Breast Cancer Research and Treatment

, Volume 82, Issue 2, pp 113–123

Studies of the Potential Utility of Ki67 as a Predictive Molecular Marker of Clinical Response in Primary Breast Cancer


  • L. Assersohn
    • Breast UnitRoyal Marsden Hospital
  • J. Salter
    • Academic BiochemistryRoyal Marsden Hospital
  • T.J. Powles
    • Breast UnitRoyal Marsden Hospital
  • R. A'hern
    • Academic BiochemistryRoyal Marsden Hospital
  • A. Makris
    • Marie Curie Research WingMount Vernon Hospital
  • R.K. Gregory
    • Southampton University Hospitals Trust
  • J. Chang
    • Breast CenterBaylor College of Medicine
  • M. Dowsett
    • Breast UnitRoyal Marsden Hospital

DOI: 10.1023/B:BREA.0000003968.45511.3f

Cite this article as:
Assersohn, L., Salter, J., Powles, T. et al. Breast Cancer Res Treat (2003) 82: 113. doi:10.1023/B:BREA.0000003968.45511.3f


Introduction. Objectives were to characterise the relationship of the proliferation marker Ki67 with response to systemic treatment in early breast cancer and to assess its clinical utility, using fine needle aspirates.

Materials and methods. Hundred and six women were treated with primary tamoxifen (n = 33), chemotherapy (n = 33) or chemotherapy and tamoxifen (n = 40). Treatment was not randomised and response was assessed clinically. Ki67 was evaluated prior to treatment and at Day 14 or 21 after commencing treatment. To assess reproducibility, Ki67 was evaluated in repeat FNAs taken from 37 untreated patients.

Results. The percentage change in Ki67 in first 21 days was different between responders and non-responders for patients treated with tamoxifen (p = 0.007) and chemotherapy (p = 0.005) but not for chemoendocrine treatment (p = 0.062). The reproducibility study indicated that a decrease to 36% or less of the pre-treatment Ki67 value in an individual patient was required for it to be regarded as a statistically significant change. A significant decrease in Ki67 was seen in responding patients treated with chemotherapy (p = 0.026) and chemoendocrine treatment (p = 0.041). Positive and negative predictive values for response were 85 and 59% for chemotherapy patients and 88 and 54% for chemoendocrine patients, respectively.

Conclusion. Ki67 is unlikely to be useful as a predictive marker in individual patients. Further molecular markers that predict lack of response continue to be required.

breast cancerKi67markerpredictiveresponse

Copyright information

© Kluwer Academic Publishers 2003