Hair relaxer use and breast cancer risk by tumor molecular subtypes.

10596 Background: Women of West African ancestry, including US populations, are disproportionately diagnosed with aggressive breast cancers including triple negative tumors. Previously, hair relaxer use has been associated with increased breast cancer risk in the Ghana Breast Health study (GBHS), with former users at an over 2-fold increased risk compared to never users. We aimed to determine if relaxer use risk associations differed by tumor molecular subtypes suggesting possible biological mechanisms of carcinogenesis. Methods: The GBHS is a population-based case-control study conducted 2013–2015 in Accra and Kumasi, with 1,071 pathologically confirmed invasive breast cancer cases (87% had available estrogen receptor (ER) status) and 2,106 matched controls, aged 18-74 years. Molecular subtypes were defined using St Gallen’s criteria using molecular marker data for ER, progesterone receptor, human epidermal growth factor receptor-2 (HER2), and grade (G). Molecular tumor subtypes defined as: Lum A (ER/PR+, HER2- and G1/G2, N = 260), Lum B- (ER/PR+, HER2- and G3, N = 468), Lum B+ (ER/PR+, HER2+ and G1/G2, N = 195), HER2-Enriched (ER-, PR- and HER2+, N = 59), Triple-Negative (ER, PR, and HER2-, N = 129). We imputed missing data using chained equations. Polytomous logistic regression models were used to estimate odds ratios (ORs) with 95% confidence intervals (CIs) for associations of relaxer use with breast cancer subtypes. Models were adjusted for matching factors (age, hospital site), and potential confounders: education, parity, age at first birth, and median breastfeeding months per child. Results were combined across imputed datasets. Heterogeneity (p-het) in breast cancer risk by tumor molecular subtypes using polytomous logistic regression case-only models. Results: Relaxer use was common (ever-use of relaxers 96% of cases and 94% of controls). Compared to never users, increased risk for former users compared to never users was observed across subtypes. While magnitudes of risk differed (OR LumA = 2.59, 95% CI 1.01—6.60, OR LumB+ = 2.69, 95% CI 0.47—15.47, OR LumB- = 1.74, 95% CI 0.33—6.93, OR HER2-Enriched = 3.66, 95% CI 0.75—17.98, OR Triple-Negative = 1.21, 95% CI 0.37—4.00), there was no statistical evidence of etiologic heterogeneity by subtype (p-het = 0.59). Similarly, other measures of relaxer use (duration, age at first use, lye or non-lye) did not show any evidence of heterogeneity. Conclusions: In this population with a high proportion of aggressive tumor subtypes, breast cancer risk associated with relaxer use did not support etiologic heterogeneity by molecular subtype, however, power was limited. These findings suggest that relaxer exposure is unlikely to explain subtype-specific breast cancer risk patterns. Future investigations should pool studies for improved power, evaluate other molecular tumor characteristics and relaxer products’ formulations to clarify carcinogenic mechanisms to reduce risk.