Nephrol Dial Transplant (2001) 16: 1163±1169
Original Article
Prognostic value of simple measurement of chronic damage
in renal biopsy specimens
Alexander J. Howie, Maria Alice S. Ferreira and Dwomoa Adu
Department of Pathology, University of Birmingham, and Department of Nephrology, Queen Elizabeth Hospital,
Birmingham, UK
Abstract
Background. A simple method of measurement of
chronic damage in renal biopsy specimens would be
useful in clinical management, prognosis, comparisons
between different centres and trials.
Methods. An interactive image analysis system was
used to outline and measure areas of chronic damage
in 247 renal biopsy specimens to give an index of
chronic damage, expressed as a percentage of cortical
cross-sectional area. Prognostic value was analysed
by the Kaplan±Meier method to study time between
biopsy and onset of permanent dialysis.
Results. There was no signi®cant bias between measurements by the same observer or different observers.
The index of chronic damage ranged from 0 to 90%.
Increasing severity of chronic damage was associated
with shortened renal survival. Each increase of 10%
in the index increased the hazard ratio of risk of permanent dialysis by 1.5 times (95% con®dence interval
1.4±1.7, P-0.001).
Conclusions. A simple measure of chronic damage was
a powerful indicator of prognosis. This is likely to be
clinically useful in routine practice and trials.
Keywords: chronic damage; morphometry; prognosis
Introduction
Progression to chronic renal failure is in¯uenced by the
extent of irreversible damage w1x. Such damage has
been assessed in many ways since the work of Risdon
et al. in 1968 w2x, e.g. by Striker et al. w3x and Bohle
et al. w4x. Subjective grading systems have usually been
used rather than morphometric methods. Grading
systems such as the chronicity and activity indices in
lupus nephritis are easy to use but may be unreliable
w5,6x. Objective measures are more dif®cult but more
Correspondence and offprint requests to: Dr A. J. Howie, Department
of Pathology, The Medical School, Birmingham B15 2TT, UK.
#
reliable w7x. The hypothesis behind this study was that
the amount of chronic damage in renal biopsy specimens could be measured by a simple method and used
to predict renal survival.
Subjects and methods
Renal excretory function and renal survival
Serum creatinine concentration was measured at biopsy.
The interval was noted between biopsy and the end point,
which was introduction of permanent haemodialysis or peritoneal dialysis, or return to permanent dialysis in patients
with a renal allograft. No patient died before onset of
permanent dialysis.
Selection of renal biopsy specimens
Specimens were selected retrospectively to allow several
years of follow up. Diagnoses were made based on conventional criteria w8x, using vasculitic glomerulonephritis
as a synonym for focal segmental necrotizing glomerulonephritis, crescentic glomerulonephritis and similar terms w9x
(Table 1). Those with IgA nephropathy, Henoch Schoenlein
nephritis, vasculitic glomerulonephritis and minimal change
nephropathy were consecutive specimens with those diagnoses. Two patients with anti-glomerular basement membrane antibody disease (Goodpasture's syndrome) were
included in the vasculitic group. A few patients with minimal change nephropathy had been treated with cyclosporin
and a biopsy was performed to see if the drug had damaged
the kidney. The other groups consisted of all other consecutive specimens taken over a period of several weeks.
The 21 allograft specimens were taken 1 week to 5 years after
transplantation (median 2 weeks). All patients were adults.
Measurement of amount of chronic damage
Figures 1 and 2 illustrate the method used. One routinely
prepared, formalin ®xed, paraf®n embedded, 2 mm section of
each specimen stained by periodic acid-methenamine silver
was examined under a microscope with a video camera.
The section was selected only on the grounds that it was
2001 European Renal Association±European Dialysis and Transplant Association
1164
A. J. Howie et al.
technically satisfactory and as completely representative as
possible of the size of the specimen, including all pieces if
there were more than one. Images at magni®cation 3 10
were captured with an Aequitas IDA image database and
image archive management system (Dynamic Data Links,
Cambridge, UK). For analysis, each image was converted
from Aequitas IDA to Aequitas IA image analysis software (Dynamic Data Links). The threshold was adjusted to
highlight everything in a de®ned area. Cortex was de®ned
as the part inside the renal capsule and outside the medulla,
whose outer limit was de®ned by the cortical aspect of
arcuate vessels and the medullary aspect of the deepest
glomeruli. All cortex was outlined using the freehand
drawing facility. The size of this area was measured by the
system in arbitrary units. Freehand drawing was then used
to outline areas of chronic damage. This included glomeruli
showing global sclerosis but not segmental sclerosis, areas of
interstitial ®brosis, which appeared more solid and deeply
stained than normal or oedematous interstitial tissues, and
atrophic tubules, de®ned as tubules smaller than normal,
with thickened basement membranes, or tubules larger than
normal, with thin epithelium, including those large enough
to be considered cysts w10x. Arteries and arterioles were
not judged to have chronic damage unless they were completely occluded. The total area outlined was measured.
Areas of cortex and chronic damage were summed for each
specimen, and the percentage of chronically damaged cortex
was calculated to the nearest integer to give the index of
chronic damage. To test agreement between measurements
of the index, 20 biopsy specimens were measured twice at
intervals by one observer, and 20 other specimens measured
separately by two observers.
Statistical analysis
Agreement between measurements was assessed by the Bland
and Altman method after log transformation, since differences between measurements were proportional to the mean
w11x. This method gives the bias, or mean difference between
measurements, and limits of agreement, or two standard
deviations either side of the mean, with con®dence intervals (CIs) for these. The Spearman rank correlation coef®cient was used to investigate the correlation between the
index and serum creatinine concentration at biopsy. This
non-parametric test was selected because the distribution of
variables was not assumed to be normal. To study survival
to the end point, serum creatinine concentration at biopsy
was divided arbitrarily into four groups: -120 mmolul,
120±249 mmolul, 250±499 mmolul and P500 mmolul, and the
index of chronic damage was divided arbitrarily into
®ve levels: -10%, 10±19%, 20±39%, 40±59% and P60%. Survival rates were examined by the Kaplan±Meier product
limit method. The equality of survivor function across
groups was tested by the log rank test. The Cox proportional hazards model was used to study the contribution of
initial serum creatinine concentration and index of chronic
damage to the relative risk of end-stage renal failure (ESRF).
This was done by performing a Cox regression analysis
with censored data to estimate the roles of serum creatinine
Table 1. Details of 247 renal biopsy specimens on which the index of chronic damage was measured, with correlation between index and
serum creatinine concentration at biopsy (Index v cr bx)
Group
n
n cr bx
IgA nephropathy
58
58
Henoch Schoenlein nephritis
46
46
Vasculitic glomerulonephritis
36
36
Minimal change nephropathy
30
17
Allograft
21
21
Segmental sclerosing diseases
12
12
9
9
8
8
Late non-glomerulonephritic
damage
Normal
7
7
4
4
Acute tubular damage
4
4
12
12
247
234
Thin glomerular basement
membrane disease
Membranous nephropathy
Miscellaneous
Overall
Index v cr bx
n end and
index
n fup and
index
Median and
range fup
n others and
index
rs0.76
P-0.001
rs0.29
Ps0.05
rs0.15
ns
rs0.42
ns
rs 0.24
ns
rs0.89
P-0.001
rs0.56
ns
rs0.88
Ps0.004
rs0.89
Ps0.007
rs0.11
ns
rs 0.63
ns
rs0.61
Ps0.04
rs0.43
P-0.001
21
40% (1±89)
11
13% (3±61)
11
28% (1±88)
0
30
8% (0±77)
34
2% (0±23)
22
7% (1±57)
26
2% (1±31)
11
1% (0±12)
4
6% (1±33)
5
2% (0±3)
6
3% (2±11)
4
28% (2± 44)
3
3% (1±3)
2
5% (1, 8)
3
5% (3±20)
150
3% (0±77)
10 years
4 months±13 years
4 years
1±7 years
9 years
1 month±12 years
2 years
1 month±10 years
7 years
7 months±8 years
7 years
7±8 years
7 years
18 months±8 years
7 years
6±8 years
4 years
2 months±7 years
1 year
1±7 years
4 years
1±7 years
7 years
6±7 years
6 years
1 month±13 years
7
1% (0±72)
1
0%
3
2% (1±5)
4
2% (1±21)
9
5% (1±28)
6
10% (1± 47)
4
2% (1±3)
2
5% (4, 6)
2
83% (75, 90)
1
0%
2
2% (1, 2)
6
11% (0±31)
47
3% (0±90)
1
2%
2
46% (29, 63)
0
0
1
74%
0
0
3
51% (38±65)
50
37% (1±89)
Abbreviations: n, number in group; n cr bx, number on which serum creatinine concentration was known at biopsy; n end and index, number
reaching end point, with median and range of index; n fup and index, number not reaching end point with follow up information, with median
and range of index; Median and range fup, median length of follow up in years of those not reaching end point, with range in months or years;
n others and index, number without follow up, with median and range of index; ns, not signi®cant (P)0.05).
Measurement of chronic renal damage
concentration, categorized so that a change of 1 unit meant
a change of 100 mmolul in serum creatinine concentration,
and the initial index of chronic damage, categorized so that
a change of 1 unit represented a change of 10% in the index,
in the relative risk of reaching the end point (Stata Statistical
Software, Release 5.0, College Station, TX, USA).
1165
All images of cortex were analysed on every specimen.
The median number of images was six (range 1±15).
Analysis of an image took between 2 and 3 min. There
was no signi®cant bias between measurements by the
same observer (mean 0.97; 95% CI 0.92±1.02) or measurements by two observers (mean 1.0; CI 0.94±1.06).
The limits of agreement of measurements by the
same observer were 0.72 (CI 0.61±0.81) and 1.21 (CI
1.12±1.31), and the limits of agreement of measurements by two observers were 0.73 (CI 0.62±0.83)
and 1.27 (CI 1.17±1.38).
The overall index in 247 specimens ranged from 0 to
90%, but was 100% in some images. For all specimens,
there was a weak signi®cant correlation between the
Fig. 1. An image on a computer of part of a renal biopsy specimen
stained by periodic acid-methenamine silver. Figure 2 shows the
preparation of this image for morphometric determination of the
index of chronic damage.
Fig. 2. The image seen in Figure 1 after freehand drawing to illustrate in principle how the index of chronic damage is calculated.
The cortex is outlined to exclude the capsule, and areas of chronic
damage are outlined and ®lled, ready for automatic measurement of
areas. In practice, the cortex would be outlined and measured before
the areas of chronic damage.
Results
1166
index and serum creatinine concentration at biopsy
(rs0.43, P-0.001) (Table 1). There was a strong
signi®cant correlation (rP0.76) between the index and
serum creatinine concentration in four groups, a weak
signi®cant correlation in two groups, and no signi®cant
correlation in the remaining six.
Table 1 shows that the overall median index was
37% in those specimens associated with progression to
the end point, whereas the overall median index was
3% in both those that had not reached the end point
at follow up and those without follow up. By the log
rank test, the index was a strong predictor of rate
of progression to ESRF ( x 2s138.9, four degrees of
freedom (df ), P-0.001; Figure 3). Only three patients
with an index-10% reached the end point within
5 years of biopsy, one with Goodpasture's syndrome,
one with vasculitic glomerulonephritis and one with
Henoch Schoenlein nephritis. Graphs of renal survival related to the index were similar to those of the
whole series in the two largest groups, namely IgA
nephropathy ( x 2s44.9, 4 df, P-0.001) and Henoch
Schoenlein nephritis ( x 2s53.6, 4 df, P-0.001). This
was also true of the combination of the other three
groups, each with at least two patients who reached
A. J. Howie et al.
the end point, namely vasculitic glomerulonephritis,
segmental sclerosing diseases and miscellaneous
conditions ( x 2s19.3, 4 df, P-0.001), and the combination of the other seven groups that individually
had either one patient or no patient who reached
the end point ( x 2s19.6, 4 df, P-0.001). For simplicity, Figure 4 illustrates the graph of survival for
IgA nephropathy and Henoch Schoenlein combined
(x 2s85.1, 4 df, P-0.001), and Figure 5 illustrates
that for the other groups combined (x2s50.7, 4 df,
P-0.001).
Serum creatinine concentration at biopsy predicted
the development of ESRF less strongly than the index
of chronic damage ( x 2s56.9, 3 df, P-0.001). Several
patients with a raised serum creatinine concentration
recovered. Using the Cox proportional hazards model,
after controlling for initial serum creatinine concentration, each 10% increase in the index resulted in
reduced survival with a hazard ratio of 1.5 (CI 1.4±1.7,
P-0.001). With this model, after controlling for the
index of chronic damage, each increase of 100 mmolul
in the initial serum creatinine concentration resulted in
reduced survival with a hazard ratio of 1.3 (CI 1.2±1.4,
P-0.001).
Fig. 3. Renal survival after biopsy related to the index of chronic damage for all specimens (ns200). Upper solid line, index -10%; upper
interrupted line, index 10±19%; middle solid line, index 20±39%; lower interrupted line, index 40±59%; lower solid line, index P60%.
Measurement of chronic renal damage
Discussion
This study con®rms that a major determinant of
prognosis of a kidney that is biopsied is the amount
of chronic renal damage at diagnosis w1x. Concepts of
chronic renal damage in relation to renal function
began to emerge when Risdon and others established
the connection between tubular atrophy and renal
function w2x. The group of Striker emphasized the
importance of interstitial changes such as ®brosis and
correlated tubular disease and interstitial disease w3x.
The group of Bohle gave evidence to support the role
of interstitial ®brosis in prognosis w4x, and they and
Seron et al. w12x showed the relation between obliteration of intertubular capillaries and renal function.
Seron et al. also related the obliteration of intertubular capillaries to the proportion of globally sclerosed
glomeruli w12x. Alexopoulos and others correlated the
chronicity index in lupus nephritis with the amount
of interstitial in¯ammatory in®ltrate w13x. Various
aspects of chronic damage have been investigated in
these previous studies, but the method described now
1167
allows simple assessment of all changes by amalgamation of tubular, interstitial, vascular and glomerular
abnormalities.
Other methods of measurement of chronic damage
have been described but have disadvantages. Point
counting is tedious w14±16x. Measurements of the renal
content of myo®broblasts w17,18x and collagen w19x
require specially prepared immunostained sections,
and sometimes require sampling and manual counting.
Image analysis methods such as texture analysis w20x
and Fibrosis HR w21x require extensive and complicated manipulation of images. These methods, like
the collagen method w19x, give an extent of interstitial
®brosis of ;15% in normal kidneys and of much
less than 100% in abnormal kidneys. Cysts and
thyroidized tubules cause most of these methods to
underestimate the amount of damage.
In contrast, the index of chronic damage described
here is measured on routine sections, allows the whole
of the cortex in a section to be assessed easily, and is
reliable and reproducible. Tests of agreement showed
that there was no signi®cant bias and that the limits
Fig. 4. Renal survival after biopsy related to the index of chronic damage for specimens of IgA nephropathy and Henoch Schoenlein nephritis
(ns96). Upper solid line, index -10%; upper interrupted line, index 10±19%; middle solid line, index 20±39%; lower interrupted line, index
40±59%; lower solid line, index P60%.
1168
of agreement were about one quarter above and below
the initial measurement. The index was 100% in some
images and could have the maximum range from 0 to
100% to allow greater precision than methods with a
restricted range w17±21x. Cysts and thyroidized tubules
are not a problem. The index requires interaction
between observer and computer, but much less than
in automatic measurement w20,21x.
Dif®culties are more related to the material than to
the method. Only one section was examined, as is usual
in morphometric studies of renal biopsy specimens
w14±21x. The small size of the sample meant that it
may not have been representative of the kidney and
that an asymmetrical distribution of abnormalities in
the kidney may not have been re¯ected in the sample.
Variation within the sample was thought to be of
little importance compared with any possible difference between the sample and the kidney. Such a difference probably explains why not every specimen with
an index of P40% was associated with progression to
ESRF (Figures 3±5). One problem with any method
of measurement of chronic damage is that as atrophic
tubules disappear, the kidney shrinks and surviving
tubules often show compensatory enlargement w10x.
A. J. Howie et al.
Methods that measure extent of chronic damage as
a proportion of cortical cross sectional area, which
itself changes as the damage progresses, are likely to
underestimate the real extent of the damage.
The index of chronic damage is a powerful predictor
of the rate of development of ESRF (Figures 3±5).
This is so even when controlled for serum creatinine
concentration at biopsy, which is overall only weakly
correlated with the index. Serum creatinine concentration is strongly correlated with the index in groups
that have mainly chronic damage to explain renal
impairment, such as IgA nephropathy and late nonglomerulonephritic damage. In those with mainly
acute damage, tubules and renal impairment may
recover and the index is not correlated with serum
creatinine concentration. This applies to groups such
as patients with vasculitic glomerulonephritis. Progression to ESRF is much more related to the amount
of chronic damage at the time of biopsy than to the
serum creatinine concentration at biopsy. The index
can be used to predict the outcome and likely rate of
progression, although the amount of chronic damage
at biopsy is not the only factor that in¯uences rate of
progression. The underlying disease is also important
Fig. 5. Renal survival after biopsy related to the index of chronic damage for specimens excluding IgA nephropathy and Henoch Schoenlein
nephritis (ns104). Upper solid line, index -10%; upper interrupted line, index 10±19%; middle solid line, index 20±39%; lower interrupted
line, index 40±59%; lower solid line, index P60%.
Measurement of chronic renal damage
because, for instance, the only three patients with an
index-10% who reached the end point within 5 years
of biopsy all had vasculitic illnesses.
The present study was retrospective to allow an
adequate length of follow up. Whether treatment had
any in¯uence on renal survival independently of the
amount of chronic damage would be dif®cult to determine, but now the amount of chronic damage can
be controlled for prospectively in any treatment trial.
The index avoids the problems of subjective grading
systems w6,7x and most image analysis systems should
be able to be used to calculate it. There are potential
applications in individual patients, clinical studies and
trials.
Acknowledgements. We are grateful to Mr T Marshall for
statistical advice and to our colleagues in the Department of
Nephrology.
1169
7.
8.
9.
10.
11.
12.
13.
14.
References
1. El Nahas AM. Mechanisms of experimental and clinical
renal scarring. In: Davison AM, Cameron JS, Grunfeld JP,
Kerr DNS, Ritz E, Winearls CG, eds. Oxford Textbook of
Nephrology. Oxford University Press, Oxford, 1998: 1749±1788
2. Risdon RA, Sloper JC, de Wardener HE. Relationship between
renal function and histological changes found in renal biopsy
specimens from patients with persistent glomerular nephritis.
Lancet 1968; 2: 363±366
3. Striker GE, Schainuck LI, Cutler RE, Benditt EP.
Structural±functional correlations in renal disease. Part 1: a
method for assaying and classifying histopathologic changes in
renal disease. Human Pathol 1970; 1: 615±630
4. Bohle A, Mackensen-Haen S, Gise H et al. The consequences of
tubulointerstitial changes for renal function in glomerulopathies.
A morphometric and cytological analysis. Pathol Res Pract 1990;
186: 135±144
5. Austin HA, Muenz LR, Joyce KM, Antonovych TT, Balow JE.
Diffuse proliferative lupus nephritis: identi®cation of speci®c
pathologic features affecting renal outcome. Kidney Int 1984;
25: 689±695
6. Schwartz MM, Lan SP, Bernstein J, Hill GS, Holley K, Lewis
EJ. Irreproducibility of the activity and chronicity indices limits
15.
16.
17.
18.
19.
20.
21.
their utility in the management of lupus nephritis. Am J Kid Dis
1993; 21: 374±377
Furness PN. The use of digital images in pathology. J Path 1997;
183: 253±263
Jennette JC, Olson JL, Schwartz MM, Silva FG. Heptinstall's
Pathology of the Kidney, 5th edn. Lippincott-Raven,
Philadelphia, 1998
Adu D, Howie AJ. Vasculitis in the kidney. Curr Diagnost Pathol
1995; 2: 73±77
Pirani CL. Evaluation of kidney biopsy specimens. In:
Tisher CC, Brenner BM, eds. Renal Pathology with Clinical and
Functional Correlations, 2nd edn. J. B. Lippincott, Philadelphia,
1994: 85±115
Bland JM, Altman DG. Statistical methods for assessing
agreement between two methods of clinical measurement.
Lancet 1986; 1: 307±310
Seron D, Alexopoulos E, Raftery MJ, Hartley RB, Cameron JS.
Number of interstitial capillary cross-sections assessed by
monoclonal antibodies: relation to interstitial damage. Nephrol
Dial Transplant 1990; 5: 889±893
Alexopoulos E, Seron D, Hartley RB, Cameron JS. Lupus
nephritis: correlation of interstitial cells with glomerular
function. Kidney Int 1990; 37: 100±109
Bennett WM, Walker RG, Kincaid-Smith P. Renal cortical
interstitial volume in mesangial IgA nephropathy. Lab Invest
1982; 47: 330±335
Nagy J, Trinn C, Deak G, Schmelzer M, Burger T. The role
of the tubulointerstitial changes in the prognosis of IgA
glomerulonephritis. Klin Woch 1984; 62: 1094±1096
Nicholson ML, McCulloch TA, Harper SJ et al. Early
measurement of interstitial ®brosis predicts long term renal
function and graft survival in renal transplantation. Br J Surg
1996; 83: 1082±1085
Roberts ISD, Burrows C, Shanks JH, Venning M, McWilliam
LJ. Interstitial myo®broblasts: predictors of progression in
membranous nephropathy. J Clin Pathol 1997; 50: 123±127
Badid C, Desmouliere A, McGregor B et al. Interstitial alpha
smooth muscle actin: a prognostic marker in membranous
nephropathy. Clin Nephrol 1999; 52: 210±217
Nicholson ML, Bailey E, Williams S, Harris KPG, Furness PN.
Renal allograft survival can be predicted by histomorphometric
assessment of extracellular matrix in six month protocol
biopsies. Transplant Proc 1998; 30: 1305±1306
Moreso F, Seron D, Vitria J et al. Quanti®cation of interstitial
chronic renal damage by means of texture analysis. Kidney Int
1994; 46: 1721±1727
Masseroli M, O'Valle F, Andujar M et al. Design and validation
of a new image analysis method for automatic quanti®cation
of interstitial ®brosis and glomerular morphometry. Lab Invest
1998; 78: 511±522
Received for publication: 11.5.00
Accepted in revised form: 9.1.01