Atherosclerosis, 87 (1991) 119-128 119

,c 1991 Elsevier Scientific Publishers Ireland, Ltd. 0021-9150/91/$03.50

ADONIS 002191509100090K

ATHERO 04610

The ratio of ankle and arm arterial pressure as an independent

predictor of mortality

Matthew McKenna I3* , Sidney Wolfson * and Lewis Kuller ’

‘Department of Epidemiology, University of Pittsburgh, Graduate School of Public Health, Pittsburgh, PA (U.S.A.),
and _ Department of Neurological Surgery, University of Pittsburgh School of Medicine, Peripheral Vascular Diagnostic Laboratory,
Montefiore Hospital, Pittsburgh, PA (U.S.A.)

(Received 6 July, 1990)

(Revised, received 2 October and 16 November, 1990)
(Accepted 20 November, 1990)

Summary

We carefully ascertained deaths for a cohort of 744 patients who had undergone noninvasive testing for
lower extremity peripheral arterial disease (PAD) in a university affiliated, community hospital. Using a
ratio of the ankle and brachial blood pressures (ABI) of less than 0.85 as the criteria, the relative risk (RR)
for total mortality associated with PAD was 2.36 (95% CL = 1.60, 3.48) after adjusting for baseline
covariates in a proportional hazards model. There was a strong trend for increasing risk with decreasing
ABI (P < 0.0001). Specific causes of death for which survival was directly related to the magnitude of ABI
were myocardial infarctions, and deaths other than vascular disease and cancer. There was no relationship
between ABI and the risk of mortality from cancer. The mortality experience of those with normal ABI
was very similar to that of the general U.S. population (age adjusted RR = 1.14, 95% CL = 0.78, 1.61)
whereas the risk for those with an ABI less than 0.4 was markedly elevated in comparison to the U.S.
population (RR = 4.49, 95% CL = 3.52, 5.64). Our results suggest that the relatively technically simple
measure of the ratio of ankle to brachial blood pressures, if low, carries a very poor prognosis and should
prompt investigation and treatment of atherosclerotic disease in other vascular systems.

Key words: Vascular diseases; Arterial pressure; Prognosis; Adult

Introduction

“I say, that this (angina pectoris) exactly corresponds to the

sense of weakness and want of muscular power which exists in
persons who have the arteries of the legs obstructed or ossified.”
Correspondence and present address: Dr. Matthew MC- _ These words spoken by Brodie over 140 years
Kenna, MD, Division of Chronic Disease Control and Com-
munity Intervention, Centers for Disease Control, MS K52,
ago demonstrate that clinicians have long recog-
1600 Clifton Rd., Atlanta, GA 30333. U.S.A. Phone: (404) nized lower extremity peripheral arterial disease
458-5485. (PAD) as a sentinel for potentially lethal coronary
120

atherosclerosis [l]. In order to document this fact has been performing lower extremity vascular
many epidemiologic studies have used the clinical studies to detect PAD in referred inpatients and
criteria developed by Geoffrey Rose [2] to di- outpatients. Prior to testing, each patient has un-
agnose PAD. These investigations suggest a strong dergone a detailed general medical history, includ-
univariate association between PAD, stroke, and ing questions concerning cardiovascular risk fac-
diseases of coronary insufficiency such as myocar- tors, performed by a trained technologist. Infor-
dial infarction, angina, and congestive heart failure mation is based on the patient’s report, and any
[3-71. However, after adjusting for the signs and available medical records. All data have been re-
symptoms of coronary artery atherosclerosis with corded on a standardized form. In 1985 this proc-
multivariate statistical methods, clinically di- ess was automated with the entry of all historical
agnosed PAD had no independent effect in one and test result information directly into a com-
study [3]. The inability to demonstrate an inde- puterized database. For this study we entered into
pendent predictive effect may be due to the mis- the database the microfiched records of the first
classification of subjects who have leg symptoms lower extremity arterial study of patients seen
from nonvascular causes as victims of PAD, since prior to 1985 in a sequential, alphabetical order.
the lack of precision associated with the clinical The subjects of this study are the first 1038 pa-
history and exam in diagnosing PAD has been tients entered who were less than 90 years of age
well documented [8,9]. at the time of testing for PAD.
There have been a limited number of prospec- One-hundred and fifty-six subjects (15.0%) were
tive studies using noninvasive methods of diagno- excluded from this analysis because of missing
sis to assess the risk of mortality and morbidity data, or because they were not residents of Penn-
associated with PAD. Two European studies found sylvania. Another 131 (12.6%) patients, who had
elevated relative risks, but the significance of PAD undergone a previous vascular surgical procedure
as a measure of atherosclerotic risk, independent of the aorto-iliac vessels, were removed from the
of other factors, was not evaluated [lO,ll]. In a analysis. Finally, 7 ( < 1%) patients with results
study of patients with claudication, a direct and which were considered indicative of definite
independent relationship between the severity of calcific arteriosclerosis were excluded. All seven of
PAD and increasing risk for cardiovascular death these patients were diabetics. Seven-hundred and
was found, and this hazard was independent of forty-four people ranging in age from 19 to 89
other antecedent cardiovascular conditions or risk years of age with an average of 66 years remained.
factors [12]. No comparison with an asymptomatic
control group was made in this study. An analysis Diagnostic methods
of a sample from the Lipid Research Clinics cohort PAD was evaluated using resting ratios of ankle
did include multivartiate methods, but had too to brachial systolic blood pressure, which will be
few deaths to adequately evaluate cause-specific referred to as the ankle/ bra&al index (ABI).
mortality risks [13]. This is a standard method for diagnosing occlu-
In order to evaluate the relationship of PAD to sion or stenosis of the proximal arteries of the leg
total mortality, and its specific relationship to the [8,14], and was the measure recommended for
risk of cardiovascular death, we conducted a epidemiologic studies of peripheral arterial disease
mortality study of a cohort of patients evaluated by a concensus panel of experts [15]. While in a
over a lo-year period at a peripheral vascular supine position the systolic blood pressure was
laboratory specializing in noninvasive methods. determined in the ankle by Doppler readings from
We also compared the mortality experience of this the right and left posterior and anterior tibia1
cohort with that of the general United States arteries, and then in the bra&al artery. For each
population during the same period. leg the higher of the readings from either anterior
or posterior tibia1 artery was used as the value for
Subjects and methods
that ankle. The ABI was calculated using the
Since 1977 the peripheral vascular laboratory of lower of the left or right ankle readings, and the
Montefiore Hospital in Pittsburgh, Pennsylvania, higher of the left or right bra&al measures. In
 121

patients who had undergone an amputation of a difference between two survival curves determined
lower extremity above the ankle, the value from by the logrank test [18]. Adjustment for baseline
the unamputated leg was used. An ABI of I 0.85 characteristics was performed with a Cox propor-
was considered indicative of PAD. Patients with tional hazards model [19]. Age and systolic blood
ABI of greater than 1.5 were considered to have pressure were entered as continuous variables. All
calcific arteriosclerosis which results in spuriously baseline variables were included in the model.
high ABI readings, and were excluded from the In order to evaluate the specificity of ABI as a
analysis. predictor of total, and cause-specific mortality, we
categorized this variable into three levels adapted
Ascertainment of deaths from the schema described by Bernstein and
All deaths were ascertained through the State Fronek [14]. To test for linear trend the ABI value
of Pennsylvania index of deaths for all residents. was entered as a continous variable [20]. To
A printed copy of this index, sorted by year of evaluate the ABI category-specific, adjusted rela-
death, county of residence and name, was manu- tive risk the categories of ABI were entered into
ally checked against a listing of patients in the the model as “dummy” variables.
cohort without any knowledge of their PAD status. In comparing the mortality rates of the cohort
The death certificate for each potential match of to the general population, data were derived from
names was then reviewed to assure the accuracy of the 1984 U.S. population [21]. Death rates were
identification according to birthdate, social secur- determined for 10 year intervals and multiplied by
ity number and county of residence. The underly- the same age specific person years in the study
ing cause of death was then coded by one of the population to calculate an expected number of
authors (M.T.M.) according to ICD-9 rules for deaths. The indirect method of age-adjustment
each subject [16]. The accuracy of death certifi- using the standardized mortality ratio (SMR) as
cates from this community for the cause of death the estimate of the relative risk was used in this
has been validated in other studies [17]. Follow-up analysis.
was completed through 1987. The median time of P-values for the differences in baseline char-
follow-up was 3.28 years. acteristics between subjects with PAD and those
Baseline characteristics without PAD were calculated using a two-tailed
All information concerning baseline variables Fisher’s exact test or &i-square statistic with a
were derived from recorded interviews of the pa- continuity correction where appropriate [20], and
tients or from medical records if they were avail- the two-tailed t-test for continuous variables. P-
able at the time of the noninvasive ABI measure- values less than 0.05 were considered statistically
ment. Patients with a clinical history of angina or significant.
coronary artery heart disease were defined as
“angina only”, whereas patients with a history of Results
myocardial infarction, coronary artery bypass
grafting, percutaneous transluminal coronary an- Table 1 shows the baseline characteristics of
gioplasty, or an angiogram diagnostic of coronary patients with, and without, PAD. The PAD sub-
atherosclerosis were defined as “definite ischemic jects were older, more likely to have a history of
heart disease.” Transient ischemic attacks were smoking, and have hypertension. In addition they
defined by the Rose criteria [2]. were more likely to have a history of definite
Patients who had undergone amputations of ischemic heart disease and congestive heart failure.
the lower extremity below the ankle joint were There was also a slight tendency for non-whites to
categorized as minor amputees. More proximal experience PAD more than whites.
amputations were classified as major amputations. The survival experience for people without PAD
was substantially better than those with PAD (P
Statistical analysis < 0.001, Fig. 1). The 5 and 10 year survival expe-
Lifetable analysis of survival was done using rience associated with all of the baseline character-
the actuarial method and the probability for the istics, and PAD were calculated and are presented
 than 50 years), congestive heart failure (difference
= 0.42), and stroke (difference = 0.31) were asso-
ciated with larger values than the difference in
Without PAD
survival probabilities for those with PAD versus
%_
--.._-...__ those without PAD (difference = 0.27).
-I____
----..__ After adjustment with the proportional hazards
----------.- .______________
-.. model factors other than age found to be indepen-
With PAD dently associated with mortality were PAD, con-
1-L""' /
0 12 3 4 5 6 7 6 9 (0 gestive heart failure, stroke and diabetes mellitus
YEAR
(Table 3). Of these covariates, only congestive
Fig. 1. Survival curves for patients with PAD, and without
heart failure had a relative risk larger than that of
PAD.
PAD.
Because of the concern that diabetes could cause
in Table 2. In addition to PAD clinical manifesta- substantial misclassification of PAD status be-
tions of cardiovascular disease such as CHF and cause of the poorly compressible arteries which
stroke (P -c0.001) exhibited stronger effects on can result from diabetically induced calcific
mortality than classic risk factors such as male sex arteriosclerosis, and hence result in spuriously high
(P = 0.95), systolic blood pressure (P = 0.20), or a ABI readings, mortality rates were calculated for
history of hypertension (P = 0.17). When ex- diabetics and non-diabetics by age group and PAD
pressed as the difference in the 5 year probability status. The death rates for those with PAD was
of survival only age (difference of 0.32 between consistently higher regardless of the presence or
those greater than 70 years of age, and those less absence of diabetes (Fig. 2). The age adjusted

TABLE 1

BASELINE CHARACTERISTICS OF PATIENTS WITH AND WITHOUT PERIPHERAL ARTERIAL DISEASE (PAD)

Characteristic With PAD Without PAD P-value

(n=482) (n = 262)

Age (yrs) a 68*10 63*14 < 0.00001

Men 46% 46% 0.94
Race (white) 70% 77% 0.03
Ever smoked 67% 56% 0.004
Systolic BP (mm Hg) a 151 f 28 140+23 < 0.00001
Diabetes mehitus 34% 28% 0.12
Hypertension 53% 39% 0.0003
Congestive heart failure 10% 6% 0.03
Non-ischemic heart disease 8% 6% 0.47
Major amputation b 4% 1% 0.03
Minor amputation ’ 2% 1% 0.21
Ischemic heart disease 40% 29% 0.003
Definite disease d 24% 16% 0.01
Angina pectoris only 16% 13% 0.24
Cerebrovascular disease 15% 13% 0.38
Stroke 10% 6% 0.10
TIA 4% 6% 0.24
Carotid surgery 3% 2% 0.43

’ Mean + SD.
b Major amputations = at any joint proximal to, and including, the ankle.
c Minor amputations = below the ankle.
d Definite ischemic heart disease = history of a myocardial infarction, coronary artery bypass grafting, or positive angiogram.
TIA = Transient &heroic attacks.
 123

relative risk of death for PAD in diabetics was ABI with cutpoints at 0.4 and 0.85. There was a
slightly less than that in non-diabetics, namely definite and powerful relationship between
2.33 versus 2.84. However, this difference was far survival, and the severity of arterial obstruction
from statistically significant with a P-value for (Table 4 and Fig. 3). The 5 year actuarial survival
heterogeneity of the relative risks of 0.62. for a subject with an ABI < 0.4 was only 44%.
In order to evaluate the risk of death associated versus 90% for those with an ABI > 0.85. This
with the severity of arterial disease we calculated trend remained significant even after adjustment
survival curves for patients at 3 separate levels of in the Cox model (P -c0.001)and the relative

TABLE 2

LIFE TABLE SURVIVAL ESTIMATES BY BASELINE CHARACTERISTICS

Variable Number of 5 Year survival 10 year survival P

deaths estimate estimate value

< 50 yrs. 6 0.91 0.85 Reference

50-60 yrs. 88 0.78 0.59 i 0.05
70+ 121 0.59 0.45 i 0.01
Ischemic heart disease
None 121 0.75 0.60 Reference
Angina only 36 0.67 0.53 0.06
Definite disease 58 0.62 0.44 < 0.001
Non-ischemic heart + 20 0.59 0.32
disease _ 195 0.72 0.57 < 0.01
Congestive heart + 39 0.33 0.27
failure _ 176 0.75 0.58 < 0.001
Peripheral + 183 0.63 0.46
artery disease _ 32 0.90 0.77 i 0.001
Cerebrovascular disease
None 178 0.73 0.58 Reference
TIA 7 0.79 0.62 0.43
Stroke 30 0.44 0.27 < 0.001
Diabetes mellitus + 90 0.59 0.44
_ 125 0.77 0.61 i 0.001
Ever smoked + 123 0.66 0.58
- 88 0.74 0.50
Lower extremity amputation
None 198 0.72 0.57 Reference
Minor amputation 6 0.53 0.32 0.32
Major amputation 11 0.51 0.33 < 0.05
Carotid surgery + 7 0.64 0.54
_ 208 0.71 0.56 0.67
Systolic BP (mm Hg)
> 140 + 140 0.68 0.55
140 or less _ 75 0.77 0.60 0.20
History of + 110 0.68 0.51
hypertension - 105 0.74 0.59 0.17
Male 99 0.70 0.53
Female 116 0.72 0.58 0.95
White race 153 0.72 0.55
Other race 62 0.70 0.55 0.98

TIA = transient ischemic attacks.

124

TABLE 3 TABLE 4

ADJUSTED RELATIVE RISKS, 95% CONFIDENCE CAUSE SPECIFIC LIFE TABLE SURVIVAL ESTIMATES
LIMITS AND P VALUES FOR TOTAL MORTALITY FOR BY LEVEL OF ABI, AND ADJUSTED TEST FOR TREND
THOSE VARIABLES WHICH REMAIN SIGNIFICANT IN RESULTS
THE PROPORTIONAL HAZARDS MODEL
Adjusted with proportional hazards model (see Methods).
Adjusted for all baseline variables (see Methods).
Cause of Death 10 year survival estimates
Variables Relative risk 95% confidence limits P value (no. of deaths)

Diabetes 1.64 (1.22, 2.20) 0.001 Ankle/brachial index Trend

Stroke 1.88 (1.26, 2.82) 0.002 P
< 0.4 0.4-0.85 > 0.85
CHF 2.38 (1.62, 3.47) < 0.001
PAD 2.36 (1.60, 3.48) < 0.001 Ischemic heart disease 0.53 0.72 0.86 < 0.01
(ICD-9, 410-415) (39) (47) (15)
Other cardiovascular 0.81 0.88 0.96 0.72
disease (17) (25) (9)
(ICD-9, 390-409.99,
risks for death were significantly higher than 1.00 416-448)
for ABI < 0.4 and ABI values between 0.4 and Total cardiovascular 0.43 0.63 0.83 c: 0.05
0.85 (see Table 5). disease (56) (72) (24)
The causes of death were grouped into 4 major Cancer 0.98 0.89 0.97 0.13
(ICD-9 140-239) (2) (18) (3)
categories: ischemic heart disease (ICD-9 codes
Other 0.79 0.90 0.97 c 0.001
410-415), cardiovascular diseases other than
(15) (20) (5)
ischemic heart disease (390-409.99, 416-448)
cancer (140-239), and other causes. The majority Total mortality 0.33 0.51 0.77 < 0.001
(73) (110) (32)
(71.0%) of the 215 deaths in this cohort were
attributable to cardiovascular diseases. The associ-
ation of ABI with specific causes of mortality was TABLE 5

evaluated by estimating the 10 year actuarial ADJUSTED RELATIVE RISKS FOR MORTALITY, WITH
survival at different levels of ABI for the four 95% CONFIDENCE LIMITS AND P VALUES, FOR
LEVELS OF ABI
categories of mortality. As shown in Table 4 there
was a definite trend in increasing risk of death Adjusted for all baseline variables (see Methods).

from cardiovascular diseases, particularly ischemic

Categories Relative 95% confidence P value
heart disease, with decreasing levels of ABI even of ABI risk limits
after adjustment in the Cox model (P < 0.05).
ABI < 0.4 3.35 (2.16, 5.20) < 0.001
ABI 0.4-0.85 2.02 (1.34, 3.02) < 0.001
ABI > 0.85 1 .oo Reference
vi ??No Diabetes dr No PAD ? ?Diabetes dr No PAD
$
x
250
• gi No Dlobetea & PAD 69 Diobetss h PAD

20 ' I I I I I
0 2 4 6 0 IO
Agegroups
Fig. 2. Mortality rates for patients with PAD and without YEAR
PAD, by age group and diabetes status. Fig. 3. Survival curves for 3 levels of ABI measures.
 125

I
larly strong, linear, and inverse relationship with
ABI < 0.4
-;
1’
the risk of death from myocardial infarction. These
1 results agree with the findings of other studies
.~ which have used univariate methods of analysis,
and two other studies which used mulltivariate
j \.,, , ,/ /$;;;;;
methods.
1 ,/’
/’
Kallero studied 312 symptomatic men and wo-
__.--’
. ..___ ____..-.---
-...__ _ iBI > 0.85 men who underwent non-invasive testing and
-~ .._.__......_ ,.,..<..I. .._
~--%_____ ______.~~~~: ,. ..... .. .. -.. _ _ _
~~.~~~~~.~~~~~~~_-------
/ found a 5 year mortality 2.5 times higher in those
50 Yrs. 50-59 Yrs. 60-69 Yrs. 70-79 Yrs. 80-89 YrC
with PAD than those without [ll]. This excess was
Agegroups primarily related to deaths from myocardial in-
Fig. 4. Mortality rates by age groups in patients with 3 differ- farctions. He did not adjust for other covariates in
ent levels of ABI, and the 1984 U.S. population. The respective his analysis, nor did he find any correlation be-
standard mortality ratios (SMR) and 95% confidence limits for
tween the degree of reduction of blood flow and
each level of ABI to the U.S. population are 4.49 (3.52, 5.64)
with an ABI of i 0.4, 2.51 (2.06, 3.02) with an ABI between
mortality rate. DaSilva and Widmer [lo] found a
0.4 and 0.85, and 1.14 (0.78, 1.61) for those with ABI P- 0.85. similar relative risk in a longitudinal study of 2630
men who were diagnosed with PAD by oscillogra-
phy. However, they did not evaluate the indepen-
Though the survival probabilities for cardiovascu- dence of this diagnosis as a predictor for mortal-
lar diseases other than ischemic heart diseases ity.
suggest a trend, the P-value was not significant Jonason and Ringqvist did analyze data from a
after adjustment in the Cox model (P = 0.72). follow-up study of 224 patients with PAD using
There was no suggestion of a consistent trend for multivariate methods and found a direct and inde-
cancer mortality with the lowest survival probabil- pendent relationship between the degree of arterial
ity being in the intermediate category of ABI. obstruction, and risk for death from myocardial
To estimate the generalizability of our cohort infarction, but not cerebrovascular disease [12].
findings we compared mortality rates for our No comparison with a control population was
cohort with those for the U.S. population for 10 attempted in this study.
year age groups, starting at age 50 years, for the 3 Using a cohort of 567 hyperlipidemic men and
levels of ABI and the 1984 United States popula- women evaluated by noninvasive methods, Criqui
tion (see Fig. 4). The mortality rates for subjects et al. calculated a relative risk of 4.51 for total
with low ABI were generally much higher in com- mortality associated with PAD [13]. Though they
parison to the U.S. population, whereas the group found a large cause-specific relative risk for
without PAD experienced rates very similar to the cardiovascular diseases of 11.8. the numbers of
U.S. population. The age adjusted relative risk for death were too few to assert the statistical signifi-
those with an ABI < 0.4 in comparison to the U.S. cance of this result.
populations was 4.49 with 95% confidence limits Our results suggest that PAD is also associated
of (3.52, 5.64). However, the cohort free of PAD with deaths from causes of death other than car-
had a relative risk of mortality of only 1.14, and diovascular disease. This is not surprising since
95% confidence limits which overlapped with unity other studies have documented the association of
(0.78. 1.61). This suggests that our results may PAD with atherosclerotic pathology in the
have applicability to other populations. coronary and cerebrovascular vessels [23,24]. It is
likely that generalized atherosclerosis would hasten
the demise of a patient with any number of un-
Discussion derlying maladies. This may explain the very
strong relationship with the category of mortality
This study supports the independence of PAD, other than cardiovascular or cancer. There was no
as measured by the ankle/ bra&al index, as a suggestion of a trend between ABI and cancer
predictor of mortality. A low ABI has a particu- mortality risk.
126

Using a criteria of an ABI less than 0.85 for ily than are those without diabetes. This results in
PAD we found a relative risk for total mortality the over-representation of diabetics in the group
between 2 and 3 for patients with PAD versus without PAD. Another possible explanation is that
those without. This value is in general agreement diabetics frequently have calcified vessels which
with some studies [11,12], but is somewhat less can result in spuriously elevated blood pressure
than the estimate found by Criqui et al. [13] of readings, and hence a falsely high ABI. In fact all
4-5. This discrepancy from the Criqui et al. may 7 patients with unnacceptably high ABI were di-
be due to the remarkably good survival experience abetic. However, the decreased survival of patients
of the subjects without PAD in that study. Apply- with decreased ABI, regardless of their diabetes
ing life-table estimates from the general U.S. status, suggests that ABI has a definite role in
population [25] to the demographic information in evaluating diabetics for PAD.
that study one would have expected about 10% of We have no information about serum lipids in
the people without PAD to have died, whereas the these subjects. Elevated cholesterol is a powerful
proportion who died during follow-up was only risk factor for ischemic heart disease mortality
2.7%. [35-371. However, in order for lipid level to con-
Our data indicate that the subjects in this study found the significant relationship between mortal-
without PAD had a mortality rate more similar to ity and PAD found in this study, hyperlipidemia
that of the general U.S. population. This situation would have to be related to both PAD and mortal-
existed even though the study involves a clinical ity. The relationship of serum triglycerides and
population consisting of individuals referred for cholesterol to PAD is not completely clear. Some
arterial examination, many of whom had co- studies have found an independent association
morbid clinical diseases and risk factors. Prospec- between lipids and PAD [23,27,34-361, whereas
tive studies with more rigorous measurements of others have not [2,26,37]. Additionally, this is a
baseline data, in particular covariate risk factors relatively elderly cohort with a mean age of 66
for coronary disease, are obviously needed. One years. The effect of lipids on mortality at this age
such investigation which recently began is the is probably not as strong as that found in younger
Atherosclerotic Risk in Communities (ARIC) age groups [38]. Therefore, it is unlikely that our
study. The investigators have included ABI as a results are the consequence of any confounding by
baseline measure in this large, prospective, multi- lipid level on the relationship between PAD and
center study [26]. mortality.
Since it has been demonstrated that PAD is Another weakness of this study is the use of
correlated with atherosclerotic ischemic heart dis- death certificates as the sole source for ascertain-
ease, it is not surprising that PAD has also been ment of vital status. Theoretically, deaths may be
strongly associated with several classic risk factors missed because of out-migration of people from
for coronary insufficiency, especially diabetes mel- the area of coverage, or matches may be missed
litus, cigarette smoking, and hypertension [10,27- because of inaccurate identifiers in the cohort or
311. Our analysis of baseline characteristics agrees death certificate records. The latter problem should
with these findings except that there was an equal be minimal in our study because a number of
proportion of diabetics in the groups with, and identifiers were screened including name, social
without, PAD. This is at odds with the clinical security number, and county of residence. We
impression, substantiated by most research, that then manually checked all possible matches with
diabetes mellitus is a very powerful risk factor for death certificate data using date of birth and
PAD [32-341. This is probably explained by the address to confirm the match.
tendency of patients with diabetes mellitus to ex- The possibility that deaths could have been
perience non-specific leg symptoms which may missed is more difficult. The cohort was limited to
arise from a number of causes including neuropa- residents of Pennsylvania. However, people still
thy. Since diabetes is widely recognized as a risk could have left the state, resulting in an under-
factor for PAD, diabetics may be referred for estimation of the mortality rates. In general this
evaluation of arterial obstruction much more read- dynamic should have little effect on the relative
 127

risk estimates since it is more likely that healthy References

individuals, who are the least likely to die, would
be the most likely to move. Therefore, we should 1 Brodie, B.C., Lectures on mortification delivered at the
be missing few deaths due to this effect. Addition- medical theatre of St. George’s Hospital, Lecture v. In:
ally, the age-specific death rates in our results Gazette S. (Ed.), New Series of London Med., 52 (1841)
714.
generally parallel the U.S. population, suggesting
2 Rose. G., Blackbum, H., Gillium, R.F. and Prineas R.J.,
no gross underestimation of rates. Cardiovascular survey methods. WHO Monogr. Ser., 56
The issue as to the strength and independence (1982). ed 2.
of PAD as a risk factor for mortality is an im- 3 Reunanen. A., Heikki, T. and Aromaa, A., Prevalence of
portant one since the danger of both PAD and intermittent claudication and its effect on mortality. Acta.
Med. Stand., 211 (1982) 249.
coronary artery disease rise as the number of
4 Boyd, A.M.. The natural course of Arteriosclerosis of the
atherosclerotic risk factors accumulate [39]. Yet, lower extremities. Proc. R. Sot. Med.. 55 (1962) 591.
even in those patients with the most lethal profile 5 Agner, E., Natural history of angina pectoris, possible
as assessed by traditional atherosclerotic risk fac- previous myocardial infarction and intermittent claudica-
tors, the absolute proportion who will imminently tion during the eighth decade. Acta Med. Stand.. 210
(1981) 271.
develop CAD is small [40-421. Theoretically, the
6 Gordon, S. and Kannel, W.B., Predisposition to atheroscle-
presence of atherosclerosis in arteries such as the rosis in the head, heart. and legs. J.A.M.A., 221 (1972) 661.
tibia1 vessels, which would be suggested by the 7 Hughson, W.G., Mann, J.I., Tibbs, D.J.. Woods, H.F. and
easily performed procedure of measuring a low Walton I.. Intermittent claudication: factors determining
ABI, could provide important clues as to how outcome. Br. Med. J., 1 (1978) 1377.
8 Marinelli, M.R., Beach, K.W., Glass. M.J.. Primozich, J.F.
aggressively and urgently other therapeutic modal-
and Strandness, D.E., Noninvasive testing vs clinical
ities should be pursued. Such treatments could evaluation of arterial disease. J.A.M.A.. 241 (1979) 2031.
involve pharmacological interventions on risk fac- 9 Criqui, M.H., Fronek, A., Klauber, M.R.. Barrett-Connor.
tors such as hyperlipidemia and hypertension, or E. and Gabriel, S. The sensitivity, specificity, and predictive
more intensive diagnostic procedures to evaluate value of traditional clinic evaluation of peripheral arterial
disease: results from noninvasive testing in a defined popu-
the necessity of surgical procedures such as
lation. Circulation, 71 (1985) 516.
coronary artery bypass grafts. The results of our 10 DaSilva, A. and Widmer. L.K., Occlusive Peripheral Artery
study suggest that a patient with a very low ABI. Disease: Early diagnosis, incidence. course. significance,
such as < 0.4, has a 5 year probability of survival Hans Huber Publishers, Bern, 1980.
of only 44%. This survival is lower than that seen 11 Kallero, K.S.. Mortality and morbidity in patients with
intermittent claudication as defined by venous occlusion
in colon cancer or non-Hodgkin’s lymphoma [43],
plethysmography. A ten-year follow-up study. J. Chron.
which are conditions where aggressive surgical and Dis., 34 (1981) 455.
systemic therapies are freguently recommended 12 Jonason, T. and Ringqvist. I., Mortality and morbidity in
[44-461. This suggests that a simple measure of patients with intermittent claudication in relation to the
blood pressures in the arm and ankle, with readily location of the occlusive atheroscerosis in the leg. Angiol-
ogy, 1 (1985) 310.
available Doppler probes, could provide inval-
13 Criqui, M.H.. Cot&in, S.S. and Fronek. A., Noninvasively
uable information concerning an individual’s over- diagnosed peripheral arterial disease as a predictor of
all cardiovascular status, and assist in the develop- mortality: results from a prospective study. Circulation 72
ment of a systemic approach to the treatment of (1985) 768.
atherosclerosis. 14 Bernstein, E.F. and Fronek, A., Current status of non-inva-
sive tests in the diagnosis of peripheral arterial disease.
Surg. Clin. North Am., 62 (1982) 473.
Acknowledgements 15 Prineas. R.J.. Harland, W.R., Janzon, L. and Kannel, W..
Recommendations for use of non-invasive methods to de-
Thanks must be given to Janet McDonald who tect atherosclerotic peripheral arterial disease - In popula-
provided a great deal of her time, computer exper- tion studies. Circulation, 65 (1982) S1156.
16 U.S. Department of Health and Human Services, The In-
tise. and moral support during the course of this
ternational Classification of Diseases. 9th revision, Clinical
study. This study was supported in part by grant Modification, DHHS Publ#-80-1260, 1980. 2nd ed.
HL-07011-13 from the National Institutes of 17 Kuller, L.H., Traven. N.D., Rutan. G.H.. Perper. J.A. and
Health, Public Health Service. Ives, D.G.. Marked decline of coronary heart disease
128

mortality in 35-44 year-old white men in Allegheny County, Vascular laboratory assessment. Mayo Clin. Proc., 56 (1981)
Pennsylvania. Circulation, 80 (1989) 261. 223.
18 Kahn, H.A. and Sempos, C., Statistical methods in epide- 33 Beach, K.W., Brunzell, J.D. and Strandness, D.E., Preva-
miology, Oxford University Press, Inc., New York, 1989, p. lence of severe arteriosclerosis obliterans in patients with
168. diabetes mellitus: relation to smoking and form of therapy.
19 Cox, D.R., Regression models and life tables (with discus- Arteriosclerosis, 2 (1982) 275.
sion). J. R. Stat. Sot., 34 (Series B) (1972) 187. 34 Beach, K.W., Bedford, G.R., Bergelin, R.O. et al., Progres-
20 Breslow, N.E. and Day, N.E., Statistical Methods in Cancer sion of lower-extremity arterial occlusive disease in type II
Research: Vol II, The design and analysis of cohort studies, diabetes mellitus. Diabetes Care, 11 (1988) 464.
IARC Scientific Publications, Lyon No. 82, 1987, p. 48. 35 U.S. Department of Health and Human Services, Report of
21 U.S. Department of Health and Human Services: National the expert panel on detection, evaluation and treatment of
Center for Health Statistics. Vital statistics of the United high blood cholesterol in adults. D.H.H.S., Publ. No. -89-
States 1984, Mortality, 2, 1984. 2925 (1989).
22 Breslow, N.E. and Day, N.E., Statistical Methods in Cancer 36 Stamler, J., Wentworth, D. and Neaton, J.D., Is the rela-
Research, Vol I, The design and analysis of case-control tionship between serum cholesterol and risk of premature
studies, IARC Scientific Publications, Lyon, No. 32, 1983, death from coronary heart disease continuous and graded?
p. 192. J.A.M.A., 256 (1986) 2823.
23 Hertzer, N.R., Young, J.R., Kramer, J.R. et al., Routine 37 Rose, G. and Shipley, M., Plasma cholesterol concentration
coronary angiography prior to elective aortic reconstruc- and death from coronary heart disease: ten-year results of
tion. Arch. Surg., 114 (1979) 1336. the Whitehall Study. Br. Med. J., 293 (1986) 306.
24 Sutton, K.C., Wolfson, S.K. and Kuller, L.H., Carotid and 38 Kuller, L.H., Are risk factors for CHD the same at differ-
lower extremity arterial disease in elderly adults with iso- ent ages. J. Clin. Epidemiol., 42 (1989) 91.
lated systolic hypertension. Stroke, 18 (1987) 817. 39 Janzon, L., Bergentz, S.E., Ericsson, B.F. and Lindell, S.E.,
25 U.S. Department of Health and Human Services, National The arm-ankle pressure gradient in relation to cardiovascu-
Center for Health Statistics. U.S. decennial life tables for lar risk factors in intermittent claudication. Circulation, 63
1979-1981.DHHS Pub1 No., (PHS) 85-1150-L 1985. (1981) 1339.
26 The ARIC Investigators, The atherosclerosis in communi- 40 Laupacis, A., Sackett, D.F. and Roberts, R.A., An assess-
ties (ARIC) study: Design and objectives. Am. J. Epide- ment of useful measures of the consequences of treatment.
miol., 129 (1989) 687. N. Engl. J. Med., 318 (1988) 1728.
27 Schroll, M. and Munck, O., Estimation of peripheral 41 Malenka, D.J. and Baron, J.A., Cholesterol and coronary
arteriosclerotic disease by ankle blood pressure measure- heart disease: the importance of patient-specific attributa-
ments in a population study of 60-year-old men and wo- ble risk. Arch. Intern. Med., 148 (1988) 2247.
men. J. Chron. Dis., 134 (1981) 261. 42 Rose, G., Strategy of prevention: lessons from cardiovascu-
28 Hughson, W.G., Mann, J.I. and Garrod, A., Intermittent lar disease. Br. Med. J., 282 (1981) 1847.
claudication: prevalence and risk factors. Br. Med. J., 1 43 U.S. Department of Health and Human Services, Cancer
(1978) 1379. Statistics Review, 1973-1906. D.H.H.S. Pub1 No. -89-2789
29 Gofin, R., Kark, J.D., Frielander, Y. et al., Peripheral (1989).
vascular disease in a middle-aged population sample. Isr. J. 44 Wolmark, N., Fisher, B., Rockette, H. et al., Postoperative
Med. Sci., 23 (1987) 155. adjuvant chemotherapy for colon cancer: results from
30 Criqui, M.H., Fronek, A., Barrett-Connor, E., Klauber, NSABP protocol C-01. J. Natl. Cancer Inst., 80 (1988) 30.
M.R., Gabriel, S. and Goodman, D., The prevalence of 45 Mayer, R.J., Does adjuvant therapy work in colon cancer?
peripheral arterial disease in a defined population. Circula- New Engl. J. Med., 322 (1990) 399.
tion, 71 (1985) 510. 46 DeVita, V.T., Jaffe, E.S. and Hellman, S., Hodgkins disease
31 Hill, D.A. and McGrath, M.A., The epidemiology of lower and the non-Hodgkins lymphomas. In: DeVita, V.T., Hell-
limb ischemia in a new south wales rural community. Aust. man, S., Rosenberg, S.A. (Eds.), Cancer: Principles and
N.Z. J. Surg., 51 (1981) 157. Practice of Oncology, 2nd edn., J.B. Lippincott Co., Phila-
32 Osmudson, P.J., Chesebro, J.H., O’Fallon, W.M., Zimmer- delphia PA, 1985, p. 1623.
man, B.R., Kazmier, F.J. and Palumbo, P.J., A prospective
study of peripheral occlusive arterial disease in diabetes. II.