| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Technical Advances |
From the Departments of Pathology,*
Internal
Medicine–Division of Hematology/Oncology,
and
Urology,
University of Michigan Comprehensive
Cancer and Geriatrics Center, Ann Arbor, Michigan; and the Department
of Pathology,
Johns Hopkins University School
of Medicine, Baltimore, Maryland
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
TMA-based technology prompted the need for a system that effectively managed data generated from this high-throughput approach. For example, the typical prostate specimen of a single patient consists of up to 50 individual paraffin blocks (range, 12 to 50; median, 20) containing multiple foci of tumor in addition to benign prostate, prostatic intraepithelial neoplasia, and other potential tissue types (eg, stroma, atrophy, benign prostatic hyperplasia). The use of a large spreadsheet has been the standard solution to handle voluminous amounts of data. This approach is useful for experiments on a one-time basis. However, this approach has lead to the loss of an enormous amount of molecular data because of an inability to return data in a convenient manner back into the database. As part of on-going collaborations between the Specialized Programs in Research Excellence for prostate cancer, the three groups from the University of Michigan (Ann Arbor, MI), Johns Hopkins Hospital (Baltimore, MD), and Baylor College of Medicine (Houston, TX) sought a solution that would allow us to provide data to translational researchers and provide a mechanism to return data back to the database. This approach should ultimately lead to a rich database and prevent duplication of effort. Adequately detailed data sets as described above would contain a high degree of repetitive data in tabular format, requiring more time for manual data entry and the potential for data entry errors.
The use of a database composed of multiple interrelated tables minimizes the potential pitfalls that arise from using single large spreadsheets to record TMA data. The use of unique identifiers to link fields of data between multiple tables and databases is a central concept to the effectiveness of relational databases. Unique identifiers are values appearing only once within a given field in a single table of data, but may be replicated within another table serving as a link between the two data sets. We present a model that links all TMA-related data and corresponding image histology results with clinical and pathology information currently available at the University of Michigan on >1300 PCA patients. The key components are 1) the TMA database (TMA DB), 2) the TMA-image database TMA-I DB), and 3) the prostate pathology and clinical information databases.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The first step in a typical TMA study is the acquisition and preparation of relevant donor specimen blocks. Matching hematoxylin and eosin (H&E) reference slides are cut from all donor blocks, graded, and targeted tissue types of interest are circled in ink. These grading markers serve as a template for sampling discrete diagnosis from the donor block. Information such as patient number, specimen number, specimen block label, tissue, and targeted diagnosis are routinely recorded into our TMA DB at this point or after completion of the TMA block. Targeted TMA sample diagnosis refers to the presumed pathology diagnosis of tissue cores being transferred to the recipient TMA block. Because of tissue heterogeneity and potential sampling inaccuracies, these diagnoses are verified at a later time (confirmed TMA sample diagnosis).
TMAs are constructed using a manual tissue puncher/array (Beecher
Instruments, Silver Spring, MD) as previously described (Figure 1)
.1
TMAs are typically
constructed with 400 to 600 individual tissue samples of 0.6 mm
diameter. Samples are arrayed in a recta-linear pattern with 0.8 mm
between the centers of each sample. Once the TMA study block has been
prepared, the block can be sectioned into 6 mm thick sections,
counterstained with H&E for use as a reference slide to confirm
histology, and treated with different immunoperoxidase stains.
|
A relational TMA DB was created in Microsoft Access 2000
(Microsoft, Redmond, WA) to organize and link all levels of
sample-related information and to allocate an array ID for each core in
a given TMA block (Figure 2)
. This type of database solution eliminates
having to enter redundant information as one would in a single
spreadsheet. The TMA DB is comprised of four main tables containing
patient, block, diagnosis, and array location information linked to
each other via unique autogenerated numbers (ie, primary keys) that are
hidden beneath the user interface during data entry process (Table 1
, column 2). Although certain database
fields are unique to prostate cancer-related research, this data
structure could be adapted to include other tissue-specific data.
|
|
All data entry is performed manually on a main form comprised of fields from three subforms. Specifically, the main form is based on a query that extracts data from the patient table. The three subforms originate from queries that extract designated data fields from the block, diagnosis, and array location tables, respectively. Basing the forms on sorted query tables instead of original data tables allows for more efficient data handling and speeds up searches. Using Microsoft Access 2000 as opposed to Microsoft Access 97 has the advantage of visually presenting hierarchically organized subnested data sets. This feature allows one to expand a single row of data in a datasheet and view all other related data. Before the single expandable datasheet was used, separate forms were used, making data entry less efficient. With a single form, all relevant patient, block, diagnosis, and array location information can be entered in one location, improving the efficiency of data entry. Microsoft Access 2000 also offers the data access page (DAP) feature that accommodates the visual assessment of stained TMA sample images through a web-based interface (described below).
TMA Sample Image Acquisition and Management
Digital images are acquired using the Bacus Labs Incorporated
imaging system (BLISS; Bacus Labs. Inc., Lombard, IL). This system
utilizes the matrix array scan segment with customizable features that
were developed in collaboration with the Prostate SPORE Tissue
Microarray Working Group (University of Michigan, Baylor College of
Medicine, and Johns Hopkins University School of Medicine). This
imaging system overlays a grid on each TMA slide for which the user can
specify the slide size. The system automatically captures
high-resolution composite images at the vertices of all grid lines that
correspond to the location of TMA samples on the slide. The system
permits the operator to reposition imaging targets for TMA samples out
of alignment the regular grid pattern. The final composite images are
comprised of six separate x20 objective microscope fields stitched
together to form a single image of 
x100 final magnification. Images
are saved in jpeg format and they are between 200 and 300 KB in size.
The BLISS software automatically assigns a unique name to each image of
each sample. This naming convention consists of numeric variable for
lab ID, TMA block number, X position, Y position, treatment code (ie,
H&E = 1), and sequential image number. Hence, the image name
390_01_05_06_1_105.jpeg comes from lab number 390, TMA block
01, positioned at x-coordinate 05 and y-coordinate 05, treated with H&E
(numerical code 1), and that was the 105th composite image acquired
from the array. A directory is created on a networked image file server
that holds these jpeg images and would be named as follows:
TMA01_treatment_date_operator initials.
Final composite images can be evaluated in any number with image review utility applications such as I-View MultiMedia (Crystal Bay, NV; www.scriptsoftware.com). Using this utility, a list of all composite image file names can be generated and used to link images to other data fields.
TMA-I DB and DAP
The TMA-I DB is used to capture all pathological visual
assessments made from TMA sample images. This DB utilizes the latest
Microsoft Access 2000 feature called a DAP. This component is
essentially a database form prepared as an HTML document for publishing
over the Internet. Each TMA image file name is given a URL extension
that specifies the computer path to the image file location. This
permits image viewing, entering, and editing data over the Internet.
The DAP opens a live link to the TMA-I DB when accessed by an
authorized user. It presents images and database fields over the
Internet and permits the pathology expert to enter data into these
fields that is automatically returned to the database on the
home-server (Figure 3)
.
|
. Because the TMA-I DB maintains live links to all
four tables in the TMA DB, an update query can be run to fill in the
corresponding tissue information in this table by linking the array ID
fields. Images of individual TMA samples are presented on the computer
screen, usually on a 19 to 21 inch monitor for optimal high resolution.
Figure 3
shows TMA 01 stained with H&E. Because the DAP is available on the Internet, two levels of security are in place to access the page. The first level of security involves Windows NT authentication where the user is prompted for a login name and password. The next level of security follows Access 2000 database-level security where the specified login name has distinct privileges for entering or reading data. For instance, a designated pathologist will have full data-entry rights whereas a co-worker may have read-only rights. For a detailed discussion on how to make a DAP secure and Internet ready, refer to DAP articles in the Microsoft Knowledge Base (article ID no. Q253754). Notably, all data presented on the World-Wide Web is anonymous. There is no information that can be traced back to patients.
Prostate Clinical and Pathology Database
Thus far it has been shown that the TMA DB and TMA-I DB can communicate with one another via linking of the array ID. Hence, for any given array ID, the patient, block, tissue, or diagnosis origin can easily be obtained by use of a simple query. In addition, the histology diagnosis information for a given array ID and stain is attainable. However, to do a worthwhile study, the clinical and pathology information needs to also be accessible for the patients in the TMA construct. The true value of the TMA DB lies in the ability to link clinical outcome to TMA data.
The Prostate Clinical and Pathology Microsoft Access 1997 Databases have existed since 1995 and contain clinical and pathology information on >1300 prostate cancer radical prostatectomy patients. These databases can obtain information from one another because of the commonality of the patient registration number in all their tables. Live links of all of the tables between the databases are maintained to allow for query extraction of desired information. Note that each database is maintained separately and it is the collaboration of the pathology and urology department that allows such sharing of information to be possible. There are plans to convert the Access 1997 databases into Access 2000 format in the near future.
Because we are only interested in specific information, additional
fields were created in the radical prostatectomy table in the pathology
database to download relevant clinical information from the clinical
database. Examples include digital rectal examination results,
preoperative prostate-specific antigen (PSA) results, hormone and
radiation treatment, postsurgical PSA follow-up data, and race (column
1 of Table 1
). This eliminates the need to create several queries every
time such information is desired. More importantly, this allows direct
access of both relevant clinical and pathology data for the TMA studies
in a single step. The surgical specimen number entered in the patient
table in the TMA DB can be linked to the surgical specimen number
entered in the radical prostatectomy table. A full diagram showing the
interaction between all of the databases is shown in Figure 2
. Note that neither surgical specimen
number nor registration number are ever shown in the query results.
They are used for linking purposes only. As a final step, the query
results are given to a statistician for analysis.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
is an example of a
medium-sized array made up of 324 biopsy cores from 11 prostate cancer
patients. Only the first two rows of this TMA map will be used for
subsequent explanation purposes. This row encompasses biopsy cores
taken from two prostate tissue blocks (1D; 1F) of a single patient
(P1). Each sample has been assigned a targeted tissue diagnosis (T
= tumor, N = normal, PIN = prostatic
intraepithelial neoplasia) and an array identification number to
describe its location within the TMA. An X and Y coordinate system is
used to identify samples within a TMA block such that an array ID
(array number + x-coordinate + y-coordinate) is assigned to each biopsy
core. A sample at position x = 1, y = 1would be located in
the upper left. Hence, array ID’s 01-01-01 through 01-01-06 represent
the first six cores assembled in the upper left hand corner of the TMA
01 block (Figure 1C)
. The concept of the TMA sample identification
code, or array ID, is key to being able to effectively track the
identity of cores within a TMA and link relevant clinical data to final
histological results.
There are 19 fields in the TMA DB (column 2 of Table 1
). Prostate
cancer patients (n = 336) are entered in the TMA
DB, which translates to 19 TMA blocks with 5451 TMA biopsy cores. There
are 1695 normal, 3171 tumor, 464 prostatic intraepithelial neoplasia,
and 121 miscellaneous (eg, stroma, atrophy) cores.
The TMA-I DB contains 11 fields (column 3 of Table 1
).
All 19 TMA blocks have been analyzed with several molecular
markers with E-CAD, MIB-1, and PSA. Image stain selection depends on
the particulars of the study (Figure 4)
.
|
(Figure 4)
. Of the 1300 prostate cancer patients available for statistical analysis, 336 pertain to the TMA study sets. The mean age was 61.5 years; the mean pre-operative PSA level was 8.7 ng/ml; 60.7% had positive digital rectal examinations; 13.1% are African-American, 74.5% Caucasian, and 12.5% are of unknown ethnic background; 92.1% of these cases had a Gleason score of 6 or 7; and 73.4% had multifocal tumors. The majority of cases had organ-confined disease (pT2 56.4%). Nearly all (96.7%) cases had surgery as a monotherapy (eg, no hormone or radiation treatment). Of the 336 patients, 325 can be evaluated for PSA recurrence data after surgery with a median follow-up time of 3.7 years for all patients. Seventy-two of the 253 patients had a PSA elevation of >0.2 ng/ml (biochemical failure) after radical prostatectomy.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A further benefit of this system is that the review of TMA sample images can be done on the World-Wide Web with adequate high-speed Internet access. With the increasing availability of bandwidth, this mode of pathology review is becoming increasingly more used. This web-based review feature frees the pathologist not only from the microscope, but also from a single computer workstation and permits the assessment of TMA images from any suitable terminal on the Internet. This is critical for performing multi-institutional studies. This system as described is being used to share and evaluate research data. This makes collaborations between laboratories in different buildings, cities, and countries practical. The only limitation to this has been the speed of Internet connections. In practice the rate of image transfer speed depends on the time of day. The quality of the images has recently been assessed in an intra- and interobserver study with the goal of Gleason grading PCA over the Internet.2 The image quality was subjectively evaluated as excellent. We have also been evaluating immunostained tissues with a similar level of viewer satisfaction.3-6
TMA data has been successfully used in combination cDNA expression array data in a high-throughput manner to help validate biomarkers.7,8 We believe that linking cDNA to TMA data will be valuable. As the TMA data grows, queries could be performed based on cDNA expression results. For example, if a set of tumor cases cluster together discretely from other tumors, in addition to looking for similarities based on clinical or pathology results, one could also display TMA images to look for a similar phenotype. TMA samples that have no overlap with the ones used for cDNA analysis can also serve as a validation step for cDNA expression array findings. In recent work, we have used the TMA database to confirm cDNA expression array data (unpublished results). The same clustering algorithms used for cDNA microarray data can be used for TMA data.
We have also recently used this approach to validate putative biomarkers.4,5 Using the described system we are able to perform outcome analysis, looking for associations between protein expression levels as determined by immunohistochemistry of TMAs and adverse pathological features (extraprostatic extension, positive surgical margins, Gleason pattern 4/5 PCA) or PSA recurrence. This powerful tool allows us to perform efficient translation outcomes studies that will facilitate the evaluation of multiple biomarkers while conserving tissue-using TMAs.
One concern regards the use of an Internet browser to review
immunohistochemistry as opposed to performing standard light
microscopy. Light microscopy is the gold standard allowing for the
absolute best optical evaluation with the best detail. However, one
draw back is the potential to make mistakes as to the sample location
on the TMA. This is especially true when the pathologist reviewing the
slide has limited familiarity with TMAs, which may be the case in a
collaborative research setting. By evaluating a linked digital image,
there is a built-in set of quality controls, which obviates the need to
determine the x-y coordinates of a 0.6-mm diameter tissue sample among
400 to 600 samples. Another important byproduct of this system is that
we have established a true databank. This structure facilitates the
return of research data into the database. This will prevent
duplication of studies on the same samples. It will also allow for
computer-based experiments, where research questions can be asked by
analyzing data from multiple studies. As the database becomes rich with
experimental data, we will be able to search for associations that may
not be apparent by examining data from individual studies. The use of
digital images compromises image quality for flexibility in review and
facility in data entry. However, we feel that the loss of resolution is
minimal. As discussed above, the quality of the H&E images has recently
been assessed in an intra- and interobserver study with the goal of
Gleason grading prostate cancer over the Internet.2
Our
subjective assessment is that most immunostains can be evaluated
over the Internet. Examples from two recent projects are presented in
Figure 4
. Limitations to an Internet review would include monitor
quality and Internet speed.
The eventual development of common data elements will facilitate the sharing or results between research groups and the expansion of results to evaluate. We would anticipate that working groups would come up with common naming conventions to facilitate the sharing of data. The National Cancer Institute has set up a commondata elements web-site tool to help with this type of work (http://ciiserver4.nci.nih.gov:85/cde_tool/cde_tool.home).
The complexity of these data sets contributes to further complication for statistical analysis. Standardized solutions for the biostatistical analysis of these complex final data sets is yet another frontier for potential exploration. This DB is extremely valuable for dealing with the increasing amount of data generated by high-throughput technologies. Perhaps the use of bar code labels on such items as donor-specimen blocks and slides and as codes for various tissue diagnoses is one way to further increase the efficiency of entering voluminous data. Further robotic tissue arrayers, automated digital imaging systems (slide scanners), and automated slide stainers may offer integrated solutions for such studies.
|
Acknowledgements |
|---|
|
Footnotes |
|---|
Supported by the National Cancer Institute Specialized Program of Research Excellence in Prostate Cancer (grant P50CA69568 to University of Michigan) and Specialized Program of Research Excellence in Prostate Cancer (grant P50CA58236 to Johns Hopkins University School of Medicine).
Accepted for publication May 16, 2001.
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  R. J. Marinelli, K. Montgomery, C. L. Liu, N. H. Shah, W. Prapong, M. Nitzberg, Z. K. Zachariah, G. J. Sherlock, Y. Natkunam, R. B. West, et al. The Stanford Tissue Microarray Database Nucleic Acids Res., January 11, 2008; 36(suppl_1): D871 - D877. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. H. Ostrander, A. R. Daniel, K. Lofgren, C. G. Kleer, and C. A. Lange Breast Tumor Kinase (Protein Tyrosine Kinase 6) Regulates Heregulin-Induced Activation of ERK5 and p38 MAP Kinases in Breast Cancer Cells Cancer Res., May 1, 2007; 67(9): 4199 - 4209. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Efstathiou, P. Troncoso, S. Wen, K.-A. Do, C. A. Pettaway, L. L. Pisters, T. J. McDonnell, and C. J. Logothetis Initial Modulation of the Tumor Microenvironment Accounts for Thalidomide Activity in Prostate Cancer Clin. Cancer Res., February 15, 2007; 13(4): 1224 - 1231. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. C. Cunningham, F. Kamangar, M. P. Kim, S. Hammoud, R. Haque, C. Iacobuzio-Donahue, R. Ashfaq, S. E. Kern, A. Maitra, R. E. Heitmiller, et al. MKK4 Status Predicts Survival After Resection of Gastric Adenocarcinoma Arch Surg, November 1, 2006; 141(11): 1095 - 1099. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. C. Cunningham, F. Kamangar, M. P. Kim, S. Hammoud, R. Haque, C. A. Iacobuzio-Donahue, A. Maitra, R. Ashfaq, S. Hustinx, R. E. Heitmiller, et al. Claudin-4, mitogen-activated protein kinase kinase 4, and stratifin are markers of gastric adenocarcinoma precursor lesions. Cancer Epidemiol. Biomarkers Prev., February 1, 2006; 15(2): 281 - 287. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Mehra, S. Varambally, L. Ding, R. Shen, M. S. Sabel, D. Ghosh, A. M. Chinnaiyan, and C. G. Kleer Identification of GATA3 as a Breast Cancer Prognostic Marker by Global Gene Expression Meta-analysis Cancer Res., December 15, 2005; 65(24): 11259 - 11264. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. J. Dimitroff, L. Descheny, N. Trujillo, R. Kim, V. Nguyen, W. Huang, K. J. Pienta, J. L. Kutok, and M. A. Rubin Identification of Leukocyte E-Selectin Ligands, P-Selectin Glycoprotein Ligand-1 and E-Selectin Ligand-1, on Human Metastatic Prostate Tumor Cells Cancer Res., July 1, 2005; 65(13): 5750 - 5760. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A. Rubin, T. A. Bismar, O. Andren, L. Mucci, R. Kim, R. Shen, D. Ghosh, J. T. Wei, A. M. Chinnaiyan, H.-O. Adami, et al. Decreased {alpha}-Methylacyl CoA Racemase Expression in Localized Prostate Cancer is Associated with an Increased Rate of Biochemical Recurrence and Cancer-Specific Death Cancer Epidemiol. Biomarkers Prev., June 1, 2005; 14(6): 1424 - 1432. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
V. J. Assikis, K.-A. Do, S. Wen, X. Wang, J. H. Cho-Vega, S. Brisbay, R. Lopez, C. J. Logothetis, P. Troncoso, C. N. Papandreou, et al. Clinical and Biomarker Correlates of Androgen-Independent, Locally Aggressive Prostate Cancer with Limited Metastatic Potential Clin. Cancer Res., October 15, 2004; 10(20): 6770 - 6778. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Koopmann, N. S. Fedarko, A. Jain, A. Maitra, C. Iacobuzio-Donahue, A. Rahman, R. H. Hruban, C. J. Yeo, and M. Goggins Evaluation of Osteopontin as Biomarker for Pancreatic Adenocarcinoma Cancer Epidemiol. Biomarkers Prev., March 1, 2004; 13(3): 487 - 491. [Abstract] [Full Text]
|
|
|
|
|
|
R. Kuefer, M. D. Hofer, J. E. Gschwend, K. J. Pienta, M. G. Sanda, A. M. Chinnaiyan, M. A. Rubin, and M. L. Day The Role of an 80 kDa Fragment of E-cadherin in the Metastatic Progression of Prostate Cancer Clin. Cancer Res., December 15, 2003; 9(17): 6447 - 6452. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. R. Rhodes, M. G. Sanda, A. P. Otte, A. M. Chinnaiyan, and M. A. Rubin Multiplex Biomarker Approach for Determining Risk of Prostate-Specific Antigen-Defined Recurrence of Prostate Cancer J Natl Cancer Inst, May 7, 2003; 95(9): 661 - 668. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. T. van Heek, A. K. Meeker, S. E. Kern, C. J. Yeo, K. D. Lillemoe, J. L. Cameron, G. J. A. Offerhaus, J. L. Hicks, R. E. Wilentz, M. G. Goggins, et al. Telomere Shortening Is Nearly Universal in Pancreatic Intraepithelial Neoplasia Am. J. Pathol., November 1, 2002; 161(5): 1541 - 1547. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. Kuefer, S. Varambally, M. Zhou, P. C. Lucas, M. Loeffler, H. Wolter, T. Mattfeldt, R. E. Hautmann, J. E. Gschwend, T. R. Barrette, et al. {alpha}-Methylacyl-CoA Racemase: Expression Levels of this Novel Cancer Biomarker Depend on Tumor Differentiation Am. J. Pathol., September 1, 2002; 161(3): 841 - 848. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. A. Rubin, M. Zhou, S. M. Dhanasekaran, S. Varambally, T. R. Barrette, M. G. Sanda, K. J. Pienta, D. Ghosh, and A. M. Chinnaiyan {alpha}-Methylacyl Coenzyme A Racemase as a Tissue Biomarker for Prostate Cancer JAMA, April 3, 2002; 287(13): 1662 - 1670. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Luo, S. Zha, W. R. Gage, T. A. Dunn, J. L. Hicks, C. J. Bennett, C. M. Ewing, E. A. Platz, S. Ferdinandusse, R. J. Wanders, et al. {alpha}-Methylacyl-CoA Racemase: A New Molecular Marker for Prostate Cancer Cancer Res., April 1, 2002; 62(8): 2220 - 2226. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D Gancberg, A Di Leo, G Rouas, T Jarvinen, A Verhest, J Isola, M J Piccart, and D Larsimont Reliability of the tissue microarray based FISH for evaluation of the HER-2 oncogene in breast carcinoma J. Clin. Pathol., April 1, 2002; 55(4): 315 - 317. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. Zha, W. R. Gage, J. Sauvageot, E. A. Saria, M. J. Putzi, C. M. Ewing, D. A. Faith, W. G. Nelson, A. M. De Marzo, and W. B. Isaacs Cyclooxygenase-2 Is Up-Regulated in Proliferative Inflammatory Atrophy of the Prostate, but not in Prostate Carcinoma Cancer Res., December 1, 2001; 61(24): 8617 - 8623. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
