Quantitation of Hypertensive Nephrosclerosis on an Objective
Rational Scale of Measure in Adults and Children
RICHARD E. TRACY, M.D., PH.D., DONALD E. MERCANTE, M.S.,
ARMANDO MONCADA, M.D., AND GERALD BERENSON, M.D.
The development of a precise, mathematical relationship between
blood pressure and renal microvascular abnormalities would be
highly desirable. Such a relationship would require that abnormalities be quantitative on a rational scale. The dominant abnormality in nephrosclerosis occurs in arcuate and cortical arteries of 50 to 400 Mm outer diameter. This abnormality consists
of acquired inner layers of fibroplastic tissue accompanied by
variable fibrosis or withering of the preexisting vessel wall. It
is this pathologic variable of interest, the amount of fibroplasia,
that can be measured by its thickness in a direction perpendicular
to the arterial axis. A method for quantitating the fibroplasia is
described. Use of this method in a series of 93 autopsies suggests
two tentative conclusions. (1) The outer diameter of 141 fim
marks the size of artery in which fibroplasia best correlates with
blood pressure. (2) The linear function, mean blood pressure
= 1.60 X microvascular lesions + 79.7, with correlation coefficient 0.698, governs a relationship similar at all ages. This relationship can be used to compute expected blood pressure from
measurements of microvascular abnormalities in kidneys obtained at autopsy. Because all ages include ages 14 to 21 years,
the observations imply that the initial tissue changes of hypertensive disease occur early in life. (Key words: Hypertension;
Nephrosclerosis; Aging) Am J Clin Pathol 1986; 85: 312-318
THE KIDNEYS OF PATIENTS with hypertension examined at autopsy typically have severe nephrosclerosis.
A mathematic model of the quantitative relationship between blood pressure and renal microvascular abnormalities would be very useful. Such a model, however,
would require that magnitudes be measured on some kind
of rational quantitation scale. A satisfactory scale of measuring microvascular features of nephrosclerosis has not
been described previously. The recent availability of microcomputers allows fresh approaches to the previously
unsolved morphometric problem. Accordingly, we reevaluated the question using updated quantitative
methods.
Received April 9, 1985; received revised manuscript and accepted for
publication May 23, 1985.
Supported in part by Grant No. HL08974, National Heart and Blood
Institute, National Institutes of Health, Bethesda, Maryland. Photographs
by Gene Wolfe.
Address reprint requests to Dr. Tracy: Department of Pathology, LSU
Medical School, 1901 Perdido Street, New Orleans, Louisiana 70112.
Departments of Pathology and Medicine, Louisiana State
University Medical Center, New Orleans, Louisiana
Materials and Methods
Collection of Cases
Subjects examined at autopsy in the Oregon and New
Orleans series were screened for inclusion in this study.9
If the hospital chart contained at least one acceptable
(outpatient) blood pressure reading in each of the six years
preceding death, and at least ten readings altogether, then
the subject was accepted. At the University of Oregon
from 1965 to 1967, 33 subjects were collected, and from
the Louisiana State University service of Charity Hospital
in New Orleans during 1967 to 1971, 60 subjects were
assembled.
In the Bogalusa Heart Study, approximately 5,000
children have been observed for blood pressure levels by
cross-sectional and longitudinal study, often more than
once.1 A system for obtaining tissues at autopsy from those
subjects who died has yielded kidney samples from 35
cases. Of these, eight have been processed for tissue sections and measured for use in this report. They ranged in
age from 14 to 21 years, and included five white males,
two black males, and one white female; deaths were the
result of violence in all but one white male who had histiocytic lymphoma.
Blood Pressure Records: Variables Pa and Pb
In the Oregon and New Orleans series, blood pressure
readings were rejected if taken on an inpatient basis; only
outpatient records were accepted. For each recorded
measurement of systolic (Sy) and diastolic (Di) pressure,
the mean of the pulse wave (M) was calculated as M = (Sy
+ 2Di)/3. In the sequence of values of M, a series of averages were computed and subjected to Mest. The average
of the first two was compared with the average of all remaining values; the average of the first three was compared
with all remaining values; and so on. In this sequence of
Mests, the t having the smallest P value was called the
312
Vol. 85 • No. 3
MORPHOMETRY OF MICROVASCULATURE
"transition point." The inherent properties of a series of
blood pressure readings that justify this identification of
the transition point have been extensively discussed elsewhere.9 The average of the pulse wave means before (Pb)
and after (Pa) transition and the duration of P a were tested
for relationships with lesion measures and age.
The Bogalusa cases were assessed at evaluation sessions
designed to obtain blood pressure measurement as free
from known sources of error as possible.1 The most recently obtained reading was used in this analysis.
Processing of Tissues
Six pieces of tissue were processed from the best preserved regions of kidney having no scars or focal lesions.
Sections of 6 pm were mounted one to a slide and stained
in a modified Mowry's alcian blue-periodic acid-Schiff
(PAS) routine, in which picric acid is replaced by metanil
yellow.9
Measuring Microvascular Abnormalities:
Variables Hy, S, and I
The hyalinization of arterioles having outer diameters
less than 50 ^m was evaluated separately from the fibroproliferative-mediodestructive features of small arteries
having outer diameters of 50 to 400 jtm. Hyalinization
was graded by counting the numbers of glomeruli and
hyalinized arterioles; the ratio expressed as hyalinized arterioles per thousand glomeruli is called Hy. In the small
arteries, departures from normal are in the form of an
acquired fibroplastic layer within the preexisting muscular
media, often obliterating and replacing the media (Fig.
1). An eyepiece ruler is used to measure the outer diameter
of the artery with the X10 objective lens, and the thickness
of the acquired layer is measured with the X40 objective.
Most of the arteries are presented to view in elliptic section
(Fig. 1). Measurements are made on the least axis of the
ellipse. Arteries sectioned on a branch are excluded from
measurement, along with all of those that are artifactually
distorted, most commonly because of oblique or tangential sections.
The measurements are keyed into a microcomputer
beside the microscope. The paired measurements of outer
diameter (D) and fibroplastic thickness (R) in approximately 40 to 60 arteries are used to fit the equation, R
= SD + I, with empirical parameters S and I. The method
of orthogonal regression is used to fit the major axis of
the confidence ellipse because of the fact that both R and
D are random variables.3 The major axis is less prone to
spurious effects from large or small error variation than
is the regression coefficient. Elsewhere throughout this
paper, however, the usual least-squares regression equa-
313
tions are used, because in those applications a designation
of independent and dependent variables is appropriate.
Results
Within Kidneys
The thickness of the acquired inner layer of the arterial
wall (R) can be expressed as a linear function of the vessel
size measured by its outer diameter (D). This linear function can be stated R = SD + I, where S (slope) and I
(intercept) are empirical parameters that can be estimated
by a variety of statistical technics. When a typical hypertensive subject is compared with a typical nonhypertensive
subject (Fig. 2), differences between them seem mainly to
be in the parameter I and not in S. In the comparison of
these two kidneys, vessels of all sizes have, on average,
the same amount of added R in the hypertensive compared with the nonhypertensive (about 20 nm vertical
displacement between the almost parallel lines). In the
nonhypertensive subject, all of the larger arteries had an
acquired layer, but the appearance of most smaller arteries
remained normal. In the hypertensive subject, the acquired layers in larger arteries were further thickened by
a certain increment, and that same-sized increment was
newly added to the inner walls of smaller arteries. This
might suggest that abnormalities of larger arteries do not
relate well to blood pressure while the smaller vessels are
of greater importance in this regard. The statistical testing
of this impression is, however, not entirely straightforward.
Each of the 93 cases from New Orleans and Oregon
yielded a set of measurements of R and D, usually numbering about 40 to 60 data pairs. These data sets were
used to estimate I and S for each case by the method of
principal axis. On the average, in the 93 cases, the two
variables held an inverse linear relationship expressed by
the regression equation, I = -117.6S+ 14.3 (r = —0.708).
Separation of the hypertensive from nonhypertensive
subjects was done by using 115 mmHg in the average
mean pressure after transition (Pa) as the cut-off point.
The two types of subjects generally tend to be separated
from each other in the plot of I versus S (Fig. 3). However,
whether or not the hypertensive subjects are higher on S
or I, or both, is not immediately apparent.
This matter was examined in the following manner.
The 93 cases from New Orleans and Oregon were separated into quartiles on S, the cut-off points being 0.213,
0.170, and 0.143. Mean P A values in the quartiles were
112.3, 110.6, 113.2, and 102.6; the lowest quartile was
significantly different from each of the other three at P
< 0.05 (multiple range test). Similarly, quartiles on I were
constructed with cut-off points of —0.37, —6.50, and
-11.00. Pa values averaged 117.7, 108.5, 105.7, and 105.9
314
TRACY ET AL.
AJ.C.P. • March 1986
MORPHOMETRY OF MICROVASCULATURE
Vol. 85 • No. 3
315
FIG. 1. Commonplace elliptic profiles of arteries about 141 ^m in outer diameter (D) are shown. An acquired layer offibroplasiainterior to the
preexisting "media" is the pathologic variable of interest (R). The media (M) may thicken by fibrosis (B), or more often wither and vanish (C).
Examples of hyalinized arterioles (H) are shown in C. PAS-Alcian blue (X260).
in these quartiles, and the highest quartile significantly
differed from each of the other three. These results imply
two immediate conclusions. (1) The cases having the
smallest slopes had lower average blood pressures. (2) The
cases having the highest intercepts had the higher blood
pressures.
Multiple Regression
Thefindingsshown in Figure 3 suggest that a weighted
average of S and I might better relate to blood pressure
than either variable alone. An optimal weighted average
computed by linear regression for the New Orleans and
Oregon cases was Pa = 227S + 1.601 + 79.7 (r = 0.600).
Various curvilinear expansions of this relationship did
not significantly improve on the linear fit. The change of
variable R14l = 141S + I, which factors from the regression
equation, has an interesting geometric meaning in Figure
2. If a vertical is erected at 141 on the abscissa, then that
vertical will be intersected at the ordinate R]41 by the line
that was fitted through data points, SD + I = R. Setting
D = 141, the thickness of the acquired layer for an artery
of 141 micrometers outer diameter is derived. Abnormalities in arteries of this size are better related to blood
pressure than are those in larger or smaller arteries. This
result is illustrated in Figure 4. If a value other than 141
was used, then the scatter plot would be more circular
and the correlation less.
Hyalinized Arterioles
The variable Hy, based upon enumeration of hyalinized
arterioles in a unit area of tissue section, correlates with
blood pressure, although not as well as Ri41 (r = 0.439
and 0.600, respectively). In combination, the multiple r
is 0.616, which is a significant improvement (0.001 < P
0
Mean
(A
«
10 - \
c
o
\
> 115= •
< 115= o
\
i
0-
£
Z
a
y
Case a = .
Cats b= o
jS
60141
o
V •
X
•
/
4 0
'
.*'
*«
£
|
yS
oy
yT
•
o
• •
•
•• «/
• o
o
O
•
20-
X ' v
/ n
°f
o
. • -
/
•
Ss
''
*
•
y
.?
o
°
°„\
° \
.\
° °
°o° ' ° o \ °0
10 -
.\ p\ ° .*o:
o °
(O
\oo
bo\
«
E
o
° If
o
\
°
o
o
o
I
N
o
20 -
a
4)
O
0)
,'
•
•
\
0° ° \
0)
y f
3
o
* *
••
\
(0
«^
o
•
o •
oo
i*
80-
o
\
•a
0)
'5
a
u
E
a.
i."
Blood
Pressure
0
• "b
0
-
\
c
o
' ' ' o
\
30s
0
100
200
300
400
A r t e r i a l O u t e r D i a m e t e r ; urn
FIG. 2. The thicknesses of the acquired layers offibroplasia(R) are
given in relation to the outer diameters (D) of each of the arteries observed
in two representative cases arbitrarily called a and b. Linesfittedto the
points are the principal axes through the two scatters. The slopes (S) and
intercepts (I) of these lines are the summary statistics that embody the
pathologic information of interest. The positions of these two cases relative
to all others is noted in Figure 3. On the abscissa, the vertical at 141 is
marked to aid discussion.
0.1
0.2
0.3
S = Slope; micrometers per micrometer
FIG. 3. Slopes and intercepts derived as noted in Figure 2 are plotted
for the 93 cases of the retrospective study: Reasons for taking 115 mmHg
for the average mean blood pressure as the arbitrary value separating
hypertensive from normotensive subjects are given in reference 9. Cases
a and b are those in Figure 2. The equation 14IS + 1 = 18 yields the
plotted line that appears to separate the two kinds of cases.
TRACY ET AL.
316
0
2
I
between observers is 12.8, which is about one-third as
large as the variance of R14) in the 93 case total sample.
Figure 5 relates the R14i values from the two observers to
each other.
"
* New O r l » « m
Ortgon
& Bogalusa
0
r=0.600
140-
Children Samples
*
!
o
_"
°
o
o
•
o
120 -
•
<
o
0
•
0
Off
•
/
"
i
So
O
,
•
^
O
•/>
• o
. .
100 ° / •
*•
S
a
-/
-/
"o • •
•
°
O
A.J.C.P. • March 1986
° •
o
*
0
80 •
10
20
30
R 1 4 1 = Acquired Wall T h i c k n e s s at 1 4 1 Intercept: urn
FIG. 4. Average mean blood pressure after transition (P„) is related to
a summary statistic (Ri4,) which optimally estimates the pathologic variable of interest. The individual points represent the 93 cases of the retrospective study, largely in the age range of 50 to 80 years, and the eight
cases aged 14 to 21 years from the prospective study. The regression line
fitted to the 93 retrospective points is drawn. The position of 18 i>m
noted on the abscissa represents the diagonal line in Figure 3.
< 0.05), and the equation is Pa = 1.32R141 + 0.023Hy
+ 82.0. The coefficient for R141 alone was 1.60; hence,
the fall to 1.32 implies a multicolinearity effect, so that
both variables may be measuring the same underlying
property of the kidney. Subsequent entry of age into this
equation did not significantly improve the model fit.
Comparison between Observers
Two observers independently measured 22 cases. Letting RD = R, - R2 = (S, - S2)D + (I, - I2) be the difference
between the findings of the two observers in each of the
22 cases, we can ask what choice of D would minimize
the variance of RD. Setting a = Si - S2 and /3 = l{ - I2,
their covariance and variances, V„ = 0.007849, Va/3
= -1.1872, and V^ = 192.329, are computed from the
data. The variance of RD is VD = VaD2 + 2V„„D + V^.
Putting the derivative of VD equal to zero yields D0
= -Va/}/Va = 151 for the desired choice. Hence, the average size of the acquired wall thickness can best be expressed in terms of arteries with outer diameter of 151
micrometers, which is coincidentally close to 141 already
derived. Using this choice, the variance of the difference
Eight cases of ages 14 to 21 years from the Bogalusa
Heart Study were combined with the 93 hospital-derived
cases to explore an enlarged age range. In the expanded
data set, linear regression yielded Pa = 238S + 1.651 + 77.8
(r = 0.698), from which R)45 = 145S + I is factored; age
did not significantly improve the correlation. However,
in the equation R145 = 0.238Pa + 0.164 Age - 18.24,
entry of age substantially improved the correlation to
0.788. In comparison with the restricted data set composed
of 93 hospital cases, the expanded data set, including eight
young subjects from a different study in which blood
pressure was measured in a different way under different
circumstances, showed these contrasts: (1) Correlations
are improved. (2) The intercept shifts from 79.7 to 77.8.
(3) While age continues to fail in reducing residual Pa,
age has a greater effect in reducing residual R)45. In both
data sets, no curvature was found in the relationship between blood pressure and microvascular abnormalities;
they emerge with a strictly linear function. And the vessel
sizes that best predict blood pressure, 141 in one data set
and 145 in the other, seem closely similar.
Discussion
The technics used for arterial mensuration have been
reviewed recently by Cook and Yates.2 They favored the
30-
20-
10r=0.862
10
20
30
R141 by Observer One; micrometers
FIG. 5. Measurements of R14, in 22 cases obtained by two independent
observers are shown. The line representing equality between the two
observers is drawn. The star gives the position of the mean values.
Vol. 85 • No. 3
MORPHOMETRY OF MICROVASCULATURE
4
approach introduced by Furuyama. In that method, the
medial area of the vessel wall, excluding the intima, and
the length of the internal elastic lamina are measured on
circular cross-sections. Wall thickness and radius are
computed using the formulas of circular measure. The
ratio of these computed variables is taken to reflect a
pathologic feature often called "medial hypertrophy."
Another approach suggested by Nishi and colleagues is to
measure intimal and medial areas, and to express "intimal
hypertrophy" as a ratio.6 Thus, one group of workers attends to the media to the exclusion of the intima, while
the other group is concerned only with intima using media
as a kind of baseline. Both of these approaches require
time-consuming planimetric or computer methods and
highly selected, perfectly circular cross-sections. In our
experience, arterial profiles in a tissue section are seldom
presented in perfect circular outline and without artefactual distortion. If all of both kidneys are used, then perhaps
100 to 150 sections may be prepared to provide enough
samples to yield statistically meaningful results. In contrast, the method used in this study requires, at most, four
to six sections to include ample corticpmedullary tissue
offering sufficient numbers of arcuate and cortical vessels
of the optimum sizes.
Older methods of mensuration also often called for use
of only circular cross-sections.5 As noted by Short and
co-workers, this requirement is not necessary.7 Elliptic
sections always have shortest diameters that present cuts
that are perpendicular through the cylindrical arteries.
This allows valid measurements to be made quickly and
easily with an eyepiece ruler on most elliptic or longitudinal sections, even among some minor artefacts.
A further problem not previously considered concerns
the simultaneous contributions of both "intima" and
"media" to the wall thickness. In vessels with outer diameters over 500 fim, pathologic changes of interest are
chiefly medial, and perhaps best approached by the Furuyuma system. In vessels from 50 to 400 nm, the conspicuous abnormalities of predominant interest consist of
layers of acquired tissue added interior to the preexisting
normal vessel wall. The acquired layer is bulky, with dense
interstitial connective tissue materials and poor in cells,
and the cells tend to be longitudinally oriented. The acquired layer is considered "intimal" and might better be
called "fibroplasiav rather than hypertrophy. The amount
of acquired intimal fibroplastic tissue is the pathologic
variable of principal interest in relation to blood pressure.
The thickness of this tissue in micrometers is a continuous
quantity with a zero baseline, which means that it expresses the magnitude of the pathologic variable on a rational scale. The preexisting "media" responds to the fibroplasia in various ways. At times the cells expand their
pericellular matrix to increase the medial bulk, a circumstance sometimes called "fibrosis" (Fig. IB). More often,
317
the cells and their matrix wither and vanish, leaving an
indistinct or absent media (Fig. 1C). By measuring the
acquired fibroplastic layer to the exclusion of a normal,
expanded, or withered media, one gains some advantages
conceptually and also in the speed and ease of making
measurements.
The method described is easily reproduced by different
observers (Fig. 5). The independent measurements of 22
cases by two observers were made without consensus or
training sessions to improve agreement. Materials were
not selected for large sample size or freedom from artefacts. The result, therefore, does not reflect the maximum
agreement that might be achieved, but rather one that
untrained observers might be expected to achieve independently.
The use of perfusion fixation also was examined in this
study. Hypothetically, one might wish to restore living
dimensions to the postmortem material by fixing with
formalin perfused under pressure into the renal artery.
Our material from a previous study of perfusion fixation
was found unsuitable for this investigation because of
profound artifact formation. The acquired fibroplastic
layers became swollen and variably separated by clear
spaces, thus largely obliterating differences between kidneys.
The overallfindingsof this study can be presented conceptually. It is likely that arteries of all sizes begin in youth
with little or no intima, and, therefore, in Figure 2 points
fall along the abscissa with zero or near zero magnitudes
for R. As the intima begins to appear in aging arteries,
the points begin to rise upward perhaps at a constant rate
like bubbles rising in water. The growth of intima was
found in this study to begin early in the largest arteries
and affect the smallest much later. Once the growth of an
acquired layer begins in the artery wall it appears to progress at the same rate in vessels of all sizes, the difference
in artery size affecting only the time that the process begins. Blood pressure was found to be elevated only when
the process had spread into the smallest arteries. Hence,
blood pressure tended to settle at a level largely governed
by the average magnitude of the abnormal fibroplasia in
arteries around the size of 141 micrometers outer diameter, in accordance with the linear function Pa = 1.60 Rui
+ 79.7. The correlation coefficient that relates these variables was 0.600. Measurement errors that tend to attenuate the correlation have been discussed in detail elsewhere.10
Combining data from a retrospective hospital-based
study and a prospective community-based study allows
us to represent the extremes of high and low of the observed variables, but such findings should be viewed as
tentative. Interestingly, whether or not the eight young
subjects are included, the results are not greatly different.
The regression line drawn through the retrospective data,
3 J8
TRACY ET AL.
AJ.C.P. • Match 1986
Table 1. Variable Names and Symbols, Mean, Standard Deviations (SD) and Correlation Coefficients for 93 Cases
Variable name
Mean
SD
Pb
S
I
RMi
Age
Hy
RHI
P„: Mean pressure after transition
Pb: Mean pressure before transition
S: Slope
I: Intercept
R14,:141S + I
Age .
Hy: Hyalinized arterioles
RHI: Ordinal Index
109.4
110.1
0.175
-6.27
18.5
67.9
139
1.34
16.5
16.6
0.052
8.55
6.2
12.4
151
0.65
0.553
0.120
0.115
0.330
0.176
-0.708
0.600
0.380
0.196
0.554
0.003
0.171
0.152
0.034
0.227
0.439
0.163
-0.028
0.346
0.459
-0.011
0.660
0.416
0.016
0.550
0.801
0.143
0.520
Correlation coefficients in italic type are significantly different from zero. P < 0.01.
when extended, passes close to the center of the prospective data points (Fig. 4). These findings strengthen our
provisional impression that blood pressure is, from these
results, seen to be a linear function of renal microvascular
abnormalities as measured by the described methods.
Conclusion
The hypertensive kidney is most typically characterized
by structural abnormalities in the cortical and arcuate
arteries in the size range 50 to 400 fim. T h e predominant
abnormality consists of acquired layers of intimal fibroplastic tissue accompanied by variable fibrosis or withering
of the preexisting medial vessel wall. The a m o u n t of fibroplasia, measured by its thickness in a direction perpendicular to the arterial axis, is the pathologic variable
of interest. T h e largest size range of arteries is the first to
acquire a fibroplastic layer, and this initial alteration is
not accompanied by any measurable change in the blood
pressure. When the fibroplasia later extends into smaller
arteries, a tendency for blood pressure to rise is observed.
This tendency is expressed by the linear function mean
blood pressure = 1.60 X microvascular abnormalities
+ 79.7. The youngest observed subjects already showed
early tissue changes of the kind that relate to blood pressure levels in later life. A microcomputer placed beside
the microscope can be used to process raw measurements
of fibroplasia and arterial diameter. Using practical
a m o u n t s of time and tissue sample, this method allows
quantitation of microvascular nephrosclerosis on a rational scale of measure, and these measurements can easily
be reproduced by independent observers.
APPENDIX
T H E FORMER SUBJECTIVE METHOD: T H E VARIABLE RHI
The slides from New Orleans and Oregon were used previously
and evaluated on an ordinal scale of measure based on averaging
about 100 estimates of lesion severity from 0 to 4+. 9 The relationships of the former measurements (RHI) to the quantitative
measurements obtained by the newly introduced eyepiece ruler
method (R| 4 i) were examined by linear regression. Between RHI
and Rut, a correlation of 0.801 was found and a quadratic
equation was not significantly better than a linear one to relate
the two variables. Pa related to RHI and R 14| with correlations
of 0.660 and 0.600, respectively. The regression equations Pa =
16.6RHI + 87.3 (r = 0.660) and Pa = 17.1 RHI - 0.188 Age
+ 99.4 (r = 0.674) were not significantly improved by entry of
Ri4i. The two measures of vascular abnormality appear to be
interchangeable estimates of the same renal property, differing
only in random measurement error. The rational scaled value
R,4i is preferred chiefly because it is much easier to standardize
between observers.
BLOOD PRESSURE BEFORE AND AFTER TRANSITION
The average mean blood pressure values before (Pb) and after
(Pa) transition both correlated with R,41 (r = 0.380 and 0.600,
respectively), although for P b , the correlation was significantly
lower (0.01 > P > 0.005). In combination, P b did not significantly
improve the multiple regression fit. Age, however, did enter with
significant improvement of r to 0.640, yielding R141 = 0.223Pa
+ 0.11 Age - 13.59. The asymetric behavior of age may relate
to the way the cases were selected. Autopsied subjects with long
blood pressure records included many young hypertensives.
Consequently, blood pressure does not correlate with age, but
R,4i does (Table 1).
References
1. Berenson GS, Voors AW, Gard P, Newman HI WP, Tracy RE:
Clinical and anatomic correlates of cardiovascular disease in
children from the Bogalusa heart study. Proceedings of the 6th
International Symposium of Atherosclerosis. Edited by FG
Schettler, AM Gotto, G MiddelhofF, AJR Habenicht, KR Jurutka.
New York, Springer Verlag, 1983, pp 60-65
2. Cook TA, Yates PO: A critical survey of techniques for arterial mensuration. J Pathol 1972; 108:119-127
3. Diem K (ed): Documenta Geigy Scientific Tables, 6th ed. Ardsley,
New York, Geigy Chemical Corporation, 1962, p 181
4. Furuyama M: Histometrical investigations of arteries in reference
to arterial hypertension. Tohoku J Exp Med 1962; 76:388-415
5. Kernohan JW, Anderson EW, Keith NM: The arterioles in cases of
hypertension. Arch Intern Med 1929; 66:395-423
6. Nishi T, Bond Jr C, Brown G, Solez K, Heptinstall RH: A morphometric study of arterial intimal thickening in kidneys of dialyzed patients. Am J Pathol 1979; 95:597-610
7. Short D; Morphology of the intestinal arterioles in chronic human
hypertension. Br Heart J 1966; 28:184-192
8. Tracy RE, Overll EO: Arterioles of perfusion-fixed hypertensive and
aged kidneys. Arch Pathol 1966; 82:526-534
9. Tracy RE, Tabares Toca V: Nephrosclerosis and blood pressure I.
Rising and falling patterns in lengthy records. Lab Invest 1974;
30:20-29
10. Tracy RE, Johnson WD, Lopez CR, Toca VT: Hypertension and
arteriolar sclerosis of the kidney, pancreas, adrenal gland, and
liver. Virchows Arch 1981;391:91-106