| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
1
Referencelaboratory, National Institute of Occupational Health, Lersø Parkallé 105, DK-2100 Copenhagen, Denmark.
2
Department of Biostatistics, University of Copenhagen,
Blegdamsvej 3, DK-2200 Copenhagen N, Denmark.
a Author for correspondence. Fax 45-39-16-52-01; e-mail ahg{at}ami.dk
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Methods: A total of 21 healthy female subjects participated in the study. Using a random effects analysis of variance, we estimated CVg and total within-subject variation (CVti), i.e., the combined within-subject and analytical variation, from logarithmically transformed data. Analytical variation was subtracted from CVti to give CVi. CVi was estimated from samples taken monthly during 1 year (CViy), weekly during 1 month (CVim), and six times within 1 day (CVid).
Results: A cyclic seasonal variation was demonstrated for total cholesterol, DHEA-S, HbA1c, prolactin, and free testosterone. Within-day variation was shown for prolactin and free testosterone. The overall mean values for the group and the variability (CViy and CVg) were: 5.1 mmol/L, 5.5%, and 5.0% for total cholesterol; 6.6 µmol/L, 7.1%, and 21% for DHEA-S; 4.3%, 2.6%, and 3.3% for HbA1c/hemoglobintotal; 2.1 g/L, 5.9%, and 13% for IgA; 136 mIU/L, 23%, and 27% for prolactin; and 5.4 pmol/L, 21%, and 29% for free testosterone.
Conclusions: Collecting samples at specific hours of the day or times of the year may reduce high biological variation. Alternatively, the number of individuals may be increased and a paired study design chosen to obtain adequate statistical power.
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The effect of interventions, e.g., prevention of disease by changing the working environment or lifestyle, may be evaluated by measurement of early physiological response variables, and thereby supplement the traditional measures of effect such as reduction of mortality. The inclusion of early clinical response variables may provide a quicker evaluation of the effect of an intervention and make it easier to optimize future interventions. The planning and interpretation of measurements of early response variables in intervention requires knowledge of biological variation. The response variables selected in the present study, total cholesterol, dehydroepiandrosterone sulfate (DHEA-S), hemoglobin A1c (HbA1c), IgA, prolactin, and free testosterone, were included because the increased psychosocial stress of contemporary society may contribute to changes in concentrations of these response variables.
Diurnal variation has been described previously for DHEA-S, prolactin, and testosterone in healthy elderly women (5). Seasonal variation has been shown previously in women for DHEA-S (5) and prolactin (6), whereas other studies have failed to demonstrate seasonal variation for prolactin (5)(7) and testosterone (7). Data on within- and between-subject variation for healthy individuals for several clinically relevant response variables have been compiled by Sebastian-Gambaro et al. (8). However the data compiled for DHEA-S, HbA1c, IgA, and prolactin cover only a time span of up to 24 weeks, and data for free testosterone are not available.
The aim of the present study was to provide data on biological variation including diurnal and seasonal variation for total cholesterol, DHEA-S, HbA1c, IgA, prolactin, and free testosterone in healthy females, and in addition, to present within-subject variation for time spans of 1 day, 1 month, and 1 year, together with other useful statistics derived from measures of variation: Ii, prediction intervals, and power calculations necessary for planning future studies. The intention is to provide tools useful for design of studies of individuals in clinical settings and studies of groups of individuals.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
design
For investigation of within-year variation, a total of 12 blood
samples from each subject in study group A (n = 11) were obtained
between 0900 and 1100 once a month during 1 year (June 1995 to May
1996). For the same study group (A), a total of five samples were
collected once a week within 1 month for investigation of within-month
variation. For investigation of within-day variation, a total of six
blood samples from each subject in study group B (n = 10) were
obtained on 1 day between 0900 and 1800 (May 1996 to June 1996). The
subjects in this group were seated the whole day with a short walk once
every hour.
sample preparation
Blood was collected from the antecubital vein by venipuncture in
10-mL Vacutainer® tubes (Becton
Dickinson). Samples for HbA1c analysis
were collected in tubes containing EDTA. Samples for total
cholesterol, DHEA-S, IgA, prolactin, and free testosterone were
collected in plain tubes with no additives and after 2 h at room
temperature were centrifuged 10 min at 1731g. Serum
was separated from the sediment and stored frozen (-20 °C) for
2–12 months until assayed.
chemical analysis
Samples for analysis of total cholesterol, DHEA-S, and IgA were
held and analyzed in a few assays within 6 weeks. Analysis of
HbA1c, prolactin, and free testosterone was
carried out within 2 months of sample collection because the response
variables were stable for that period. In all cases, samples were
analyzed in the order they were collected.
The HPLC system for the analysis of HbA1c consisted of a Waters 625 LC system together with a Waters photodiode array detector model 996 and a WISP 717 autosampler for automatic injection of the samples. Millennium chromatography software was used for calculation of concentrations (Waters Associates). A cation-exchange Mono S HR 5/5 column from Pharmacia Biotech was used to separate HbA1c from other components in the samples. A competitive enzyme immunoassay in 96-well MaxiSorp Nunc-immuno plates (Nunc) was used to determine DHEA-S in serum. Each well was incubated overnight at room temperature with a saturated solution of DHEA-albumin (Steraloids) diluted (1:10 000) in phosphate-buffered saline, pH 7.3. The plates were washed in wash buffer (phosphate-buffered saline containing 0.5 mL/L Tween 20), and 50 µL of diluted (1:2000) DHEA-S calibrators in the range of 1.4–24 µmol/L or samples were applied to wells in duplicate together with 50 µL of diluted (1:5000) sheep anti-DHEA-S antiserum (Guildhay) and incubated for 1 h at room temperature with shaking. After washing, the retained antibody was incubated 1 h with peroxidase-conjugated anti-sheep immunoglobulin antibody (Dako A/S). 2,2'-Azino-di-(3-ethyl-benzthiazoline)sulfonat-6-diammonium salt (Boehringer Mannheim) was used as a chromophore (9)(10), and color development was measured with a microtiter plate photometer (R-400 Sfc; SLT Labinstruments). MultiCalc, Ver. 2.0, software from Wallac was used for calculation of concentrations. According to the manufacturer, the anti-DHEA-S antibody exhibits full cross-reactivity to DHEA, some cross-reactivity to androstenedione (<10%), and <1% cross-reactivity to all other steroids tested. DHEA-S purchased from Sigma was used for calibration. The RIAs used for determination of free testosterone and prolactin in serum were Coat-a-Count kits purchased from Diagnostic Products. A 1470 Wizard gamma counter from Wallac was used for measurement of radioactivity. Immunoturbidimetric analysis for determination of IgA and colorimetric determination of total cholesterol were carried out in a COBAS Mira Plus (Roche Diagnostic Systems). The assays used were UNIMATE 3 for IgA (cat. no. 07 3696 1) and UNIMATE 5 for total cholesterol (cat. no. 07 3663 5), both from Roche Diagnostic Systems.
The following commercially available quality-control materials were used: Con6 Immunoassay Tri-level Controls from Diagnostic Products (DHEA-S, prolactin, and free testosterone); Lyphochek Diabetes Control from Bio-Rad for HbA1c; Human Serum High and Low Control from Dako (IgA); and Human Serum Control N for Cobas from Roche (total cholesterol).
quality assurance
The analytical methods for measurement of DHEA-S and IgA had been
evaluated by a method evaluation function design according to
Christensen et al. (11) to estimate the random and
systematic effects. The method evaluation function was based on a
linear least-squares regression analysis of the measured concentration
vs the conventional true concentration of a series of method evaluation
samples containing the physiological response variable in the linear
range of the method. No statistically significant systematic effects
were found for the DHEA-S method, whereas the IgA method had a bias of
4.7%. DHEA-S purchased from Sigma was used for method evaluation.
Different batches were used for calibration and method evaluation.
According to the manufacturer, the water content (Karl Fischer method)
of both batches was 8.2%, and the purity as determined by thin layer
chromatography was >99.5%. The HbA1c method had
been evaluated by interlaboratory comparison based on 17 patient
samples. The functional model
E(Yi) =
b x E(Xi)
+ a, where b denotes the slope and a
denotes the intercept, was estimated and the approximate SD of the
estimates of a and b were calculated
(12). The analysis allowed for adjustment of differences in
variation of the two different analytical methods by use of the factor
=
y2/x2.
Based on the functional model, there was no statistically significant
difference between the two laboratories. The kit for free testosterone
had been compared by the manufacturer to a conventional equilibrium
dialysis method using 42 serum samples from female subjects. Linear
regression analysis yielded the following components: (Diagnostic
Products kit) = 0.79(dialysis) + 1.0 ng/L. No information
on traceability was given. According to the manufacturer, the kit for
measurement of prolactin was traceable to the WHO First International
Reference Preparation of Human Prolactin for Radio-immunoassay, no.
75/504 (1st IRP 75/504) and Third International Standard for Prolactin,
no. 84/500 (3rd IS 84/500). The following conversion factor was given:
1 µg = 26 mIU of prolactin. An evaluation based on the "Guide
to the Expression in Uncertainty of Measurement" (GUM) concept
has been carried out for the prolactin RIA method (13).
To show equivalence between different analytical runs, commercially available control samples for the specific physiological response variables were analyzed together with samples. Westgaard control charts were used to document that the analytical methods remained in analytical and statistical control, i.e., the precision and the trueness of all the analytical methods remained stable. An ongoing monitoring of the analytical performance against the norm of other laboratories was provided by participation in the following external quality assessment schemes by Labquality (Helsinki, Finland): Hormones and other immunochemical determination (DHEA-S, IgA, prolactin, and free testosterone), and Glycated hemoglobin (HbA1c). The target values used for evaluation of the laboratory performance were calculated as the means of all methods after exclusion of outliers (greater than ± 3 SD). Furthermore, an evaluation was provided based on grouping of the analytical methods.
statistics
Statistical analysis was carried out using
SAS® SystemTM, Ver. 6.12
(SAS Institute). AMIQAS was used for method evaluation and internal
quality control (14).
Test for outliers.
Cochran’s criterion test was used to test
for outliers in the variances within the subjects (15). In
six cases, a high within-subject variance was identified. The
single-Grubb’s test and double-Grubb’s test (15) were
applied to the measurements of the subjects with high variance to
detect any outlier observations. In three cases, the high
within-subject variance was explained by a single outlier, which was
subsequently excluded from the analysis. In three cases, no
measurements could be identified as outliers and the whole set of
measurements was included in the analysis.
Statistical model.
The data were tested for log-normal
distribution and consequently were logarithmically transformed before
statistical analysis. Data were analyzed using a variance component
model with subject as random effect by use of the Mixed procedure in
the SAS System. Variance components describing variation between
individuals (Vg), and total within-subject
variation (Vti), i.e., variation within
individuals combined with analytical variation, were estimated. The
latter includes unexplained variation. Pearson correlation coefficients
were calculated using averages for each individual.
Seasonal and within-day variation.
For analysis of periodic
variation over the year (study group A), a parametric model, cyclic
over the year using sine and cosine, was included. In this case, the
within-subject variation may be reduced depending on the fit of the
parametric function. Data were analyzed for periods of 6 and 12 months.
For description of the variation between 0800 and 1800, a
repeated-measures general linear model (GLM) was used for analysis of
within-day variation after grouping measurements into morning
(0900–1000), midday (1200–1400), and afternoon (1500–1800).
Biological variation.
Variance components were calculated for
blood samples taken within 1 year, within 1 month, and within 1 day.
Where nothing else is stated, the calculations are based on results
from samples taken during 1 year. The total within-subject variability
(CVti) and the between-subject variability
(CVg) were estimated as the square roots of the
respective variance component estimates. Estimates of the analytical
variation (CVa) were based on logarithmically
transformed data of control samples from the analytical runs in which
the samples were measured. The concentrations of the response variables
in the control samples were within the range of individual average
concentrations. The within-subject variability
(CVi) was calculated for 1 day
(CVid), 1 month (CVim), and
1 year (CViy) by subtraction of the analytical
variation using the general formula: CVi =
.
CVti includes, and therefore ideally should
exceed, the analytical variation. In case the estimated
CVti was less than CVa,
CVi was given as less than the upper value of the
99% confidence interval for the estimate of CVi.
We have followed common practice in this report, and calculated Ii as
CVti/CVg, despite the
semantic problem of the index being small when there are large
differences between individuals, as pointed out by Fraser and Harris
(3). Prediction intervals for the concentration of a future
sample taken from the same woman based on a single measurement were
calculated using the formula: 1.96 x 
x
CVti. Prediction intervals are expressed in
percentages because the underlying distribution is log-normal. The
minimal detectable differences (MIDEDIF) obtainable at the 5%
significance cutoff with the power 0.8 were estimated as:
 |
and
 |
where 1.96 and 0.84 represent the fraction of the normal
distribution [N (0,1)] corresponding to the desired significance
(
= 0.05) and the power (P = 1 - ß = 0.8). n
denotes the number of subjects in each group, and SD or
SDdif indicates the standard deviation of the
underlying normal distribution: SD =
and SDdiff =
.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
, the average concentrations ± SD of the measured
physiological response variables are plotted for all subjects in study
group A. The average concentrations of DHEA-S and free testosterone
were positively correlated between subjects [correlation coefficient
(
) = 0.64; n = 11]. Concentrations of DHEA-S were
negatively correlated with concentrations of HbA1c
( = -0.73). No systematic relationships were found when
concentrations of the selected response variables or residuals from the
model were evaluated against the number of days since the last
menstrual cycle.
|
The seasonal variations for total cholesterol, DHEA-S,
HbA1c, prolactin, and free testosterone are shown
in Fig. 2
. The concentrations of HbA1c
(P <0.001) and DHEA-S (P = 0.002) exhibited
a cyclic variation with a period of 6 months. The
HbA1c concentrations were highest in
September–November and March–May (4.4% of total hemoglobin) and
lowest in June–August and December–February (4.1% of total
hemoglobin). The DHEA-S concentrations were highest in August–October
and February–April (6.3 µmol/L) and lowest in May–July and
November–January (5.7 µmol/L). The concentrations of total
cholesterol (P <0.001), prolactin (P <0.001),
and free testosterone (P = 0.025) exhibited a cyclic
variation with a period of 12 months. Total cholesterol was highest in
January–March (5.4 mmol/L) and lowest in July–September (4.8 mmol/L),
whereas prolactin was highest in March–May (153 mIU/L) and lowest in
September–November (98 mIU/L). Free testosterone was highest in
July–September (3.9 pmol/L) and lowest in January–March (3.0 pmol/L).
The difference between the 3 months with the highest and the lowest
concentrations was 12% for total cholesterol, 9% for DHEA-S, 7% for
HbA1c, 44% for prolactin, and 31% for free
testosterone based on the predicted values. No seasonal effects
were observed for IgA.
|
Average concentrations for three periods of the day are given for the
measured physiological response variables together with data from
repeated-measures GLM in Table 1
. A statistically significant within-day variation was
demonstrated for concentrations of prolactin (P =
0.044) and free testosterone (P = 0.003). For both
physiological response variables, the concentrations were highest in
the morning. No diurnal variation was found for DHEA-S
(P = 0.37), whereas there was a tendency for IgA
(P = 0.059) to decrease during the day.
|
Table 2
shows the overall average concentration for the group and range
of individual averages, together with analytical
(CVa), within-subject
(CVi), and between-subject
(CVg) variability for the measured physiological
response variables. CVi has been estimated for
samples collected during 1 day (CVid), during 1
month (CVim) and during 1 year
(CViy).
|
The CVa/CViy ratio, the Ii,
and prediction intervals for two samples from the same woman are given
in Table 3
. The minimal detectable difference at a significance of
= 0.05 with the power P = 1 - ß = 0.80 is
shown for group sizes of 1–100 in Fig. 3
.
|
|
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Concentrations of the anabolic hormones, DHEA-S and free testosterone, were found to be positively correlated in healthy women. A corresponding correlation has been found previously by others (20). The correlation between DHEA-S and free testosterone in women may possibly be explained by a parallel excretion of DHEA-S, testosterone, and the testosterone precursor, androstenedione, because it has been reported previously that, in women, most of the DHEA-S and one-half of the testosterone and androstenedione originate from the adrenal cortex (21). Moreover, concentrations of the anabolic hormone, DHEA-S, were negatively correlated with concentrations of HbA1c, a marker for catabolic processes. The combination of low anabolic activity and high catabolic activity may be indicative of increased risk of cardiovascular disease (19).
Hormones and other physiological response variables measured in
biological fluids often exhibit a multifrequency time structures with
significant circadian or seasonal periodicity. Seasonal variations were
demonstrated in healthy women for total cholesterol, DHEA-S,
HbA1c, prolactin, and free testosterone in the
present study (Fig. 2
). Cholesterol concentrations are known to be
affected by diet, which may offer an explanation for the seasonal
variation (5). Seasonal variation with a period length of 12
months has been reported previously for HbA1c(22). The low frequency of sampling in the previous study
(four samples per individual per year) may explain the contrast to our
finding of a period of 6 months. The seasonal variations in serum
concentrations of DHEA-S and prolactin in the present study correspond
to previous findings by others
(5)(6)(23). A seasonal variation of
testosterone in men is well established, but seasonal variation has, to
our knowledge, not been examined for free testosterone in healthy
women. Circannual variation of total testosterone has been studied in
women in three studies, one of which demonstrated a statistically
significant seasonal variation (24). However, no seasonal
variation could be demonstrated in women for prolactin or total
testosterone in two recent studies, both of which were conducted with a
high degree of standardization, although there was a tendency for total
testosterone to be highest in summer and autumn
(5)(7). The main differences from our study were
the inclusion of postmenopausal women in the study groups together with
ethnic and geographic differences. The finding of seasonal variation
implies that standardizing the time of year for collection of samples
must be considered when planning a study with measurement of these five
physiological response variables. This is particularly important for
prolactin where the difference between the seasonally highest and
lowest concentrations was 44%.
A circadian rhythm has been described previously for prolactin and
total testosterone in women (5), which is in accordance with
our findings of within-day variation for the two physiological response
variables (Table 1
). For IgA, a circadian rhythm with maximal values in
the early afternoon was documented in humans; however, Halberg
et al. (25) found the highest concentration in the early
afternoon, whereas we found a tendency for concentrations of IgA to
decrease during the day. In the study by Nicolau et al. (5),
a circadian rhythm was demonstrated for DHEA-S in spring and summer but
not in winter and autumn. Investigations were carried out with the same
study group for all four seasons, indicating that the presence of a
circadian rhythm for DHEA-S may depend on the season. Because the
present study was carried out in the spring, the inability to find a
circadian rhythm is more likely related to differences in sampling
strategy, e.g., the number of hours per day or because the subjects in
the present study were seated the whole day.
Concentrations of HbA1c had very low
within-subject variation (Table 2
). In fact, the
CVti based on measurement of samples collected
within 1 month was less than CVa, and
CVim was therefore estimated to be less than the
upper limit of the 99% confidence interval for
CVim. An explanation may be that the uncertainty
of the estimates of CVti and
CVa were too high compared with the relatively
small CVi.
The within-subject variation (Table 2
) for samples collected within 1
year for prolactin (CViy = 23%) demonstrated in
this study was smaller than the corresponding within-subject variation,
CVi = 39.2%, found by Maes et al.
(7). The most likely reason is that the
CVi found by Maes et al. (7) was
estimated from data that were not logarithmically transformed. An
analysis of our data without logarithmic transformation revealed a
CViy for prolactin of 39%. In the present study,
the within-subject variation for total cholesterol for samples
collected within 1 year (CViy = 5.5%) is similar
to the within-subject variation (7.2%) reported by Costongs et al.
(26). We present new data for within-subject variation for
healthy women for free testosterone and data over a longer time span (1
year) than studies compiled by Sebastian-Gambaro et al. (8)
for DHEA-S, HbA1c, IgA, and prolactin in healthy
women.
The estimated within-subject variation may be used to set analytical goals for methods used for monitoring individual patients in clinical settings. It has been suggested that a desirable imprecision could be defined as CVa < 0.5(CVi), and an optimum performance could be defined as CVa < 0.25(CVi) (27). In the present study, CVa/CViy was <0.25 for methods for IgA and prolactin and <0.5 for total cholesterol and free testosterone. Thus, the analytical performance for these methods was adequate for clinical purposes. The CVa/CViy for methods for DHEA-S and HbA1c was >0.5, and a reduction in analytical variation appears desirable. However, both methods are within the criterion for minimum performance (CVa/CVi < 0.75); for HbA1c the analytical variation was very low (CVa = 1.8%), and it appears that for this analyte, the desirable performance standards are not obtainable with current technology and methodology.
The between-subject variations (Table 2
) for DHEA-S
(CVg = 21%) and prolactin
(CVg = 27%) found in our study are lower than
the between-subject variation, 31% and 43–65%, respectively,
presented by others (1)(7). We present new data
for healthy women for between-subject variation for free testosterone,
CVg = 29%. Previously, only data for total
testosterone have been presented (CVg = 26.9%)
(1).
The Ii was <1.4 for all physiological response variables (Table 3
).
According to a recent report by Petersen et al. (28), this
has no significance for the use of population-based reference
intervals, when a single measurement is evaluated. However, the low Ii
makes it important to stratify populations to obtain separate reference
intervals for subpopulations and to accumulate data from samples from
the same individual where possible (28).
The prediction intervals presented in Table 3
indicate the intervals
within which the concentration of a physiological response variable in
a future sample taken from the same woman is expected to be. For
example, if a second sample taken from the same healthy woman within 1
year gives a result that is within 69–145% of an earlier measurement
of total cholesterol, it may simply reflect the underlying biological
and analytical variation.
When planning a study on groups of individuals, larger groups are
required when using a non-paired design compared with a paired design,
if the same minimal detectable difference is desired (Fig. 3
). This is
partly because the between-subject variation is included in the power
calculations for non-paired designs. The effect is observed when
comparing the curves for IgA and total cholesterol. Although the same
number of subjects are required for a given minimal detectable
difference for a paired design, in a non-paired design, more subjects
are required for measurement of IgA, which exhibits large
between-subject variation compared with total cholesterol. Measurement
of prolactin and free testosterone requires the largest number of
subjects in each group for a given minimal detectable difference,
mainly because of high biological variation for prolactin and high
analytical variation for free testosterone.
In conclusion, in a study that includes physiological response variables on groups of individuals, optimal study design is important to reduce both the analytical and the biological variation. As shown in the present study, one way to reduce the biological variation is to restrict sample collection to specific hours of the day (prolactin and free testosterone) or times of the year (total cholesterol, DHEA-S, HbA1c, prolactin, and free testosterone) to avoid within-day and seasonal variation of the selected physiological response variables. However, such variation often has to be included in the study design. Correspondingly, the number of individuals in each group often is increased and a paired study design is chosen to obtain adequate statistical power.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  P. Desmeules, J. Cousineau, and P. Allard Biological variation of glycated haemoglobin in a paediatric population and its application to calculation of significant change between results Ann Clin Biochem, January 1, 2010; 47(1): 35 - 38. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W S A Smellie What is a significant difference between sequential laboratory results?Calf muscle pain can indicate localised vasculitis J. Clin. Pathol., April 1, 2008; 61(4): 419 - 425. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Bjornerem, B. Straume, P. Oian, and G. K. R. Berntsen Seasonal Variation of Estradiol, Follicle Stimulating Hormone, and Dehydroepiandrosterone Sulfate in Women and Men J. Clin. Endocrinol. Metab., October 1, 2006; 91(10): 3798 - 3802. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Luraschi, S. Brambilla, R. Mozzi, G. Cattozzo, and C. Franzini Monitoring Analytical Quality in Routine Glycohemoglobin Measurements Clin. Chem., September 1, 2002; 48(9): 1594 - 1597. [Full Text] [PDF]
|
|
|
|
 |
|
 A. K. Cheung, G. Yan, T. Greene, J. T. Daugirdas, J. T. Dwyer, N. W. Levin, D. B. Ornt, G. Schulman, and G. Eknoyan Seasonal Variations in Clinical and Laboratory Variables among Chronic Hemodialysis Patients J. Am. Soc. Nephrol., September 1, 2002; 13(9): 2345 - 2352. [Abstract] [Full Text] [PDF]
|
|
Read all eLetters
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
,
DHEA-S; 
, HbA1c; 
, IgA, +, prolactin; 
, free
testosterone.