was shown that the magnitude of intradialytic change in BP displays a nonlinear association with mortality.^51,52 Among 113,255 prevalent patients from a large dialysis chain in the United States, Park et al.⁵² reported that large intradialytic reductions in systolic BP of >30mm Hg and rise in systolic BP from predialysis to postdialysis of any degree were both associated with increased mortality risk. Contrary to the above mentioned USRDS data,⁴⁹ the association of intradialytic hypertension with mortality was greater among patients with predialysis systolic BP of 120mm Hg or higher.⁵² This U-shaped curve was confirmed in a subsequent study enrolling 98,577 patients treated in facilities operated by a large dialysis care provider in the United States during 2001–2006. Over 134,814 patient-years of at-risk time, 16,054 (16%) patients died, yielding a crude all-cause mortality rate of 119/1000 patient-years.⁵¹ In models adjusted for either predialysis systolic BP or mean arterial pressure, declines in PP of >25mm Hg (HR: 1.26; 95% CI: 1.18–1.4) as well as elevations in PP of >25mm Hg (HR: 1.27; 95% CI: 1.18–1.35) were both associated with increased risk of death.⁵¹ This U-shaped curve persisted in the case-mix and malnutrition-inflammation-complex syndrome adjusted hazard models.

Whether the excess mortality risk among patients experiencing intradialytic hypertension is solely attributable to the pattern of hemodynamic response within dialysis or whether background systemic hypertension is also of prognostic significance remains unknown. This issue could be clarified by modeling the risk of intradialytic hypertension in analyses adjusted for the degree of 44-hour interdialytic ambulatory BP. If the predictive power of intradialytic hypertension for all-cause or cardiovascular mortality is mitigated after such an adjustment, background interdialytic hypertension would be the prominent contributor to the excess risk of death associated with this condition;⁵³ such an analysis is currently missing.

CONCLUSION

The inverse association of BP recorded before or after dialysis with all-cause and cardiovascular mortality had for long generated the notion that hypertension among people on dialysis is not harmful and should not be adequately controlled. Although, it is clear that systolic BP <110mm Hg measured just prior to the initiation of dialysis, especially in a patient not being treated with antihypertensive agents, is a harbinger of poor outcomes, the theory of “reverse” epidemiology of hypertension in dialysis has fostered complacence in not treating those with volume excess and severe hypertension. It is now time to revisit reverse epidemiology in light of a growing body of evidence that indicates—strongly and consistently—that elevated BP recorded outside of dialysis provides a direct mortality signal. In dialysis patients more so than in any other population, the wide use of home and ABPM is expected to advance the prognostication of cardiovascular morbidity and mortality and improve management of hypertension. Given that currently available guideline recommendations for BP targets used in the general population as well as in people with chronic kidney disease not yet on dialysis are not applicable to those on maintenance dialysis, a high priority goal of future research is to identify the optimal thresholds at which BP should be targeted in these individuals. Since observational data do not provide cause-and-effect associations, randomized clinical trials using out-of-dialysis BP monitoring are urgently warranted to fully elucidate this critical gap in the existing knowledge.

DISCLOSURE

R.A. has served as an advisor for several pharmaceutical companies that make antihypertensive drugs including Merck, Takeda, Novartis, Daiichi Sankyo, Abbvie, Bayer, and Johnson and Johnson; P.I.G. has no conflicts of interest to disclose.