An analysis of the effect of selection bias on the association of hormone replacement therapy and breast cancer risk.

Abstract

A sensitivity analysis was conducted to determine the impact on measures of effect of a suspected differential participation response rate between hormone replacement therapy (HRT) users and nonusers, among controls recruited to a population-based case-control study of breast cancer. The age-specific prevalence of current HRT use among controls was compared to data from the 1996 Canadian National Population Health Survey (NPHS). Control women identified as current HRT users were randomly re-sampled to replicate the prevalence of HRT use reported by the NPHS. Unconditional logistic regression was conducted to estimate odds ratios (OR) and 95 percent confidence intervals (CI) for the use of HRT and breast cancer risk before and after re-sampling. Multivariate adjusted ORs for breast cancer and estrogen-only and estrogen-progestin formulations were 0.76 (0.53-1.10) and 0.94 (95% CI: 0.64 - 1.38), respectively, using the original case-control controls and 0.99 (0.77-1.27) and 1.57 (95% CI: 1.02 - 2.40), respectively, following re-sampling of the controls. This sensitivity analysis illustrates the extent to which differential participation rates between HRT users and nonusers may affect estimates of measures of effect.

• Authors: Ilona Csizmadi, Christine M Friedenreich, Heather E Bryant, Kerry S Courneya
• Journal: Chronic diseases in Canada (Chronic Dis Can) 2005 Spring-Summer Vol. 26 Issue 2-3 Pg. 73-9 ISSN: 0228-8699 [Print] Canada
• PMID: 16251013 (Publication Type: Journal Article, Research Support, Non-U.S. Gov't)

Topics

• Aged
• Alberta (epidemiology)
• Breast Neoplasms (epidemiology)
• Case-Control Studies
• Female
• Hormone Replacement Therapy (statistics & numerical data)
• Humans
• Middle Aged
• Odds Ratio
• Risk Factors
• Selection Bias