Second Breast Cancer Events in Relation to Frequent Antibiotic Use: An Assessment of Risk and the Development of a Framework to Evaluate Detection Bias in Breast Cancer Pharmacoepidemiology Studies

Abstract

<bold>Background:<bold> Frequent antibiotic use is common and associated with increased primary breast cancer risk, but no study to date has evaluated the relation between antibiotic use and breast cancer outcomes among breast cancer survivors. Detection bias is commonly cited as a potential source of confounding in observational studies but rarely explored. We sought to evaluate the relation between frequent antibiotic use and second breast cancer events (SBCEs) and to develop a framework for investigating unexplored detection bias in breast cancer pharmacoepidemiology studies. <bold>Methods:<bold> Retrospective cohort study among female health plan enrollees aged ≥18 years when diagnosed with stage I or II breast cancer between 1990 and 2008 via the Surveillance, Epidemiology, and End Results registry. We used Cox regression models to estimate the relation between frequent antibiotic use (≥4 dispensings in 12 months) and SBCEs (first of recurrence or second primary). We evaluated disease free survival (DFS) in secondary analyses. We used generalized estimating equations models to estimate associations of surveillance mammography over 10 years with exposure to two disparate medication classes, antibiotics and statins. We categorized antibiotic users as infrequent (1-3 dispensings in 12 months) and frequent (≥4 dispensings in 12 months), and statin users as less adherent (1 dispensing in 6 months) and more adherent (≥2 dispensings in 6 months). <bold>Results:<bold> Among the 4,216 women in the SBCE cohort, 40% were frequent antibiotic users and 13% developed a SBCE. Frequent antibiotic use was associated with a non-significant higher risk of SBCE (HR 1.15, 95% CI 0.88-1.50) and a significantly higher risk of DFS events (HR 1.25, 95% CI 1.03-1.51). Of the 3,965 women evaluated for use of surveillance mammography over time, 79% received surveillance in year 1 and this decreased to 63% in year 10. Frequent antibiotic users were slightly less likely to receive mammograms during follow-up, though not significant (OR 0.93, 95% CI 0.84-1.02). More adherent statin users were more likely to receive mammograms (OR 1.15, 95% CI 1.03-1.28). <bold>Conclusions:<bold> Breast cancer prognosis may be altered by frequent antibiotic use. Receipt of surveillance procedures may confound associations between medication exposure and breast cancer outcomes through detection bias.