Ankle-Brachial Blood Pressure Index Predicts All-Cause and Cardiovascular Mortality in Hemodialysis Patients

Home , Hemodialysis

J Am Soc Nephrol 14: 1591–1598, 2003 Ankle-Brachial Blood Pressure Index Predicts All-Cause and Cardiovascular Mortality in Hemodialysis Patients

KUMEO ONO,* AKIYASU TSUCHIDA,† HIRONOBU KAWAI,ʈ HIDENORI MATSUO,‡ RYOUJI WAKAMATSU,§ AKIRA MAEZAWA,¶ SHINTAROU YANO,¶¶ TOMOYUKI KAWADA,# and YOSHIHISA NOJIMA@ for the GUNMA Dialysis and ASO Study Group *Kan-etsu Chuo Hospital, †Toho Hospital, ʈMaebashi Saiseikai Hospital, ‡Hidaka Hospital, §Nishikatakai Clinic, ¶Wakaba Hospital, ¶¶ Hirosegawa Clinic, #Department of Public Health, Gunma University School of Medicine, and @Third Department of Internal Medicine, Gunma University School of Medicine, Maebashi, Japan.

Abstract. A reduction in ankle-brachial BP index (ABPI) is icantly higher in patients with lower ABPI than those with associated with generalized atherosclerotic diseases and pre- ABPI Ն 1.1 to Ͻ1.3. During the study period, 77 cardiovas- dicts cardiovascular mortality and morbidity in several patient cular and 41 noncardiovascular fatal events occurred. On the populations. However, a large-scale analysis of ABPI is lack- basis of Cox proportional hazards regression analysis, ABPI ing for hemodialysis (HD) patients, and its use in this popula- emerged as a strong independent predictor of all-cause and tion is not fully validated. A cohort of 1010 Japanese patients cardiovascular mortality. After adjustment for confounding undergoing chronic hemodialysis was studied between Novem- variables, the hazard ratio (HR) for ABPI Ͻ 0.9 was 4.04 (95% ber 1999 and May 2002. Mean age at entry was 60.6 Ϯ 12.5 yr, confidence interval, 2.38 to 6.95) for all-cause mortality and and duration of follow-up was 22.3 Ϯ 5.6 mo. Patients were 5.90 (2.83 to 12.29) for cardiovascular mortality. Even those stratified into five groups (Ͻ 0.9, Ն 0.9 to Ͻ 1.0, Ն 1.0 to Ͻ with modest reductions in the ABPI (Ն 0.9 to Ͻ1.1) appeared 1.1, Ն 1.1 to Ͻ 1.3, and Ն 1.3) by ABPI measured at entry by to be at increased risk. Patients having abnormally high ABPI an oscillometric method. The frequency distribution of ABPI (Ն 1.3) also had poor prognosis (HR, 2.33 [1.11 to 4.89] and was 16.5% of patients Ͻ 0.9, 8.6% of patients Ն 0.9 to Ͻ 1.0, 3.04 [1.14 to 8.12] for all-cause and cardiovascular mortality, 16.9% of patients 1.0 Ն to Ͻ 1.1, and 47.0% of patients Ն 1.1 respectively). Thus, the present findings validate ABPI as a to Ͻ 1.3, whereas 10.9% of patients had an abnormally high powerful and independent predictor for all-cause and cardio- ABPI (Ն 1.3). The relative risk of a history of diabetes mellitus vascular mortality among hemodialysis patients. (DM), cardiovascular, and cerebrovascular disease was signif-

Growing numbers of patients are now on hemodialysis as a strategies should ultimately improve their survival and quality result of end-stage renal disease (ESRD). ESRD is associated of life. with a substantially reduced life expectancy, of which cardio- Peripheral arterial occlusive disease (PAOD) has recently vascular disease (CVD) is the leading cause (1–3). Several attracted much attention as a risk factor for adverse outcomes. cardiovascular risk factors are applicable to both the general Epidemiological and clinical studies of the general population and the hemodialysis population. In contrast, some cardiovas- have clearly shown that PAOD is a strong predictor for sub- cular risk factors are present to a greater extent in the hemo- sequent cardiovascular and overall mortality (5–8). Limited dialysis population than in the general population, while others available data also suggest that PAOD is prevalent in hemo- appear irrelevant in hemodialysis patients (4). Therefore, it is dialysis patients (4,9–11) and is associated with poor outcomes important to define risk factors critical for overall and cardio- (9,10). However, evaluation of PAOD receives relatively little vascular mortality in this patient population. Identification of attention among hemodialysis patients. Hence they are less patients who require aggressive preventive and interventional likely to receive appropriate treatment than are those, for example, with coronary artery disease. Ankle-brachial BP index (ABPI; the ratio of ankle to brachial systolic BP) is a simple, non-invasive, and reliable Received September 25, 2002. Accepted February 10, 2003. method to access PAOD. Not only useful in diagnosing PAOD, Correspondence to Dr. Yoshihisa Nojima, Third Department of Internal Medicine, Gunma University School of Medicine, Maebashi, Gunma 371-8511, Japan. Phone: large-scale studies showed that ABPI is a strong predictor for 81-27-220-8161; Fax: 81-27-220-8173; E-mail: [email protected] CVD and mortality (5–8). However, the measurement of ABPI 1046-6673/1406-1591 has not been fully validated in the ESRD population. There- Journal of the American Society of Nephrology fore, we sought in the current study to clarify the diagnostic Copyright © 2003 by the American Society of Nephrology and prognostic value of ABPI in hemodialysis patients. The DOI: 10.1097/01.ASN.0000065547.98258.3D ABPI was measured at the baseline in a cohort of Japanese 1592 Journal of the American Society of Nephrology J Am Soc Nephrol 14: 1591–1598, 2003 patients undergoing chronic hemodialysis. All subjects had ABPI Measurement detailed assessment of other risk factors and were followed-up ABPI was determined in all participants and control individuals for 2 yr on average. We found that there is a graded inverse using ABI-form (Colin, Japan), which simultaneously measures bi- relationship between the ABPI and both cardiovascular and lateral arm and ankle (brachial and posterior tibial arteries, respec- all-cause mortality. Multivariate analysis indicated that ABPI tively) BP by an oscillometric method. Before this study, the agree- is independent from other risk factors, including advanced age, ment was assessed between the oscillometric and the conventional Doppler methods in measuring ankle BP in 238 volunteers, including hypertension, and diabetes mellitus. Thus, the measurement of healthy subjects and patients with PAOD (unpublished observations, ABPI effectively identifies high-risk hemodialysis patients, a Mano N. et al.). The variables determined by the two methods were target population requiring intensive follow-up. highly correlated with one another (R ϭ 0.985; P Ͻ 0.001) even in the range of systolic pressure Ͻ 100 mmHg. The mean Ϯ SD of differences was Ϫ0.69 Ϯ 6.47 mmHg, and the limits of agreement were Materials and Methods 12.2 and Ϫ13.6 mmHg, respectively. The BP was measured after Study Design and Patients completion of the dialysis treatment and after allowing patients at rest This prospective cohort study was conducted at 15 dialysis centers in supine position at least for 5 min. Some patients needed more than in the Gunma and Saitama districts of Japan. To be eligible for this 15 min for the BP to stabilize. ABPI was calculated by the ratio of the study, patients had to have received regular hemodialysis at least for ankle systolic pressure divided by the arm systolic pressure. The 3 mo just before entry. Moreover, patients had to be clinically stable systolic pressure of the arm without dialysis access and the lower for 3 mo before entry and specifically lack acute cardiovascular, value of the ankle pressure were used for the calculation. The ABPI cerebrovascular, infectious, or other active diseases. A total of 1010 measurement was done once in each patient. patients were recruited from November 1999 to July 2001. All of them agreed to participate in the follow-up study. The mean Ϯ SD age of Outcome Data Collection the cohort was 60.6 Ϯ 12.5 yr. The observation ended in May 2002. At the end of the follow-up, the status of all patients was assessed During the follow-up period, one patient received a renal transplant. and data on mortality were obtained for the entire cohort. We recorded This patient was followed until the date of transplantation and then 118 deaths, including 77 fatal cardiovascular events, 36 of which were censored. Twenty-eight patients moved away from the study dialysis attributed to heart failure, 14 to cerebral infarction, 11 to the myocar- centers. Among these, outcome data from ten patients could be dial infarction, 8 to cerebral hemorrhage, 2 to pulmonary embolism, obtained. The remaining 18 patients were censored at departure to and 2 to ruptured aneurysms. Other cardiovascular events included another dialysis unit. The mean patient follow-up was 22.3 Ϯ 5.6 mo. aortic valve stenosis, ventricular fibrillation, ischemic gangrene of the Demographic and medical data were obtained from medical foot, and sudden cardiac arrest of unknown cause. The 41 fatal records and interviews with patients and/or the patient’s primary noncardiovascular causes were infectious disease (n ϭ 16), cancer (n nephrologists at study entry. These include age, gender, smoking ϭ 13), gastrointestinal bleeding (n ϭ 3), traffic accident (n ϭ 2), and ϭ history (ever versus never), body mass index (BMI, weight/height2), other (n 7). comorbid conditions, serum creatinine, albumin, cholesterol, and KT/V. Comorbid conditions were defined as follows. Coronary artery Statistical Analyses disease: a history of exertional angina, a history of ischemic electro- Data are expressed as mean Ϯ SD. The ␹2 test for trend was used cardiogram change followed by medication of vasodilator, previous to test for a dose-response relation of variables between ABPI cate- angiogram showing significant occlusive disease, a history of a past gories. Differences between mean values were assessed by ANOVA. myocardial infarction, or a history of coronary artery bypass surgery Bivariate associations between the ABPI and discrete variables were or angioplasty. Cerebrovascular disease: a history of cerebrovascular evaluated with ␹2 and unpaired t test in comparison with a group with accident including cerebral bleeding and infarction. Hypertension: ABPI Ն1.1 to Ͻ1.3. Univariate and multivariate analyses used Cox systolic BP Ն 160 mmHg and/or diastolic pressure Ն 90 mmHg proportional hazards model. Survival curves were estimated by the and/or taking anti-hypertensive medication. Diabetes mellitus (DM): Kaplan-Meier method followed by log-rank test. Statistical signifi- the use of insulin or the presence of chronic renal failure due to DM cance was defined as P Ͻ 0.05. All statistical analyses were per- nephropathy at the start of hemodialysis. Blood samples for creati- formed using a computerized statistical package (SPSS for Windows nine, albumin, and cholesterol were obtained at predialysis as a part of version 10.0, SPSS Inc). routine clinical work within one month of enrollment. KT/V was determined according to the procedure of Shinzato et al. (12). Results An age- and gender-matched control group consisted of 299 The distribution of measured ABPI in control and hemodi- healthy volunteers, 112 women and 187 men. Mean Ϯ SD age of the alysis subjects is shown in Figure 1. The mean Ϯ SD of ABPI Ϯ control was 59.8 10.2 yr. Criteria for inclusion in the control group of 229 control subjects was 1.14 Ϯ 0.08. The ABPI of male included normal renal function, normal pedal pulses, no symptoms of subjects was significantly higher than that of females (1.15 Ϯ claudication, no evidence of ischemic ulcers, gangrene, or amputa- 0.08 versus 1.12 Ϯ 0.08, respectively; P ϭ 0.0047 by unpaired tions of the lower extremity and no previous history of cardiovascular t test). The frequency of subjects with an ABPI Ͻ 0.9 was disease. Subjects were excluded from the control group if they had Ն one or more risk factors for vascular disease, including DM (fasting 0.67%, which was equal to that of subjects with an ABPI plasma glucose Ն 140 mg/dl, hypoglycemic medications, or the 1.3. In contrast, ABPI of hemodialysis patients was distributed presence of DM nephropathy), hypertension (systolic BP Ն 160 more broadly ranging from 0 to 1.75. The frequency distribu- mmHg and/or diastolic pressure Ն 90 mmHg and/or taking anti- tion of ABPI was 16.5% of patients Ͻ 0.9, 8.6% of patients Ն hypertensive medication), or hypercholestemia (fasting cholesterol Ն 0.9 to Ͻ 1.0, 16.9% of patients 1.0 Ն to Ͻ 1.1, and 47.0% of 260 mg/dl). patients Ն 1.1 to Ͻ 1.3, whereas 10.9% of patients had an J Am Soc Nephrol 14: 1591–1598, 2003 ABPI Predicts Mortality in Hemodialysis Patients 1593

Figure 1. Bar graph showing frequency distribution of the ankle-brachial pressure index (ABPI) in control individuals (A) and hemodialysis patients (B). abnormally high ABPI (Ն 1.3). The mean value for the 1010 artery disease, history of cerebrovascular disease, increased hemodialysis patients was 1.07 Ϯ 0.24, which is significantly pulse pressure, decreased diastolic pressure, decreased serum lower than that of control subjects (P Ͻ 0.0001). The differ- albumin, and decreased creatinine levels (P Ͻ 0.0001). Smok- ence was also significant between ABPI of male and female ing history (ever versus never), hypertension, systolic BP, and hemodialysis patients (1.08 Ϯ 0.24 versus 1.05 Ϯ 0.23, re- serum cholesterol levels had no relation to the level of ABPI. spectively; P ϭ 0.027 by unpaired t test). During a 2-yr follow-up, 118 deaths were recorded. Table 3 The characteristics of study population at the time of inclu- shows a Cox proportional hazards analysis of the covariates to sion are shown in Table 1. The age at inclusion was 60.6 Ϯ predict all-cause mortality. In the univariate regression analy- 12.5 yr (range, 18 to 96 yr), and patients were on hemodialysis sis, a hazard ratio (HR) of ABPI Ͻ 0.9 versus Ն 1.1 to Ͻ1.3 for 6.5 Ϯ 5.8 yr (range, 0.3 to 27.5 yr). Patients were stratified was 7.09 (95% CI, 2.28 to 11.80). The HR was also signifi- by ABPI into five subgroups (Ͻ 0.9, Ն 0.9 to Ͻ 1.0, Ն 1.0 to cantly increased with even a mild reduction in ABPI (ABPI Ն Ͻ 1.1, Ն 1.1 to Ͻ 1.3, and Ն 1.3). The characteristics of 0.9 to Ͻ 1.0, 4.83 [2.59 to 9.01]; ABPI Ն 1.0 to Ͻ 1.1, 2.43 patients according to stratified ABPI are shown in Table 2. [1.32 to 4.49]). Moreover, an ABPI Ն 1.3 was also associated Decreased ABPI level was found to be significantly associated with a significant increase in HR (2.20 [1.07 to 4.54]; P ϭ with increased age, prevalence of DM, history of coronary 0.033), indicating that ABPI Ն 1.3 is another risk factor for mortality. Other predictive variables for mortality included male gender, DM, age, coronary artery disease, cerebrovascu- Table 1. Patients characteristics of studied population lar disease, diastolic BP, pulse pressure, serum albumin, and Characteristic All Patients (n ϭ 1010) creatinine levels (P Ͻ 0.0001). Multivariate Cox analysis identified the ABPI, age, cerebrovascular and coronary artery dis- Ϯ Age (yr) 60.6 12.5 ease, and serum albumin level as variables that independently Male gender (%) 63.5 predicted all-cause mortality. The impact of diabetes mellitus, Ϯ Duration of dialysis (yr) 6.5 5.8 diastolic BP, and pulse pressure on all-cause mortality was no Diabetes mellitus (%) 33.8 longer significant. Smoking (ever versus never) (%) 41.4 Seventy-seven cardiovascular deaths were documented dur- Coronary artery disease (%) 18 ing the follow-up period. Table 4 shows a Cox analysis of the Cerebrovascular disease (%) 13.4 ability of various parameters to predict cardiovascular mortal- Hypertension (%) 79 ity. Crude HR of ABPI Ͻ 0.9, Ն 0.9 to Ͻ 1.0, and Ն 1.0 to Ͻ Systolic BP (mmHg) 141.6 Ϯ 22.9 1.1 versus ABPI Ն 1.1 to Ͻ 1.3 were 10.63 [5.24 to 21.58], Diastolic BP (mmHg) 80.6 Ϯ 12.9 8.19 [3.64 to 18.44], and 3.65 [1.60 to 8.32], respectively. Pulse pressure (mmHg) 61.0 Ϯ 16.3 Thus, there was an inverse correlation between low ABPI and BMI 20.7 Ϯ 3.0 cardiovascular mortality. Other predictive covariates entering Albumin level (g/dl) 3.8 Ϯ 0.4 the univariate regression model were age, DM, a history of Cholesterol level (mg/dl) 155.3 Ϯ 33.5 cardiovascular disease, diastolic BP, pulse pressure, albumin Creatinine level (mg/dl) 10.8 Ϯ 3.1 level, creatinine level, cholesterol level, and Kt/V. In multivar- Kt/V 1.3 Ϯ 0.3 iate Cox analysis, low ABPI was a strong independent predic- Hematocrit (%) 30.3 Ϯ 3.5 tor of cardiovascular mortality along with a history of coronary 1594 Journal of the American Society of Nephrology J Am Soc Nephrol 14: 1591–1598, 2003

Table 2. Characteristics of patients at inclusion according to stratified ankle-brachial BP index (ABPI)

ABPI P value Characteristics Ͻ ϭ Ն Ͻ Ն Ͻ Ն Ͻ Ն ϭ ANOVA or 0.9 (n 0.9 to 1.0 1.0 to 1.1 1.1 to 1.3 1.3 (n ␹2test (trend) 167) (n ϭ 87) (n ϭ 171) (n ϭ 475) 110)

Age (yr) 65.4 Ϯ 10.9d 63.2 Ϯ 12.0d 60.8 Ϯ 12.3 58.9 Ϯ 12.7 58.0 Ϯ 11.9 Ͻ0.0001 Male gender (%) 59.9 57.5 61.4 63.2 78.2a 0.012 Duration of dialysis (yr) 5.9 Ϯ 4.7 5.9 Ϯ 5.5 7.7 Ϯ 6.6 6.4 Ϯ 5.8 6.2 Ϯ 6.1 0.034 Diabetes mellitus (%) 61.7b 33.3 31 25.5 31.8 Ͻ0.0001 Smoking history (%) 49.1 35.6 40.9 40.6 38.2 0.209 Coronary artery disease (%) 27.5b 27.5b 22.8b 12.6 11.8 Ͻ0.0001 Cerebrovascular disease 25.7b 18.4a 11.1 9.5 10.9 Ͻ0.0001 (%) Hypertension (%) 80.8 87.4 77.8 78.9 71.8 0.108 Sytolic BP (mmHg) 143.8 Ϯ 23.4 142.4 Ϯ 27.1 141.6 Ϯ 23.6 141.7 Ϯ 22.0 137.6 Ϯ 20.0 0.279 Diastolic BP (mmHg) 77.3 Ϯ 12.7d 77.7 Ϯ 13.9d 80.6 Ϯ 11.9 82.3 Ϯ 128 80.4 Ϯ 12.9 Ͻ0.0001 Pulse pressure (mmHg) 66.5 Ϯ 16.2d 64.7 Ϯ 19.8d 61.0 Ϯ 16.9 59.4 Ϯ 15.6 56.9 Ϯ 12.1 Ͻ0.0001 BMI 20.7 Ϯ 3.1 20.2 Ϯ 2.9 20.3 Ϯ 3.6 20.7 Ϯ 2.7 21.5 Ϯ 2.7c 0.016 Albumin level (g/dl) 3.7 Ϯ 0.4d 3.7 Ϯ 0.4d 3.8 Ϯ 0.4d 3.9 Ϯ 0.4 3. Ϯ 80.5 Ͻ0.0001 Cholesterol level (mg/dl) 157.7 Ϯ 37.3 154.9 Ϯ 33.2 153.5 Ϯ 30.8 154.4 Ϯ 34.0 149.5 Ϯ 28.8 0.394 Creatinine level (mg/dl) 9.7 Ϯ 2.8d 10.0 Ϯ 3.1d 10.9 Ϯ 3.0 11.3 Ϯ 3.0 11.1 Ϯ 3.3 Ͻ0.0001 Kt/V 1.21 Ϯ 0.28 1.32 Ϯ 0.29 1.2 Ϯ 70.28 1.26 Ϯ 0.27 1.21 Ϯ 0.27 0.027 Hematocrit (%) 30.2 Ϯ 3.9 30.3 Ϯ 3.4 30.8 Ϯ 3.5 30.2 Ϯ 3.4 30.0 Ϯ 3.6 0.394

a P Ͻ 0.05, ␹2 test in comparison with ABPI Ն1.1 to Ͻ1.3. b P Ͻ 0.01, ␹2 test in comparison with ABPI Ն1.1 to Ͻ1.3. c P Ͻ 0.05, Unpaired t test in comparison with ABPI Ն1.1 to Ͻ1.3. d P Ͻ 0.01, Unpaired t test in comparison with ABPI Ն1.1 to Ͻ1.3.

artery disease and advanced age. Although DM was a strong In the current study, we performed a large-scale, prospective predictor in univariate analysis, the statistical impact on car- analysis of hemodialysis patients. We found that a low ABPI is diovascular mortality was no longer significant in multivariate strongly associated with a past history of atherosclerotic vas- analysis. Low levels of serum cholesterol paradoxically pre- cular disease and predicts both all-cause and cardiovascular dicted cardiovascular mortality in univariate and multivariate mortality. Multivariate analysis revealed that the role of low analyses. A high ABPI (Ն 1.3) was also a significant parameter ABPI was independent of other risk factors known to affect independently predicting cardiovascular mortality (HR, 3.04 mortality in hemodialysis patients, namely advanced age, DM, [1.14 to 8.12]). preexisting cardiovascular disease, BP, and serum albumin To further demonstrate the predictive value of a low ABPI, levels. Although a low ABPI is strongly associated with a we separately analyzed patients who had no previous cardio- history of atherosclerotic vascular disease, its power to predict vascular events (n ϭ 728). The frequency distribution of ABPI survival is still remarkable even when patients with no preced- in this subpopulation was 12.5% of patients Ͻ 0.9, 7.7% of ing cardiovascular events were analyzed separately. patients Ն 0.9 to Ͻ 1.0, 16.2% of patients 1.0 Ն to Ͻ 1.1, and ABPI was determined using ABI-form (Colin, Japan), which 51.8% of patients Ն 1.1 to Ͻ 1.3, whereas 11.8% of patients simultaneously measures arm and ankle BP by an oscillometric had an abnormally high ABPI (Ն 1.3). As shown in Table 5, a method. This automatic devise allowed us to measure ABPI low ABPI still provides a strong predictor for all-cause mor- quickly, non-invasively, and reproducibly. Previous studies tality in this patient population by multivariate analysis. validated the oscillometric method for measuring arterial BP in Figure 2 shows the Kaplan-Meier curves of all-cause and hemodialysis (13,14) or PAOD patients (15). However, Jon- cardiovascular survival as a function of ABPI values. Compar- sson et al. (16) recently argued that ankle BP by the oscillo- isons between survival curves by log-rank test were highly metric method tended to record values higher than those mea- significant. sured by the conventional Doppler method. The discrepancy was significant when systolic ankle BP was Ͻ 100 mmHg. Discussion Such overestimation of ankle BP by the oscillometric method It has been reported that ABPI is a valuable tool not only for might overlook PAOD by overestimating ABPI. However, the diagnosis of PAOD but also for predicting mortality and mor- ABI-form solves this problem by introducing a double cuff bidity among elderly or hypertensive patients (5,6,8). How- configuration, which is composed of an occlusion cuff and a ever, studies in patients with ESRD have been limited thus far. sensor cuff. As described in Materials and Methods, the agree- J Am Soc Nephrol 14: 1591–1598, 2003 ABPI Predicts Mortality in Hemodialysis Patients 1595

Table 3. Cox proportional hazards regression analysis for all-cause mortality

Univariate Multivariate Parameter Hazard Ratios (95% CI) P Value Hazard Ratios (95% CI) P Value

ABPI (versus Ն1.1 to 1.3Ͻ) Ͻ0.9 7.09 (2.28 to 11.8) Ͻ0.0001 4.04 (2.38 to 6.95) Ͻ0.0001 Ն0.9 to Ͻ1.0 4.83 (2.59 to 9.01) Ͻ0.0001 3.24 (1.70 to 6.20) 0.0004 Ն1.0 to Ͻ1.1 2.43 (1.32 to 4.49) 0.005 1.92 (1.02 to 3.59) 0.042 Ն1.3 2.20 (1.07 to 4.54) 0.033 2.33 (1.11 to 4.89) 0.025 Age (per 1yr) 1.06 (1.04 to 1.07) Ͻ0.0001 1.03 (1.01 to 1.06) 0.003 Male versus female 1.64 (1.09 to 2.48) 0.018 1.67 (1.02 to 2.75) 0.042 Duration of dialysis (per 1 yr) 0.98 (0.95 to 1.01) 0.272 1.02 (0.98 to 1.06) 0.472 Diabetes mellitus 1.88 (1.31 to 2.69) 0.001 1.23 (0.80 to 1.88) 0.351 Smoking (ever versus never) 1.55 (1.08 to 2.22) 0.018 1.22 (0.81 to 1.83) 0.346 Coronary artery disease 2.93 (2.02 to 4.26) Ͻ0.0001 1.76 (1.19 to 2.61) 0.045 Cerebrovascular disease 3.67 (2.50 to 5.37) Ͻ0.0001 2.10 (1.40 to 3.17) 0.0004 Hypertension 0.92 (0.59 to 1.47) 0.692 0.95 (0.58 to 1.56) 0.839 Systolic BP (per 1 mmHg) 1.01 (0.997 to 1.01) 0.232 —— Diastolic BP (per 1 mmHg) 0.98 (0.96 to 0.99) 0.002 0.99 (0.98 to 1.01) 0.446 Pulse pressure (per 1 mmHg) 1.02 (1.01 to 1.03) Ͻ0.0001 1.00 (0.997 to 1.02) 0.157 BMI 0.91 (0.86 to 0.98) 0.006 0.93 (0.87 to 1.00) 0.05 Creatinine level (per 1 mg/dl) 0.87 (0.82 to 0.92) Ͻ0.0001 0.98 (0.90 to 1.06) 0.978 Cholesterol level (per 1 mg/ 0.99 (0.99 to 1.00) 0.009 0.99 (0.99 to 1.00) 0.066 dl) Albumin level (per 1 g/dl) 0.28 (0.19 to 0.43) Ͻ0.0001 0.47 (0.29 to 0.78) 0.003 Kt/V (per 1.0) 0.38 (0.19 to 0.76) 0.006 0.60 (0.27 to 1.31) 0.198 Hematocrit (per 1%) 0.98 (0.93 to 1.03) 0.411 1.02 (0.97 to 1.08) 0.411 ment was excellent between ankle BP determined by the con- erate reduction of ABPI (Ն 0.9 to Ͻ 1.1) as well as with an ventional Doppler method and the oscillometric ABI-form, abnormally high ABPI should be monitored carefully. even when systolic ankle pressure was Ͻ 100 mmHg. Thus, the Previous studies identified several risk factors for PAOD ABI-form reliably measures ankle BP and is validated as a among hemodialysis patients, including advanced age, diabetes method to assess ABPI. mellitus, a history of cardiovascular and cerebrovascular dis- ABPI values of Ͻ 0.9 have been established as a reliable eases, smoking history, low diastolic BP, and low albumin diagnostic marker for PAOD with high sensitivity and speci- levels (4,24). In the current study, we confirmed most of these. ficity (17–19), indicating that at least 16.5% of our studied An interesting and novel finding is a strong correlation of low population had PAOD. However, the fact that patients with ABPI with high pulse pressure. The pulse pressure depends on ABPI Ն 0.9 to Ͻ 1.0 had a significantly increased mortality ventricular ejection, arterial stiffness, and timing of wave re- rate suggests that a substantial part of this group has arterial flections. In this regard, patients with ESRD exhibit vascular obstruction as well. ABPI values can be falsely elevated among abnormalities that contribute elevated pulse pressure, including patients with extensive vascular calcification of the lower arterial stiffness and early wave reflection (25,26). Thus, the extremities (20,21). Because many of our patients had diabetes association between elevated pulse pressure and low ABPI is mellitus with possible vascular calcification, the prevalence of not surprising. PAOD may be underestimated when the criteria of ABPI Ͻ 0.9 In addition to ABPI, we analyzed multiple conventional risk was used. It should also be stressed that ABPI Ն 1.3 is more factors for mortality in hemodialysis patients. Although DM prevalent in hemodialysis patients than in control subjects. was a strong predictor in univariate analysis, there was no This abnormally high ABPI has been interpreted as a sign of statistical significance in multivariate analysis on all-cause and medial wall calcification (21). Long-term hemodialysis pa- cardiovascular mortality. Thus, the impact of DM might be tients have advanced arterial calcification, a process in which mediated by atherosclerotic vascular disease represented by a abnormal calcium and phosphate metabolism plays a pivotal low ABPI. Low levels of serum albumin, serum creatinine, and role. The high prevalence of ABPI Ն 1.3 in our patient pop- BMI are parameters indicating malnutrition, and have been ulation may reflect such ESRD-related pathologic states. Toe repeatedly regarded as powerful predictors for the poor prog- brachial BP index (TBBP) is an alternative method to solve this nosis of hemodialysis patients (27–30). On multivariate anal- issue, which overcomes the false elevation of ABPI due to ysis, however, low ABPI was a more powerful predictor than calcification (22,23). Thus, hemodialysis patients with a mod- these other parameters. As described, a low ABPI was signif- 1596 Journal of the American Society of Nephrology J Am Soc Nephrol 14: 1591–1598, 2003

Table 4. Cox proportional hazards regression analysis for cardiovascular mortality

Univariate Multivariate Parameter Hazard Ratios (95% CI) P Value Hazard Ratios (95% CI) P Value

ABPI (versus Ն1.1 to 1.3Ͻ) Ͻ0.9 10.63 (5.24 to 21.58) Ͻ0.0001 5.90 (2.83 to 12.29) Ͻ0.0001 Ն0.9 to Ͻ1.0 8.19 (3.64 to 18.44) Ͻ0.0001 5.30 (2.30 to 12.24) Ͻ0.0001 Ն1.0 to Ͻ1.1 3.65 (1.60 to 8.32) 0.019 2.82 (1.22 to 6.54) 0.016 Ն1.3 3.08 (1.17 to 8.09) 0.027 3.04 (1.14 to 8.12) 0.025 Age (per 1 yr) 1.06 (1.04 to 1.08) Ͻ0.0001 1.04 (1.01 to 1.07) 0.014 Male versus female 1.46 (0.89 to 2.40) 0.132 1.44 (0.79 to 2.64) 0.239 Duration of dialysis (per 1 yr) 0.98 (0.95 to 1.02) 0.435 1.02 (0.97 to 1.08) 0.386 Diabetes mellitus 2.17 (1.38 to 3.39) 0.0007 1.43 (0.84 to 2.42) 0.19 Smoking (ever versus never) 1.56 (0.998 to 2.44) 0.0512 1.16 (0.69 to 1.93) 0.576 Coronary artery disease 4.30 (2.75 to 6.74) Ͻ0.0001 2.52 (1.57 to 4.02) 0.0001 Cerebrovascular disease 3.21 (1.98 to 5.20) Ͻ0.0001 1.85 (1.10 to 3.12) 0.021 Hypertension 0.99 (0.57 to 1.71) 0.965 1.19 (0.64 to 2.21) 0.588 Systolic BP (per 1 mmHg) 1.00 (0.99 to 1.01) 0.878 —— Diastolic BP (per 1 mmHg) 0.97 (0.96 to 0.99) 0.003 0.99 (0.97 to 1.01) 0.512 Pulse pressure (per 1 mmHg) 1.02 (1.01 to 1.03) 0.008 1.00 (0.99 to 1.02) 0.913 BMI (per 1 kg/m2) 0.95 (0.87 to 1.02) 0.157 0.97 (0.89 to 1.05) 0.429 Creatinine level (per 1 mg/dl) 0.86 (0.80 to 0.93) Ͻ0.0001 0.96 (0.87 to 1.06) 0.39 Cholesterol level (per 1 mg/dl) 0.99 (0.98 to 0.996) 0.003 0.99 (0.98 to 0.998) 0.011 Albumin level (per 1 g/dl) 0.29 (0.17 to 0.49) Ͻ0.0001 0.54 (0.28 to 0.99) 0.049 Kt/V (per 1.0) 0.36 (0.15 to 0.83) 0.016 0.51 (0.19 to 1.35) 0.174 Hematocrit (per 1%) 0.99 (0.94 to 1.07) 0.973 1.04 (0.97 to 1.11) 0.286

Table 5. Cox proportional hazards analysis for all-cause mortality in patients without previous cardiovascular disease at baseline

Univariate Multivariate Parameter Hazard Ratios (95% CI) P Value Hazard Ratios (95% CI) P Value

ABPI (versus Ն1.1 to 1.3Ͻ) Ͻ0.9 5.48 (2.57 to 11.71) Ͻ0.0001 3.02 (1.35 to 6.76) 0.009 Ն0.9 to Ͻ1.0 4.69 (1.92 to 11.47) 0.0007 4.04 (1.61 to 10.16) 0.003 Ն1.0 to Ͻ1.1 1.82 (0.73 to 4.73) 0.19 1.65 (0.65 to 4.21) 0.297 Ն1.3 2.24 (0.84 to 5.96) 0.11 2.09 (0.77 to 5.64) 0.146 Age (per 1 yr) 1.06 (1.04 to 1.09) Ͻ0.0001 1.04 (1.01 to 1.08) 0.01 Male versus female 1.63 (0.86 to 3.01) 0.13 2.04 (1.03 to 4.02) 0.04 Diabetes mellitus 2.08 (1.18 to 3.66) 0.0115 1.42 (0.64 to 2.32) 0.548 Diastolic BP (per 1 mmHg) 0.98 (0.98 to 0.99) 0.0344 0.99 (0.96 to 1.01) 0.276 Pulse pressure (per 1 mmHg) 1.03 (1.01 to 1.04) 0.0013 1.02 (1.00 to 1.04) 0.123 Creatinine level (per 1 mg/dl) 0.84 (0.76 to 0.92) 0.0002 0.95 (0.85 to 1.07) 0.403 Albumin level (per 1 g/dl) 0.21 (0.11 to 0.39) Ͻ0.0001 0.26 (0.12 to 0.55) 0.0015

icantly associated with a past history of cardiovascular athero- Japanese hemodialysis population. The oscillometric method sclerotic diseases. However, we demonstrated that a low ABPI proved to be valuable in screening such high-risk patients. still provides a strong predictor for all-cause mortality in pa- Effective screening will allow us to identify patients who are tients with no preceding cardiovascular events. This further likely to benefit from close follow-up and multiple interven- supports the validity of ABPI measurement in hemodialysis tions, which are of proven value in the general population. patients. Limitations of measuring ABPI in hemodialysis patients were In summary, low ABPI is associated with poor prognosis in also discussed and should be overcome in the future. Such J Am Soc Nephrol 14: 1591–1598, 2003 ABPI Predicts Mortality in Hemodialysis Patients 1597

Clinic: T. Sekihara; Nishikatakai Clinic: T. Matsumoto; Hidaka Hospital: H. Ishida.

References 1. U.S. Renal Data System: USRDS 1998 Annual Data Report, Bethesda, The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases, 1998 2. Mallick NP, Jones E, Selwood N: The European (European Dialysis and Transplantation Association-European Renal Asso- ciation) Registry. Am J Kidney Dis 25: 176–187, 1995 3. Shinzato T, Nakai S, Akiba T, Yamagami S, Yamazaki C, Kitaoka T, Kubo K, Maeda K, Morii H: Report of the annual statistical survey of the Japanese Society for Dialysis Therapy in 1996. Kidney Int 55: 700–712, 1999 4. Cheung AK, Sarnak MJ, Yan G, Dwyer JT, Heyka RJ, Rocco MV, Teehan BP, Levey AS: Atherosclerotic cardiovascular disease risks in chronic hemodialysis patients. Kidney Int 58: 353– 362, 2000 5. Vogt MT, Cauley JA, Newman AB, Kuller LH, Hulley SB: Decreased ankle/arm pressure index and mortality in elderly woman. JAMA 270: 465–469, 1993 6. Newman AB, Sutton-Tyrrell K, Vogt MT, Kuller LH: Morbidity and mortality in hypertensive adults with a low ankle/arm blood pressure index. JAMA 270: 487–489, 1993 7. Leng GC, Lee AJ, Fowkes FG, Whiteman M, Dunbar J, Housley E, Ruckley CV: Incidence, natural history and cardiovascular events in symptomatic and asymptomatic peripheral arterial disease in the general population. Int J Epidemiol 25: 1172–1181, 1996 Figure 2. Probabilities of overall (A) and cardiovascular survival (B) 8. Newman AB, Shemanski L, Manolio TA, Mittelmark M, Polak in the study population according to the level of stratified ABPI. JF, Powe NR, Siscovick D: Ankle-arm index as a predictor of ␹2 ϭ Ͻ Log-rank test for overall survival: 81.46; P 0.0001. Log-rank cardiovascular disease and mortality in Cardiovascular Health ␹2 ϭ Ͻ test for CV survival: 68.95; P 0.0001. Study. Arterioscler Thromb Vasc Biol 19: 538–545, 1999 9. Fishbane S, Youn S, Flaster E, Adam G, Maesaka JK: Ankle-arm blood pressure index as a predictor of mortality in hemodialysis efforts might help ultimately improve survival and quality of patients. Am J Kidney Dis 27: 668–672, 1999 10. Testa A, Ottavioli JN: Ankle-arm blood pressure index (AABPI) life in hemodialysis patients. in hemodialysis patients. Arch Mal Coeur Vaiss 91: 963–965, 1998 Acknowledgments 11. O’Hare AM, Johansen KL: Lower-extremity peripheral arterial The nursing and technical staffs of the 15 dialysis units, who helped disease among patients with end-stage renal disease. J Am Soc to collect the clinical data, are gratefully acknowledged for their Nephrol 12: 2838–2847, 2001 assistance. We thank Nobue Mano (First Department of Surgery, 12. Shinzato T, Nakai S, Fujita Y, Takai I, Morita H, Nakane K, Nagoya University) for measuring ABPI by Doppler method in com- Maeda K: Determination of Kt/V and protein catabolic rate using parison with oscillometric method. We also thank Kenji Kawachi pre and postdialysis blood urea nitrogen concentrations. Nephron (Kawachi Cardiovascular Clinic), Hisao Kumakura (Cardiovascular 67: 280–290, 1994 hospital of Central Japan), and David Rothstein (Yale University, 13. Fagugli RM, Vecchi L, Valente F, Santirosi P, Laviola MM: New Haven, CT) for their helpful suggestions. Comparison between oscillometric and auscultatory methods of ambulatory blood pressure measurement in hemodialysis pa- Appendix: the Gunma Dialysis and ASO Study tients. Clin Nephrol 57: 283–288, 2002 Group Members 14. Cavalcanti S, Marchesi G, Ghidini C: Validation of automated Fujioka General Hospital: K. Fukasawa; Toho Hospital: Y. oscillometric sphygmomanometer (HDBPM) for arterial pres- Abe, K. Amemiya, K. Kaji, S. Tokisawa; Usui Hospital: H. sure measurement during haemodialysis. Med Biol Eng Comput Koyanagi, M. Kobayashi; Fukaya Red Cross Hospital: Y. Ishii, 38: 98–101, 2000 15. Lee BY, Campbell JS, Berkowitz P: The correlation of ankle T. Yamamoto, M. Sakai; Zen-shukai Hospital: J. Hayashi; oscillometric blood pressures and segmental pulse volumes to Shibukawa Central Hospital: M. Shinohara, Y. Yamada; Mae- Doppler systolic pressures in arterial occlusive disease. J Vasc bashi Saiseikai Hospital: C. Takagi, Y. Shimizu, T. Kitahara; Surg 23: 116–122, 1996 Komoro Kousei Hospital: T. Noguchi, M. Yamashita; Wa- 16. Jonsson B, Lindberg LG, Skau T, Thulesius O: Is oscillometric tanabe Clinic: Y. Watanabe; Tone Central Hospital: A. Nada, ankle pressure reliable in leg vascular disease? Clin Physiol 21: S. Kobayashi; Shimada Memorial Hospital: H. Kanai; Shirane 155–163, 2001 1598 Journal of the American Society of Nephrology J Am Soc Nephrol 14: 1591–1598, 2003

17. Laing S, Greenhalgh RM: The detection and progression of asymp- 25. London GM, Marchais SJ, Safar ME, Genest AF, Guerin AP, tomatic peripheral arterial disease. Br J Surg 70: 628–630, 1983 Metivier F, Chedid K, London AM: Aortic and large artery 18. Fowkes FG: The measurement of atherosclerotic peripheral arterial compliance in end-stage renal failure. Kidney Int 37: 137–42, disease in epidemiological surveys. Int J Epidemiol 17: 248–254, 1988 1990 19. Fiegleson HS, Criqui MH, Fronek A, Langer RD, Molgaard CA: 26. Barenbrock M, Spieker C, Laske V, Heidenreich S, Hohage H, Screening for peripheral arterial disesase: The sensitivity, spec- Bachmann J, Hoeks AP, Rahn KH: Studies of the vessel wall ificity, and predictive value of noninvasive tests in adefined properties in hemodialysis patients. Kidney Int 45: 1397–400, popoulation. Am J Epidemiol 140: 526–534, 1994 1994 20. Richbourg MJ: Whatever happened to foot care? Preventing 27. Merkus MP, Jager KJ, Dekker FW, de Haan RJ, Boeschoten EW, amputations in patients with end stage renal disease. EDTNA Krediet RT: The Netherlands Cooperative Study on the Ade- ERCA J 24: 4–10, 1998 quacy of Dialysis. The NECOSAD Study Group: Predictors of 21. Orchard TJ, Strandness DE Jr: Assessment of peripheral vascular poor outcome in chronic dialysis patients. Am J Kidney Dis 35: disease in diabetes. Report and recommendations of an interna- 69–79, 2000 tional workshop sponsored by the American Diabetes and the 28. Leavey SF, Strawderman RL, Jones CA, Port FK, Held PJ: American Heart Association September 18–20, 1992 New Or- Simple nutritional indicators as independent predictors of mor- leans, Louisiana. Circulation 88: 819–828, 1993 22. Ramsey DE, Manke DA, Sumner DS: Toe blood pressure: A tality in hemodialysis patients. Am J Kidney Dis 31: 997–1006, valuable adjunct to ankle pressure measurement for assessing pe- 1998 ripheral arterial disease. J Cardiovascular Surg 24: 43–48, 1983 29. Marce´n R, Teruel JL, de la Cal MA, Gamez C, Spanish Coop- 23. Leskinen Y, Salenius JP, Lehtimaki T, Huhtala H, Saha H: The erative Study of Nutrition in Hemodialysis: The mpact of mal- prevalence of peripheral arterial disease and medial arterial cal- nutrition in morbidity and mortality in stable haemodialysis cification in patients with chronic renal failure: Requirements for patients. Nephrol Dial Transplant 12: 2324–2331, 1997 diagnostics. Am J Kidney Dis 40: 472–479, 2002 30. Owen WF Jr, Lew NL, Liu Y, Lowrie EG, Lazarus JM: The urea 24. O’Hare AM, Hsu CY, Bacchetti P, Johansen KL: Peripheral reduction ratio and serum albumin concentration as predictors of vascular disease risk factors among patients undergoing hemo- mortality in patients undergoing hemodialysis. N Eng J Med 329: dialysis. J Am Soc Nephrol 13: 497–503, 2002 1001–1006, 1993

Access to UpToDate on-line is available for additional clinical information at http://www.jasn.org/