Prediction of the therapeutic response to paclitaxel by gene expression profiling in neoadjuvant chemotherapy for breast cancer

Reviewer: S. Jack Wei, MD
Abramson Cancer Center of the University of Pennsylvania
Last Modified: June 5, 2004

Share article


Presenter: M. Yoshimoto
Presenter's Affiliation: Cancer Institute Hospital, Tokyo, Japan
Type of Session: Scientific

Background

  • Many breast cancer patients receive chemotherapy as part of their treatment program
  • A common chemotherapeutic agent used in breast cancer is paclitaxel
  • Response to paclitaxel varies greatly from patient to patient
  • Identification of patients who are more likely to respond to paclitaxel would help to customize treatment of breast cancer and avoid unnecessary treatment with chemotherapy of patients who are unlikely to respond to treatment

Materials and Methods

  • Needle samples from 75 patients with primary breast cancer (>3 cm) were taken prior to treatment
  • Patients were divided into 5 groups depending on their response to neoadjuvant  paclitaxel.
    • Group 1: extremely resistant (n=6)
    • Group 2: resistant (n=5)
    • Group 3: moderate responder (n=11)
    • Group 4: responder (n=5)
    • Group 5: high responder (n=7)
  • RNA was extracted from the needle samples and profiled on cDNA microarrays of 23,000 human transcripts
  • RNA expression was validated using semi-quantitative RT-PCR
  • Differentially expressed genes between the extremely resistant (Group 1) and high responder (Group 5) groups were selected by Mann-Whittney U-test (p<0.05)
  • A high score predictive set of gene was selected from these genes using a machine-learning method comparing the non-responders (Groups 1,2) and the responders (Groups 3,4,5)
  • The predictive set of genes was tested using leave-one-out cross-validation on the original research set of patients
  • Single nucleotide polymorphism (SNP) analysis was performed to identify genes that predict for side effects from treatment

Results

  • Of the 75 patients initially included, 24 patients were excluded due to lack of sufficient material to perform the microarray analysis
  • An additional 17 patients were removed from the research set to form a validation set, leaving 34 total patients from which the gene set was selected
  • Overall, clinical response rate to treatment was 74% with a pathologic complete response rate of 3%
  • Grade III/IV toxicity to chemotherapy treatment was seen in 7% of patients
  • 66 differentially expressed genes were selected that discriminated between Groups 1 and 5
  • A predictive set of 7 genes was selected from these genes using the machine-learning repeats method
  • All 33/34 patients (97%) were correctly classified on cross-validation analysis
  • 2716 SNPs from 298 genes on 54 patients were analyzed for prediction of side effects, and 2 genes (CYP2C8 and BUB1b) were found to be closely related to granulocytopenia

Author's Conclusions

  • The expression of 7 genes can predict response to paclitaxel chemotherapy with an accuracy of 97%
  • Two genes identified by SNP analysis can predict for granulocytopenia with treatment

Clinical/Scientific Implications

This study describes a gene expression profile that may predict for response to treatment with paclitaxel of breast cancer patients.  Compared to similar studies using microarrays to screen large numbers of genes to find a predictive gene set, this study has the advantage of holding a high predicitive value using a small number of genes.  Like many studies using microarray technology, this study suffers from a lack of validation of the gene set against an independent set of patients.  It is not surprising that testing the gene set against the group of patients from which that set is derived would accurately group those patients.  An independent set of 17 patients have been removed from the original patient set in this study; unfortunately, the analysis of these patients was not yet completed at the time of this report.   Despite these shortcomings, this study holds interesting promise in predicting an individual's response to a paclitaxel therapy.  In theory, studies such as this may one day lead to the development of clinical test that would help tailor therapy to individual patients.  Currently, the clinical applicability of gene profiling studies is extremely limited; however, the future potential for these studies is high.

Oncolink's ASCO Coverage made possible by an unrestricted Educational Grant from Bristol-Myers Squibb Oncology.