Predictive models for response on neoadjuvant treatment in breast cancer

From Master Projects
Jump to: navigation, search

Predictive models for response on neoadjuvant treatment in breast cancer
status: finished
Master: project within::Bioinformatics
Student name: student name::Marc Jan Bonder
Start start date:=2012/01/16
End end date:=2012/07/16
Supervisor: Jorma de Ronde
Second supervisor: Lodewyk Wessels
Second reader: has second reader::Sanne Abeln
Company: has company::NKI-AVL
Thesis: has thesis::Media:Thesis.pdf
Poster: has poster::Media:Posternaam.pdf

Signature supervisor



Although it is known that treatment with neoadjuvant agents reduces the mortality of breast cancer patients, it is also known that a substantial portion of the patients only experience the downside of the treatment. Neoadjuvant therapy is the administration of therapeutic agents before the main treatment, usually surgery to remove the breast tumor. Neoadjuvant treatments can include chemotherapy, hormonal therapy, a targeted drug or a combination of these. It has been shown that neoadjuvant treatment has the same overall survival benefits as adjuvant treatment, with the added benefit of being able to monitor response and, in some cases, enabling breast conserving surgery. Because the treatments can be extremely toxic and not all patients benefit from it, it would be very beneficial if these non-responders can be accurately separated from the responders.

In addition to tumor size, grade, lymph node involvement and patient demographics, there are now three main molecular markers that are used to stratify patients into subgroups. These molecular markers are: progesterone receptor (PR), human epidermal growth factor 2 (HER2) and the estrogen receptor (ER). The status of these markers are most commonly determined by immunohistochemistry (IHC). An example of a subgroup that gets specific treatment is the group of patients where HER2 is amplified in the tumor. If this is the case, the patients receive specific treatment consisting of trastuzumab (a targeted drug). While the subgroups show different response rates to neoadjuvant treatment, there is no stratification of patients into those that do and do not benefit from treatment.

In recent years multiple different predictive models have been proposed to predict the response of breast cancer tumors to neoadjuvant therapy1-6. Most predictive models are based on gene microarray expression data. In the paper by Borst et al.7 an overview of the available models is given and the authors comment on the usefulness and the accuracy of the available models. In this project we will select models that have previously been published and test whether these models can be validated on data available at the NKI-AVL. After testing the individual models we will try to improve the models. There are several different options we will explore to improve the models. First, we will combine the separate models in a single predictor, following a strategy similar to Zhao et al.8. Second, we will include another data type, array-comparative genomic hybridization (aCHG) data in the model. For most patients treated at the NKI-AVL both microarray expression data and aCHG data are available and by using both data types the model could possibly be improved. In addition to these approaches, we will also create our own model and validate it on publicly available expression data.

1. Ayers, M. et al. Gene expression profiles predict complete pathologic response to neoadjuvant paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide chemotherapy in breast cancer. J Clin Oncol 22, 2284-93 (2004).

2. Hess, K.R. et al. Pharmacogenomic predictor of sensitivity to preoperative chemotherapy with paclitaxel and fluorouracil, doxorubicin, and cyclophosphamide in breast cancer. Journal of clinical oncology : official journal of the American Society of Clinical Oncology 24, 4236-44 (2006).

3. Cleator, S. et al. Gene expression patterns for doxorubicin (Adriamycin) and cyclophosphamide (cytoxan) (AC) response and resistance. Breast cancer research and treatment 95, 229-33 (2006).

4. Bertucci, F. et al. Gene expression profiling for molecular characterization of inflammatory breast cancer and prediction of response to chemotherapy. Cancer research 64, 8558-65 (2004).

5. Sorlie, T. et al. Gene expression profiles do not consistently predict the clinical treatment response in locally advanced breast cancer. Molecular cancer therapeutics 5, 2914-8 (2006).

6. Gyorffy, B. et al. Gene expression profiling of 30 cancer cell lines predicts resistance towards 11 anticancer drugs at clinically achieved concentrations. International journal of cancer. Journal international du cancer 118, 1699-712 (2006).

7. Borst, P. & Wessels, L. Do predictive signatures really predict response to cancer chemotherapy? Cell Cycle 9, 4836-40 (2010).

8. Zhao, X. et al. Combining gene signatures improves prediction of breast cancer survival. PloS one 6, e17845 (2011).