original contributions
Reverse Epidemiology of Hypertension-Mortality
Associations in Hemodialysis Patients: A Long-Term
Population-Based Study
Chih-Chiang Chien1,2, Chun-Sheng Yen1, Jhi-Joung Wang3, Hung-An Chen4, Ming-Ting Chou5,6, Chin-Chen
Chu3, Chung-Ching Chio7, Jyh-Chang Hwang1,6, Hsien-Yi Wang1,6, Yi-Hua Lu1 and Wei-Chih Kan1,8
Background
Although hypertension (HTN) is a predictor of mortality, recent
data have questioned the link between baseline HTN and mortality
in incident hemodialysis (HD) patients. We used Taiwan’s National
Health Insurance claim data (NHRI-NHIRD-99182) to investigate
the association.
Methods
In 1999, this longitudinal cohort study enrolled 5752 new HD
patients. Follow-up began from the initiation of HD until death, the
end of HD, or the end of 2008. A Kaplan–Meier survival analysis was
done. Cox proportional hazard analysis was used to identify the risk
factors for mortality.
Results
The prevalence of baseline HTN was 75.47%. Patients with HTN had
a higher prevalence of diabetic mellitus (DM) and cardiovascular
diseases. The 1-, 5-, and 9-year cumulative survival rates were 95.5, 63.7
and 41.8% in patients with HTN, and 95.5, 71.0, and 52.0% in those
Introduction
Hypertension (HTN) is an important public health problem
and affects 24% of the general population.1 The prevalence of
HTN in patients with chronic kidney disease (CKD) signifi-
cantly exceeds that seen in the general population; it is esti-
mated to be more than 70% in patients with end-stage renal
disease (ESRD) and undergoing hemodialysis (HD).2,3
Patients with CKD are in the highest risk group for cardio-
vascular events.4 Although HTN is an important predictor of
1Department of Nephrology, Chi-Mei Medical Center, Tainan, Taiwan;
2Department of Food Nutrition, Chung Hwa University of Medical Technology,
Tainan, Taiwan; 3Department of Medical Research, Chi-Mei Medical Center,
Tainan, Taiwan; 4Department of Rheumatology, Chi-Mei Medical Center,
Tainan, Taiwan; 5Department of Cardiology, Chi-Mei Medical Center, Tainan,
Taiwan; 6Department of Sport Management, College of Leisure and Recreation
Management, Chia Nan University of Pharmacy and Science, Tainan, Taiwan;
7Department of Neurological Surgery, Chi-Mei Medical Center, Tainan, Taiwan;
8Department of Medical Laboratory Science and Biotechnology, Chung Hwa
University of Medical Technology, Tainan, Taiwan. Correspondence: Wei-Chih
Kan (rockiekan@ntu.edu.tw)
Received 2 December 2012; first decision 11 February 2012; accepted 13 April 2012.
© 2012 American Journal of Hypertension, Ltd
900 august 2012 | VOLUME 25 NUMBER 8 | 900-906 | AMERICAN JOURNAL OF HYPERTENSION
nature publishing group
Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020
without HTN (log-rank test: P <0.001). Multivariate analysis showed
that patients with baseline HTN may have a higher survival rate (hazard
ratio (HR) 0.901, 95% confidence interval (CI): 0.819–0.992). After
stratification by age and DM, only elderly (≥65) patients without DM
had a significantly higher survival rate (HR 0.769, 95% CI: 0.637–0.927).
HTN predicts lower mortality with increasing age in patients with
congestive heart failure (CHF) or coronary artery disease (CAD).
Conclusions
There is a reverse (counterintuitive) association between baseline
HTN and mortality in elderly HD patients without DM and a clear
tendency for a reverse association with increasing age in patients
with CHF or CAD. Further study of the association between HTN and
mortality in older HD patients may be warranted.
Keywords: blood pressure; hemodialysis; hypertension; mortality;
reverse epidemiology
American Journal of Hypertension, advance online publication 31 May 2012.
doi:10.1038/ajh.2012.60
cardiovascular mortality, several studies1,5,6 have questioned
the link between HTN and mortality in HD patients. Some
studies7–11 report a counterintuitive association between HTN
and mortality in HD patients: a normal-to-low blood pressure
(BP) is associated with increased all-cause and cardiovascular
death, whereas a high BP confers survival advantages.7–10 The
phenomenon is referred to as “reverse epidemiology”,12 which
has been proposed to address the observation of a counterin-
tuitive association between an exposure and outcome. Foley et
al.6 found that elevated BP was associated with an increase in
left ventricular mass index, heart failure, and coronary artery
disease (CAD), but that low BP was independently associated
with increased mortality. However, there are still conflicting
effects on the impact of HTN. Tomita et al.11 reported that
the survival rate was higher in patients with a systolic BP <
160 mm Hg than those with a systolic BP ≥ 160 mm Hg.
The primary goal of HD is to prolong patient survival.
Age, diabetes mellitus (DM) and heart disease are important
factors.13 We hypothesized that baseline HTN has a differ-
ent impact on subsequent mortality in HD patients stratified
by age, DM, and heart disease. Publications on this subject
Hypertension-Mortality Associations in HD Patients
are still inconclusive,14 especially epidemiological data for
national cohort studies of incident HD patients. To explore
this hypothesis, we used the Taiwan National Health Insurance
(NHI) database to investigate the association between baseline
HTN and mortality in HD patients.
Methods
Database
NHI program has provided compulsory universal health
insurance in Taiwan since 1995. With the exception of prison
inmates, all citizens are enrolled in the program. All contracted
medical institutions must submit standard computerized claim
documents for medical expenses. Patients with ESRD are
eligible for any type of renal replacement therapy free of any
charge; all chronic dialysis patients are covered by NHI.
Data were obtained from the National Health Insurance
Research Database15 and released for research by the Taiwan
National Health Research Institute. The National Health
Insurance Research Database covers nearly all (99%) inpatient
and outpatient medical benefit claims for Taiwan’s 23 million
residents, is one of the largest and most comprehensive data-
bases in the world, and has been used extensively in various
studies.15 Patient identification numbers, gender, birthday,
dates of admission, and discharge, medical institutions provid-
ing the services, the ICD-9-CM (International Classification of
Diseases, 9th Revision, Clinical Modification) diagnostic and
procedure codes (up to five each), and outcomes are encrypted.
The dataset was released with de-identified secondary data
for public research purposes. All types of personal identifi-
cation on files connected with the present study were scram-
bled using surrogate identification numbers to secure patient
privacy. The present study was exempted from ethical review.
The Bureau of National Health Insurance approves the appli-
cation (NHRI-NHIRD-99182) after reviewing all the required
medical documents. We used the National Health Insurance
Research Database for ambulatory care claims, all inpatient
claims, and the updated registry for beneficiaries from 1998
to 2008 for this study. All datasets can be interlinked through
each individual’s unique personal identification number.
Patient selection and definition
We designed a longitudinal cohort study and selected all adult
ESRD patients (≥18 years old) on maintenance dialysis who
began renal replacement therapy between 1 January 1999, and
31 December 1999 (n = 6,170). ESRD patients on maintenance
dialysis were defined as having undergone dialysis for more
than 90 days. Patients who had undergone renal transplantation
before beginning dialysis were excluded (n = 36). Patients were
followed-up from the first reported date of HD to the date of
death, a switch to peritoneal dialysis, renal transplantation, or 31
December 2008. Data on patients were censored if they switched
to peritoneal dialysis, underwent renal transplantation, or 31
December 2008. The primary outcome (event) was death from
any cause. The predictor was HTN. The determination of modal-
ity switches after the initial modality choice followed the “60-day
rule” used by the USRDS (i.e., any change in modality lasting less
AMERICAN JOURNAL OF HYPERTENSION | VOLUME 25 NUMBER 8 | august 2012 901
original contributions
than 60 days is not recorded as a “switch” in the database).16 A
total of 5,752 HD patients were analyzed in this study.
Ascertaining the demographic and comorbid variables
We linked to the diagnostic codes through the inpatient and
outpatient claims databases of the NHI. Our research included
not only survival status, but also date of death, patient demo-
graphics, and baseline comorbidities. Baseline comorbidi-
ties—such as HTN, DM, congestive heart failure (CHF), CAD,
cerebrovascular accident, peripheral vascular disease, chronic
Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020
lung disease, chronic liver disease, and cancer—were impor-
tant factors affecting mortality and were assessed at the start of
dialysis. These characteristics are consistent with those in pre-
vious studies and demonstrate the need to adjust when com-
paring mortality rates among dialysis patients. The ICD-9-CM
codes used to define each condition are shown in Appendix 1.
An individual was classified as an HTN patient if he or she had
essential HTN, secondary HTN, hypertensive heart disease,
hypertensive renal disease, or hypertensive retinopathy. An
individual was classified as a DM patient if he or she had type 1
DM, type 2 DM, DM without mention of complication, or DM
with mention of complications (ketoacidosis, hyperosmolarity,
coma, nephropathy, neuropathy, peripheral circulatory disor-
ders, hypoglycemia, retinopathy, and cataract). An individual
was classified as a CAD patient if he or she had acute myocar-
dial infarction, old myocardial infarction, angina pectoris or
another acute, subacute, or chronic form of ischemic heart dis-
ease. An individual was classified as a CHF patient if he or she
had left heart failure, right heart failure, systolic heat failure,
diastolic heart failure, rheumatic heart failure, acute myocar-
ditis, cardiomyopathy, hypertensive heart failure, or any other
form of heart failure. An individual was classified as a periph-
eral vascular disease patient if he or she had atherosclerosis,
aortic aneurysm and dissection, other aneurysms, arterial
embolism and thrombosis, phlebitis and thrombophlebitis,
portal vein thrombosis, or other peripheral vascular disease.
Comorbidities, which the patients had, were determined by
whether they fit one of the following two definitions: (i) diag-
nostic codes from outpatient visits if the patient had an initial
diagnosis at any time in the 1 year before the start of dialysis
and then had one or more additional diagnoses within the
subsequent 12 months. The first and last outpatient visit within
1 year had to have been >30 days apart to avoid accidental
inclusion of miscoded patients; (ii) diagnostic codes in hospi-
talization databases at least once in the 1 year before the start
of dialysis.
Statistical analyses
The data were analyzed using the Statistical Package for Social
Sciences for Windows 17.0 (SPSS, Chicago, IL). Baseline char-
acteristics of groups of patients with and without HTN were
compared using Pearson χ2 tests. Age was entered as a categor-
ical variable (18–44, 45–64, and 65 years or older). Significance
was set at P < 0.05. Overall patient survival was described using
the Kaplan–Meier method based on dialysis modality. Hazard
ratios (HRs) and 95% confidence intervals (CIs) were derived
original contributions Hypertension-Mortality Associations in HD Patients

from Cox proportional hazards models. Cox models met the
assumption of proportionality of risks. To adjust for potential
confounding in the association between comorbidities and the
risk of mortality, multivariate analyses were used to model all-
cause mortality. To test the hypothesis that being in different
age groups and having DM have a different effect on mortality
in HD patients with and without HTN, we further stratified
these HD patients by age-group and DM.
Results
female patients. With increasing age, the proportion of HTN
increased. Patients with baseline HTN tended to have more
comorbidities than those without baseline HTN (Table 1).
Cumulative survival rate
During the follow-up period, 2,775 patients died (Figure 1).
The mean follow-up time alive on dialysis was 87.13 months
(95% CI: 84.94–89.32) in patients without HTN and 80.45
months (95% CI: 79.18–81.72) for patients with HTN.
The cumulative survival rate for patients without HTN

Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020

Demographics and clinical characteristics was 95.5% at 1 year, 71.0% at 5 years, and 52.0% at 9 years,
Of 5,752 HD patients, 4,341 patients (75.47%) had baseline and for patients with HTN was 95.5% at 1 year, 63.7% at 5
HTN. There were no significant differences between male and years, and 41.8% at 9 years. The log-rank test showed a sig-

Table 1 | Patient characteristics and association with and without hypertension on hemodialysis
HD without HTN HD with HTN P value
n (%) n (%)
Gender 0.958
  Female 751 (53.2) 2307 (53.1)
  Male 660 (46.8) 2034 (46.9)
Age at start of HD <0.001
  18-44 years old 328 (23.20) 566 (13.00)
  45-64 years old 578 (41.00) 1991 (45.90)
  ≥ 65 years old 505 (35.80) 1784 (41.10)
Baseline comorbidity
Diabetes mellitus <0.001
  No 999 (70.8) 1666 (38.4)
  Yes 412 (29.20) 2675 (61.60)
Congestive heart failure <0.001
  No 1286 (91.10) 3445 (79.40)
  Yes 125 (8.90) 896 (20.60)
Coronary artery disease <0.001
  No 1275 (90.40) 3172 (73.10)
  Yes 136 (9.60) 1169 (26.90)
Cerebrovascular disease <0.001
  No 1349 (95.60) 3728 (85.90)
  Yes 62 (4.40) 613 (14.10)
Peripheral vascular disease <0.001
  No 1373 (97.3) 4141 (95.4)
  Yes 38 (2.7) 200 (4.6)
Chronic lung disease <0.001
  No 1335 (94.60) 3931 (90.60)
  Yes 76 (5.40) 410 (9.40)
Chronic Liver Disease 0.208
  No 1278 (90.60) 3980 (91.70)
  Yes 133 (9.40) 361 (8.30)
Cancer <0.001
  No 1320 (93.60) 4211 (97.0)
  Yes 91 (6.40) 130 (3.0)
HD, hemodialysis; HTN, hypertension.

902 august 2012 | VOLUME 25 NUMBER 8 | AMERICAN JOURNAL OF HYPERTENSION

Hypertension-Mortality Associations in HD Patients original contributions

nificant difference in survival rates between the two groups
(log-rank: P < 0.001). Although crude survival was better in
patients without HTN, there was a reverse outcome in long-
term mortality between the two groups after adjusting for
age, gender, and comorbidities (HR: 0.901, 95% CI: 0.819–
0.992) (Table 2).
Risk factors for all-cause mortality in HD patients
A multivariate Cox proportional hazards analysis of baseline
data showed that male gender, being ≥45 years old, DM, CHF,
in patients 18–44 years old, 0.934 in those 45–64 years old,
and 0.697 in those ≥65 years old. There was a clear tendency
toward HTN’s predicting lower mortality with increasing age
in patients with CHF.
Risk factors for all-cause mortality in HD patients stratified by
CAD and different age groups
We further stratified patients by CAD and different age groups
(Table 5). After a multivariate analysis, baseline HTN was an
independent predictor for lower mortality only in nonCAD

Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020

CAD, cerebrovascular accident, peripheral vascular disease, HD patients ≥65 years old (HR: 0.840, 95% CI: 0.721–0.97).
chronic lung disease, chronic liver disease, and cancer were The HR decreased with increasing age in patients with CAD;
associated with a significantly higher mortality rate (Table 2). the HR was 1.134 in patients 18–44 years old, 1.029 in those
However, the mortality rate for patients with baseline HTN 45–64 years old, and 0.74 in those ≥65 years old. There was a
was lower than for those without HTN. clear tendency toward HTN’s predicting lower mortality with
increasing age in patients with CAD.
Risk factors for all-cause mortality in HD patients stratified by
DM and different age groups Discussion
We further stratified patients by DM and different age groups This study used the Taiwan NHI database to investigate the
(Table 3). After multivariate analysis, baseline HTN was an association between baseline HTN and subsequent long-term
independent predictor for lower mortality only in nonDM mortality in chronic HD patients. The population studied was
HD patients ≥ 65 years old (HR: 0.769, 95% CI: 0.637–0.927). nationwide and representative. There was a high prevalence of
Mortality was higher for male than female patients only in baseline HTN (75.47%) in incident HD patients in our registry.
the nonDM group. A history of CHF was a significant predic- Patients with a history of HTN also had a higher rate of DM
tor of mortality in the DM group and the elderly (≥65 years
old).
Table 2 | Adjusted hazard ratio a for all-cause mortality in
Risk factors for all-cause mortality in HD patients stratified by hemodialysis patients
CHF and different age groups Covariate Univariate analysis Multivariate analysis
We further stratified patients by CHF and different age groups HR (95% CI) HR (95% CI)
(Table 4). After a multivariate analysis, baseline HTN was an Sex (male vs. female) 1.228 (1.140–1.323)* 1.175 (1.090-1.266)*
independent predictor for lower mortality in elderly patients Age at initiation of
both with (HR: 0.849, 95% CI: 0.733–0.983) and without hemodialysis
(HR: 0.697, 95% CI: 0.489–0.995) CHF. The HR decreased   18–44 years old Referent Referent
with increasing age in patients with CHF; the HR was 5.492   45–64 years old 2.659 (2.300–3.075)* 2.142 (1.849–2.482)*
  ≥ 65 years old 5.637 (4.878–6.515)* 4.384 (3.780–5.084)*
1.0
Log-rank P < 0.001 Diabetes mellitus (yes vs. no) 2.197 (2.033–2.375)* 1.776 (1.636–1.928)*
Hypertension (yes vs. no) 1.312 (1.199–1.436)* 0.901 (0.819–0.992)*
0.8
Congestive heart failure 1.988 (1.817–2.175)* 1.514 (1.374–1.668)*
Cumulative survival rate

Without HTN (yes vs. no)

0.6
With HTN Coronary artery disease 1.860 (1.711–2.022)* 1.139 (1.040–1.248)*
(yes vs. no)
0.4 Cerebrovascular disease 1.862 (1.674–2.071)* 1.354 (1.216–1.509)*
(yes vs. no)
0.2 Peripheral vascular disease 2.141 (1.567–2.926)* 1.648 (1.203–2.256)*
(yes vs. no)

0.0 Chronic lung disease 1.805 (1.599–2.037)* 1.278 (1.130–1.446)*

(yes vs. no)
0 12 24 36 48 60 72 84 96 108 120
Chronic liver disease 1.467 (1.293–1.663)* 1.334 (1.176–1.513)*
Number at risk Survival time (months) (yes vs. no)
With HTN patients 4341 3960 3443 2908 2561 2209 1933 1733 1552 1371
Cancer (yes vs. no) 1.521 (1.268–1.825)* 1.273 (1.059–1.532)*
Without HTN patients 1411 1288 1133 996 906 827 758 684 636 588
CI, confidence interval; HR, hazard ratio.
aThe analyses were adjusted for age, gender, diabetic mellitus, hypertension, congestive
Figure 1 | Crude overall Kaplan–Meier survival curves for incident end-stage heart failure, coronary artery disease, cerebrovascular disease, peripheral vascular disease,
renal disease hemodialysis patients stratified by the presence and absence of chronic lung disease and cancer. *P < 0.05.
hypertension (HTN) at the start of dialysis.

AMERICAN JOURNAL OF HYPERTENSION | VOLUME 25 NUMBER 8 | august 2012 903

original contributions Hypertension-Mortality Associations in HD Patients

Table 3 | Numbers and adjusted hazard ratioa for all-cause mortality in hemodialysis patients stratified by diabetes mellitus and
age-group
Covariate Age 18–44 Age 45–64 Age ≥65
NonDM DM NonDM DM NonDM DM
(n = 640) (n = 254) (n = 1039) (n = 1530) (n = 986) (n = 1303)
Sex (male vs. 1.777 (1.218–2.592)* 1.122 (0.734–1.713) 1.725 (1.385–2.149)* 1.094 (0.954–1.254) 1.182 (1.000–1.398)* 0.994 (0.865–1.142)
female)
Hypertension 0.692 (0.478–1.002) 1.293 (0.714–2.343) 0.964 (0.769–1.208) 0.958 (0.779–1.179) 0.769 (0.637–0.927)* 0.879 (0.718–1.078)
(yes vs. no)

Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020

Congestive heart 1.431 (0.656–3.121) 1.799(1.070–3.025)* 1.352 (0.946–1.932) 1.427 (1.204–1.691)* 1.416 (1.132–1.771)* 1.633 (1.382–1.929)*
failure (yes vs. no)
Coronary artery 0.811 (0.254–2.585) 1.220 (0.714–2.085) 1.213 (0.881–1.671) 1.305 (1.107–1.538)* 1.241 (1.011–1.523)* 0.966 (0.831–1.123)
disease (yes vs. no)
Cerebrovascular 0.580 (0.134–2.512) 1.845 (0.983–3.461) 1.476 (0.980–2.221) 1.468 (1.224–1.761)* 1.196 (0.907–1.578) 1.271 (1.066–1.514)*
disease (yes vs. no)
Peripheral vascular 25.655 (2.951–223.04)*10.209 (1.356–76.862)*2.360 (0.323–17.242) 1.226 (0.785–1.916) 2.300 (0.951–5.670) 2.127 (1.193–3.793)*
disease (yes vs. no)
Chronic lung 0.755 (0.098–5.831) 1.212 (0.487–3.020) 1.631 (1.093–2.436)* 0.895 (0.693–1.155) 1.361 (1.069–1.735)* 1.434 (1.171–1.757)*
disease (yes vs. no)
Chronic liver 2.530 (1.394–4.592)* 0.912 (0.429–1.937) 1.887 (1.349–2.639)* 1.302(1.043–1.626)* 1.286 (0.941–1.757) 1.118 (0.880–1.420)
disease (yes vs. no)
Cancer (yes vs. no) 4.371 (1.583–12.072) 2.570 (0620–10.657) 3.300 (2.141–5.088)* 1.093 (0.743–1.608) 0.976 (0.681–1.400) 1.060 (0.739–1.521)
DM, diabetic mellitus; n, the number of patients in each subgroup.
aThe analyses were adjusted for gender, hypertension, congestive heart failure, coronary artery disease, cerebrovascular disease, peripheral vascular disease, chronic lung disease

and cancer. All data are hazard ratio (95% confidence interval).
*P < 0.05.

Table 4 | Numbers and adjusted hazard ratioa for all-cause mortality in hemodialysis patients stratified by congestive heart failure
and age-group
Covariate Age 18–44 Age 45–64 Age ≥65
NonCHF CHF NonCHF CHF NonCHF CHF
(n = 823) (n = 71) (n = 2133) (n = 436) (n = 1775) (n = 514)
Sex (male vs. female) 1.613 (1.202–2.165)* 0.989 (0.407–2.403) 1.276 (1.120–1.454)* 1.149 (0.894–1.476) 1.076 (0.952–1.216) 1.037 (0.829–1.298)
Hypertension 0.733 (0.539–1.013) 5.492 (0.709–42.537) 0.964 (0.818–1.137) 0.934 (0.618–1.410) 0.849 (0.733–0.983)* 0.697 (0.489–0.995)*
(yes vs. no)
Diabetic mellitus 2.420 (1.782–3.288)* 2.786 (1.000–7.766) 2.204 (1.906–2.548)* 2.236 (1.546–3.232)* 1.392 (1.225–1.581)* 1.531 (1.203–1.948)*
(yes vs. no)
Coronary artery 0.819 (0.438–1.528) 2.286 (0.974–5.369) 1.344 (1.128–1.601)* 1.251 (0.968–1.618) 0.973 (0.839–1.127) 1.244 (0.991–1.561)
disease (yes vs. no)
Cerebrovascular 1.522 (0.819–2.830) 0.599 (0.163–2.198) 1.411 (1.161–1.714)* 1.600 (1.158–2.212)* 1.329 (1.118–1.579)* 1.092 (0.818–1.457)
disease (yes vs. no)
Peripheral vascular 21.855 (5.186– – 1.121 (0.659–1.906) 1.630 (0.758–3.509) 2.306 (1.350–3.939)* 2.226 (0.698–7.104)
disease (yes vs. no) 92.105)*
Chronic lung disease 1.385 (0.563–3.410) 0.848 (0.097–7.409) 1.317 (1.020–1.700)* 0.688 (0.465–1.018) 1.481 (1.221–1.798)* 1.344 (1.034–1.747)*
(yes vs. no)
Chronic liver disease 1.869 (1.151–3.033)* 0.636 (0.126–3.215) 1.534 (1.247–1.888)* 1.162 (0.770–1.752) 1.161 (0.936–1.440) 1.175 (0.777–1.776)
(yes vs. no)
Cancer (yes vs. no) 3.636 (1.589–8.319)* – 1.663 (1.225–2.258)* 1.210 (0.531–2.757) 0.934 (0.696–1.252) 1.274 (0.761–2.132)
CHF, congestive heart failure; n, the number of patients in each subgroup.
aThe analyses were adjusted for gender, hypertension, congestive heart failure, coronary artery disease, cerebrovascular disease, peripheral vascular disease, chronic lung disease

and cancer. All data are hazard ratio (95% confidence interval).
*P < 0.05.

904 august 2012 | VOLUME 25 NUMBER 8 | AMERICAN JOURNAL OF HYPERTENSION

Hypertension-Mortality Associations in HD Patients original contributions

Table 5 | Numbers and adjusted hazard ratioa for all-cause mortality in hemodialysis patients stratified by coronary artery disease
and age-group
Covariate Age 18–44 Age 45–64 Age≥65
NonCAD CAD NonCAD CAD NonCAD CAD
(n = 830) (n = 64) (n = 2046) (n = 523) (n = 1571) (n = 718)
Sex (male vs. 1.513 (1.132–2.023)* 1.617 (0.480–5.443) 1.284 (1.123–1.469)* 1.176 (0.934–1.480) 1.094 (0.961–1.246) 1.003 (0.829–1.213)
female)
Hypertension 0.787 (0.578–1.071) 1.134 (0.220–5.580) 0.944 (0.801–1.113) 1.029 (0.679–1.561) 0.840 (0.721–0.978)* 0.749 (0.551–1.019)
(yes vs. no)

Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020

Diabetic mellitus 2.323 (1.715–3.145)* 4.074 (1.058–15.679)* 2.208 (1.903–2.562)* 2.227 (1.611–3.079)* 1.508 (1.318–1.726)* 1.226 (1.003–1.498)*
(yes vs. no)
Congestive heart 1.219 (0.722–2.059) 4.106 (1.630–10.343)* 1.470 (1.205–1.794)* 1.365 (1.078–1.727)* 1.364 (1.128–1.651)* 0.966 (0.831–1.123)*
failure (yes vs. no)
Cerebrovascular 1.651 (0.874–3.120) 0.649 (0.207–2.032) 1.503 (1.226–1.843)* 1.375 (1.031–1.833)* 1.187 (0.979–1.440) 1.387 (1.097–1.752)*
disease (yes vs. no)
Peripheral vascular 22.091 (5.268–92.641)* – 1.030 (0.580–1.828) 1.677 (0.859–3.273) 2.785 (1.316–5.897)* 1.966 (1.041–3.712)*
disease (yes vs. no)
Chronic lung 1.330 (0.576–3.072) – 1.309 (1.005–1.703)* 0.724 (0.501–1.046) 1.548 (1.270–1.887)* 1.216 (0.947–1.562)
disease (yes vs. no)
Chronic liver 1.571 (0.967–2.551) 3.490 (0.544–22.406) 1.729 (1.399–2.137)* 0.950 (0.654–1.380) 1.237 (0.986–1.552) 1.000 (0.708–1.413)
disease (yes vs. no)
Cancer (yes vs. no) 3.515 (1.583–8.033)* – 1.512 (1.094–2.090)* 1.889 (1.027–3.474)* 0.986 (0.732–1.329) 1.060 (0.643–1.747)
CAD, coronary artery disease; n, the number of patients in each subgroup.
aThe analyses were adjusted for gender, hypertension, congestive heart failure, coronary artery disease, cerebrovascular disease, peripheral vascular disease, chronic lung disease

and cancer. All data are hazard ratio (95% confidence interval).
*P < 0.05.

and cardiovascular diseases, including CHF, CAD, cerebrov-
ascular accident, and peripheral vascular disease. We found a
tendency of reverse epidemiology between baseline HTN and
subsequent lower mortality; however, it was significant in HD
patients ≥65 years old without DM. There was a clear tendency
for a reverse association with increasing age in patients with
CHF or CAD.
We, like Kimura et al.,17 found that baseline normal BP was
associated with poor outcomes in nonDM HD patients ≥65
years old. Kimura et al. also reported that HTN had no signifi-
cant effect on elderly patients. Elderly dialysis patients have a
much higher mortality rate.13 In the presence of strong factors
other than BP that affect mortality, the effect of HTN might nei-
ther be discovered nor be as important. We also found there was
a clear tendency for a reverse association with increasing age
in patients with CHF or CAD. It is also possible that a higher
BP could be “protective” in the elderly, especially with CHF or
CAD, by improving cardiac perfusion or overcoming the resist-
ance of stiff arteries.9 In addition, the normal-to-low BP may
be a marker of the already poor survival rate of elderly dialysis
patients,18,19 especially with CHF or CAD. Older patients with
no HTN might have latent heart failure and thus be more likely
to die.17 The association of normal-to-low BP with mortality
was a marker for having had cardiac failure before death.6,19 A
recent study20 also showed that the association between low BP
and death was especially evident in the short term. To evaluate
the hypothesis that elderly dialysis patients without DM, but with
one or more other serious conditions, have a poor survival rate,
we analyzed mortality, but excluded death in the initial stage,
during follow-up. Patients who died in the initial stage (within
90 days after initiation of HD) were excluded from our study; the
HR was 0.769 (95% CI: 0.637–0.927) (Table 3). We performed
an additional analysis of mortality excluding death within 1 year
after the initiation of HD. A multivariate analysis showed that
baseline HTN was also an independent predictor for lower mor-
tality only in nonDM HD patients ≥65 years old (HR: 0.784, 95%
CI: 0.642–0.956). The HR increased from 0.769 to 0.784, which
indicated that serious health conditions may be partially respon-
sible for mortality in elderly patients; other reasons for mortality
are yet to be determined. Thus, the reason for an inverse associa-
tion between baseline HTN and mortality may be that a higher
BP “protective” in the elderly; in contrast, normal-to-low BP may
reflect the fact that organ failure or other events leading to hypo-
tension had occurred before death.
Some reports7 refer to the tendency of an inverse associa-
tion between baseline HTN and subsequent mortality in dial-
ysis patients as “reverse epidemiology”, a term first used in
1999 by Josef Coresh (cited in Levin et al. 2007)12. However,
the term appears to have generated confusion and inaccurate
usages over the past decade. Levin et al.12 clarified that the
epidemiology is not reversed. Most ESRD dialysis patients
have subclinical manifestations of the disease process and
often they and their physicians are unaware, until very late in
the course of the disease, of how their disease and treatment
are affecting them. There appears to be a counterintuitive
association between baseline BP and subsequent mortality
partly because of unmeasured baseline disease manifesta-
tions of risk factors and partly because of the methods used

AMERICAN JOURNAL OF HYPERTENSION | VOLUME 25 NUMBER 8 | august 2012 905

original contributions
to analyze disease progression and treatment efficacy7; there
may be other reasons.
There are several limitations to this study. First, although we
stratified our patients by age and DM status, like other stud-
ies that have used administrative data, we did not control for
some unmeasured confounding variables: antihypertensive
treatment, BP levels in hypertensive patients, etc. Whether this
is based on latent confounding or whether there is a biologi-
cal basis remains to be determined. Second, the comorbidi-
ties relied on the claim data and ICD-9-CM diagnosis codes,
but it is possible that some of the diseases were misclassified.
Third, we were unable to take into account the severity of the
diseases, which reduced our chances of showing the severity-
related effects of comorbidities. Fourth, our study lacked spe-
cific data on dialysis adequacy, patient compliance, nutritional
status, biochemical data, and socioeconomic characteristics.
Finally, it would be better to describe the causes of death and
analyze the mortalities of cardiovascular and noncardiovascu-
lar diseases; however, the Taiwan Bureau of National Health
Insurance does not afford the cross-link information between
this and the database with “causes of death”.
In conclusion, maintenance of HD patients have a high
prevalence of HTN. Using the Taiwan NHI claim data, we
found a counterintuitive reverse association between base-
line HTN and mortality in nonDM chronic HD patients
≥65 years old. There was a clear tendency for a reverse asso-
ciation with increasing age in patients with CHF or CAD.
Additional studies of this counterintuitive reverse associa-
tion between HTN and mortality in elderly HD patients may
be warranted.
Appendix 1 | ICD-9-CM codes used to identify clinical
conditions
Conditions ICD-9-CM code
Diabetes mellitus 250.a, 357.2, 362.0a, 366.41
Hypertension 362.11, 401.a–405.a, 437.2
Congestive heart failure 428, 398.91, 422, 425, 402.a1, 404.a1, 404.a3
Coronary artery disease 410.a–414.a
Cerebrovascular accident/TIA 430–438
Peripheral vascular disease 440–444, 447, 451–453, 557
Chronic lung disease 490–496a, 500-505a, 506.4a
Chronic liver disease 570, 571, 572.1, 572.4, 573.1–573.3
Cancer 140–208; 230–231; 233–234
ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification;
TIA, Transient ischemic attack.
aCan be any number or missing.
906 august 2012 | VOLUME 25 NUMBER 8 | AMERICAN JOURNAL OF HYPERTENSION
Hypertension-Mortality Associations in HD Patients
Disclosure: The authors declared no conflicts of interest.
Acknowledgment: The study was supported by grant CMFHR10078 from
Chi-Mei Medical Center and grant NHRI-NHIRD-99182 from the National
Health Research Institutes in Taiwan. We are grateful to the staff in Professor
Jhi-Joung Wang’s Department of Medical Research, especially Chin-Li Lu and
Shih-Feng Weng, who helped with statistical analyses.
1. Burt VL, Whelton P, Roccella EJ, Brown C, Cutler JA, Higgins M, Horan MJ,
Labarthe D. Prevalence of hypertension in the US adult population. Results
from the Third National Health and Nutrition Examination Survey, 1988-1991.
Hypertension 1995; 25:305–313.
2. VanDeVoorde RG, Mitsnefes MM. Hypertension and CKD. Adv Chronic Kidney Dis
Downloaded from https://academic.oup.com/ajh/article/25/8/900/2282145 by Joongbu University user on 13 August 2020
2011; 18:355–361.
3. United States Renal Data System <http://www.usrds.org>. Accessed 25 October
2011.
4. Anavekar NS, McMurray JJ, Velazquez EJ, Solomon SD, Kober L, Rouleau JL,
White HD, Nordlander R, Maggioni A, Dickstein K, Zelenkofske S, Leimberger JD,
Califf RM, Pfeffer MA. Relation between renal dysfunction and cardiovascular
outcomes after myocardial infarction. N Engl J Med 2004; 351:1285–1295.
5. Fernández JM, Carbonell ME, Mazzuchi N, Petruccelli D. Simultaneous analysis of
morbidity and mortality factors in chronic hemodialysis patients. Kidney Int 1992;
41:1029–1034.
6. Foley RN, Parfrey PS, Harnett JD, Kent GM, Murray DC, Barre PE. Impact of
hypertension on cardiomyopathy, morbidity and mortality in end-stage renal
disease. Kidney Int 1996; 49:1379–1385.
7. Agarwal R. Hypertension and survival in chronic hemodialysis patients–past
lessons and future opportunities. Kidney Int 2005; 67:1–13.
8. Zager PG, Nikolic J, Brown RH, Campbell MA, Hunt WC, Peterson D, Van Stone J,
Levey A, Meyer KB, Klag MJ, Johnson HK, Clark E, Sadler JH, Teredesai P. “U” curve
association of blood pressure and mortality in hemodialysis patients. Medical
Directors of Dialysis Clinic, Inc. Kidney Int 1998; 54:561–569.
9. Kalantar-Zadeh K, Kilpatrick RD, McAllister CJ, Greenland S, Kopple JD. Reverse
epidemiology of hypertension and cardiovascular death in the hemodialysis
population: the 58th annual fall conference and scientific sessions. Hypertension
2005; 45:811–817.
10. Li Z, Lacson E Jr, Lowrie EG, Ofsthun NJ, Kuhlmann MK, Lazarus JM, Levin NW. The
epidemiology of systolic blood pressure and death risk in hemodialysis patients.
Am J Kidney Dis 2006; 48:606–615.
11. Tomita J, Kimura G, Inoue T, Inenaga T, Sanai T, Kawano Y, Nakamura S, Baba S,
Matsuoka H, Omae T. Role of systolic blood pressure in determining prognosis of
hemodialyzed patients. Am J Kidney Dis 1995; 25:405–412.
12. Levin NW, Handelman GJ, Coresh J, Port FK, Kaysen GA. Reverse epidemiology: a
confusing, confounding, and inaccurate term. Semin Dial 2007; 20:586–592.
13. Vonesh EF, Snyder JJ, Foley RN, Collins AJ. The differential impact of risk factors on
mortality in hemodialysis and peritoneal dialysis. Kidney Int 2004; 66:2389–2401.
14. Lacson E Jr, Lazarus JM. The association between blood pressure and mortality
in ESRD-not different from the general population? Semin Dial 2007; 20:510–517.
15. Bureau of National Health Insurance <www.doh.gov.tw/statistic/index.htm>
[In Chinese]; <http://www.doh.gov.tw/EN2006/index_EN.aspx> [In English].
Accessed 25 September 2011.
16. United States Renal Data System. Treatment history (chapter V). In: Researcher’s
Guide to the USRDS Database 1999. <http://www.usrds.org>. Accessed 25
October 2011.
17. Kimura G, Tomita J, Nakamura S, Uzu T, Inenaga T. Interaction between
hypertension and other cardiovascular risk factors in survival of hemodialyzed
patients. Am J Hypertens 1996; 9:1006–1012.
18. Loyke HF. Disease states in which blood pressure is lowered. South Med J 1989;
82:864–867.
19. Covic A, Goldsmith DJ. Ambulatory blood pressure measurement in the renal
patient. Curr Hypertens Rep 2002; 4:369–376.
20. Bos WJ, van Manen JG, Noordzij M, Boeschoten EW, Krediet RT, Dekker FW. Is the
inverse relation between blood pressure and mortality normalized in ‘low-risk’
dialysis patients? J Hypertens 2010; 28:439–445.