|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2008.17.3567 on October 20 2008 © 2008 American Society of Clinical Oncology.
A Decade of Tissue Microarrays: Progress in the Discovery and Validation of Cancer Biomarkers
From the Department of Pathology, Yale University School of Medicine, New Haven, CT Corresponding author: David L. Rimm, MD, PhD, Department of Pathology, Yale University School of Medicine, 310 Cedar St, PO Box 208023, New Haven, CT 06520-8023; e-mail: david.rimm{at}yale.edu
This year, 2008, marks the 10-year anniversary of the development of the modern tissue microarray (TMA). During the last decade, the use of TMAs has grown steadily and accounts for a small but increasing percentage of all cancer biomarker studies performed. The growing popularity of TMA-based studies attests to their benefits in the discovery and validation of new biomarkers. This review will focus on these benefits, but also on the faults of TMAs and the challenges of TMA studies that have been overcome in the last decade. We will also discuss the role of TMAs in the latest revolution in cancer treatment, the use of targeted drug therapy.
Tissue microarray (TMA) technology was first described by Wan et al1 in 1987. However, it was not until 10 years later, when Kononen et al2 developed a device that could rapidly and reproducibly produce quality TMAs, that the technology took off (Fig 1). Early on, interest in TMA technology benefited from the (then) recent development of nucleic acid arrays and the anticipated potential of multiplexed biomarker analysis. The key benefit underlying TMA technology is the ability to assay hundreds of patient tissues arrayed on a single microscope slide. In its most common form, a core of tissue is lifted from a formalin-fixed, paraffin-embedded sample and placed in a predrilled hole in a paraffin recipient block. On sectioning, each sample is represented as a small (0.6- to 2-mm diameter) histologic section (histospot) arrayed in a grid that allows easy linkage to clinicopathologic data. The result is a single slide that contains samples from 40 to 800 patients (depending on core size). A newer alternative, the cutting edge matrix assembly array, is produced by cutting and stacking sections in a serial manner to produce arrays that represent thousands of specimens.3 Other researchers have adapted TMA technology to frozen tissues,4 cell lines,5,6 and needle biopsies.7 Most studies assay biomarkers on TMAs using immunohistochemical (IHC) techniques; although some researchers have successfully used other methods, including in situ hybridization (eg, fluorescent in situ hybridization). Here, we will examine the advantages and disadvantages of TMA technology and its use in the evaluation of tissue biomarkers.
The power of TMA technology is most clearly demonstrated when you consider the alternative. Before the TMA, researchers wishing to study biomarkers on human tissues would collect archived paraffin blocks from potentially hundreds of patients and cut histologic sections from each. These sections would be stored until they were immunostained in manageably sized batches, often over a period of weeks. A pathologist would then examine each stained slide, identify areas of tumor, and assess the level of staining—a process that most pathologists would agree is quite tedious and forces them to rely on the inherently poor human ability to assess subtle differences in intensity. From start to finish, the whole process would generally take several months, with each subsequent marker to be examined requiring a complete repetition of the process. Furthermore, the time elapsed during all the steps of the process makes it difficult to standardize many of the variables inherent in tissue management. TMAs, in contrast, provide a significantly different and ultimately far more efficient workflow. With TMAs, the work of the pathologist occurs only once and up front. Routine hematoxylin and eosin–stained sections of tumors in the cohort are analyzed, and areas of tumor are circled by the pathologist. Thus, the pathologist‘s diagnostic skills are used in a prospective manner, avoiding the potential bias of post hoc analysis that can occur with conventional whole-section analysis. These slides are then used as a template for deciding where tissue cores are taken from the corresponding paraffin blocks. Each core is embedded in a recipient paraffin block in an array, so that the identity of the specimen is retained. Once constructed, the TMA block can be sectioned potentially hundreds of times (thousands of times for cutting edge matrix assembly arrays), with each section providing a fully annotated cohort of tumors that is ready for biomarker analysis. Because the cores are taken from pathologist-identified regions of interest, IHC staining can often be scored reliably by individuals with only rudimentary training (including an unpublished study comparing an expert pathologist with his 12-year-old son; G. Sauter, personal communication, March 2008), or as discussed later, by machine. After the up-front effort of producing the TMAs for the first time, each subsequent biomarker analysis, including staining and reading, takes only a few days. In addition to the savings in time, TMAs provide several additional benefits. First, because each TMA uses only a small core from the donor blocks, each block can be used in dozens (or potentially hundreds) of newly created TMAs. This effectively amplifies the number of assays performed per patient tumor from 75 to 100 to 75,000 to 100,000.8 Second, TMAs can drive significant cost savings both in terms of reagents and technician time required to stain one slide instead of hundreds. Third, because of the inherent efficiency in processing hundreds to thousands of tumors at one time, TMAs can dramatically increase the number of tumors that can be analyzed in a given cohort compared with traditional whole-section studies. In our laboratory, TMA cohorts routinely include 350 to 700 individual patient samples, and others have published TMA studies using thousands of samples.9,10 Larger cohorts positively impact the quality, significance, and reliability of the resulting biomarker studies. This is particularly true for biomarkers that are present in only a small percentage of tumors (eg, HER2/neu), where smaller cohorts would provide insufficient numbers for analysis. Because TMAs are arrayed on a single slide, all of the tumor specimens are stained consistently, at the same time, under the same conditions, and with exactly the same antibody dilution. Arrays may be stained with histochemical stains or, more commonly, IHC stains with either chromogenic or fluorescent visualization. Example histospots using each staining method are shown in Figure 2. The standardization and uniformity is a marked contrast to traditional methods where batch-to-batch variability (and intervening time between batches) can have a profound effect on staining.11,12 TMAs can easily incorporate internal standards such as cell lines,13 purified proteins or peptides,14 non-neoplastic control tissue, and tumor replicates, which can provide a means of standardizing results across experiments. Finally, because the format of the TMA is so portable, TMAs can easily be shared, providing a new outlet for performing multi-institutional studies using training and validation sets, which represent perhaps the best methods of assuring statistical rigor.
Although there are clear benefits to the use of TMAs in prognostic biomarker discovery and validation, the technique has some significant weaknesses. TMA studies suffer from many of the same issues that plague traditional whole-section analyses. In particular, they are dependent on good quality tissue, appropriately validated antibodies, and standardized laboratory techniques. In addition, the technical skill required to produce and cut TMAs is substantially greater than that required to embed traditional blocks and cut whole sections. Although automated TMA construction devices exist, high-quality TMAs require the skills of a trained technician. TMAs are only as good as the cohorts from which they are created. Because TMA technology facilitates the development of large cohorts, researchers often use archival tissue from greater time spans, increasing the chance that variations in tissue processing techniques over time can confound results. Although formalin-fixed paraffin-embedded material can retain its antigenicity for many decades,15 researchers must ensure that there is no correlation between staining intensity and the age of archival tissue. Arguably, the most significant issue facing TMA technology is the trade-off between maximizing the number of slides cut per TMA and minimizing the loss of tissue antigenicity. Because of the initial effort required to create a TMA, each block, and subsequently each cut slide, becomes a precious resource. The use of tape transfer techniques,16 as opposed to floating sections in a water bath, permits the collection of virtually every section from a TMA block, dramatically increasing the number of TMAs that can be cut from a single block. This technique has the added advantage that it maintains the orientation of the TMA grid by preventing individual histospots from floating off the slide. Tape transfer is particularly important when cores are taken from thin (eg, nearly exhausted) blocks or biopsy material, where the potential number of TMA sections is small. Many researchers cut multiple TMA sections at one time to avoid section loss caused by need to reface the block every time it is remounted on a microtome. Unfortunately, tissue oxidation and its deleterious effects on antigenicity begin at the moment TMA slides are cut.12,17 We and others have developed methods to slow the oxidation process so that multiple slides can be cut from a block at one time without the loss of sections.17,18 In general, our laboratory prefers to cut only 10 to 20 sections at a time, paraffin coat the slides and store them in a nitrogen desiccator, and use them within 1 to 2 months of cutting. Anecdotally, slides maintained for a year using this preservation method show significant and irreversible loss of antigenicity for some antibodies, whereas they are fine for others. This variable is difficult to control and argues for limited storage of arrays once they are cut.
Because TMAs examine only a fraction of the tumor that is analyzed using traditional methods, many researchers were initially concerned that TMA cores would not adequately assess biomarkers that exhibited tissue heterogeneity. Subsequently, multiple groups have demonstrated strong correlations between TMA histospots and whole-tissue sections. The general consensus of these studies is that, in most instances, two 0.6-mm histospots adequately represent the staining seen on an entire histologic section.15,19 However, for some biomarkers that exhibit significant heterogeneity or location-dependent expression (eg, markers of hypoxia), this number could be substantially higher.20,21 In large TMA studies, sampling error is diluted by the size of the cohort, and some high-quality studies have used a single histospot for investigation of important biomarkers.10 The TMA format provides an inherent method for assessing tumor heterogeneity; small cohort test arrays can be constructed with two-fold redundancy, and heterogeneity can be assessed using a simple linear regression of the matched cores. This practice serves both as a method of validating antibody reproducibility as well as assessing tumor heterogeneity. In our experience, most biomarkers show excellent core-to-core correlations (Pearson r > 0.6; Fig 3), including some markers that are generally thought to be heterogeneous (eg, the proliferation marker Ki-67 or tumor microvessel density). Biomarkers exhibiting significant core-to-core variability can be analyzed by including additional histocores. For studies examining intratumoral expression, TMAs can be constructed to include different tumor areas (eg, the peripheral leading edge v central areas, or radial v vertical phases of melanoma) in separate histocores, thus providing an unbiased method of assessing local differences in biomarker expression.
In addition to heterogeneity within a tumor, the issue of heterogeneity between histospots taken from different depths within a tissue core is also a significant concern. Even with careful initial examination of hematoxylin and eosin–stained sections by a pathologist, there is no guarantee that sections cut from deeper levels within a TMA block will still contain tumor. This issue is of greater concern when analyzing in situ lesions, which are generally small and admixed with normal stroma. In such cases, researchers may wish to stain every fifth to tenth section for visual examination. For invasive tumors, which are generally larger with minimal intervening normal epithelium, the issue is less of a concern. Individuals with rudimentary training, as well as automated readers, are generally good at discounting nonepithelial staining, be it stroma, lymphocytes, or necrotic tumor. In cases where there is significant contamination with normal epithelium, such as with prostate carcinoma, double staining with a tumor-specific antigen may be helpful.22
Although the size of TMA histospots presented challenges to assessing tumor heterogeneity, they also provided a new opportunity for developing automated methods of analysis. Indeed, histospots are sufficiently small that rigorous molecular quantification is feasible. Because TMAs are prevalidated by a pathologist during construction, automated systems would only have to assess staining intensity and would not have to determine whether the staining was from areas of tumor. Because of this potential, a number of companies developed machines for automated TMA analysis.23 In addition to eliminating the tedious and subjective task of manually scoring immunostaining on each histospot, automated analysis permits the quantification of biomarkers in a way that matches their biologic expression, that is, on a continuous scale. Before the development of automated analysis, manual-scoring systems such as the H score attempted to represent the continuous nature of biomarker expression, but this usually resulted in relatively subjective nominal categorizations akin to clinical scales for characteristics such as pain or performance status. The inaccuracy that is an artifact of this method was illustrated in a commentary and response in Journal of Clinical Oncology.24,25 However, this approach has been accepted because it is easily achieved by pathologists without expensive tools. However, as standardization becomes more important and as targeted therapy becomes more complex, we will likely see markers where accurate analysis of quantitative expression is required. The accurate analysis of biomarker expression on TMA histospots is not a trivial endeavor, considering that many tumors have abundant and variable amounts of desmoplastic stroma admixed with tumor cells. Our analyses have demonstrated that for melanoma and breast cancer, the amount of a histospot covered by tumor cells (v stroma) is itself associated with tumor aggressiveness and is, therefore, a significant confounder in efforts to predict outcome. The automated system developed in our laboratory, the AQUA system (HistoRx, New Haven, CT), uses multicolor immunofluorescent histochemistry to distinguish between tumor and stromal elements (Fig 4). Other systems address this issue using feature extraction–and/or artificial intelligence–based methods to define tumor from stroma stained with chromogenic reagents (eg, hematoxylin). These nonfluorescent methods have the advantage of using staining techniques that are both familiar to pathologists and that can be validated by eye using a traditional light microscope. Fluorescent analysis provides other important benefits, including the ability to define and analyze different elements within tumors (eg, nuclei, Golgi, stroma, microvessels, and so on) based on colocalization with associated markers (eg, DAPI, 58K, collagen, and CD31, respectively). This type of subcellular localization can provide important information, particularly for biomarkers that function differently in different cellular compartments.26 Fluorescent-based systems also provide a wider dynamic range for analysis because of the inherent benefits in using emissive rather than absorptive detection methods.27
The time savings inherent in TMA technology exponentially increase the number of studies that a single individual can perform, making it easy to become trapped under a mountain of paraffin blocks, stained histospots, and raw data. Large TMA facilities must collect, archive, and organize many thousands of paraffin blocks and must ensure that cores taken from each block are correctly linked back to the appropriate patient. The TMA facility at Yale (New Haven, CT) uses radiofrequency identification (RFID) tags embedded in the paraffin blocks and scanned when each core is taken during array construction, minimizing transcriptional errors. Once arrays are cut and stained, the reader—be it man or machine—must ensure that expression scores match to the correct coordinates; being off by even one spot will destroy any linkage between patient and histospot. Automated TMA readers that store images for scoring help prevent transcriptional errors. Several laboratories have developed image databases that help organize images from TMA slides and permit manual scoring by individuals using remote Web-based access.28-31 Our laboratory uses a scalable relational database called Cruella (an homage to Cruella de Vil, the collector of spotted Dalmatians) to store both the expression scores from TMA spots as well as patient and tumor demographics.32 The final information hurdle is developing a coordinated approach to analyze the potentially hundreds of biomarkers arising from a single TMA cohort. The solutions to this issue are many and diverse and fall beyond the scope of this review; however, several researchers have used clustering algorithms developed for nucleic acid microarrays,33-35 random forest models,36 and genetic algorithms.37
TMAs and the normalization provided by the single-slide format provided the opportunity to collect continuous quantitative biomarker data from entire cohorts. However, this opportunity presented several new challenges. First among these was the realization that different titers of the same antibody could profoundly impact associations between recorded biomarker expression and clinical outcome. McCabe et al38 revealed that low antibody concentrations were better at discriminating between high and very high levels of expression, whereas high antibody concentrations were better at the lower end of the scale. This phenomenon has as much to do with antibody titer as it does with the fact that IHC generally requires enzymatic amplification (eg, horseradish peroxidase or alkaline phosphatase) to achieve signal intensities that are readily visible above background. In the case of the oncogene product HER-2, low anti–HER-2 titers are adept at separating the 10% to 15% of patients with marked overexpression. As expected, these tumors exhibit an aggressive phenotype. High anti–HER-2 titers, in contrast, identify a subset with low expression that exhibits equally poor outcome.39 These tumors represent the so-called triple-negative tumors—estrogen receptor, progesterone receptor, and HER-2 negative. With careful antibody titer selection, both subsets can be visualized, but the fact remains that antibody titer significantly impacts IHC results. In our lab, we routinely use small test arrays of 20 to 30 representative tumors to titer antibodies to achieve the broadest dynamic range of IHC staining. The ability to assay different antibody titers in parallel on patient cohorts is a significant advantage of the TMA format and would be extraordinarily difficult to replicate using whole-slide studies. In fact, TMAs have now been used to discover the optimal titer for a given antibody in a completely automated manner using AQUA-based analysis (M. Gustavson, personal communication, March 2008). Another challenge in using continuous rather than nominal data is the inability to quickly visualize the data as it relates to clinicopathologic variables such as patient outcome. With nominalized data (eg, manual scoring on a 0 to 3+ scale), there are discrete categories of patients. The divisions between these subsets are observer dependent and usually highly arbitrary and overlapping; however, they do subset patients for subsequent analyses. In contrast, TMAs can provide quantitative data on a continuous scale, providing the opportunity to discover biologically meaningful cut points, rather than relying on arbitrary divisions like tertiles or quartiles. To help visualize continuous data and their association with patient outcome, we developed a software program called X-tile, which permits real-time cut point visualization.40 Use of this software or other similar visualization methods (eg, spline fit analysis) can identify the low-, intermediate-, and high-expression subsets of HER-2 described earlier.41 They also provide the user with a sense of how biomarker data correlate with outcome. For instance, are the data evenly distributed throughout the range of expression, or are their distinct high and low subsets that may represent different tumor phenotypes or subclasses? Identifying clinically relevant and statistically robust cut points with continuous data represents an additional challenge. Consider a 300-tumor cohort stained for a particular biomarker. As a continuous variable, there are 299 possible cut points into high and low expression subsets. By random chance, at least one of these cut points will provide a statistically significant P value (P > .05) more than 50% of the time. Of course, even manually scored studies are not immune to similar dramatic inflations of the type I error rate because researchers often reinterpret the divisions between staining categories. It is likely that the inappropriate subsetting of data and the aforementioned issues with antibody titer account for the vast number of published biomarkers that are subsequently never validated.42 The easiest resolution to cutting continuous data is simply not to do it.43,44 Several statistical methods exist to deal with continuous data, including the widely available Cox continuous univariate analysis for survival, all of which prevent that dilution in statistical power that is inherent to cutting data.44,45 Of course, clinical decision making is dependent on patient dichotomization (ie, to treat or not to treat), and in such cases, the P value inflation associated with determining the best cut point can be corrected statistically.43,46 Ultimately, cut points are best validated by defining a single optimal cut point on a training cohort and testing it on a validation cohort. Here, TMAs can have a significant impact by easily allowing different researcher to share their cohorts with others for the purposes of validating biomarker discoveries.
In addition to biomarker discovery, TMAs can help standardize established diagnostic tests both within and between clinical laboratories. Historically, the lack of such standards (TMA based or otherwise) has hindered the development of IHC-based companion diagnostics. The development of HercepTest (Dako, Carpinteria, CA) is perhaps the best known example of this. HercepTest, which is celebrating the 10-year anniversary of its US Food and Drug Administration approval, is an IHC assay for HER-2 that helps determine which patients should receive trastuzumab. The HercepTest immediately became the standard of care. However, only rudimentary standards (positive/negative cell lines) were used as controls, and issues of reproducibility quickly prompted the development of competing assays, ultimately leading to significant controversy over which test was best. The controversy escalated to the point that a national conference was convened to define standards for this test, now referred to as the American Society of Clinical Oncology/College of American Pathologists guidelines.47 Here again, the lack of standardization and lack of published studies on standardization meant that the guidelines could only advocate proficiency testing rather than true standardization using absolute, accurate, and reproducible controls. The HercepTest experience provides a valuable lesson in the requirement for truly standardized IHC tests including objective, automated methods of analysis. New standards for clinical assays are evolving, driven both by the National Institute of Standards and Technology, the National Cancer Institute, and others.48 These standards include the ability to produce quantitative measurements of in situ biomarker concentration. Although there are skeptics that claim IHC can never be a rigorously quantitative method, others, including some studies from our lab, suggest that, when carefully controlled, automated IHC analysis can quantify biomarkers on a molecules-per-cell basis.49 As with any quantitative test (eg, enzyme-linked immunosorbent assay [ELISA]), accuracy requires the use of controlled standards performed with each sample tested and the subsequent calculation of coefficients of variation for the test. ELISA assays routinely achieve coefficients of variation of less than 10%, so it is reasonable to expect the same from automated IHC. TMAs provide a format for IHC standardization because they can closely parallel an ELISA assay. Investigators have created TMAs that are composed of cell lines expressing a range of known levels of a particular biomarker (eg, HER-2). By including cells that exhibit a broad range of expression, a TMA can be constructed that serves as a standard dilution series, akin to an ELISA standard (Fig 5). These TMA controls can then be embedded within or cut along side traditional whole sections, permitting accurate and reproducible quantification.50 Of course, TMAs can only provide an extrinsic control mechanism for stain processing. They cannot control for issues that are intrinsic to the tissue itself, such as cold ischemic time, fixation time, and quality of fixative.
The last 10 years have provided an opportunity to invent and refine new techniques in production, staining, and analysis that will help TMA technology with its next big challenge—the discovery of biomarkers that predict response to therapy. Dozens of novel targeted therapies for cancer are already in clinical trials, with perhaps hundreds more to be developed in the coming decade. Although these therapies are largely targeted to known regulatory proteins, biomarker profiles that can predict treatment response are essentially unknown. These profiles may consist of one or multiple markers that are most likely downstream from the biotherapy target or in alternative signaling pathways. Given the sheer number of potential targets, this task is ideally suited for analysis using TMAs. Therefore, it is not surprising that many cooperative groups and pharmaceutical companies are making TMA construction a major component of their clinical trials. The path from prospective marker to clinically approved diagnostic provides multiple opportunities for TMA technology. Pepe51 has defined the following five phases for biomarker development: candidate biomarker discovery, validation of antibodies directed against these markers, analysis of biomarkers in large retrospective cohort analyses, validation of the results on independent cohorts, and finally, confirmation using randomized, prospective clinical trials. As detailed in this review, TMA technology can dramatically speed the progress of each of these steps. TMAs are ideally suited to rapidly triage hundreds or thousands of potential biomarkers, permitting researchers to focus on a few likely candidates.52 The use of cell line and control TMAs has, and will, play a key role in validating antibodies for IHC. The portability of TMAs permits the easy transfer of large cohorts between institutions for biomarker validation. Finally, the ability of TMAs to provide a standard control between and within reference laboratories can facilitate the prospective confirmation of companion diagnostics. During the next 10 years, we anticipate that TMA technology will come of age. Building on the lessons of the last decade, in particular the advances made in automated quantitative analysis of IHC, TMAs are ready to play an important role in defining predictive biomarkers for future biospecific therapies.
Although all authors completed the disclosure declaration, the following author(s) indicated a financial or other interest that is relevant to the subject matter under consideration in this article. Certain relationships marked with a "U" are those for which no compensation was received; those relationships marked with a "C" were compensated. For a detailed description of the disclosure categories, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and the Disclosures of Potential Conflicts of Interest section in Information for Contributors. Employment or Leadership Position: None Consultant or Advisory Role: Robert L. Camp, HistoRX (C); David L. Rimm, HistoRx (C) Stock Ownership: Robert L. Camp, HistoRx; David L. Rimm, HistoRx Honoraria: None Research Funding: None Expert Testimony: None Other Remuneration: None
Conception and design: Robert L. Camp, David L. Rimm Financial support: David L. Rimm Collection and assembly of data: Veronique Neumeister Data analysis and interpretation: Robert L. Camp, David L. Rimm Manuscript writing: Robert L. Camp, Veronique Neumeister, David L. Rimm Final approval of manuscript: Robert L. Camp, David L. Rimm
published online ahead of print at www.jco.org on October 20, 2008. Authors’ disclosures of potential conflicts of interest and author contributions are found at the end of this article.
1. Wan WH, Fortuna MB, Furmanski P: A rapid and efficient method for testing immunohistochemical reactivity of monoclonal antibodies against multiple tissue samples simultaneously. J Immunol Methods 103:121-129, 1987[CrossRef][Medline] 2. Kononen J, Bubendorf L, Kallioniemi A, et al: Tissue microarrays for high-throughput molecular profiling of tumor specimens. Nat Med 4:844-847, 1998[CrossRef][Medline] 3. LeBaron M, Crismon H, Utama F, et al: Ultrahigh density microarrays of solid samples. Nat Methods 2:511-513, 2005[CrossRef][Medline] 4. Schoenberg Fejzo M, Slamon DJ: Frozen tumor tissue microarray technology for analysis of tumor RNA, DNA, and proteins. Am J Pathol 159:1645-1650, 2001 5. Li R, Ni J, Bourne PA, et al: Cell culture block array for immunocytochemical study of protein expression in cultured cells. Appl Immunohistochem Mol Morphol 13:85-90, 2005[CrossRef][Medline] 6. Montgomery K, Zhao S, van de Rijn M, et al: A novel method for making "tissue" microarrays from small numbers of suspension cells. Appl Immunohistochem Mol Morphol 13:80-84, 2005[CrossRef][Medline] 7. Datta MW, Kahler A, Macias V, et al: A simple inexpensive method for the production of tissue microarrays from needle biopsy specimens: Examples with prostate cancer. Appl Immunohistochem Mol Morphol 13:96-103, 2005[CrossRef][Medline] 8. Rimm DL, Camp RL, Charette LA, et al: Tissue microarray: A new technology for amplification of tissue resources. Cancer J 7:24-31, 2001[Medline] 9. Andersen CL, Monni O, Wagner U, et al: High-throughput copy number analysis of 17q23 in 3520 tissue specimens by fluorescence in situ hybridization to tissue microarrays. Am J Pathol 160:73-79, 2002 10. Cheang MC, Treaba DO, Speers CH, et al: Immunohistochemical detection using the new rabbit monoclonal antibody SP1 of estrogen receptor in breast cancer is superior to mouse monoclonal antibody 1D5 in predicting survival. J Clin Oncol 24:5637-5644, 2006 11. Pozner-Moulis S, Cregger M, Camp RL, et al: Antibody validation by quantitative analysis of protein expression using expression of Met in breast cancer as a model. Lab Invest 87:251-260, 2007[CrossRef][Medline] 12. Fergenbaum JH, Garcia-Closas M, Hewitt SM, et al: Loss of antigenicity in stored sections of breast cancer tissue microarrays. Cancer Epidemiol Biomarkers Prev 13:667-672, 2004 13. Andersson AC, Strömberg S, Bäckvall H, et al: Analysis of protein expression in cell microarrays: A tool for antibody-based proteomics. J Histochem Cytochem 54:1413-1423, 2006 14. Vani K, Sompuram SR, Fitzgibbons P, et al: National HER2 proficiency test results using standardized quantitative controls: Characterization of laboratory failures. Arch Pathol Lab Med 132:211-216, 2008[Medline] 15. Camp RL, Charette LA, Rimm DL: Validation of tissue microarray technology in breast carcinoma. Lab Invest 80:1943-1949, 2000[Medline] 16. Rimm DL, Camp RL, Charette LA, et al: Amplification of tissue by construction of tissue microarrays. Exp Mol Pathol 70:255-264, 2001[CrossRef][Medline] 17. DiVito KA, Charette LA, Rimm DL, et al: Long-term preservation of antigenicity on tissue microarrays. Lab Invest 84:1071-1078, 2004[CrossRef][Medline] 18. Dhir R: Tissue microarrays: An overview. Methods Mol Biol 441:91-103, 2008[CrossRef][Medline] 19. Torhorst J, Bucher C, Kononen J, et al: Tissue microarrays for rapid linking of molecular changes to clinical endpoints. Am J Pathol 159:2249-2256, 2001 20. Iakovlev V, Pintilie M, Morrison A, et al: Effect of distributional heterogeneity on the analysis of tumor hypoxia based on carbonic anhydrase IX. Lab Invest 87:1206-1217, 2007[CrossRef][Medline] 21. Linderoth J, Ehinger M, Akerman M, et al: Tissue microarray is inappropriate for analysis of BCL6 expression in diffuse large B-cell lymphoma. Eur J Haematol 79:146-149, 2007[CrossRef][Medline] 22. Rubin MA, Zerkowski MP, Camp RL, et al: Quantitative determination of expression of the prostate cancer protein alpha-methylacyl-CoA racemase using automated quantitative analysis (AQUA): A novel paradigm for automated and continuous biomarker measurements. Am J Pathol 164:831-840, 2004 23. Giltnane JM, Rimm D: Technology insight: Identification of biomarkers with tissue microarray technology. Nat Clin Pract Oncol 1:104-111, 2004[Medline] 24. Schnitt SJ: Estrogen receptor testing of breast cancer in current clinical practice: What's the question? J Clin Oncol 24:1797-1799, 2006 25. Rimm DL, Giltnane J, Moeder C, et al: Bimodal population or pathologist artifact? J Clin Oncol 25:2487-2488, 2007 26. Kluger HM, Siddiqui S, Angeletti C, et al: Classification of renal cell carcinoma (RCC) based on expression of VEGF and VEGF receptors in both tumor cells and endothelial cells. Lab Invest [epub ahead of print on July 14, 2008] 27. Rimm D: What brown cannot do for you. Nat Biotechnol 24:914-916, 2006[CrossRef][Medline] 28. Bova GS, Parmigiani G, Epstein JI, et al: Web-based tissue microarray image data analysis: Initial validation testing through prostate cancer Gleason grading. Hum Pathol 32:417-427, 2001[CrossRef][Medline] 29. Manley S, Mucci NR, De Marzo AM, et al: Relational database structure to manage high-density tissue microarray data and images for pathology studies focusing on clinical outcome: The prostate specialized program of research excellence model. Am J Pathol 159:837-843, 2001 30. Chen W, Reiss M, Foran DJ: A prototype for unsupervised analysis of tissue microarrays for cancer research and diagnostics. IEEE Trans Inf Technol Biomed 8:89-96, 2004[CrossRef][Medline] 31. Faith D, Isaacs W, Morgan J, et al: Trefoil factor 3 overexpression in prostatic carcinoma: Prognostic importance using tissue microarrays. Prostate 61:215-227, 2004[CrossRef][Medline] 32. Cowan JD, Rimm D, Tuck DP: Cruella: Developing a scalable tissue microarray data management system. Arch Pathol Lab Med 130:817-822, 2006[Medline] 33. Makretsov NA, Huntsman DG, Nielsen TO, et al: Hierarchical clustering analysis of tissue microarray immunostaining data identifies prognostically significant groups of breast carcinoma. Clin Cancer Res 10:6143-6151, 2004 34. Au NH, Cheang M, Huntsman DG, et al: Evaluation of immunohistochemical markers in non-small cell lung cancer by unsupervised hierarchical clustering analysis: A tissue microarray study of 284 cases and 18 markers. J Pathol 204:101-109, 2004[CrossRef][Medline] 35. Siddiqui SF, Pawelek J, Handerson T, et al: Coexpression of beta1,6-N-acetylglucosaminyltransferase V glycoprotein substrates defines aggressive breast cancers with poor outcome. Cancer Epidemiol Biomarkers Prev 14:2517-2523, 2005 36. Shi T, Seligson D, Belldegrun A, et al: Tumor classification by tissue microarray profiling: Random forest clustering applied to renal cell carcinoma. Mod Pathol 18:547-557, 2005[CrossRef][Medline] 37. Dolled-Filhart M, Ryden L, Cregger M, et al: Classification of breast cancer using genetic algorithms and tissue microarrays. Clin Cancer Res 12:6459-6468, 2006 38. McCabe A, Dolled-Filhart M, Camp RL, et al: Automated quantitative analysis (AQUA) of in situ protein expression, antibody concentration, and prognosis. J Natl Cancer Inst 97:1808-1815, 2005 39. Camp RL, Dolled-Filhart M, King BL, et al: Quantitative analysis of breast cancer tissue microarrays shows that both high and normal levels of HER2 expression are associated with poor outcome. Cancer Res 63:1445-1448, 2003 40. Camp RL, Dolled-Filhart M, Rimm DL: X-tile: A new bio-informatics tool for biomarker assessment and outcome-based cut-point optimization. Clin Cancer Res 10:7252-7259, 2004 41. Jensen KC, Turbin DA, Leung S, et al: New cutpoints to identify increased HER2 copy number: Analysis of a large, population-based cohort with long-term follow-up. Breast Cancer Res Treat [epub ahead of print on January 9, 2008] 42. Hilsenbeck SG, Clark GM, McGuire WL: Why do so many prognostic factors fail to pan out? Breast Cancer Res Treat 22:197-206, 1992[CrossRef][Medline] 43. Altman DG, Lausen B, Sauerbrei W, et al: Dangers of using "optimal" cutpoints in the evaluation of prognostic factors. J Natl Cancer Inst 86:829-835, 1994 44. Altman D, Royston P: The cost of dichotomising continuous variables. BMJ 332:1080, 2006 45. Royston P, Altman D, Sauerbrei W: Dichotomizing continuous predictors in multiple regression: A bad idea. Stat Med 25:127-141, 2006[CrossRef][Medline] 46. Miller R, Siegmund D: Maximally selected chi square statistics. Biometrics 38:1011-1016, 1982 47. Wolff AC, Hammond ME, Schwartz JN, et al: American Society of Clinical Oncology/College of American Pathologists guideline recommendations for human epidermal growth factor receptor 2 testing in breast cancer. J Clin Oncol 25:118-145, 2007 48. Xiao Y, Gao X, Gannot G, et al: Quantitation of HER2 and telomerase biomarkers in solid tumors with IgY antibodies and nanocrystal detection. Int J Cancer 122:2178-2186, 2008[CrossRef][Medline] 49. Dolled-Filhart M, McCabe A, Giltnane J, et al: Quantitative in situ analysis of beta-catenin expression in breast cancer shows decreased expression is associated with poor outcome. Cancer Res 66:5487-5494, 2006 50. Chung GG, Zerkowski MP, Ghosh S, et al: Quantitative analysis of estrogen receptor heterogeneity in breast cancer. Lab Invest 87:662-669, 2007[CrossRef][Medline] 51. Pepe MS: Evaluating technologies for classification and prediction in medicine. Stat Med 24:3687-3696, 2005[CrossRef][Medline] 52. Persson A, Hober S, Uhlén M: A human protein atlas based on antibody proteomics. Curr Opin Mol Ther 8:185-190, 2006[Medline] Submitted March 27, 2008; accepted August 7, 2008.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2008 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
