| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Original Studies |
Departments of Medicine (R.D.H., K.D.B., G.G.H.), Community Medicine and Statistics (G.R.H.), Neurology (J.E.R.), and Mathematics (I.C., G.G.), West Virginia University, Morgantown, West Virginia 26506-9159; and Department of Medicine (S.M.M., A.L.), University of Washington, Seattle Washington 98195-7710
Address correspondence and requests for reprints to: Robert D. Hoeldtke, M.D., Ph.D., Department of Medicine, West Virginia University Medical School, P.O. Box 9159, Morgantown, West Virginia 26506-9159. E-mail: rhoeldtke{at}hsc.wvu.edu
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Patients with high GAD65Ab had worse glycemic control and higher insulin requirements. Patients with high GAD65Ab had slower motor nerve conduction velocities in the median, ulnar, and peroneal nerves (P < 0.025 for each nerve). The mean motor nerve conduction velocity Z scores at the time of the third evaluation was 0.341 ± 0.25 for the low GAD65Ab patients and -0.600 ± 0.25 for the high GAD65Ab patients (P < 0.01). Similar differences between the low and high GAD65Ab groups were observed for F wave latencies, thermal threshold detection, and cardiovascular autonomic function. The composite peripheral nerve function Z scores in the low GAD65Ab patients were 0.62 ± 11, 0.71 ± 0.19, and 0.21 ± 0.14 at the first, second, and third evaluations, significantly different from those in the high GAD65Ab patients in whom they were -0.35 ± 0.15, -0.46 ± 0.18, and -0.42 ± 0.16 (P < 0.001).
In summary, GAD65Ab in patients with recent onset type 1 diabetes are associated with worse glycemic control and slightly worse peripheral nerve function. Although the latter remained within normal limits and none of the patients had clinical neuropathy, the GAD65Ab-related differences in composite peripheral nerve function were highly significant (P < 0.001) and could not be attributed to GAD65Ab-related differences in glycemic control.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Thirty-seven patients (10 males and 27 females) with type 1 diabetes enrolled 2–22 months after diagnosis in a longitudinal study of peripheral nerve function. Patients with symptoms of autonomic or somatosensory neuropathy, other systemic illnesses, a history of alcoholism, or psychiatric disease were excluded. Patients who needed to take vasodilator or other drugs that might interfere with autonomic function testing were also excluded. All patients were taught home glucose monitoring and instructed to adjust their insulin dose as necessary to maintain optimal glycemic control. Thirty-six of the patients underwent three annual evaluations; one patient withdrew from the study after the second year. The time interval between evaluations was 1 yr for most patients. Delays in scheduling prolonged the study in four patients, but the time interval between the first and third evaluation was always less than 3 yr.
Autonomic function tests and quantitative sensory testing were also performed in 41 age- and sex-matched controls to provide a basis of comparison with the diabetic patients. Nerve conduction velocity measurements and related electrodiagnostic studies were not performed in the controls because the limits of normality were established.
The research protocol was approved by the Institutional Review Board of West Virginia University Hospital, and informed consent was obtained from the participants.
Annual evaluations
The diabetic patients were admitted to a bed designated for research at West Virginia University Hospital to control their dietary intake, activity, and glucose prior to and during the neurological testing. Glucose was monitored before each meal and snack and at 0300 h, and insulin adjustments were made, as needed. Patients were administered a weight-maintaining diet containing 130 meq sodium daily. Caffeine, aspirin, and cigarette smoking were not allowed on the morning of the tests because of possible effects on autonomic function. The control subjects were also admitted to the hospital, administered the same diet, and subjected to the same restrictions.
Large fiber somatosensory function
Nerve conduction studies were performed with a TD-20 TECCA electromyograph (TECCA Corp., Pleasantville, NY). Skin temperature was measured and maintained above 31 C. Motor nerve conduction velocities, compound action potentials, distal latencies, and F wave latencies were measured in the median, ulnar, and peroneal nerves. Sensory nerve amplitudes and latencies were measured in the median, ulnar, and sural nerves.
Cardiovascular autonomic function
Beat to beat variation with deep breathing. Patients were studied in the supine posture after relaxing comfortably for at least 10 min. Heart rate was monitored electrocardiographically while they breathed slowly (5 sec inspiration/5 sec expiration) as deeply as possible for 5 min. The difference between the maximum and minimum instantaneous heart rates (max-min) reflects the integrity of the parasympathetic innervation of the heart (8). In addition, vector analysis of the instantaneous heart rate was performed, and the mean circular resultant was determined. This alternative index of heart rate variability minimizes error introduced by variation in intrinsic heart rate or ectopic cardiac beats (9).
Heart rate response to the valsalva maneuver. The heart rate was monitored electrocardiographically while the patients were supine and instructed to expire into a sphygmomanometer until a pressure of 40 mm Hg was maintained for 20 sec. The Valsalva ratio was calculated by dividing the maximal instantaneous heart rate during the maneuver by the minimal heart rate observed after release. The test was performed twice, and the average result was calculated. A normal response (ratio greater than 1.15) indicates that the baroreceptor reflex and the efferent limb of the sympathetic nervous system are intact.
Power spectral analysis. Instantaneous heart rate was measured with a Hokanson electrocardiograph monitor, which allows each R-R interval to be recorded into a computer program (DE Hokanson Inc., Bellevue, WA). Rhythm strips were run simultaneously to exclude premature contractions and other rhythm disturbances. The R-R interval data file was then interpolated and averaged, and power spectral analysis was performed using the Fast Fourier Transform (10). Respiration was monitored simultaneously with a thermistor so that spurious low frequency spectra resulting from slow breathing or sighing could be eliminated. High-frequency spectra (0.15–0.40 Hz) indicate parasympathetic cardiac innervation, whereas intermediate frequency spectra (0.04–0.15 Hz) were interpreted to signify predominately sympathetic modulation.
Small fiber somatosensory function. Quantitative sensory
testing was used to assess small and thinly myelinated A 
fibers,
which convey cold sensation, and C fibers, which convey heat
(11). The hot and cold stimuli were applied to the dorsal
aspect of the feet and the wrist, and participants were asked to
distinguish between progressively small thermal stimuli until they were
no longer able to detect the change in temperature. Specific thermal
thresholds were then determined by a microprocessor-controlled forced
choice technique (Neurolink, East Lyme, CT). Thresholds were determined
on two separate days, and the average performance was calculated for
each of the four parameters of interest (heat threshold feet, cold
feet, heat wrist, and cold wrist).
Biochemical measurements
Glycosylated hemoglobin. Glycosylated hemoglobin (hemoglobin A1) was measured by agar gel electrophoresis one to four times a year (12). The reference range for the nondiabetic population was 4.7–7.3%.
GAD65Ab. GAD65Ab was measured in serum samples frozen at -80 C and coded before the analysis in a radiobinding immunoassay (13, 14). Briefly, in the assay, the 35S-GAD65 was produced in an in vitro-coupled transcription and translation system with SP6 RNA polymerase and nuclease treated rabbit reticulocyte lysate (Promega, Madison, WI). In each analysis, 20,000 cpm and trichloroacetic acid precipitable 35S-GAD65, in triplicate, was diluted in 60 µL immunoprecipitation buffer [20 mM Tris, 150 mM NaC1, 0.15% (v/v) Tween 20, 0.1% (w/v) aprotinin, 10 mM benzamidine, and 0.1% (w/v) BSA (pH 7.4)] before the addition of 2.5 µL human serum (final serum dilution, 1:25). Free 35S-GAD65 was separated from the antibody-bound tracer by protein A-Sepharose and several washes in 20 mM Tris, 150 mM NaC1, 0.1% (v/v) Tween 20, and 0.1% (w/v) BSA buffer (pH 7.4) using 96-well plates, containing 0.65 µM hydrophilic PVDF filters (Millipore, Bedford, MA). The immunoprecipitated radioactivity was counted after transferring the filters to glass scintillation vials using a Multiscreen Multiple Punch System (Millipore).
The levels of GAD65Ab were expressed as a relative (GAD65 index) using one positive serum (Juvenile Diabetes Foundation World Standard for islet cell antibody) and three negative standard sera from healthy subjects, as described previously (13, 14). The upper level of normal was evaluated as mean + 3 SD sera from healthy individuals included in each assay. The interassay coefficient or variation was 14.7%, as evaluated in the positive standard serum in 53 consecutive assays. In the First and Second International GAD Autoantibody Workshops our GAD65Ab assay showed 100% and 82% sensitivity and 100% and 96% specificity, respectively.
IA-2Ab. Antibodies to the islet cell antigen IA-2 were measured under identical conditions as described for GAD65Ab (15, 16). The plasmid containing the cDNA for the cytoplasmic portion of islet antigen 512 was kindly donated by Dr. G. Eisenbart (Barbara Davis Research Center, Denver, CO). The same Juvenile Diabetes Foundation standard serum and control sera as in the GAD65Ab assay was used to correct the interassay coefficient of variation by calculating a 1A-2Ab index for each sample. The interassay coefficient of variation for the positive standard serum was 19.0% as determined in 38 consecutive assays.
Antibodies to the autonomic nervous system. Antibodies to the vagus nerve and sympathetic ganglia were assayed with a complement fixation indirect immunofluorescence technique using sections of the vagus nerve and superior cervical/ganglia of the rabbit (2). Following the incubation of the nerve tissue with human serum and subsequent washing, incubation with human complement was performed and immunofluorescence was measured using a fluoresceinated antibody to the C-fragment of human complement C3 (Calbiochem, La Jolla, CA).
HLA phenotypes. HLA-DR typing was performed using a microlymphocytotoxicity assay (17). Separated B lymphocytes are incubated for 1 h at room temperature with specific antisera and then for 30 min at 37 C. Pooled DR rabbit complement was then added, and a second incubation was performed. Those cells positive for a specific DR antiserum in the presence of complement demonstrated a damaged membrane, as indicated by intracellular staining of eosin Y. Negative cells have intact membranes and no intracellular staining.
Statistical methods. Data were analyzed using ANOVA methods (18). In this study, individuals were classified as diabetic or nondiabetic and were assigned to other groups such as "high-GAD" or "low-GAD" based on whether they were above or below the median value for the measured value of GAD. Many analyses used split-plot techniques, and appropriate error terms were used for individual tests. In some cases, other factors such as age of the patients and HgbA1 were treated as continuous variables and an analysis of covariance was performed. In those cases, the group effects were sequentially assessed after controlling for the covariates. The JMP statistical program was used. (JMP is a registered trademark of the SAS Institute, Cary, NC).
Because many tests of peripheral nerve function are affected by age, we calculated age-adjusted Z scores. The effect of age on test performance was assessed in control subjects, and then the predicted age-adjusted performance was subtracted from the actual performance; Z scores were calculated by dividing the result by the root mean square error for test performance in controls.
Principle component analysis was used to calculate a composite Z score, which reflected motor conduction velocity, F wave latency, thermal threshold detection, and cardiovascular autonomic function. The composite index was then used as a response variable in a repeated measure ANOVA comparing high and low GAD65Ab and high and low HgbA1. In addition, the latter variables were treated as continuous measurements and analysis of covariance was performed to access associations with Z scores for the various categories of peripheral nerve function and composite peripheral nerve function.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). There were a few exceptions. Two
patients shifted from the high to the low group after yr 1, and one
patient shifted from high to low after yr 2. Only one patient shifted
from low to high, and this occurred after yr 2. IA-2Ab was detected at
the time of the initial evaluation in 21 patients and present in 20
patients at the time of the third evaluation. Antibodies to the
sympathetic and parasympathetic nervous systems were present in 18 and
15 patients, respectively, at the time of the first evaluation (see
Table 2). Three control subjects had antibodies to the sympathetic
nervous system, and two had antibodies to the parasympathetic. Eleven
patients had the HLA-DR3/4 haplotypes, 6 were DR3, 11 were DR4, and the
remainder had other haplotypes.
|
|
). Patients in good control had the same
age and sex distribution as those in poor control. As previously
reported (19), patients with high GAD65Ab had higher HgbA1
than those with low GAD65Ab (Table 2
).
|
), and there was a negative correlation between blood pressure and
creatinine clearance (P < 0.01). Electrodiagnostic studies
Many electrophysiologic parameters were significantly affected by
age, so all data were corrected accordingly in the assessment of
differences between the low and high GAD65Ab groups. Patients with high
GAD65Ab had slower motor nerve conduction velocities in the median,
ulnar, and peroneal nerves (Table 3), and
the means of the age-adjusted conduction velocity Z scores for the
three nerves were different (P < 0.01), (Fig. 2). Motor nerve conduction velocity was
slower in the median nerve in patients with poor glycemic control
compared with those with good glycemic control (P <
0.05) (Table 4). There was a significant
correlation between motor nerve conduction velocity and HgbA1
(P < 0.05) for each nerve (Table 4). The mean Z scores
for motor nerve conduction velocities were lower in the patients with
poor control (P < 0.025).
|
|
|
). The differences in mean conduction velocity Z scores for the low
and high GAD65Ab patients also remained significant after correcting
for the effect of glycemia.
F wave latencies were prolonged in the high GAD65Ab vs. the
low GAD65Ab patients for the median (P < 0.05), ulnar
(P < 0.01), and peroneal nerves (P <
0.025) (Table 3), and the differences in mean age adjusted F wave
latency Z scores were different (P < 0.025) (Fig. 3). Glycemia, however, correlated with
the F wave latency only for the ulnar nerve (P < 0.05)
(Table 4). Analysis of covariance with respect to HgbA1 revealed that
the differences in F wave latencies between the low and high GAD65Ab
groups remained significant (P < 0.05) for the ulnar
and peroneal nerves after correcting for the effect of glycemia (Table 3).
|
The thermal thresholds of the patients with high HgbA1 were not
different from those with low HgbA1. Nevertheless, there was a positive
correlation between the mean HgbA1 and the detection of thresholds for
cold in the wrist (P < 0.001), heat in the wrist
(P < 0.025), and cold in the feet (P
< 0.025) (Table 4).
Patients with high GAD65Ab had higher thermal threshold for heat and
cold in the feet than those with low GAD65Ab (P <
0.01) (Table 5) (Fig. 4). A similar trend in the wrist was not
statistically significant. The mean thermal thresholds Z scores were
higher for the GAD65Ab patients than with low GAD65Ab, even after the
data were corrected for the effects of age and HgbA1 (P
< 0.05).
|
|
Heart rate variability with deep breathing. At the time of
their initial evaluation, heart rate variability with deep breathing
was slightly greater in the diabetic patients than the control
subjects. The maximum-minimum heart rate was 19% greater in the
diabetic patients (P < 0.025), and the mean circular
resultant was 20% greater (P < 0.05). The
maximum-minimum results showed a significant negative correlation with
age (P < 0.05), whereas the mean circular resultant
did not. Glycemia did not significantly affect either parameter (Table 4).
The high GAD65Ab patients had decreased heart rate variability as
assessed by both the maximum-minimum test (P < 0.025)
and the mean circular resultant (P < 0.05) (Table 5).
Post-Valsalva R-R interval. The post-Valsalva R-R interval
tended to be lower in the diabetic patients than the controls,
significantly so at the time of the third evaluation (P
< 0.05). Patients with low HgbA1 had higher post-Valsalva R-R
intervals than did those with high HgbA1 (P < 0.05),
and there was a significant correlation between test performance and
HgbA1 (P < 0.01). Patients with high GAD65Ab had lower
post-Valsalva R-R intervals than did those with low GAD65Ab
(P < 0.05), although the differences were no longer
significant after data were corrected for the effect of glycemia
(P = 0.065) (Table 3).
Power spectral analysis. A dramatic effect of age was observed
on all power spectra, which was corrected for statistically in
all group comparisons. The power spectra of the diabetic patients as a
group were no different from controls spectra. Spectra tended to be
decreased in the patients with high GAD65Ab, but the results were not
variable than the other tests and only the changes in total power and
high-frequency power, both upright posture, were significant
(P = 0.008; Table 5).
Mean cardiovascular autonomic Z scores. To integrate the
performance on the multiple cardiovascular autonomic function tests we
chose those tests that correlated the most with each other, the mean
circular resultant, the post-Valsalva R-R interval, and the
high-frequency supine power spectra. The mean age-adjusted Z scores
were interpreted to reflect global cardiovascular autonomic function.
Patients with high GAD65Ab had lower mean Z scores than those with low
GAD65Ab (P = 0.04; Table 5 and Fig. 5). There was no effect of glycemia on
the mean Z scores. After correcting the data for the effect of
glycemia, however, the Z score differences were of marginal
significance (P = 0.055).
|
).
Patients with low HgbA1 did not have significantly different peripheral
nerve function than those with high HgbA1. The duration of diabetes did
not correlate with the individual tests of peripheral nerve function or
composite peripheral nerve function (Fig. 7).
|
|
). This analysis revealed, not
surprisingly, a disproportionate number of patients who had low
GAD65Ab, low HgbA1, and good peripheral nerve function (group 1 in
Table 6) or were the opposite for all three parameters (group 4a).
Fourteen of the 18 patients with high GAD65Ab had poor peripheral nerve
function whereas only 10 of 17 patients with high HgbA1 had poor
function. Only 3 of 19 patients with low GAD65 had poor nerve function
whereas 8 of 20 patients with good glycemia had poor function. The
relatively large number of patients in group 2a and the presence of
only one in group 3a illustrates how GAD65Ab was strongly associated
with peripheral nerve function (P < 0.001), whereas
HgbA1 was not (P = 0.7). When these parameters were
analyzed as continuous variables, however, HgbA1 had a stronger
association with nerve function (P < 0.001) than
GAD65Ab (P < 0.02).
|
Antibodies to the autonomic nervous system. There was no correlation between GAD65Ab and the antibodies to the sympathetic or parasympathetic nervous system. Peripheral nerve function was not different in patients with or without antibodies to the autonomic nervous system.
HLA phenotypes. Peripheral nerve function did not differ between patients with the HLA DR3/4 phenotype and those with other phenotypes.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The second major finding in this study is that patients with high
GAD65Ab have worse peripheral nerve function than those with low
GAD65Ab. Kaufman et al. (5) originally reported
an association between GAD65Ab and peripheral nerve function in type 1
diabetes. This relationship was investigated because GAD65Ab had been
shown to be associated with central nervous system dysfunction in the
stiff man syndrome, a rare neurological syndrome linked to a central
deficiency of the product of GAD enzyme, 
-aminobutyric acid
(25). The observations of Kaufman et al.
(5) on the effect of GAD65Ab on peripheral nerve function
in type 1 diabetes, however, were not confirmed in a series of
subsequent reports (26, 27). Our study is different from
these others, however, because we studied patients during the early
natural history of the disease. We reasoned that multiple risk factors
(hyperglycemia, hypertension, macrovascular disease, and smoking) have
been linked to neuropathy in patients with chronic diabetes
(28) and, therefore, any effect of GAD65Ab might be
difficult to identify in patient with longstanding disease. Moreover,
GAD65Ab may mediate autoimmune tissue destruction only transiently
before and during the first few years of type 1 diabetes, processes
that may not be accurately reflected by antibody titers measured years
later. Thus, patients whose GAD65Ab titers decrease over time could
easily be incorrectly categorized as GAD65Ab negative in
cross-sectional studies of long-standing disease.
Could the poorer peripheral nerve function in the high GAD65Ab patients
simply reflect their worse glycemic control? Previous studies of small
numbers of patients has shown, and we have confirmed, that glycemic
control impacts on peripheral nerve function even in the first few
years after diagnosis (29, 30). Thus, it seems that the
effect of GAD65Ab on glycemic control contributes to its effect on
peripheral nerve function. Nevertheless, correcting the peripheral
nerve test results for the impact of glycemia generally had only a
minor effect on the statistical differences between the low and high
GAD65Ab patients (Tables 3 and 5). It has previously been recognized
that hyperglycemia is not the only factor involved in the initiation
and early progression of diabetic complications. Nathan
(31) and the Diabetes Control and Complications Research
Group (32) documented that hyperglycemia has a strong
relationship to retinopathy, a slightly weaker relationship with
nephropathy and somatosensory neuropathy, an even less robust
relationship with autonomic neuropathy (32), and little or
no correlation with macrovascular disease.
How might GAD65Ab affect peripheral nerve function by mechanisms independent of effects on ß-cell function and glycemic control? GAD is widely distributed throughout the peripheral and autonomic nervous systems, and it is possible that the function of the enzyme is compromised by circulating GAD65Ab. It is well established, for example, that circulating antibodies to the acetylcholine receptor disrupt the function of the neuromuscular junction in myasthenia gravis and the Eaton Lambert syndrome. GAD is an intracellular cytosolic enzyme, however, to which circulating GAD65Ab presumably has little access (33). An alternative explanation is that GAD65Ab reflects the presence of a systemic inflammatory process that has parallel effects on the ß cell and peripheral nervous system. There is recent evidence, for example, that GAD65Ab in patients with type 2 diabetes activates the secretion of acute phase proteins such as fibrinogen and C-reactive protein which, in turn, have been associated with inflammatory cytokines (34). This is an attractive hypothesis because of the mounting evidence that inflammation and endothelial cell dysfunction play a role in multiple diabetic complications, especially neuropathy (34, 35). We acknowledge, however, that the observed association between GAD65Ab and peripheral nerve function does not prove that GAD65Ab has a direct neurotoxic effect. Hyperglycemia-induced pancreatic or neurological damage, for example, might cause GAD65 to leak into the circulation, which would mean that GAD65Ab is then formed as a result (rather than the cause) of the neurological damage. We have shown, for example, that streptozotocin can cause GAD65Ab release from intact ß cells (36). Although it would seem unlikely that nerve damage would be severe enough in early diabetes for this to occur, this interpretation of our findings is difficult to disprove.
The potential clinical significance of the differences in peripheral nerve test performance between the low and high GAD65Ab patients needs to be addressed. None of the high GAD65Ab patients had signs or symptoms of diabetic neuropathy, nor was their performance outside the range of normal as established in control subjects. Thus, our patients did not exhibit evidence of even subclinical neuropathy, and in this regard our findings failed to confirm previous reports of early onset asymptomatic peripheral neuropathy in type 1 diabetes (37, 38). Our results are similar, however, to those of Ziegler et al. (39) as well as those of the Diabetes Control and Complications Trial, in which peripheral nerve function in early diabetes was normal (32). Nevertheless, our results show GAD65Ab-related differences in performance on a wide variety of peripheral nerve function tests. Although the changes are small, this would be expected given the short duration of diabetes. IA2-Ab and antibodies to the autonomic nervous system, by contrast, did not correlate with glycemia or peripheral nerve function.
In summary, patients with type 1 diabetes and high GAD65Ab have worse glycemic control, higher insulin requirements, and slightly decreased somatosensory and autonomic nerve function.
|
Footnotes |
|---|
Received February 10, 2000.
Revised May 18, 2000.
Accepted June 6, 2000.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
-Aminobutyric acid (GABA) metabolism in mammalian neural and
nonneural tissues. Comp Biochem Physiol. 112A:247–263.
This article has been cited by other articles:
 |
|
 |
|
  A. I. Vinik, D. Anandacoomaraswamy, and J. Ullal Antibodies to Neuronal Structures: Innocent bystanders or neurotoxins? Diabetes Care, August 1, 2005; 28(8): 2067 - 2072. [Full Text] [PDF]
|
|
|
|
 |
|
 R. D. Hoeldtke, K. D. Bryner, D. R. McNeill, S. S. Warehime, K. Van Dyke, and G. Hobbs Oxidative Stress and Insulin Requirements in Patients with Recent-Onset Type 1 Diabetes J. Clin. Endocrinol. Metab., April 1, 2003; 88(4): 1624 - 1628. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. D. Hoeldtke, K. D. Bryner, D. R. McNeill, G. R. Hobbs, J. E. Riggs, S. S. Warehime, I. Christie, G. Ganser, and K. Van Dyke Nitrosative Stress, Uric Acid, and Peripheral Nerve Function in Early Type 1 Diabetes Diabetes, September 1, 2002; 51(9): 2817 - 2825. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
