| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Test Utilization and Outcomes |
Departments of
1
Clinical Chemistry,
2
Hematology, and
3
Internal Medicine III, University Hospital Rotterdam, Dr. Molewaterplein 40, 3015 GD Rotterdam, the Netherlands.
4
Department of Epidemiology and Biostatistics, Erasmus
University, Dr. Molewaterplein 40, 3015 GD Rotterdam, the Netherlands.
a Author for correspondence. Fax Int. +31–104367894; e-mail boersma{at}ckcl.azr.nl
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
Hcy is derived from the intracellular metabolism of methionine and is exported into plasma where it circulates primarily in oxidized form (i.e., Hcy and Cys–Hcy disulfide) and bound to proteins. Concentrations of total Hcy are increased in 15–40% of patients with coronary, cerebral, or peripheral arterial diseases (2)(8). Mechanisms that may relate to the pathogenesis of atherothrombosis in hyperhomocyst(e)inemia are the change in hemostatic condition from antithrombotic to thrombogenic, the increased incorporation of Lp(a) into fibrin, and the increased oxidation of LDL (2)(8).
The oxidation hypothesis of atherosclerotic disease emphasizes the causal role of oxidized lipoproteins in atherogenesis (9). If decreased antioxidant concentrations accelerate lipoprotein oxidation and hence atherosclerotic disease, detection of a decreased total antioxidant status (TAOS), as measured by antioxidant-mediated quenching of the absorbance of a radical cation (10), may prove to be a valuable test.
To date, data on biological variation are absent for Hcy (and Cys) and TAOS, and abundant but conflicting for Lp(a) (11)(12)(13)(14)(15)(16)(17)(18)(19). In the case of Lp(a), earlier estimates of the biological intraindividual CV (CVb) showed a 7% week-to-week variation (13), whereas in the ARIC study (14) the CVb was estimated to be as low as 2.9%. More recent studies (15)(16)(17) reported average CVbs of 7.6%, 10%, and 18%, whereas Marcovina et al. (18)(19) found the estimated CVb to be highly variable (range 3–51%) and to have a systematic inverse relation with the Lp(a) concentration. Possibly, the highly skewed Lp(a) distribution and the 1000-fold interindividual spread in blood Lp(a) concentrations in Caucasians, in combination with the investigation of rather limited numbers of individuals, may have caused apparently conflicting data on intraindividual biological variation of Lp(a).
In the present study a comprehensive biological variability study for these analytes was carried out in a rather large group of healthy sex- and age-matched Caucasians. To ensure inclusion of an adequate number of individuals with high serum Lp(a), stable outpatients from the Lipid Clinic who repeatedly had Lp(a) mass concentrations >300 mg/L were enrolled. An experimental protocol that minimized preanalytical and analytical sources of variability was used. The aims were: (a) to estimate, in healthy and in chronically diseased but stable Caucasians, the biological variation of Lp(a), Hcy, Cys, and TAOS around the intraindividual homeostatic setpoints as well as the relation between the biological intraindividual variation and the analyte concentration; (b) to determine desirable analytical goals for these new or potential risk factors on the basis of intra- and interindividual biological variation (11); (c) to gain a clear understanding of the value of conventional population-based reference values for these analytes (12); (d) to gain a clear insight into significant and insignificant analyte changes in serial specimens (12)(20); and (e) to calculate the minimum number of serial specimens needed to determine the "true" analyte concentration (21).
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
Outpatients with hyper-Lp(a) lipoproteinemia.
Twelve
Caucasian outpatients (5 men and 7 women; age range 22 to 69 years)
from the Lipid Clinic of the University Hospital Rotterdam with an
Lp(a) mass >300 mg/L were included. All patients were on a
lipid-lowering diet for at least 3 months before enrollment.
study protocol
Blood was collected biweekly at each of four visits per
individual. Subjects were seen in standardized format at each occasion,
i.e., they were asked to fast for 10 to 12 h before each visit,
and to maintain their diet, life-style, and possible medication
throughout the evaluation period. The design and intention of the study
were thoroughly explained to all subjects and informed consent was
obtained. All study subjects were interrogated on each of the four
visits by one of two physicians who checked, by means of a predefined
questionnaire, whether diet, life-style, smoking and drinking habits,
and possible medication were maintained throughout the study. Height
and weight were measured at the first visit, whereas body weight was
checked at each subsequent visit. Fertile women were questioned about
possible new pregnancies.
Venous blood was collected in the upright sitting position, immediately after individuals had been seated. Sampling was done between 0800 and 1000 by a single phlebotomist. Whole blood was collected for Lp(a) and TAOS (22), whereas EDTA blood was collected for Hcy and Cys analyses (23). Except for serum TAOS, which has limited stability according to the manufacturer, all four samples from one individual were analyzed in one run at the end of the study, to omit between-run analytical variation. The study protocol was approved by the Medical Ethical Committee of the University Hospital.
specimen handling and storage
A strictly predefined protocol was used for specimen preparation:
EDTA (1.5 g/L) blood tubes were put on crushed ice immediately after
blood collection (23). Whole-blood tubes were kept at room
temperature until clotting took place. Both whole-blood and EDTA tubes
were centrifuged at 4 °C (10 min, 1500g) within 1 h
after blood drawing (23). Serum and EDTA plasma were
separated from the cells immediately after centrifugation. TAOS
determinations were performed the same day, whereas the other aliquots
were stored at -70 °C for combined analysis of all samples from one
individual at the end of the study. Blood specimens for this study were
gathered during a 3-month period.
lp(a), hcy, cys, and taos measurements
Lp(a) was measured in serum with an anti-apo(a) polyclonal capture
ELISA from Biopool [TintElize lipoprotein(a), cat. no. 610220]
(4)(5)(6). Total Hcy and Cys were measured in EDTA plasma by
using a rapid, isocratic HPLC method (24)(25).
Serum TAOS was applied on a Hitachi 911 analyzer (Boehringer Mannheim)
with the Randox kit and calibrator control material (cat. nos. NX2332
and NX2331, respectively). The TAOS assay measures the
antioxidant-mediated quenching of the absorbance of a radical cation
(10).
TAOS analyses were done in duplicate at the day of sample collection, whereas Lp(a), Hcy, and Cys analyses from one individual were performed in duplicate within one run at the end of the study. The maximum sample storage time for frozen aliquots was 5 months. To further minimize analytical variation, a single technician performed all the assays and single lots of reagents were used. The between-run CV for the TAOS control material was 4.3% (n = 15 runs), corresponding to a between-run variance of 0.0025.
statistics
If a quantity X exhibits biological variation such that
the standard deviation in an individual is proportional to the
homeostatic setpoint of the individual, then the quantity X
is said to have a constant CV. The CV is a parameter expressing the
proportionality of the standard deviation to the homeostatic setpoint.
It can be defined similarly for the analytical variation around the
true value of a specimen. When the CV of X is small, then it
is well approximated by the standard deviation (
) of lnX,
with few assumptions required regarding the distribution of
X or lnX. A proof of this is given in the
Appendix.
The smaller the CV, the better the approximation will be. From a
statistical point of view, it is better to directly estimate the CV by
estimating the of lnX as a single parameter (be it an
approximation) than as a ratio of an estimated standard deviation to an
estimated mean. For CVs <0.4 (i.e., <40%), the approximation by
of lnX is good enough for practical purposes, considering
the efficiency gained by estimating it as a single parameter. In
summary, analytical CV (CVa) and CVb can be
estimated as a and b, respectively, after
log transformation of the measured analyte values.
Hitherto, means, variances (2), and CVs were
estimated by using standard formulas. In case of TAOS, serial specimens
were analyzed in separate runs, and thus 2b
was an estimator for the total of within-subject and between-run
variance. To obtain a proper estimate of the within-subject variance
for TAOS, the between-run variance derived from the TAOS control
material was subtracted from 2b.
Indices were derived from CVa and CVb data
(11)(12)(20)(21)(26)
as follows: analytical goal for imprecision (AG CVa)
1/2CVb; analytical goal for bias (AG bias)
1/4(CVb2 +
CVg2)[1/2]
(CVg is the between-subject or interindividual biological
CV); index of individuality = (CVb2 +
CVa2)[1/2]/CVg;
reference change value (RCV) or critical difference =
2.77(CVa2 +
CVb2)[1/2]; number of
specimens required to ensure with 95% confidence that the mean result
is within ±5% of the individual's homeostatic setpoint [NS
(±5%)] = 1.962[(CVa2 +
CVb2)/25]; number of specimens required to
ensure with 95% confidence that the mean result is within ±10%
of the individual's homeostatic setpoint [NS (±10%)] =
1.962[(CVa2 +
CVb2)/100].
data analysis
Data analysis was done separately for healthy subjects and
outpatients after removing one sample from the TAOS data set because of
in vitro hemolysis. All analyte results were transformed with natural
logarithms. Variances calculated from the logarithmically transformed
data were multiplied by 10 000 to convert the estimated standard
deviations () to the CVs expressed in percent.
Differences in biological variation between men and women in each group
were tested by calculating ratios of the pooled variance of analytical
and within-subject variance from one gender to the other gender. These
calculated F-ratios were compared with the critical
F-values (
= 0.05). Differences in biological variability
between the healthy subject and patient groups were calculated in a
similar way. Contribution of analytical variability to total test
variability was calculated as:
{[(CVa2/CVb2) +
1][1/2] - 1} x 100% (11).
Concentration dependency of CVb and CVa vs the
average analyte concentration was studied with linear regression
analysis: lnCV = ln - ß lnmean + residual. The null
hypothesis checked was that the slope ß would be equal to zero.
Confidence intervals (95%) for serum Lp(a) were calculated as ±
1.96 [(analyte concentration x 0.01 x
CVa)2 + (analyte concentration x
0.01 x
CVb)2][1/2]/number
of specimens[1/2]. Overall, a significance
level of P 0.05 was adopted.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
summarizes the baseline characteristics of the study subjects
as well as the analyte concentrations at the first visit. Mean
intraindividual weight changes varied between -0.06 and +0.98 kg
between subsequent visits (data not shown). Table 2
summarizes mean CVa, CVb, and
CVg, and percentile distributions of CVb for
Lp(a), Hcy, and Cys, whereas for TAOS only mean CVb and
CVg are presented. In outpatients, similar average
CVb estimates were found, except for Lp(a) (7.5% vs
20.0%; F = 7.07; df1 = 162,
df2 = 36; P <0.0001). CVb did
not differ between men and women for any of the analytes, either in
healthy subjects (F <1.44 at = 0.05;
df1 = 81, df2 = 81) or in outpatients
(F <2.18 at = 0.05; df1 = 15,
df2 = 21 or F <2.37 at = 0.05;
df1 = 21, df2 = 15) (data not shown).
On the basis of average CVa and CVb values, we
found that <10% of the total test variability was analytical for
Lp(a), Hcy, and Cys, whereas for TAOS up to 98% of the observed test
variability was analytical. Table 3
displays the average derived indices for each analyte studied.
Analytical goals for imprecision, based on average CVb,
were achieved for all analytes except for TAOS
(11)(12)(26). All analytes had
marked individuality (index <0.6) (12), demonstrating
that the use of population-based reference values is inadequate for
their interpretation. In the healthy subject group average, applicable
differences required for two results to be significantly different at
= 0.05 were 60% for Lp(a), 28% for Hcy, 17% for Cys, and 9% for
TAOS (20). Further, Table 3
reveals that for proper
assessment of coronary artery disease risk in the population by means
of the assays used, multiple serial specimens are needed if the
observed value should be within, e.g., 5% of the true value.
|
|
|
Concentration dependency of ln(CVa) and
ln(CVb) was studied in the healthy subject group for all
analytes (data not shown). None of the parameters showed concentration
dependency, except ln[Lp(a)], the slopes being (borderline)
significantly different from zero (P = 0.04 for
CVa vs subject mean per visit; P = 0.07 for
CVb vs overall subject mean). After taking the
antilogarithm the equations were: CVb =
{[42.9/Lp(a)0.31] x exp[(0.82)2/2]}
and CVa = {[12.2/Lp(a)0.35] x
exp[(0.93)2/2]}. Notable is that the slopes were
comparable, whereas the intercept with the y-axis was 3.5
times higher for CVb compared with CVa. Fig. 1
illustrates the impact of changing CVa and
CVb values across the Lp(a) concentration range on the
Lp(a) test variability, analyzing one, three, and five serial
specimens. If at least three serial specimens are analyzed per
individual, the observed Lp(a) result is within ±10% of the true
value if the Lp(a) concentration is >500 mg/L. Below 500 mg/L the test
uncertainty runs up quickly because of increasing biological and
analytical CVs. If only one specimen is analyzed, the observed Lp(a)
value is within ±15–20% of the true value, even at Lp(a)
concentrations >500 mg/L. Below 500 mg/L the confidence limits
increase even more dramatically. At 300 mg/L, an internationally
recognized though arbitrarily defined cutpoint for Lp(a)
(1), the confidence intervals range between ±21%,
±12%, and ±9% depending on whether one, three, or five serial
specimens, respectively, were analyzed.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
In healthy Caucasians average CVbs were 20.0% for
Lp(a), 9.4% for Hcy, 5.9% for Cys, and 2.8% for TAOS (Table 2
), mean
CVbs being similar in men and women for all analytes
studied. In the outpatient group, comparable CVb estimates
were found for Hcy, Cys, and TAOS but not for Lp(a) (7.5% in
outpatients vs 20.0% in healthy controls). Moreover, in accordance
with Marcovina et al. (18)(19), a systematic
inverse relation was demonstrated between CVa and
CVb and Lp(a) concentration. For the other analytes, no
concentration dependency was found. Our data (a) illustrate
the inadequacy of using average CVb and CVa
values for Lp(a); (b) explain the controversy in the
literature regarding intraindividual biological variability of Lp(a)
(13)(14)(15)(16)(17) and corroborate the findings of Marcovina et al.
(18)(19); and (c) underscore the
fact that the intraindividual biological variability of Lp(a) is
greater than previously believed, especially in the low concentration
range (1).
Data on interindividual biological variation (CVg) are
presented if meaningful (Table 2
). In general, interperson variability
is determined by age, sex, diet, and genetics. In the case of Lp(a),
interperson variability is mainly determined by genetics
(1), whereas diet and genetics may influence plasma Hcy
concentrations (2)(8)(23). Because
the outpatient group represents a selected high-risk group, including
individuals with both increased Lp(a) and Hcy concentrations (Table 1
),
CVg data cannot be extrapolated from one study to another.
Consequently, CVg data are specific for the population
studied and therefore are of limited value.
Several indices have been derived from the biological variability
study. First, analytical goals for imprecision, having been the subject
of a variety of approaches (11)(12), were
calculated. In this study the approach of Harris (26) was
used, which states that maximum allowable analytical imprecision should
be 1/2CVb. Average goals for CVa were
met for Hcy and Cys, but not for TAOS (Table 3
), average goals for
CVa being similar for Hcy and Cys in either study group. In
contrast, for Lp(a) a single mean CVa goal was not
ubiquitously valid because of concentration dependency of the
CVb estimates. From the estimated regression lines that
described the relation between CVa and CVb and
Lp(a) concentrations, a fairly constant 3.5-fold difference between
CVa and CVb could be demonstrated across the
entire Lp(a) concentration range, signifying that the imprecision of
the Lp(a) Biopool kit used is adequate at all Lp(a) concentrations.
Therefore, we disagree with Pagani and Panteghini
(15)(21), who claimed that in practice the
analytical goal for Lp(a) cannot be achieved with current assays.
Notwithstanding the lack of international Lp(a) standardization and the
fact that a different ELISA was used by these authors, the discrepancy
with their data can be explained by the inappropriate use of average
CVb and CVa estimates. Finally, although the
CVa goals were met for Lp(a) and Hcy (and Cys), one may
object that the assays were performed under optimal conditions of
variance as between-day variation was omitted. Yet between-day
analytical CVs from routine practice in our laboratory of 7.1% at 67
mg/L, 4.2% at 213 mg/L, and 5.1% at 379 mg/L were achieved for Lp(a),
whereas between-day CVs of 4.0% at 19.5 µmol/L and 3.2% at 52.2
µmol/L were obtained for Hcy, confirming the practical attainment of
the analytical goals.
Second, desirable goals for average analytical bias were calculated
(Table 3
) (11)(12). Documenting bias of
routine assays necessitates the development of reference and (or)
definitive methods and standardization programs for the analytes
studied. So far, no international standardization has been reached.
Third, the utility of conventional population-based reference values
was assessed by calculating an index of individuality in the
healthy subject group (Table 3
). The index gives a philosophical view
on the interpretation of analyte data measured in healthy individuals
and pathological changes in relation to reference intervals
(12). If the index is <0.6, then the use of reference
intervals is of limited value in the detection of unusual individual
results; if the index is >1.4, then reference values are of
significant utility. In this study, all analytes had marked
individuality, demonstrating that the use of population-based reference
values is inadequate for their interpretation. This favors the adoption
of cutpoints based upon relative risk of coronary artery disease.
Fourth, biological in addition to analytical variation data are also
used for the critical evaluation of the significance of changes in
results obtained from analysis of serial specimens
(12)(20). To interpret serial results
objectively it is necessary to know the change that must occur before
significance can be claimed. This RCV depends on both analytical and
intraindividual biological variation, and holds only if all individuals
have the same within-subject variation and if the analytical variation
is constant across the concentration range. For Hcy, Cys, and TAOS,
average RCVs for detecting significant changes in 50% of the
individuals are presented in Table 3
. In view of the concentration
dependency of CVa and CVb for Lp(a), i.e., the
observed variance reduction with increasing Lp(a) concentrations, mean
CVa and CVb cannot be the basis for calculating
the critical difference that is generally applicable in all
individuals. So far, critical differences for Lp(a) were reported by
one group (15), the RCV being estimated as 29%, on the
basis of a mean CVa of 7.4% and a mean CVb of
7.6%. According to our findings, critical differences for Lp(a) should
be calculated on the basis of individual CVa and
CVb values.
Finally, from the variation data obtained in this study one can
estimate the number of specimens required to determine the
individual's true homeostatic setpoint value (21). Again,
simple recommendations regarding the average number of specimens needed
can be made for Hcy, Cys, and TAOS (Table 3
), while such an approach is
not valid for Lp(a) (Fig. 1
). However, in light of the enormous
interindividual concentration differences, we agree with Marcovina et
al. (18)(19) that the CVb of Lp(a)
is not likely to be an important contributor to the misclassification
of an individual's risk, unless the value is near the cutpoint of
enhanced coronary artery disease risk.
Overall, the findings in the present study demonstrate that the understanding of the magnitude of the physiological variations that occur in Lp(a), Hcy, Cys, and TAOS concentrations in serum or plasma is indispensable for proper use of these laboratory data for risk classification of patients with coronary artery disease. In essence, we demonstrated that average CVa and CVb estimates and mean derived indices are valid for Hcy, Cys, and TAOS, whereas individual values should be used for Lp(a). Second, the analytical performance of the Lp(a), Hcy, and Cys assays used is acceptable, taking into consideration the biological variation of these parameters, whereas the performance of the TAOS assay was insufficient.
|
Appendix 1 |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
. This standard deviation is assumed to be small enough (e.g.,
<0.4) that the probability of negative or zero values of U
must be zero. No further assumptions about the distribution of
U are made.
CV(X) = SD(X)/E(X) =
[(T)/T] = , which is a constant single
parameter. An approximate estimator for can be obtained by first
taking the natural logarithm: lnX = lnT +
lnU, and then applying the so-called "delta method"
(27) to the variance of lnU:
 |
2, because
[d(ln U)/dU]2 developed
for E(U) = 1 equals unity (27).
Hence, CV(X) can approximately be estimated by
SD(lnX) within the same subject or within the same specimen
under very general conditions. |
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionAppendix 1References |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  L. R. Solomon Cobalamin-responsive disorders in the ambulatory care setting: unreliability of cobalamin, methylmalonic acid, and homocysteine testing Blood, February 1, 2005; 105(3): 978 - 985. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. J. Huijgen, F. P.W. Tegelaers, C. H.H. Schoenmakers, C. J. Pronk-Admiraal, and S. Ekema Multicenter Analytical Evaluation of an Enzymatic Method for the Measurement of Plasma Homocysteine and Comparison with HPLC and Immunochemistry Clin. Chem., May 1, 2004; 50(5): 937 - 940. [Full Text] [PDF]
|
|
|
|
|
|
M. R. Fokkema, M. F. Gilissen, J. J. van Doormaal, M. Volmer, I. P. Kema, and F. A.J. Muskiet Fasting vs Nonfasting Plasma Homocysteine Concentrations for Diagnosis of Hyperhomocysteinemia Clin. Chem., May 1, 2003; 49(5): 818 - 821. [Full Text] [PDF]
|
|
|
|
 |
|
 A. De Bree, W. M. M. Verschuren, D. Kromhout, L. A. J. Kluijtmans, and H. J. Blom Homocysteine Determinants and the Evidence to What Extent Homocysteine Determines the Risk of Coronary Heart Disease Pharmacol. Rev., December 1, 2002; 54(4): 599 - 618. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Nurk, G. S. Tell, S. E. Vollset, O. Nygard, H. Refsum, and P. M. Ueland Plasma Total Homocysteine and Hospitalizations for Cardiovascular Disease: The Hordaland Homocysteine Study Arch Intern Med, June 24, 2002; 162(12): 1374 - 1381. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Voortman, A. Melse-Boonstra, J. M. Schulz, J. Burema, M. B. Katan, and P. Verhoef Optimal Time Interval between Repeated Blood Sampling for Measurements of Total Homocysteine in Healthy Individuals Clin. Chem., October 1, 2001; 47(10): 1839 - 1841. [Full Text] [PDF]
|
|
|
|
|
|
M. C. McKinley, J.J. Strain, J. McPartlin, J. M. Scott, and H. McNulty Plasma Homocysteine Is Not Subject to Seasonal Variation Clin. Chem., August 1, 2001; 47(8): 1430 - 1436. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Laskowska-Klita, E. Szymczak, and B. Radomyska Serum Homocysteine and Lipoprotein (a) Concentrations in Hypercholesterolemic and Normocholesterolemic Children Clinical Pediatrics, March 1, 2001; 40(3): 149 - 154. [Abstract] [PDF]
|
|
|
|
|
|
K. Rasmussen, J. Moller, and M. Lyngbak Within-Person Variation of Plasma Homocysteine and Effects of Posture and Tourniquet Application Clin. Chem., October 1, 1999; 45(10): 1850 - 1855. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. Thirup and S. Ekelund Day-to-Day, Postprandial, and Orthostatic Variation of Total Plasma Homocysteine Clin. Chem., August 1, 1999; 45(8): 1280 - 1283. [Full Text] [PDF]
|
|
|
|
|
|
J. B. Ubbink, R. Delport, R. Riezler, and W.J. H. Vermaak Comparison of Three Different Plasma Homocysteine Assays with Gas Chromatography–Mass Spectrometry Clin. Chem., May 1, 1999; 45(5): 670 - 675. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Causse, N. Siri, H. Bellet, S. Champagne, C. Bayle, P. Valdiguie, R. Salvayre, and F. Couderc Plasma Homocysteine Determined by Capillary Electrophoresis with Laser-induced Fluorescence Detection Clin. Chem., March 1, 1999; 45(3): 412 - 414. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
