E-Poster 63rd Endocrine Society of Australia Annual Scientific Meeting 2020

Risk of bias against premenopausal breast cancer patients in gene expression-based precision medicine (#41)

Sarah M Bernhardt 1 2 , Pallave Dasari 1 2 , Danielle J Glynn 1 2 , Joseph Wrin 1 2 , Lucy Woolford 3 , Wendy Raymond 4 , Lachlan M Moldenhauer 2 , Suzanne Edwards 5 , David Walsh 1 , Amanda R Townsend 1 6 , Timothy J Price 1 6 , Wendy V Ingman 1 2
  1. Adelaide Medical School, The University of Adelaide, Woodville, SA, Australia
  2. The Robinson Research Institute, The University of Adelaide, Adelaide, SA, Australia
  3. School of Animal and Veterinary Sciences, The University of Adelaide, Roseworthy, SA, Australia
  4. Flinders Medical Centre, Flinders University of South Australia and Clinpath Laboratories, Adelaide, SA, Australia
  5. School of Public Health, The University of Adelaide, Adelaide, SA, Australia
  6. Department of Medical Oncology, The Queen Elizabeth Hospital, Woodville, SA, Australia

Background: Gene expression-based algorithms that guide treatment decisions for breast cancer patients are at the forefront of a new era of precision medicine. However, they were developed and validated using datasets predominantly comprised of postmenopausal women. In premenopausal women, fluctuations in estrogen and progesterone during the menstrual cycle impact gene expression in hormone-responsive cancers. However, the extent to which menstrual cycling affects gene expression-based algorithms remains unclear. Here, we use mouse models and human breast cancer samples to demonstrate that the clinically-employed Oncotype DX 21-gene algorithm is critically affected by the menstrual cycle.

Methods: RNA was extracted from 25 pairs of formalin-fixed paraffin-embedded, invasive hormone receptor (HR)-positive breast cancer samples that had been collected approximately 2 weeks apart. Additionally, RNA was extracted from HR-positive mammary tumours dissected from naturally cycling Mmtv-Pymt mice at different ovarian cycle stages (n=53). A 21-gene signature analogous to the Oncotype DX platform was assessed through quantitative real time PCR and experimental recurrence scores (RS) were calculated.

Results: There was a significant inverse association between patient age and discordance in RS. For every one-year decrease in age, discordance in RS between paired samples increased by 0.08 units (95% CI: -0.14, -0.01; p=0.017). Discordances were driven primarily by proliferation and HER2-associated genes. In mice, RS were significantly increased in mammary tumours collected at diestrus, driven by genes associated with proliferation and HER2, compared to tumours dissected at estrus. Clustering analysis revealed a relationship between ovarian cycle stage and tumour gene expression.

Conclusions: These results suggest that gene expression-based algorithms are critically affected by menstrual cycle stage at the time of tissue collection. Caution in the adoption of gene expression-based algorithms is required, as their use in informing treatment decisions for premenopausal breast cancer patients could lead to unnecessary or suboptimal therapy.