| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Articles |
Departments of
1
Medical Oncology and
2
Pathology and Laboratory Medicine, University Hospital Groningen, 9700 RB Groningen, The Netherlands.
a Address correspondence to this author at: Department of Pathology and Laboratory Medicine, University Hospital Groningen, PO Box 30.001, 9700 RB Groningen, The Netherlands. Fax 31-50-3612290; e-mail i.p.kema{at}lab.azg.nl
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Indole markers were measured in 688 patients with suspected carcinoid disease. The initial values of indole markers from patients in whom a carcinoid tumor was confirmed during follow-up (n = 98) were used for ROC analysis. Two groups served as reference populations. The first consisted of 45 healthy individuals ("healthy controls"). The second was a random sample of 40 patients, drawn from the 590 (688 minus 98) patients with carcinoid-like symptoms but without a carcinoid tumor ("clinically suspected patients").
Results: ROC curve analysis showed platelet serotonin to have the highest discriminating capacity, especially in foregut carcinoids. Cutoff values for platelet serotonin obtained from ROC analysis with healthy controls as reference group (5.4 nmol/109 platelets) gave a sensitivity of 74%, specificity of 91%, positive predictive value of 63%, and negative predictive value of 95% when applied to the initial 688 patients. Using the cutoff value with the clinically suspected patients as the reference group (9.3 nmol/109 platelets) gave a sensitivity of 63%, specificity of 99%, positive predictive value of 89%, and negative predictive value of 93%. Indole markers were increased in 169 (25%) of 688 patients. In 76 (45%) of these 169 patients, a carcinoid tumor was present. Slight increases of markers were associated with non-carcinoid neuroendocrine tumors, non-neuroendocrine tumors, and disturbed bowel motility.
Conclusions: ROC curve analysis shows that platelet serotonin is the most discriminating indole marker for the diagnosis of carcinoid tumors. Platelet serotonin especially improves the diagnosis of carcinoids producing small amounts of serotonin.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
50%
(2). With recent advances in the palliative treatment of
carcinoid tumors, an early diagnosis will improve quality of life and
survival (3)(4). Enterochromaffin cells produce serotonin (5-hydroxytryptamine) as a paracrine hormone (5)(6). Markers such as chromogranins and neuropeptides (7) possess a high sensitivity for neuroendocrine tumors. However, these markers are unable to detect an enhanced serotonin metabolism, which is considered a hallmark of carcinoid tumors. Three indole markers are used in carcinoid disease: urinary 5-hydroxyindoleacetic acid (5-HIAA),1 urinary serotonin, and platelet serotonin content. Urinary 5-HIAA may be within the reference interval in 30% of carcinoids (8). The predictive value of an increased urinary 5-HIAA is low (9). Moreover, urinary 5-HIAA may increase in normal subjects after consumption of a serotonin-rich diet (10)(11). Urinary serotonin excretion mainly results from decarboxylation of circulating 5-hydroxytryptophane (5-HTP) by the renal tubular cells (12) and can unmask a carcinoid tumor that produces the serotonin precursor 5-HTP predominantly (10)(13)(14). Dietary serotonin only slightly increases the urinary free serotonin excretion (11). Platelets take up serotonin from the circulation by an active transmembrane mechanism (15)(16). In contrast to urinary 5-HIAA, platelet serotonin content is not influenced by a short-term, serotonin-containing diet (11).
The secretory activity of carcinoids is related to the primary site of the tumor (17). Carcinoids originating from the foregut region may produce several substances, such as histamine and catecholamines, in addition to serotonin and 5-HTP (8)(18). These tumors produce less serotonin than midgut carcinoids. Midgut carcinoids produce serotonin predominantly. Particularly midgut carcinoids can, after liver metastases have occurred, give rise to a carcinoid syndrome. This syndrome, comprising symptoms of diarrhea, flushing, bronchial wheezing, and carcinoid heart disease, most likely results from the release of massive amounts of serotonin in the systemic circulation (19). Primary hindgut carcinoids usually show no secretory activity (20)(21)(22). Urinary 5-HIAA is a useful diagnostic marker in midgut carcinoids with a high tumor load. However, in carcinoids with a moderate increase of serotonin production or predominant secretion of 5-HTP, platelet serotonin and urinary serotonin can be more sensitive markers (14)(21)(23).
In the present study, we evaluated the discriminating capacity of HPLC-based assays of urinary 5-HIAA, urinary serotonin, and platelet serotonin in carcinoid tumors. On the basis of ROC curves, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of these indole markers were calculated. When we applied the ROC-based cutoff values to the entire population under investigation, patients with increased indole markers were identified. For these patients, the diagnosis and symptoms were identified. Finally, recommendations for the clinical use of these indole markers are presented.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
carcinoid patients
The group of carcinoid patients consisted of patients in whom a
histological diagnosis of a carcinoid tumor was made during the study
period. Biochemical results of patients after confirmed curative
resection of a carcinoid tumor were excluded from the study. In this
way, only patients with a carcinoid tumor present at the time of
biochemical testing were included. Whether a resection was curative or
not was determined with clinical follow-up and octreotide scintigraphy.
In this way, two patients were excluded from evaluation. The group
contained carcinoid patients, regardless of the outcome of biochemical
assessment. Ninety-eight carcinoid patients were included in this
group. The 44 patients described in our previous (18) study
were included in the present study. The median age was 58 years (range,
23–91); 47 were males, and 51 were females.
The carcinoid patients were classified into subgroups according to primary tumor localization (24) as foregut tumors (upper respiratory tract, lungs, esophagus, stomach, and duodenum; n = 22), midgut tumors (jejunum, ileum, appendix, cecum, and the ascending colon; n = 51), and hindgut tumors (transverse and left colon and rectum; n = 8). Biochemical results of patients unable to be classified into one of these groups (n = 17) were not included in subgroup calculations but were used in overall computations.
clinically suspected patients
The remaining 590 patients, after exclusion of the 98 carcinoid
patients, represented a group in which biochemical evaluation was
performed because of the clinical suspicion of a carcinoid tumor, based
on the presence of symptoms such as diarrhea, flushing, wheezing, and
unexplained abdominal pain. These patients had no evidence of a
carcinoid tumor during follow-up. Of these 590, we drew a random sample
of 40 patients to serve as a reference group for ROC curve
computations. The median follow-up in this group was 8.5 years (range,
4–12 years). The median age was 51 years (range, 25–78); 18 were
males, and 22 were females.
healthy controls
Of a population of hospital workers, some of whom were retired, an
age- and sex-matched control group was recruited. Disease states
affecting indole markers were excluded by taking a medical history.
From these individuals, blood and urine were collected, after informed
consent was given. This group of healthy controls provided a second
reference group for the construction of ROC curves. In this group, the
median age was 54 years (range, 44–75); 20 were males, and 25 were
females.
patients with increased indole markers
Of the complete population of 688 patients, including the
carcinoid patients and the clinically suspected patients, we identified
those patients who had increased indole markers. For this, we
used selected cutoff values obtained from ROC curve analysis with
either the clinically suspected patients or the healthy controls as
reference group. Of the patients with increased indole markers, the
primary site of the carcinoid tumor or the non-carcinoid
diagnosis or symptoms were specified.
blood and urine samples
The 24-h urine samples from patients and controls were collected
in 2-L brown polypropylene bottles (Sarsted) containing 250 mg
each of
Na2S2O5
and EDTA as preservatives. Samples were acidified to pH 4 with acetic
acid, before freezing. Venous blood samples were collected in 10-mL
Vacutainer Tubes (Becton Dickinson) containing 0.12 mL of 0.34 mol/L
EDTA solution and were immediately put on ice. Platelet-rich plasma was
prepared from whole blood within 1 h after sampling by
centrifugation for 30 min at 120g and 4 °C.
Na2S2O5
and EDTA were added as preservatives in final concentrations of 10
g/L each. Platelet concen-trations were measured with a Coulter Counter
Model S plus 4 (Coulter Electronics). Samples were stored at -20 °C
and analyzed within 1 week after collection.
analytical methods
The serotonin contents of platelet-rich plasma and urine were
determined by HPLC with fluorometric detection according to the method
of Kwarts et al. (25) in the following way. Platelet-rich
plasma or urine were mixed with an ammonium acetate buffer, pH 7.5.
Samples were subsequently applied to Amberlite CG-50 columns and washed
with the ammonium acetate buffer and 0.01 mol/L acetic acid. Elution
was done with 1.0 mol/L acetic acid that contained ascorbic acid.
Chromatographic analysis was performed by isocratic reversed-phase HPLC
with fluorometric detection. Urinary 5-HIAA concentrations were
determined in ether extracts by HPLC with fluorometric detection
(26). Urinary creatinine concentrations were measured by a
picric acid method (Mega analyzer; Merck). Urinary serotonin and 5-HIAA
were expressed in µmol/mol and mmol/mol urinary creatinine,
respectively (27). Platelet serotonin content, expressed as
nmol serotonin/109 platelets, was calculated by
dividing the serotonin concentration in platelet-rich plasma by the
concentration of platelets in the plasma.
statistics
The discriminating capacity of urinary 5-HIAA, urinary serotonin,
and platelet serotonin was calculated from ROC curves (28).
The area under the ROC curve (AUC), indicating the capability of a
marker to discriminate between normal and abnormal groups, is the
so-called "discriminating capacity", "diagnostic accuracy", or
"performance" of a marker. In the computation of the ROC curves,
either the healthy controls or the clinically suspected patients were
used as reference groups. The numbers of individuals in both reference
groups were not based on a power analysis but were chosen to obtain
balanced numbers for comparison. This way the numbers of healthy
controls and clinically suspected patients were in the same range as
the numbers of patients in the respective subgroups of carcinoid
patients (i.e., foregut, midgut, and hindgut carcinoid). To select
clinically suspected patients, we identified those patients with all
three indole markers measured within a time interval of 3 months and
with adequate follow-up available. From this group, we selected random
samples based on an exact number of 40 cases using SPSS (SPSS 9.0;
SPSS, Inc).
We selected a cutoff value for each marker that produced the highest differential positive rate (sensitivity minus false-positive rate) using either healthy controls (low-level cutoff value) or clinically suspected patients (high-level cutoff value) as reference group. For computation and analysis of ROC curves, we used the software programs ROC 2.1 (University Hospital Groningen) and Clinical Laboratory, Ver. 1.28 (Analyze-It Software, Leeds, United Kingdom). Linear discriminant analysis (29) was used to test whether combinations of indole markers would have a better discriminating ability than a single indole marker.
For comparison of age and sex distribution between groups,
respectively, the Student t-test and Pearson’s
2 test were used. Group differences between
the biochemical values were evaluated using the Kruskal–Wallis test.
Whenever differences were found, the individual group differences were
tested with the Mann–Whitney U-test, using Bonferroni
correction. In all calculations, P <0.05 indicated a
significant difference.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
, concentrations of indole markers in the healthy controls,
clinically suspected patients, and carcinoid patients, respectively,
are depicted. Urinary 5-HIAA was higher in the carcinoid group compared
with the healthy controls and the suspected group. No difference in
urinary 5-HIAA was observed between the healthy controls and the
suspected group. Urinary serotonin was higher in the carcinoid group
compared with the healthy controls. However, no difference was found
between the carcinoid group and the suspected group. In the suspected
group, higher urinary serotonin values were found, compared with the
healthy controls. Values of platelet serotonin differed among all three
groups.
|
Fig. 2
represents the distribution of the concentrations of urinary
5-HIAA, urinary serotonin, and platelet serotonin in the carcinoid
group, subdivided according to the primary localization.
|
discriminating capacity
In Fig. 3
, the relevant ROC curves are depicted. To facilitate
comparisons of the markers, the ROC curves obtained from urinary
5-HIAA, urinary serotonin, and platelet serotonin are combined in each
plot. Two ROC curves were drawn for the complete carcinoid group,
including the 17 patients with unknown primary localization. In the
first ROC curve, the healthy controls were used as reference group
(Fig. 3A
), whereas the second ROC curve was drawn with the clinically
suspected patients as reference (Fig. 3B
). Separate ROC curves were
constructed for midgut (Fig. 3C
) and foregut (Fig. 3D
) carcinoids. In
the latter two ROC curves, the healthy controls were used as reference
group. For the hindgut group, it was not possible to construct a
reliable ROC curve because of the small number of patients (n =
8).
|
Table 1
shows the AUC calculated from the ROC curves in Fig. 3
. In the
ROC curve constructed with all carcinoid patients and the healthy
controls as a reference group, no differences between the AUCs of the
various markers were found. The difference in the AUCs for platelet
serotonin and urinary 5-HIAA did not reach significance
(P = 0.08). Analysis of the carcinoid group with the
clinically suspected patients as reference group revealed larger AUCs
for platelet serotonin vs urinary serotonin (P = 0.05)
and for urinary 5-HIAA vs urinary serotonin (P = 0.02).
With the healthy controls as a reference group, subgroup calculations
were performed. In foregut carcinoids, higher AUCs were found for
platelet serotonin vs urinary 5-HIAA (P = 0.03) and in
midgut carcinoids for platelet serotonin vs urinary serotonin
(P = 0.02). Linear discriminant analysis showed that no
possible combination of the indole markers yielded higher AUCs,
compared with the single use of platelet serotonin.
|
cutoff values
From the ROC curve with all carcinoid patients vs the healthy
controls (Fig. 3A
), the cutoff values (low-level, Table 1
) were
calculated as described in the methods for the respective indole
markers: urinary 5-HIAA, 2.8 mmol/mol creatinine; urinary serotonin,
55.0 µmol/mol creatinine; and platelet serotonin, 5.4
nmol/109 platelets. Use of the clinically
suspected patients as reference group (Fig. 3B
) produced somewhat
higher cutoff values (high-level, Table 1
): urinary 5-HIAA, 6.7
mmol/mol creatinine; urinary serotonin, 99.0 µmol/mol creatinine; and
platelet serotonin, 9.3 nmol/109 platelets.
test characteristics with selected cutoff values
The sensitivity, specificity, PPV, and NPV of the indole markers
for the presence of a carcinoid tumor are represented in Table 2
. In contrast to the discriminating capacity, these test
characteristics depend on the chosen cutoff value. In accordance with
the outcome of the ROC curve analysis, platelet serotonin showed better
test characteristics than did urinary 5-HIAA and urinary serotonin
using either the low-level or high-level for cutoff values.
|
patients with increased indole markers
When we applied the low-level cutoff values to all 688 patients in
the study, 169 patients had at least one indole marker increased. In 76
(45%) of these patients, a carcinoid tumor was present. In Table 3
, the primary site of the carcinoid tumor is specified. The
remaining 93 patients had increased indole markers in the absence of a
carcinoid tumor. The diagnosis or symptoms of these patients are also
shown in the table. Non-carcinoid neuroendocrine tumors,
non-neuroendocrine tumors, and disturbances in bowel motility accounted
for increased markers in 18, 11, and 17 patients, respectively. The 18
non-carcinoid neuroendocrine malignancies were pancreatic islet cell
tumors (n = 6), solitary medullary thyroid cancer (n = 4),
pheochromocytoma and medullary thyroid cancer associated with multiple
endocrine neoplasia IIa syndrome (n = 3), solitary
pheochromocytoma (n = 1), hepatocellular carcinoma with
neuroendocrine differentiation (n = 1), and a neuroendocrine
laryngeal tumor (n = 1). In two patients, the primary site of the
neuroendocrine tumor remained unidentified. The 11 non-neuroendocrine
malignancies were adenocarcinomas arising from the digestive tract
(n = 8), breast (n = 1), and ovary (n = 1), and a
hepatocellular carcinoma (n = 1). Three of 17 patients with
disturbances in bowel motility presented with constipation. In 14
patients, diarrhea was attributable to irritable bowel syndrome (n
= 4), microscopic colitis (n = 3), inflammatory bowel disease
(n = 2), laxative abuse (n = 1), or pancreatic insufficiency
(n = 1). In three patients, no causative mechanism for the
diarrhea was found. In the remaining 47 patients, no definite diagnosis
explaining the increased markers was established.
|
Application of high-level cutoff values identified 81 patients with
increased markers; 64 (79%) of these patients had a carcinoid tumor
(Table 3
). Of 17 non-carcinoid patients with increased markers on
high-level, 7 had neuroendocrine tumors, and 4 had non-neuroendocrine
tumors. Only six patients with nonmalignant disorders had markers above
high-level cutoff values. One of these patients had inflammatory bowel
disease; in the remaining patients, no definite diagnosis was made.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Because midgut carcinoids represent >50% of the carcinoid patients in our study group, differences between subgroups of carcinoid tumors could remain unnoticed. Therefore, an analysis based on the primary localization of the carcinoid tumor was performed.
In foregut carcinoid patients, platelet serotonin has a higher discriminating capacity compared with urinary 5-HIAA. In fact, because the AUC was 0.55, urinary 5-HIAA did not contribute to the diagnosis of foregut carcinoids. This finding is in accordance with Janson et al. (7), describing a poor sensitivity of urinary 5-HIAA in 39 foregut carcinoids. In earlier studies (10)(13)(30)(31), an overproduction of 5-HTP in foregut carcinoids was reported, tentatively resulting from a deficiency of aromatic amino acid decarboxylase in the tumor. Renal decarboxylation of 5-HTP makes urinary serotonin excretion a possible marker in foregut carcinoids. Our study, however, shows that platelet serotonin is at least equivalent to urinary serotonin in the diagnosis of foregut carcinoids.
In midgut carcinoids, all three indole markers possessed a high discriminating capacity, although platelet serotonin was superior to urinary serotonin. This reflects the high rate of serotonin secretion found in midgut carcinoids. The remarkable high urinary excretion of serotonin in midgut carcinoids probably results from serotonin clearance, rather than 5-HTP synthesis. Nevertheless, 5-HTP coproduction may occur in midgut carcinoids as described by Feldman (14), who found increased urinary 5-HTP excretion in 81% of 27 carcinoid patients with increased urinary 5-HIAA excretion.
The small number (n = 8) of patients with hindgut carcinoids in our study precludes reliable conclusions in this subgroup. In a study by Koura et al. (32), urinary 5-HIAA was increased in 4 of 22 patients with rectal carcinoids. These four patients had disseminated disease with liver metastases. In agreement with other authors (20)(22), this finding illustrates that hindgut carcinoids usually secrete limited amounts of serotonin. Therefore, the indole markers make a minor contribution to the diagnosis of hindgut carcinoids, except for patients with advanced disease.
As argued previously (21), the high discriminating capacity of platelet serotonin can be explained by the metabolism of serotonin and estimations on pool size of the respective compartments involved. Platelet serotonin content follows saturation kinetics, whereas urinary 5-HIAA represents a metabolic end-compartment with unlimited capacity. This makes urinary 5-HIAA a relevant marker for follow-up of disease activity in carcinoids with grossly enhanced serotonin production. Upper limits of normal for urinary 5-HIAA (10)(33)(34) and platelet serotonin (35) mentioned in the literature are intermediate between the low-level and high-level cutoff value determined in the present study.
Test characteristics (sensitivity, specificity, PPV, and NPV) obviously
depend on the selected cutoff value (either low-level or high-level,
Table 2
). A higher sensitivity and NPV are found when using the
low-level cutoff values than with high-level cutoff values. Specificity
and PPV were lower with low-level cutoff values compared with
high-level cutoff values. The high PPV for urinary serotonin with both
low-level and high-level cutoff values was remarkable. This could be
the result of a selection bias caused by measurement of urinary
serotonin. This test was ordered more frequently in patients in whom
increased urine 5-HIAA excretion or platelet serotonin content was
found than in patients with normal urinary 5-HIAA excretion or platelet
serotonin content. The choice for either low-level or high-level cutoff
values depends on the clinical situation. To exclude a carcinoid tumor,
the low-level cutoff values are preferred; to confirm the presence of a
carcinoid tumor, the high-level cutoff values are indicated. Applying
low-level cutoff values decreases the specificity of the indole
markers. We therefore evaluated the diagnoses and symptomatology of
patients with increase of at least one of the indole markers (Table 3
).
Evidence of increased serotonin production was found in non-carcinoid
neuroendocrine tumors. This is in agreement with other reports
describing increased urinary 5-HIAA excretion in non-carcinoid
neuroendocrine tumors (9)(10)(34).
Non-carcinoid neuroendocrine tumors share metabolic characteristics
with carcinoid tumors. In the present study, an enhanced serotonin
metabolism was also observed in patients with gastrointestinal motility
disorders, presenting with diarrhea or constipation without a carcinoid
tumor. Here, serotonin is probably synthesized in above-physiological
amounts because of its action as a neurotransmitter in the intestinal
tract. This is supported by earlier observations of increased indole
markers in patients with coeliac disease
(36)(37) and irritable bowel disease
(38). Fig. 1
and Table 3
show that only moderately increased
indole markers are found in non-carcinoid patients. In the clinically
suspected patients, platelet serotonin never exceeded a value of 10
nmol/109 platelets.
Urinary 5-HIAA excretion has a lower discriminating capacity for carcinoid tumors compared with platelet serotonin and is therefore a less reliable marker in the diagnosis of carcinoid tumors. Nevertheless, urinary 5-HIAA is of relevance in the follow-up of carcinoids with a high serotonin secretion rate.
Urinary serotonin makes no additional contribution to platelet serotonin in the detection of carcinoid tumors. Even in foregut carcinoids, which occasionally produce 5-HTP instead of serotonin, there was no advantage of urinary serotonin over platelet serotonin.
Platelet serotonin is the most discriminating indole marker for the detection of carcinoid tumors. The high sensitivity of platelet serotonin is of clinical importance, in particular in carcinoids with a low serotonin production rate. Such low serotonin production is found in foregut and hindgut carcinoids and in midgut carcinoids with a small tumor volume. This establishes platelet serotonin as a reliable marker for the early diagnosis of carcinoids and in the detection of residual tumor after operation. The amount of platelet serotonin is not affected by factors such as stress, position, and diurnal variation, and moreover, the assay can be performed with relatively simple HPLC equipment (39)(40). Developments in chromatographic analyses have moved the analysis of indoles away from the point at which they were considered complicated, cumbersome, and sensitive to interference to a position where they have become accessible to clinical chemical laboratories that are equipped with modern chromatographic facilities. We therefore recommend the use of platelet serotonin as the primary indole marker for the diagnosis of carcinoid tumors.
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  A. Pfafflin, K. Mussig, and E. Schleicher Interference of paracetamol (acetaminophen) with a commercially available high-performance liquid chromatography analysis of serotonin leading to falsely low serotonin levels Ann Clin Biochem, March 1, 2009; 46(2): 146 - 148. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Zolkowska, M. H. Baumann, and R. B. Rothman Chronic Fenfluramine Administration Increases Plasma Serotonin (5-Hydroxytryptamine) to Nontoxic Levels J. Pharmacol. Exp. Ther., February 1, 2008; 324(2): 791 - 797. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. M. Zuetenhorst and B. G. Taal Metastatic Carcinoid Tumors: A Clinical Review Oncologist, February 1, 2005; 10(2): 123 - 131. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. P. Kema, W. G. Meijer, G. Meiborg, B. Ooms, P. H.B. Willemse, and E. G.E. de Vries Profiling of Tryptophan-related Plasma Indoles in Patients with Carcinoid Tumors by Automated, On-Line, Solid-Phase Extraction and HPLC with Fluorescence Detection Clin. Chem., October 1, 2001; 47(10): 1811 - 1820. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
