Systolic Blood Pressure and Mortality in Chronic Hemodialysis Patients: Results of a Nationwide Italian Study


Address for correspondence: Attilio Losito, MD, via Dei Mille 5, 06073 Corciano, Italy



Studies on the relationship between blood pressure (BP) and mortality among hemodialysis patients have yielded conflicting results. Reports have come mostly from North America and have dealt with dialysis patients as a homogenous population and differed in methods and time of BP measurement and the optimal BP target. In a prospective nationwide study in 3674 unselected Caucasian patients with end-stage renal disease undergoing chronic hemodialysis from 73 dialysis units, the authors sought to examine the relationship between the different measurements of BP and mortality according to antihypertensive treatment. The mean age of patients was 67.2±14.1 years and the prevalence of diabetes was 19.5%. During follow-up (26.5±10.5 months), 977 deaths were recorded. In the whole cohort, BP was not associated with mortality. After grouping the patients according to antihypertensive treatment, the analysis showed that only in patients who did not take antihypertensive medications (1613) was there an inverse relationship between postdialysis systolic BP and mortality. These patients differed from the others in BP, dialysis vintage, prevalence of diabetes, and type of dialysis technique. This study suggests that with respect to the relationship of BP with mortality, dialysis patients are not a homogenous population. Differences in demographic characteristics and in dialysis technique may therefore explain the reported variability of previous results.

Arterial hypertension affects the majority of patients with end-stage renal disease treated with dialysis and its control is mostly inadequate.[1, 2] In these patients, the optimal blood pressure (BP) target has not been established and studies on mortality and hypertension have yielded conflicting results.[3, 4] While some studies failed to show that hypertension influences the survival of patients, others have suggested the presence of a “J” curve or a more complex relationship between BP and mortality.[5-7] These conflicting results have been ascribed to the differences in the modality of BP measurement or to the presence of comorbid conditions.[8, 9] Important confounding variables, such as those related to dialytic treatment or to antihypertensive therapy often have not been taken into consideration.[10] Furthermore, most studies come mainly from North America or Japan and there are no long-term studies in large numbers of patients from Europe.

The purpose of this study was to investigate the relationship between BP and antihypertensive treatment with all-cause mortality in a large population of Italian dialysis patients.

Materials and Methods

Setting and Participants

This is a prospective observational study of a nationwide sample of Italian patients who underwent hemodialysis. The 4022 unselected patients from 73 dialysis units across Italy were enrolled from May 1, 2006, to December 31, 2007; the study ended December 31, 2009. Patients underwent long-term dialysis for more than 3 months and were older than 18 years. Patients with cachexia caused by malignancies or severe malnutrition or with known severe heart disease (New York Heart Association class III and IV) were excluded. A transthoracic echocardiography was performed in 1953 unselected patients according to the recommendations of the American Society of Echocardiography.[11] A left ventricular ejection fraction <50% was considered abnormal. The dialysis schedule was 3 times a week with a prescribed dialysis time of 240 minutes and a blood flow rate of 300 mL/min. The actual dialysis time during the study period was recorded. The Na+ concentration in the dialysate was not standardized in all units. Patients were also classified according to the dialysis technique: standard diffusive or convective. Interdialytic weight gain (IDWG) was inferred from the average of predialysis–postdialysis weight of the 12 recorded sessions and expressed as a percentage of target dry weight (kg). Kt/V was calculated by the 2-point Daugirdas formula.[12]

BP Measurement

BP was measured at every dialysis unit by trained nurses according to National Kidney Foundation Outcomes Quality Initiative (KDOQI) guidelines.[13] The personnel was specifically instructed on BP measurement for the present study. The measurement was made by a mercury sphygmomanometer on the patient's arm not used for dialysis access. The cuff size was chosen according to arm circumference. BP readings were recorded to the nearest 2 mm Hg. Predialysis BP was measured with the patient sitting at least 5 minutes before the needles for dialysis access were placed. Postdialysis BP was measured at least 5 minutes after the end of the procedure for 12 consecutive sessions. Likewise, predialysis and postdialysis pulse pressure was measured. The mean of the 12 consecutive measures was taken as the baseline value. Hypertension was defined by predialysis systolic BP/diastolic BP (SBP/DBP) ≥140/90 mm Hg or postdialysis hypertension SBP/DBP ≥130/80 mm Hg or the use of antihypertensive medications.[13]Death from all causes was the outcome measure.

Data Analysis

Mean, SD, or percentage of the selected variables were calculated. Continuous variables were compared by analysis of variance. The chi-square test was used for the analysis of categoric variables. The relationship between percentage of IDWG and BP was assessed by linear regression. The association between BP and mortality was assessed by Cox proportional hazard models. Models were built with the following covariates: age, sex, dialysis vintage, presence of diabetes, serum albumin concentration, hemoglobin (Hb), serum cholesterol, parathyroid hormone (PTH), body mass index (BMI), dialysis technique (diffusive or convective), Kt/V, IDWG, and BP components. The BP components in the model were predialysis and postdialysis SBP and DBP or pulse pressure separately. The proportional hazards assumption of the models was assessed using the Shoenfeld residuals.[14] The analysis was performed in the whole cohort first, then, in patients taking antihypertensive treatment, were analyzed separately from untreated patients. The summaries of the models are represented by hazard ratios (HRs) and 95% confidence intervals (CIs). Statistical significance was defined by a P<.05. Data were analyzed with SPSS 15.0 for Windows (SPSS Inc, Chicago, IL).


At the end of the study, 3674 patients completed the follow-up (26.5±10.5 months). The 348 patients lost to follow-up were either transplanted or transferred to different units.

Demographic and clinical characteristics of patients are shown in Table 1. BP was well controlled in 1532 (41.7%) patients. Antihypertensive medications were taken by 2061 patients (56.1%). The analysis of the relationship between IDWG and BP showed an inverse correlation between IDWG and postdialysis SBP, ie, higher IDWG corresponded to lower SBP, (r=0.124, P<.001). No relationship was found with predialysis BP values.

Table 1. Demographic and Clinical Characteristics of 3674 Patients With Long-Term Hemodialysis
  1. Abbreviations: BMI, body mass index; DBP, diastolic blood pressure; IDWG, interdialytic weight gain; Δ MBP, change in mean blood pressure; PP, pulse pressure; PTH, parathyroid hormone; SBP, systolic blood pressure; SD, standard deviation.

Age, y 67.2±14.1
Men, %61.7
Diabetes, %19.5
Hemoglobin, g/dL11.4±1.2
Cholesterol, mg/dL165.1±40.9
PTH, pg/mL296.7±291.4
Albumin, g/dL3.8±0.4
Postdialysis weight, kg66.3±14.0
BMI, kg/m2 24.4±4.3
SBP predialysis, mm Hg137.1±20.7
DBP predialysis, mm Hg73.4±11.3
PP predialysis, mm Hg63.7±16.3
SBP postdialysis, mm Hg131.0±21.6
DBP postdialysis, mm Hg72.0±11.1
PP postdialysis, mm Hg59.0±17.2
Δ MBP, mm Hg2.9±9.3
Antihypertensive therapy, %56.5
Hemodialysis vintage, mo69.3±77.1
IDWG, %3.97±1.32

Dialysis was performed by a convective technique in 984 patients (26.8%). During the observation time there were 977 all-cause deaths.

All-Cause Mortality

Whole Cohort

The analysis produced a model with the following predictive variables: age, male sex, dialysis vintage, low serum albumin, low BMI, diabetes, low Kt/V, and diffusive dialysis technique (Table 2). The highest HRs for mortality were associated with low Kt/V and low serum albumin concentration. None of the BP measurements, predialysis or postdialysis, were associated with all-cause mortality.

Table 2. Variables Associated With All-Cause Mortality in the Cox Model
 No.HR95% CIP Value
  1. Abbreviations: BMI, body mass index; CI, confidence interval; HR, hazard ratio; SBP, systolic blood pressure.

  2. The models were adjusted for hemoglobin, serum cholesterol, parathyroid hormone, Kt/V, interdialytic weight gain (IDWG), and all blood pressure components.

Whole cohort3674   
Age 1.0501.043–1.057<.001
Sex 1.1811.033–1.349.015
Dialysis vintage 1.0021.001–1.002<.001
Diabetes 1.3761.178–1.609<.001
Kt/V 0.6700.503–0.892.006
Serum albumin 0.7030.608–0.813<.001
Dialysis technique 1.1691.003–1.361.045
BMI 0.9720.956–0.988<.001
Patients on antihypertensive treatment2061   
Age 1.0481.039–1.057<.001
Sex 1.2551.046–1.506.014
Dialysis vintage 1.0021.001–1.003.003
Diabetes 1.3291.089–1.622.005
Serum albumin 0.7100.583–0.865<.001
Dialysis technique 1.3181.059–1.641.014
Patients with no treatment1613   
Age 1.0551.045–1.065<.001
Dialysis vintage 1.0011.000–1.003.027
Diabetes 1.4441.127–1.851.004
Kt/V 0.6460.431–0.968.034
Serum albumin 0.6970.559–0.868<.001
IDWG 0.8660.768–0.976.019
SBP (postdialysis) 0.7400.573–0.957.022

Patients With and Without Antihypertensive Treatment

In patients taking antihypertensive treatment, mortality was associated with the following variables: age, sex, dialysis vintage serum albumin, diabetes, and diffusive dialysis technique (Table 2). In patients without antihypertensive treatment, Cox analysis produced a model with the following variables associated with mortality: age, dialysis vintage, diabetes, Kt/V, serum albumin, IDWG, and postdialytic SBP (Table 2). The Figure 1 shows the mortality curves, derived from the Cox model, of patients categorized into tertiles of postdialysis SBP (<112 mm Hg, 112–13 mm Hg, >130 mm Hg). The analysis of the differences between patients treated and not treated with antihypertensive medications showed that untreated patients were characterized by lower BP values in all measurements, longer dialysis vintage, higher Kt/V, and higher prevalence of convective dialysis technique. The comparison of characteristics of the two groups of patients are shown in Table 3. Echocardiography showed a reduced left ventricular ejection fraction (<50%) in 27.5% of patients taking pharmacologic treatment and 26.2% of those not taking therapy (P=.533). Predialysis and postdialysis pulse pressure and BP changes before and after hemodialysis did not show any role in the mortality-predicting models.

Table 3. Characteristics of Patients With and Without Antihypertensive Therapy
 No TherapyTherapyP Value
  1. Abbreviations: DBP, diastolic blood pressure; HD, hemodialysis; SBP, systolic blood pressure.

Diabetes, %15.922.6<.001
HD technique, convective, %29.924.6.001
Dialysis vintage76.0±81.065.9±75.9<.001
Serum albumin, g/dL3.7±0.43.8±0.4<.001
SBP predialysis, mm Hg127.9±20.2144.2±18.4<.001
DBP predialysis, mm Hg69.6±11.076.2±10.7<.001
SBP postdialysis, mm Hg121.0±19.6138.7±19.8<.001
DBP postdialysis, mm Hg67.7±10.475.2±10.4<.001
Figure 1.

Cox survival curves for all-cause mortality in 1613 patients not treated by antihypertensive drugs. Patients are grouped by tertiles of postdialysis systolic blood pressure. Data were appropriately adjusted for covariates included in the corresponding basic model.


In this collaborative prospective study, the first in a large population of Italian dialysis patients, we found several variables associated with all-cause mortality. The analysis of the whole cohort showed no association of BP with mortality. The separate analysis of patients not treated with antihypertensive medications showed an inverse relationship between postdialysis SBP and all-cause mortality. In this group, representing the 44% of the whole cohort, the 35% risk associated with SBP, was only slightly inferior to that associated with reduced Kt/V (54.7%) or low serum albumin (43.4%). The relationship between postdialysis SBP and mortality that we have described in our dialysis patients not treated with antihypertensive medications is a new finding. Previously, dialysis patients had been analyzed as a homogenous population. Only recently, attempts of separate analysis of different subgroups have been reported.[15] As a result, the presence of a relationship between a component of BP and mortality in select subgroups might be overlooked for a dilution of the phenomenon into the whole examined population. This may explain why our analysis has led to results different from those hitherto obtained. Yet, using this approach, a reverse association between baseline BP and mortality has been found in elderly hemodialysis patients.[16] The timing, the site, and the method of the BP measurement are critical in this type of analysis, and may be responsible for the variance of results obtained. A scarce agreement between predialysis and postdialysis measurements of BP and the interdialytic ambulatory BP has been shown.[17, 18] Unfortunately, so far neither ambulatory BP monitoring (APBM) or home BP measurement have been employed in large studies in dialysis patients. Peridialysis measurements are considered imprecise compared with ABPM, and postdialysis SBP has been shown to underestimate average ambulatory BP.[19] On the other hand, it has been shown that the average of 12 predialysis BP measurements obtained in the dialysis unit had equivalent prognostic value to ABPM among patients undergoing hemodialysis.[20] Direct clues to explain the different role of SBP in the outcome between patients with and without antihypertensive medications did not emerge from this study. The finding of the inverse relationship between IDWG and postdialysis SBP, and the association of IDWG with mortality in the Cox model for patients without therapy may raise a hypothesis. The large fluid loss during the dialysis session, subsequent to the high IDWG, may cause low BP and place patients at increased risk. The differences we found in the prevalence of diabetes, dialysis vintage, and convective treatment, together with lower BP values must be taken into account, but we might have missed other relevant characteristics of this group. The lower BP rate in this group may also be a marker of frailty in this elderly population. Comparing our results with those of previous reports, we adjusted the survival analysis for many confounding variables, particularly those related to the type of dialytic treatment and antihypertensive therapy. The contrast with other reports in our findings pointing to the predialysis BP as a risk factor might therefore be more apparent than real. The complex relationship of a predialysis SBP <120 mm Hg with mortality found in other studies does not rule out the possible association of low postdialysis BP with mortality.[21, 22] Furthermore some results were obtained from retrospective studies, mostly based on incident patients. It has been shown that low SBP is associated with increased mortality mainly in the early stage of dialysis treatment; therefore, in incident patients, this phenomenon may be more relevant than in a prevalent population.[6, 12] The mean dialysis vintage of our patients was about 5 years, longer than in other series. In addition, the age of patients and prevalence of diabetes in our cohort were quite different from the majority of reports dealing with hypertension and mortality in dialysis. The mean age in our cohort was older than 67 years compared with younger than 59 in most series. Likewise, in our cohort, the prevalence of diabetes was 19.6%, in keeping with previous surveys in Italian dialysis patients.[23] In most reports on BP and mortality in hemodialysis, it ranged from 33.1% to 44.6%.[6, 12] Furthermore, no black patients were presented in our cohort, in contrast with a prevalence >30% in most series. In our cohort, at baseline, all measures of BP, systolic and diastolic, predialysis and postdialysis were at least 10 mm Hg lower than those recorded in most studies. The mixed results yielded by studies on BP and mortality can be ascribed to the effect of confounding variables. We included in our analysis biochemical and dialytical parameters such as PTH, Kt/V, and dialysis technique not thoroughly investigated previously. In our model, all these variables appear to influence the results on mortality. Previously, this relationship had not been investigated, and we believe it represents a major confounding factor and must be taken into account in the analysis of hypertension and mortality in dialysis patients.

Strengths and Limitations

The major limitation of the present study lies in the in-unit measurement of BP. The values obtained at the dialysis unit are influenced by the setting and the dialytical procedure. Therefore, the reported BPs do not reflect the actual BP profile outside the dialysis unit. Another limitation lies in the limited characterization of the group of patients not treated for hypertension. A common underlying characteristic responsible for the relationship between SBP and mortality in this group may have been missed. On the other hand, the large number of patients, the analysis of covariates related to the dialytic treatment, and the long-term follow-up add strength to the results.


The results from this follow-up study of a large sample of dialysis patients support the hypothesis of a complex relationship between BP and mortality in this population. Demographic and dialytic characteristics define a group of patients with an inverse relationship between postdialysis SBP and mortality not present in the rest of the population.

Acknowledgment and disclosure

Results were presented in abstract form in part at the World Congress of Nephrology 2011 (April 8–12, 2011, Vancouver, Canada). All the authors declare that there are no competing financial interests in relation to the work described.

Appendix 1

List of participating centers

Afragola (NA) F. Assini, Alba (CN) G. Viglino, Anagni (FR) P. Simeoni, Ancona G. Frascà, Atri (TE) M. Tancredi, Aversa (CE) D. Russo, Bollate (MI) U. Teatini, Bracciano (RM) M. Malaguti, Cagliari 1 A. Monni, Cagliari 2 G. Sau, Campobasso M. Brigante, Capodichino (NA) D. Russo, Casoli (CH) M. Joseph, Casoria (NA) D. Russo, Castelnuovo Monti (RE) T. Lusenti, Castiglion Del Lago (PG) A. Selvi, Chieti M. Bonomini, Città di Castello (PG) L. Vecchi, Cividale del Friuli (UD) A. Irlando, Civitavecchia (RM) M. Malaguti, Codroipo (UD) A. Irlando, Como S. Mangano, Conversano (BA) M. Giannattasio, Correggio (CAL RE) T. Lusenti, Fabriano (AN) F. Ciabattini, Fermo (AP) M. Concetti, Foligno (PG) M. Timio, Frosinone F. Scaccia, Gioia Del Colle (BA) M. Giannattasio, Giulianova (TE) G. Marinangeli, Grottaglie (TA) L. Vernaglione, Gualdo Tadino L. Vecchi, Guastalla (CAL RE) T. Lusenti, Gubbio (PG) L. Vecchi, Imola (BO) A. Zuccalà, Ladispoli (RM) M. Malaguti, Lanciano (CH) M. Maccarone, L'Aquila 1 U. Giammaria, L'Aquila 2 S. Stuard, Lecco (LC) L. Del Vecchio, Legnano (MI) C. Guastoni, Macerata F. Sopranzi, Magenta (MI) C. Guastoni, Manduria (TA) V. Nosella, Marsciano (PG) A. Selvi, Massa Carrara G. Betti, Milano (S. Raffaele) C. Lanzani, Modena A. Albertazzi, Montecchio (CAL RE) T. Lusenti, Ortona (CH) A. Stingone, Ostia (RM) M. Morosetti, Palermo M. Li Vecchi, Passinara (MI) U. Teatini, Pavia C. Esposito, Penne (PE) M. Liani, Perugia A. Losito, Pescara 1 T. D'Andrea, Pescara 2 V. Di Luzio, Popoli (PE) S. Acitelli, Putignano (BA) M. Giannattasio, Ravenna R. Cocchi, Reggio Calabria F. Mallamaci, Reggio Emilia T. Lusenti, Rieti Walter Valentini, Roma (FBF) M. Chiappini, Roma (S. Giovanni) A. Balducci, Roma “Villa Anna Maria” E. De Bella, San Bonifacio (Verona) L. Gammaro, San Daniele (UD) A. Irlando, Scandiano (CAL RE) T. Lusenti, Senigallia (AN) R. Boggi, Sesto S. Giovanni (MI) S. Bertoli, Sestri Levante (GE) E. Falbo, Sorgono (NU)F. Martorana, Sulmona (AQ) F. De Meo, Teramo G. Del Rosso, Udine A. Irlando, Vallo della Lucania (SA) G. Mugnani, Vasto (CH) B. Di Paolo, Ventimiglia (GE) R. Ervo.