| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Departments of Biostatistics,
2
Pathology,
3
Laboratory Medicine, and
4
Medicine, University of Washington, Seattle, WA 98195.
a Address correspondence to this author at: University of Washington School of Medicine, Box 356174, Seattle WA 98195. Fax 206-598-6706.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Using colorimetric methods, we determined HIC in multiple large (microtome) and small (biopsy-sized) paraffin-embedded samples in 11 resected livers with end-stage cirrhosis. HIC was also measured in multiple fresh samples taken within 5 mm of each other ("local" samples) and taken at sites 3–5 cm apart ("remote" samples) from six livers with end-stage cirrhosis and two healthy autopsy livers.
Results: The within-organ SD of HIC was 13–1553 µg/g (CV, 3.6–55%) for microtome samples and 60–2851 µg/g (CV, 15–73%) for biopsy-sized samples. High variability of HIC was associated with mild to moderate iron overload, because the HIC SD increased with increasing mean HIC (P <0.002). Livers with mean HIC >1000 µg/g exhibited significant biological variability in HIC between sites separated by 3–5 cm (remote sites; P <0.05). The SD was larger for biopsy-sized samples than for microtome samples (P = 0.02).
Conclusion: Ideally, multiple hepatic sites would be sampled to obtain a representative mean HIC. © 1999 American Association of Clinical Chemistry
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In this study, we provide quantitative estimates of the variability in HIC measurements, especially among patients with ESLD. The specific goals of our study were: (a) to quantify interorgan variability in HIC; (b) to quantify total intraorgan variability in HIC; (c) to estimate the components of total intraorgan variability attributable to local variability (measurements within 5 mm of each other) and to remote variability (measurements 3–5 cm apart); (d) to determine whether remote variability is statistically significant; and (e) to quantify the influence of sampling method (biopsy-sized versus microtome sections) on HIC variability.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
75
mm2 surface) was sampled by microtome (mean, 3.8 mg; range,
0.7–9.0 mg dry weight).
The hepatic disease etiologies in series A were as follows: hepatitis C
virus (HCV) in three patients; alcoholic liver disease in one patient;
HCV plus alcoholic liver disease in two patients; HCV plus
1-antitrypsin deficiency in one patient; HCV plus
heterozygosity for HHC in one patient (C282Y heterozygous, H63D
wild-type); nonalcoholic steatohepatitis in one patient; and unknown in
two patients. HFE genotyping for C282Y and H63D mutations
was done in patients with HII 
1.9 (9).
The primary purpose of the series B experiment was to estimate the
local and remote components of total intraorgan variability in HIC.
Local variability was defined as variability among measurements within
5 mm of each other. Remote variability was defined as any excess over
local variability that derives from measuring HIC at sites 3–5 cm
apart. [This is a variance components model: total variance =
local variance + remote variance (10).] Remote
variability may be attributed solely to biological variability, whereas
local variability may be attributed to both analytic and biological
variability. To allow estimation of these two components of
variability, HIC measurements were made in fresh tissue from multiple
adjacent biopsy-sized samples at three different remote sites within
the right lobe of eight livers. Specifically, one fresh 50-g
(wet weight) specimen was taken from the right lobe of each liver.
Three sites were chosen randomly from the perimeter of the 50-g
specimen, and five discrete biopsy-sized samples were sampled
sequentially from each site (mean, 3.4 mg; range, 0.9–11.9 mg dry
weight). These fresh samples were obtained from cirrhotic livers from
six consecutive orthotopic liver transplantation patients and from two
nonexplant livers removed at autopsy. In addition, microtome samples
from the six cirrhotic livers in series B were taken in the same manner
as in series A, with two to five paraffin blocks available from each
liver (mean, 6.4 mg; range, 1.5–11.9 mg dry weight). This provided
additional data to series A for the study of sample type.
Hepatic disease etiologies in series B were as follows: HCV in one
patients; HCV plus 1-antitrypsin deficiency in
one patient; HCV plus alcoholic liver disease in one patient; alcoholic
liver disease plus heterozygosity for the H63D mutation associated with
HHC in one patient; and primary sclerosing cholangitis in two patients.
The autopsy patients were without clinical, gross, or histologic liver
disease.
iron assays
A colorimetric method was used to measure HIC in this study.
Although many investigators use atomic absorption spectrometry for the
determination of iron in tissue, colorimetry has been used in a number
of important studies (8)(11)(12)(13)(14)(15), and its comparability
with atomic absorption for measurement of HIC has been established
(16). The colorimetric method is sufficiently sensitive for
assay of the specimen sizes in this study (16), is less
costly, and is more commonly available than atomic absorption in
clinical laboratories.
Tissue from paraffin blocks was washed twice in warmed xylene to remove paraffin. Both fresh and block samples were dried to constant weights in a Savant Speed Vac drying apparatus, weighed, and then digested at 65–75 °C for 10–15 min with 0.4 mL of concentrated nitric acid. After the pH was adjusted to 4.5 with sodium hydroxide and ammonium acetate buffer, the iron concentration was determined by a bichromatic endpoint method on the Dupont Dimension automated analyzer using disodium 3-(2-pyridyl)-5,6-bis-2(5-furyl sulfonic aid)-1,2,4-triazine (Ferene®) to form the chromogen (17). The concentration of iron in the digests and the dry weight of the tissue was used to calculate µg of iron per g of dry tissue (HIC). HII was calculated as HIC in µmol/g dry weight divided by patients age in years (1).
quality control
The method was monitored for analytical variation by creation of
control charts from HIC measurements on aliquots of a healthy liver
homogenate. The intraassay CV for this control specimen was 2.7%
(n = 10; SD = 54 µg/g), and the interassay CV was 7.3%
(n = 91; SD = 125 µg/g).
statistical analyses
Paired t-tests, two-sample t-tests, Wilcoxon
signed-rank tests, and Wilcoxon rank-sum tests were used for
comparisons of HIC between groups as appropriate. Linear regression
models were fitted by the method of least squares, and
t-tests were used to determine the significance of
regression coefficients. Distributional normality was assessed by the
Kolmogorov–Smirnov test. In series B, the ANOVA method was used to
estimate the local and remote components of variability for the
biopsy-sized measurements from each liver individually. Standard
F-tests were used to determine whether the remote
variability component was significantly greater than zero
(10).
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). There was large interorgan variability in HIC, demonstrated by
the large range in mean HIC (307–5165 µg/g for microtome samples,
260–5290 µg/g for biopsy samples). Mean HIC measurements obtained by
the two sampling techniques were not significantly different
(P = 0.62, paired t-test). The distribution
of HIC measurements within each liver did not differ from a gaussian
distribution, with the exception of one case (P >0.05
for 10 livers; P = 0.001 for case 8 by the
Kolmogorov–Smirnov test.)
|
Intraorgan variability was high for some, but not all, livers.
Intraorgan variability of HIC measurements, quantified by SDs of the
measurements, varied widely across livers, ranging from 28 to 1553
µg/g (mean, 477 µg/g) for microtome samples and from 86 to 2851
µg/g (mean, 717 µg/g) for biopsy-sized samples (Table 1
and Fig. 1
). SDs were significantly higher for biopsy-sized samples than
for microtome samples (P = 0.05, Wilcoxon signed-rank
test). For both sampling methods, there were large and statistically
significant increases in HIC SD as mean HIC increased (P
<0.002 for both sample types; Fig. 1C
). The SD increased with the mean
in such a manner that the CV did not change significantly as a function
of mean HIC for either sampling method (P >0.05; Table 1
).
|
Variability in HII reflected variability in HIC. HII means and ranges
for series A livers are shown in Table 1
. Three of 11 livers had at
least one HII value 1.9, a value often used as a criterion for HHC
(1). The liver from a patient with HCV and HHC
heterozygosity had a mean HII of 2.4 (case 10). The HII range (0.1,
2.8) for biopsy-sized samples from case 11 (diagnosis of alcoholic
liver disease) was illustrative of the high variability in these
measurements from patients with high mean HIC.
series b
As with series A, there was large interorgan variability in HIC
for series B livers (Table 2
). Among the cirrhotic livers, mean HIC ranged from 352 to 8010
µg/g for the microtome samples and from 254 to 3946 µg/g for the
biopsy-sized samples. HIC means from the biopsy-sized samples for the
two healthy livers were 443 and 1741 µg/g. There was no significant
difference between HIC means determined by the microtome and biopsy
sampling methods (P = 0.6, Wilcoxon signed-rank test).
|
Again, intraorgan variability was very high for many of the livers
(Fig. 2
). SDs ranged from 13 to 532 µg/g for the microtome samples
and from 60 to 1813 µg/g for the biopsy-sized samples (Table 2
). On
average, SDs from the biopsy-sized samples were 290 µg/g higher than
those from the microtome samples, but this difference did not reach
statistical significance for series B livers along (P =
0.2, Wilcoxon signed-rank test). SDs for measurements from the
biopsy-sized samples were significantly higher than SDs from microtome
samples in the two series combined (mean difference, 261 µg/g;
P = 0.02, Wilcoxon signed-rank test). Total SDs of
fresh biopsy-sized samples varied significantly with mean HIC
(P = 0.0003) according to the regression equation:
SD = -157 + 0.47 x mean HIC. A significant linear
relationship between total SD and mean HIC also was found for the
microtome samples in Series B (P = 0.01).
|
For four of the eight livers in series B, the remote variability
component (additional variability between sites separated by 3–5 cm)
was significantly greater than zero (Table 3
). The remote SD (square root of the remote variance component)
increased significantly with increasing mean HIC (P =
0.0004), and the livers with significant remote variability all had
mean HICs 982 µg/g. Local SDs also increased with mean HIC but to a
lesser extent than did remote SDs (Fig. 2B
). The CVs for local samples
were significantly lower than CVs based on total variability
(P = 0.008 by the Wilcoxon signed-rank test; Table 3
).
|
As noted above, variability was significantly lower for microtome samples. This difference may be attributed to differences in sample weight. Dry weights of the microtome cross-sections were significantly higher than dry weights of the biopsy-sized samples in both series. (series A: mean dry weight, 3.8 and 2.5 mg, respectively; P = 0.0001; series B: 6.4 and 3.4 mg, respectively; P <0.0001). A regression of SD on mean sample weight showed that the SD decreased by 120 µg/g for every 1-mg increase in sample dry weight for dry weights between 1.4 and 7.3 mg (P = 0.008 after controlling for mean HIC).
To examine the possibility that formalin fixation and paraffin preservation has an effect on variability, SDs from the biopsy-sized samples in series A (paraffin-preserved) and series B (fresh) were compared using a regression model to control for mean HIC. Mean dry weights of the biopsy-sized samples were not significantly different in the two series, and there was no significant difference in SDs for the two kinds of sample preparations.
Means and ranges for series B HIIs are given in Table 2
. One of the six
patients had at least one HII value 1.9 (case 6, diagnosis of
alcoholic liver disease plus heterozygosity for the H63D mutation
associated with HHC).
An analysis of mean HIC or variability in HIC by diagnosis is difficult
because of the diversity of diagnoses. When series A and B were
combined to obtain a larger sample size, none of the diagnoses was
found to be associated with higher variability in HIC measurements. One
autopsy liver from a patient without hepatic disease had high HIC (1741
µg/g; Table 2
, case 7), and the SD of HIC measurements from this
liver were accordingly high, as predicted by the linear regression
results.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
These results extend the results of other reports indicating that there
often exists high intraorgan variability in hepatic iron distribution
and HIC measurements. Both Ludwig et al. (7) and Deugnier et
al. (8) noted considerable intraorgan variability in iron
staining and found wide intraorgan ranges in HII for a large proportion
of cirrhotic livers studied. Results similar to ours relating HIC means
and SDs can be reproduced from Table 2
of Villeneuve et al.
(19), who studied HIC measurements from needle-biopsy
samples from eight cirrhotic livers selected from patients with
increased serum ferritin. Villeneuve et al. found HIC means ranging
from 196 to 11 180 µg/g, with SDs ranging from 56 to 3578 µg/g.
Ambu et al. (20) also found remarkable variability in HIC in
two noncirrhotic livers from ß-thalassemic patients with iron
overload: the mean HICs were 20 631 and 13 901 µg/g with SDs of
4903 and 1976 µg/g, respectively. In an earlier study performed to
compare atomic absorption and colorimetric measurements of HIC
(16), it was noted incidentally that "sampling variation
exceeds by far the analytic variation". Intraorgan CVs ranged from
16% to 22% for healthy liver and from 50% to 80% for cirrhotic
liver measurements [estimated from Table
III of Ref.
(16)]. In another early study done to establish reference
intervals for HIC (13), it was concluded by subjective
inspection of measurements from three locations in nine healthy livers
that there was "negligible" intraorgan variability in HIC. However,
calculations from Table
II of this last report show intraorgan SDs as
high as 485 µg/g, even in healthy livers (13).
It is important to try to understand the relationship between the results above showing high HIC variability and those of Overmoyer et al. (14), who concluded that one sample of liver tissue was representative of the whole organ. Careful evaluation of the latter report shows that there is not necessarily a conflict in the results, although the conclusions drawn differ. Overmoyer et al. point out that their series included no patients with disorders that might lead to a nonuniform distribution of iron, and all had health-related concentrations of (non-heme) liver iron. An important point of the present study is that high variability is most likely to occur when liver iron is increased. Furthermore, Overmoyer et al. (14) based their conclusion that iron is uniform within livers on the finding that there was no statistically significant intraliver variation in (non-heme) iron (page 553, column 2). The absence of statistical significance does not rule out large and clinically relevant variability because confidence intervals for the variability can contain very large values if the design and analysis had low statistical power (18). Because the actual estimates of variability with confidence intervals were not reported, it is difficult to judge whether this might have been the case. Finally, Overmoyer et al. (14) found that none of the six hepatic sites at which they measured iron tends to collect more iron than in the others on average in their study population. They conclude from this that "iron is uniformly distributed within the liver". However, the lack of an average effect does not preclude random fluctuations within individual livers. Given these considerations, it appears inappropriate to make sweeping application of the conclusion by Overmoyer et al. (14) that one HIC measurement from a biopsy-sized sample is adequately representative.
Although the discussion above deals with total variability in HIC measurements, our study design in series B allowed us to separate total variability into remote and local components. Remote variability is of special interest because it can be attributed solely to biological variability, i.e., "true" heterogeneity, in iron distribution as opposed to analytic variability (10). We found that livers with HICs above ~1000 µg/g had remote variability in HIC that was both clinically and statistically significant. The sites measured here were typically 3–5 cm apart, indicating macroscopic variability within the liver. The finding of significant biological variability in HIC when mean HIC is high is consistent with the observation that abnormal iron deposition can be uneven or granular (7)(20). The increase in SD with increasing mean HIC also points to uneven iron deposition as the source of this effect. For repeated measurements where analytic variation is the main source of variation, SDs would be expected to increase little or not at all with increasing HIC (and CVs would decrease). This was not the case here.
The local variability component in this study contains analytic
variability and may contain biological variability as well. The
intraassay SD of 54 µg/g for the quality-control specimen can be
taken as an estimate of the analytic variability here. Comparison of
this figure with the local SDs in Table 3
suggests considerable local
biological variability in some cases.
One consequence of these findings is that a single high HIC measurement is especially unreliable, and the variability in the HIC measurements on any given liver cannot be predicted before some HIC measurements have been made. This latter problem might be ameliorated to some degree if a good correlate of mean HIC was known in advance of the HIC measurement. This would allow a rough estimate of mean HIC, which would allow a prediction of variability through the mean/SD relationship. Serum ferritin and serum transferrin saturation have been found to correlate poorly with HIC (21). However, biological correlation may actually be high, whereas the observed correlation is low because of unreliability of the HIC measurements (22). In this case, these serum measures may prove useful in predicting HIC variability. Further research is needed to address this question.
Variability is quantified in this study mainly in terms of the SD rather than the CV because the 95% confidence interval for the true mean HIC depends directly on the SD and the number of repeated measurements (18). Clinical interest in HIC usually centers on determining the mean for a given liver and whether it exceeds a particular threshold. The SD quantifies the variability in the measured mean and can be used directly to determine the likelihood that the true mean HIC exceeds any threshold of interest, whereas the CV cannot. In addition, comparison of CVs across laboratories is inappropriate if mean values are subject to shifts across laboratories (23), which may be the case for HIC (24). Additionally, because they are ratios of other estimated quantities, computed CVs can be very unstable estimates of the true CV, leading to low power to detect differences in CVs even when comparison is appropriate.
Microtome sampling gave more precise measurements than biopsy-sized sampling in this study, probably because of the larger size of the microtome samples. However, other differences between the microtome and biopsy-sized sampling methods could theoretically account for some of the difference found here. For example, the greater cross-sectional area sampled with the microtome might contribute. In a previous study, variability was found to be significantly lower in biopsy samples with dry weights >4 mg compared with the variability in biopsy samples with dry weights <4 mg (19). Assuming that larger weight accounts for lower variability of microtome samples, our results show that the decrease in variability has practical significance as well as statistical significance because SDs were reduced by 120 µg/g for every 1-mg increase in sample weight.
Some studies have found that formalin fixation lowers HIC (19)(13)(25), whereas others, including ours, have not (12)(15). However, our study was not designed to address this specific question. The 95% confidence interval for the percentage of difference between mean HIC from fresh and fixed samples in this study does not rule out a difference as large as 60% in either direction, indicating low statistical power to detect a difference. The high variability in HIC that we have demonstrated here can lead to very low power to detect differences in mean HIC. Use of large sample sizes and (when applicable) pairing observations from the same biopsy as in Villeneuve et al. (19) are techniques that can increase the reliability of studies comparing HIC between groups.
HIC measurements have played a major role in the diagnosis of HHC. HII
1.9 is typically used as a diagnostic criterion for HHC
(2). Other researchers have cautioned recently that HII
values can be high (e.g., 1.9) in cirrhotic livers in the absence of
HHC (7)(8). The results of this study corroborate these
previous findings, with 4 of 17 patients having at least one HII
measurement 1.9. We have shown that two factors contribute to this
finding: overall (mean) HIC concentrations can be very high in
cirrhotic livers, and high variability increases the likelihood of
finding high values for single measurements. Recently, the putative
gene associated with HHC has been identified and named HFE
(9). Two mutations associated HHC have been identified in
this gene, C282Y and H63D (9)(26). The homozygous C282Y
mutation appears to be present in 80–90% of persons of northern
European descent who have the phenotypic expression of HHC. The
availability of this genetic test has had a major impact on the
diagnosis of HHC. It should be especially useful in the setting of
end-state cirrhosis, where single HIC measurements can be highly
variable. However, it mut be remembered that the C282Y mutation may be
absent in some patients with a phenotypic expression of HHC
(9)(26). For such patients, careful clinical assessment and
family studies will remain essential to identify patients with
HHC-related hepatic iron overload.
This work was supported by NIH grants DK-38215 and DK-35816.
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  J. C. Wood, C. Enriquez, N. Ghugre, J. M. Tyzka, S. Carson, M. D. Nelson, and T. D. Coates MRI R2 and R2* mapping accurately estimates hepatic iron concentration in transfusion-dependent thalassemia and sickle cell disease patients Blood, August 15, 2005; 106(4): 1460 - 1465. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. G. St. Pierre, P. R. Clark, W. Chua-anusorn, A. J. Fleming, G. P. Jeffrey, J. K. Olynyk, P. Pootrakul, E. Robins, and R. Lindeman Noninvasive measurement and imaging of liver iron concentrations using proton magnetic resonance Blood, January 15, 2005; 105(2): 855 - 861. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. L. Nuttall, J. Palaty, and G. Lockitch Reference Limits for Copper and Iron in Liver Biopsies Ann. Clin. Lab. Sci., October 1, 2003; 33(4): 443 - 450. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L.J. Anderson, S. Holden, B. Davis, E. Prescott, C.C. Charrier, N.H. Bunce, D.N. Firmin, B. Wonke, J. Porter, J.M. Walker, et al. Cardiovascular T2-star (T2*) magnetic resonance for the early diagnosis of myocardial iron overload Eur. Heart J., December 1, 2001; 22(23): 2171 - 2179. [Abstract] [PDF]
|
|
|
|
 |
|
 G. Crisponi, R. Ambu, F. Cristiani, G. Mancosu, V. M. Nurchi, R. Pinna, and G. Faa Does Iron Concentration in a Liver Needle Biopsy Accurately Reflect Hepatic Iron Burden in {beta}-Thalassemia? Clin. Chem., August 1, 2000; 46(8): 1185 - 1188. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
and dotted
line) and biopsy-sized samples (
and solid line) in
series A. The higher position of the line for biopsy-sized samples
illustrates their higher SD relative to microtome samples.