DOES FIRST LINE THERAPY WITH AROMATASE INHIBITORS INCREASE THE RISK OF HIP FRACTURES AMONG POSTMENOPAUSAL WOMEN WITH EARLY STAGE BREAST CANCER? A META-ANALYSIS

Sunday, October 19, 2014
Poster Board # PS1-30

Ava John-Baptiste, PhD1, Peter C. Austin, Phd2, Taryn Becker, MD, MSc, FRCPC3, Paula Rochon, MD, MPH, FRCPC3, Geoffrey Anderson, MD, PhD4 and Jennifer Hughes-Large, MSc, MD Candidate 20175, (1)Western University, London, ON, Canada, (2)Institute for Clinical Evaluative Sciences, Toronto, ON, Canada, (3)Women's College Research Institute, Toronto, ON, Canada, (4)University of Toronto, Toronto, ON, Canada, (5)Schulich School of Medicine & Dentistry, Western University, London, ON, Canada
Background: Randomized controlled trials (RCTs) demonstrate that for post-menopausal women diagnosed with early stage breast cancer aromatase inhibitors reduce the risk of breast cancer recurrence compared to tamoxifen. Initial therapy with aromatase inhibitors also decreases bone mineral density and increases fracture risk compared to tamoxifen.

Purpose: To estimate the increased risk of hip fractures resulting from initial therapy with aromatase inhibitors, using data from both RCTs and observational studies. To test the sensitivity of the analysis to different specifications of the statistical model and different weakly informative prior distributions for the between-study and between study-type variances.

Methods: We performed a meta-analysis of data from RCTs and observational studies using a fixed effects model, a two-level random-effects model and a three-level random-effects model incorporating a level for study type. Sensitivity analyses were performed on the specification of prior distributions for the variance (base case: τ ~ Uniform(0,2), log(τ2) ~ Uniform(−10, 1.386) and τ ~ N(0, 1), for τ > 0). Sensitivity analyses were also performed reflecting different hypotheses about RCT and observational study estimates. In the base case, the true effects from RCTs and observational studies were considered randomly distributed around the overall true effect. In sensitivity analysis, the true effects from RCTs and observational studies differed by a fixed amount from the overall true effect.

Results: Estimates of hip fracture risk were combined from two RCTs and two observational studies. Based on the three-level random effects model, aromatase inhibitors were associated with an increased hip fracture risk, RR = 3.02, 95% Credible Interval (0.36, 11.53). Wide credible intervals signify a large degree of uncertainty about the true value of the parameter, yet the probability that aromatase inhibitors increased hip fracture risk was 85%. Estimates of increased risk were consistent for different prior distributions of the between study and between study-type variances and different assumptions about the relationship of study-type estimates to the overall true effect.

Conclusion: Different approaches to evidence synthesis with a small number of studies demonstrate that as first-line therapy, aromatase inhibitors likely increase the risk of hip fractures in postmenopausal women diagnosed with early stage breast cancer when compared to tamoxifen.

See more of: POSTER SESSION I & WELCOME RECEPTION

See more of: The 36th Annual Meeting of the Society for Medical Decision Making