Development of the 21-gene assay and its application in clinical practice and clinical trials

Predictive signatures for chemotherapy sensitivity in breast cancer: are they ready for use in the clinic?

Markers that predict the sensitivity of tumours to chemotherapy must address two questions: (a) which tumours are more likely to respond to chemotherapy? and (b) what is the optimal chemotherapy regimen for a specific tumour or group of tumours? To answer these questions will require markers of general chemosensitivity and drug-specific chemosensitivity, respectively. Beyond these fundamental questions lies an important practical question: are the predictive markers in the current literature ready for routine clinical use? The focus of this paper is to address this practical question. We will first review retrospective trials that have reported promising chemotherapy signatures, presenting in a comprehensive manner for the non bio-informatician the different methods used so far. In addition, we will summarise prospective trials (either ongoing or under development) designed to test the multigene classifiers currently thought to predict chemosensitivity. Finally, we will discuss why microarray studies have so far failed to identify new targets, and how we might be able to improve on these results through large-scale genotyping of tumours.

A radiation-derived gene expression signature predicts clinical outcome for breast cancer patients

Activation of the DNA damage response pathway is a hallmark for early tumorigenesis, while loss of pathway activity is associated with disease progression. Thus we hypothesized that a gene expression signature associated with the DNA damage response may serve as a prognostic signature for outcome in cancer patients. We identified ionizing radiation-responsive transcripts in human lymphoblast cells derived from 12 individuals and used this signature to screen a panel of cancer data sets for the ability to predict long-term survival of cancer patients. We demonstrate that gene sets induced or repressed by ionizing radiation can predict clinical outcome in two independent breast cancer data sets, and we compare the radiation signature to previously described gene expression-based outcome predictors. While genes repressed in response to radiation likely represent the well-characterized proliferation signature predictive of breast cancer outcome, genes induced by radiation likely encode additional information representing other deregulated biological properties of tumors such as checkpoint or apoptotic responses.

A kernel-based integration of genome-wide data for clinical decision support

Background

Although microarray technology allows the investigation of the transcriptomic make-up of a tumor in one experiment, the transcriptome does not completely reflect the underlying biology due to alternative splicing, post-translational modifications, as well as the influence of pathological conditions (for example, cancer) on transcription and translation. This increases the importance of fusing more than one source of genome-wide data, such as the genome, transcriptome, proteome, and epigenome. The current increase in the amount of available omics data emphasizes the need for a methodological integration framework.

Methods

We propose a kernel-based approach for clinical decision support in which many genome-wide data sources are combined. Integration occurs within the patient domain at the level of kernel matrices before building the classifier. As supervised classification algorithm, a weighted least squares support vector machine is used. We apply this framework to two cancer cases, namely, a rectal cancer data set containing microarray and proteomics data and a prostate cancer data set containing microarray and genomics data. For both cases, multiple outcomes are predicted.

Results

For the rectal cancer outcomes, the highest leave-one-out (LOO) areas under the receiver operating characteristic curves (AUC) were obtained when combining microarray and proteomics data gathered during therapy and ranged from 0.927 to 0.987. For prostate cancer, all four outcomes had a better LOO AUC when combining microarray and genomics data, ranging from 0.786 for recurrence to 0.987 for metastasis.

Conclusions

For both cancer sites the prediction of all outcomes improved when more than one genome-wide data set was considered. This suggests that integrating multiple genome-wide data sources increases the predictive performance of clinical decision support models. This emphasizes the need for comprehensive multi-modal data. We acknowledge that, in a first phase, this will substantially increase costs; however, this is a necessary investment to ultimately obtain cost-efficient models usable in patient tailored therapy.

Recent advances in systemic therapy: new diagnostics and biological predictors of outcome in early breast cancer

The key to optimising our approach in early breast cancer is to individualise care. Each patient has a tumour with innate features that dictate their chance of relapse and their responsiveness to treatment. Often patients with similar clinical and pathological tumours will have markedly different outcomes and responses to adjuvant intervention. These differences are encoded in the tumour genetic profile. Effective biomarkers may replace or complement traditional clinical and histopathological markers in assessing tumour behaviour and risk. Development of high-throughput genomic technologies is enabling the study of gene expression profiles of tumours. Genomic fingerprints may refine prediction of the course of disease and response to adjuvant interventions. This review will focus on the role of multiparameter gene expression analyses in early breast cancer, with regards to prognosis and prediction. The prognostic role of genomic signatures, particularly the Mammaprint and Rotterdam signatures, is evolving. With regard to prediction of outcome, the Oncotype Dx multigene assay is in clinical use in tamoxifen treated patients. Extensive research continues on predictive gene identification for specific chemotherapeutic agents, particularly the anthracyclines, taxanes and alkylating agents.

Current and future approach to the pathologist's assessment for targeted therapy in breast cancer

Breast cancer is a common disease in the population. Contrary to public perception, it is a heterogeneous disease with varying morphology, prognosis and response to therapy. The pathological analysis is at the heart of information provided to surgeons and oncologists to plan further management. The pathologist is increasingly asked to test for biomarkers that provide prognostic and predictive information to direct treatment. Staining cancers for ER, PgR and HER2 has become routine and it is likely that addition of other biomarkers including 'basal markers', VEGF and growth factor receptors such as HER1 (EGFR) will soon follow. Microarray based genomic, transcription and proteomic methods are changing our classification systems and identifying novel targets for the development of new therapeutics. It is important for pathologists to appreciate and embrace the new developments as they will impact on daily clinical practice and require accurate assessment of biomarkers to determine treatment options as part of multidisciplinary teams.

Prediction of breast cancer metastasis by genomic profiling: where do we stand?

Current concepts conceive “breast cancer” as a complex disease that comprises several very different types of neoplasms. Nonetheless, breast cancer treatment has considerably improved through early diagnosis, adjuvant chemotherapy, and endocrine treatments. The limited prognostic power of classical classifiers determines considerable over-treatment of women who either do not benefit from, or do not at all need, chemotherapy. Several gene expression based molecular classifiers (signatures) have been developed for a more reliable prognostication. Gene expression profiling identifies profound differences in breast cancers, most probably as a consequence of different cellular origin and different driving mutations and can therefore distinguish the intrinsic propensity to metastasize. Existing signatures have been shown to be useful for treatment decisions, although they have been developed using relatively small sample numbers. Major improvements are expected from the use of large datasets, subtype specific signatures and from the re-introduction of functional information. We show that molecular signatures encounter clear limitations given by the intrinsic probabilistic nature of breast cancer metastasis. Already today, signatures are, however, useful for clinical decisions in specific cases, in particular if the personal inclination of the patient towards different treatment strategies is taken into account.

Recent advances in systemic therapy: advances in adjuvant systemic chemotherapy of early breast cancer

Adjuvant treatment for early breast cancer is an evolving field. Since the advent of the initial cyclophosphamide, methotrexate and 5-fluorouracil (CMF) regimens, which reduced risk for recurrence and death, anthracyclines and subsequently taxanes were added to the cytotoxic armamentarium for use sequentially or in combination in the adjuvant setting. The efficacy and toxicity of each chemotherapy regimen must be viewed within the context of host co-morbidities and the specific biologic phenotype of the tumor. In the era of mammographic screening, small, node-negative breast cancer is the most frequent presentation of the disease. Patient selection for adjuvant chemotherapy has become a key issue. Traditional prognostic factors continue to be of value in determining the risk for relapse, but new and sophisticated genomic tools (such as Oncotype Dx^® and Mammaprint^®) are now available and may improve our ability to select patients. For those patients who do require adjuvant chemotherapy, the 'one size fits all' paradigm should never again feature in the treatment of early breast cancer, following the important insights yielded by biomarker research to identify those who will benefit the most from a particular drug. In this review we focus on some of the current controversies and potential future steps in adjuvant chemotherapy for treatment of early breast cancer.

Factors influencing the statistical power of complex data analysis protocols for molecular signature development from microarray data

BACKGROUND

Critical to the development of molecular signatures from microarray and other high-throughput data is testing the statistical significance of the produced signature in order to ensure its statistical reproducibility. While current best practices emphasize sufficiently powered univariate tests of differential expression, little is known about the factors that affect the statistical power of complex multivariate analysis protocols for high-dimensional molecular signature development.

METHODOLOGY/PRINCIPAL FINDINGS

We show that choices of specific components of the analysis (i.e., error metric, classifier, error estimator and event balancing) have large and compounding effects on statistical power. The effects are demonstrated empirically by an analysis of 7 of the largest microarray cancer outcome prediction datasets and supplementary simulations, and by contrasting them to prior analyses of the same data.

CONCLUSIONS/SIGNIFICANCE

THE FINDINGS OF THE PRESENT STUDY HAVE TWO IMPORTANT PRACTICAL IMPLICATIONS: First, high-throughput studies by avoiding under-powered data analysis protocols, can achieve substantial economies in sample required to demonstrate statistical significance of predictive signal. Factors that affect power are identified and studied. Much less sample than previously thought may be sufficient for exploratory studies as long as these factors are taken into consideration when designing and executing the analysis. Second, previous highly-cited claims that microarray assays may not be able to predict disease outcomes better than chance are shown by our experiments to be due to under-powered data analysis combined with inappropriate statistical tests.

Gene expression profiling and prediction of clinical outcome in ovarian cancer

Epithelial ovarian cancer is the most lethal gynaecological cancer. Despite debulking surgery and platinum/taxane-based chemotherapy, the prognosis remains poor with approximately 25% 5-year survival. Current histo-clinical prognostic factors are insufficient to capture the complex cascade of events that drive the heterogeneous clinical behaviour of the disease. There is a crucial need to identify new prognostic subclasses of disease as well as new therapeutic targets. Today, DNA microarrays allow the simultaneous and quantitative analysis of the mRNA expression levels of thousands of genes in a tumour sample. They have been applied to ovarian cancer research for predicting initial surgical resectability, survival and response to first-line chemotherapy. The first results are promising. In this review, we describe recent applications of DNA microarrays in ovarian cancer research and discuss some issues to address in the near future to allow the technology to reach its full potential in clinical practice.

Breast cancer subtypes and response to docetaxel in node-positive breast cancer: use of an immunohistochemical definition in the BCIRG 001 trial

PURPOSE

To investigate the prognostic and predictive significance of subtyping node-positive early breast cancer by immunohistochemistry in a clinical trial of a docetaxel-containing regimen.

METHODS

Pathologic data from a central laboratory were available for 1,350 patients (91%) from the BCIRG 001 trial of docetaxel, doxorubicin, and cyclophosphamide (TAC) versus fluorouracil, doxorubicin, and cyclophosphamide (FAC) for operable node-positive breast cancer. Patients were classified by tumor characteristics as (1) triple negative (estrogen receptor [ER]-negative, progesterone receptor [PR]-negative, HER2/neu [HER2]-negative), (2) HER2 (HER2-positive, ER-negative, PR-negative), (3) luminal B (ER-positive and/or PR-positive and either HER2-positive and/or Ki67(high)), and (4) luminal A (ER-positive and/or PR-positive and not HER2-positive or Ki67(high)), and assessed for prognostic significance and response to adjuvant chemotherapy.

RESULTS

Patients were subdivided into triple negative (14.5%), HER2 (8.5%), luminal B (61.1%), and luminal A (15.9%). Three-year disease-free survival (DFS) rates (P values with luminal B as referent) were 67% (P < .0001), 68% (P = .0008), 82% (referent luminal B), and 91% (P = .0027), respectively, with hazard ratios of 2.22, 2.12, and 0.46. Improved 3-year DFS with TAC was found in the luminal B group (P = .025) and a combined ER-positive/HER2-negative group treated with tamoxifen (P = .041), with a marginal trend in the triple negatives (P = .051) and HER2 (P = .068) subtypes. No DFS advantage was seen in the luminal A population.

CONCLUSION

A simple immunopanel can divide breast cancers into biologic subtypes with strong prognostic effects. TAC significantly complements endocrine therapy in patients with luminal B subtype and, in the absence of targeted therapy, is effective in the triple-negative population.

Genetics and genome-wide association studies: surgery-guided algorithm and promise for future breast cancer personalized surgery

Although personalized medicine and oncology in clinical practice is still a dream, some isolated first steps have been taken. Effective preventive and therapeutic interventions are now tailored in some individual breast cancer patients on the basis of BRCA, estrogen receptor and HER2 status. Personal genome-wide tests hold rational promise toward a true personalized management. The recent dramatic increase of more aggressive surgery by 150% in the USA is effective for preventing local recurrence and contralateral breast cancer but represents a surgical overtreatment that may harm patients and health systems. This article is based on three subpopulations: familial BRCA-positive patients and BRCA-negative patients, and sporadic breast cancer patients. Combining established classic and new risk factors, including familial BRCA susceptibility to breast cancer, an integrated surgery-guided algorithm for clinical validity is proposed. Future genome-wide association studies larger than those currently available using newer genotyping platforms with more than 1 million single-nucleotide polymorphisms and copy number variants will complete the genetic map. Due to the small effects of all these risk variants, further functional studies to explore the intracellular interactions of these variants and signaling pathways networks will be required. Ultimately, large, prospective, population-based studies recording family and medical history and genetic and environmental risk factors will lead to true personalized breast cancer local control.

Validation study of the prognostic value of cyclin-dependent kinase (CDK)-based risk in Caucasian breast cancer patients

In a Japanese study, cyclin-dependent kinase (CDK) based risk determined by CDK 1 and 2 activities was associated with risk of distance recurrence in early breast cancer patients. The aim of our study was to validate this risk categorization in European early breast cancer patients. We retrospectively analyzed frozen breast cancer specimens of 352 Dutch patients with histologically confirmed primary invasive early breast cancer. CDK-based risk was determined in tumour tissues by calculating a risk score (RS) according to kinases activity and protein mass concentration assay without the knowledge of outcome. Determination of CDK-based risk was feasible in 184 out of 352 (52%) tumours. Median follow-up of these patients was 15 years. In patients not receiving systemic treatment, the proportions of risk categories were 44% low, 16% intermediate, and 40% high CDK-based risk. These groups remained significant after univariate and multivariate Cox-regression analysis. Factors associated with a shorter distant recurrence-free period were positive lymph nodes, mastectomy with radiotherapy, and high CDK-based risk. There was no significant correlation with overall survival (OS). CDK-based risk is a prognostic marker of distance recurrence of patients with early breast cancer. More validation would be warranted to use of CDK-based risk into clinical practice.

Emerging biomarkers and new understanding of traditional markers in personalized therapy for breast cancer

The era of personalized medicine is likely to see an escalation in the use of biomarkers to ensure breast cancer patients receive optimal treatment. A combination of prognostic and predictive biomarkers should enable better quantification of the residual risk faced by patients and indicate the potential value of additional treatment. Established biomarkers such as estrogen receptor and progesterone receptor already play a significant role in the selection of patients for endocrine therapy. Human epidermal growth factor receptor 2 (HER2) is recognized as a strong predictor of response to trastuzumab whereas, more recently, the role of estrogen receptor and HER2 as negative and positive indicators for chemotherapy has also been explored. Ki67 has traditionally been recognized as a modest prognostic factor, but recent neoadjuvant studies suggest that on-treatment measurement may be a more effective predictor of treatment efficacy for both endocrine treatment and chemotherapy. The last decade has seen the emergence of numerous multigene expression profiles that aim to outdo traditional predictive and prognostic factors. The Oncotype DX assay and the MammaPrint profile are currently undergoing prospective clinical trials to clearly define their role. Other gene expression-based assays also show potential but are yet to be tested clinically. Rigorous comparison of these emerging markers with current treatment selection criteria will be required to determine whether they offer significant benefit to justify their use.

National Academy of Clinical Biochemistry Laboratory Medicine Practice Guidelines for Use of Tumor Markers in Testicular, Prostate, Colorectal, Breast, and Ovarian Cancers

Background: Updated National Academy of Clinical Biochemistry (NACB) Laboratory Medicine Practice Guidelines for the use of tumor markers in the clinic have been developed.

Methods: Published reports relevant to use of tumor markers for 5 cancer sites—testicular, prostate, colorectal, breast, and ovarian—were critically reviewed.

Results: For testicular cancer, α-fetoprotein, human chorionic gonadotropin, and lactate dehydrogenase are recommended for diagnosis/case finding, staging, prognosis determination, recurrence detection, and therapy monitoring. α-Fetoprotein is also recommended for differential diagnosis of nonseminomatous and seminomatous germ cell tumors. Prostate-specific antigen (PSA) is not recommended for prostate cancer screening, but may be used for detecting disease recurrence and monitoring therapy. Free PSA measurement data are useful for distinguishing malignant from benign prostatic disease when total PSA is &lt;10 μg/L. In colorectal cancer, carcinoembryonic antigen is recommended (with some caveats) for prognosis determination, postoperative surveillance, and therapy monitoring in advanced disease. Fecal occult blood testing may be used for screening asymptomatic adults 50 years or older. For breast cancer, estrogen and progesterone receptors are mandatory for predicting response to hormone therapy, human epidermal growth factor receptor-2 measurement is mandatory for predicting response to trastuzumab, and urokinase plasminogen activator/plasminogen activator inhibitor 1 may be used for determining prognosis in lymph node–negative patients. CA15-3/BR27–29 or carcinoembryonic antigen may be used for therapy monitoring in advanced disease. CA125 is recommended (with transvaginal ultrasound) for early detection of ovarian cancer in women at high risk for this disease. CA125 is also recommended for differential diagnosis of suspicious pelvic masses in postmenopausal women, as well as for detection of recurrence, monitoring of therapy, and determination of prognosis in women with ovarian cancer.

Conclusions: Implementation of these recommendations should encourage optimal use of tumor markers.

A 38-gene expression signature to predict metastasis risk in node-positive breast cancer after systemic adjuvant chemotherapy: a genomic substudy of PACS01 clinical trial

Currently, no prognostic gene-expression signature (GES) established from node-positive breast cancer cohorts, able to predict evolution after systemic adjuvant chemotherapy, exists. Gene-expression profiles of 252 node-positive breast cancer patients (median follow-up: 7.7 years), mostly included in a randomized clinical trial (PACS01), receiving systemic adjuvant regimen, were determined by means of cDNA custom array. In the training cohort, we established a GES composed of 38 genes (38-GES) for the purpose of predicting metastasis-free survival. The 38-GES yielded unadjusted hazard ratio (HR) of 4.86 (95% confidence interval = 2.76-8.56). Even when adjusted with the best two clinicopathological prognostic indexes: Nottingham prognostic index (NPI) and Adjuvant!, 38-GES HRs were 3.30 (1.81-5.99) and 3.40 (1.85-6.24), respectively. Furthermore, 38-GES improved NPI and Adjuvant! classification. In particular, NPI intermediate-risk patients were divided into 2/3 close to low-risk group and 1/3 close to high-risk group (HR = 6.97 [2.51-19.36]). Similarly, Adjuvant! intermediate-risk patients were divided into 2/3 close to low-risk group and 1/3 close to high-risk group (HR = 4.34 [1.64-11.48]). The 38-GES was validated on gene-expression datasets from three external node-positive breast cancer subcohorts (n = 224) generated from different microarray platforms, with HR = 2.95 (1.74-5.01). Moreover, 38-GES showed prognostic performance in supplementary cohorts with different lymph-node status and endpoints (1,040 new patients). The 38-GES represents a robust tool able to type systemic adjuvant treated node-positive patients at high risk of metastatic relapse, and is especially powerful to refine NPI and Adjuvant! classification for those patients.

Gene expression predictors in breast cancer: current status, limitations and perspectives

Breast cancer is characterised by a wide heterogeneity regarding outcome and drug sensitivity. A better prediction of these two parameters at the individual level should improve patient management and therefore also improve both the quality of life and the overall survival of the patient. Several molecular predictors for prognosis (MammaPrint or Oncotype DX) and drug prediction (DLD30, SET index) have been generated using DNA-based arrays or RT-PCR, some of these being tested in phase III trials. Although they exhibit good metric performance and should improve the quality of care in the next decade, these predictors are considered suboptimal regarding the potential of the technology. New study design and arrays should generate more powerful second generation gene signatures.

Tailoring to RB: tumour suppressor status and therapeutic response

The retinoblastoma tumour suppressor (RB) is a crucial regulator of cell-cycle progression that is invoked in response to a myriad of anti-mitogenic signals. It has been hypothesized that perturbations of the RB pathway confer a synonymous proliferative advantage to tumour cells; however, recent findings demonstrate context-specific outcomes associated with such lesions. Particularly, loss of RB function is associated with differential response to wide-ranging therapeutic agents. Thus, the status of this tumour suppressor may be particularly informative in directing treatment regimens.

RNA expression analysis from formalin fixed paraffin embedded tissues

Formalin fixation and paraffin embedding (FFPE) is the most commonly used method worldwide for tissue storage. This method preserves the tissue integrity but causes extensive damage to nucleic acids stored within the tissue. As methods for measuring gene expression such as RT-PCR and microarray are adopted into clinical practice there is an increasing necessity to access the wealth of information locked in the Formalin fixation and paraffin embedding archives. This paper reviews the progress in this field and discusses the unique opportunities that exist for the application of these techniques in the development of personalized medicine.

Review of gene-expression profiling and its clinical use in breast cancer

Despite advances in the treatment of early-stage breast cancer, physicians still lack the ability to accurately predict which individual patients will relapse and would benefit from adjuvant chemotherapy. Traditional clinicopathologic factors are important in helping to determine risk of relapse, but do not fully account for the biologic complexity of breast cancer. Gene-expression profiling has provided us with insight into the heterogeneity of breast cancer and led to the development of prognostic and predictive molecular gene signature models designed to aid in clinical decision-making. However, it remains to be determined how much refinement in prognosis genomic models provide over standard clinicopathologic features and whether these refinements translate into improvements in clinical practice. On-going large prospective multi-center clinical trials will provide us with information regarding the clinical utility of two of these assays, but for now, implementation of these models into widespread clinical practice remains limited.