Background: Previous studies have examined environmental phenol exposures as risk factors for gestational diabetes mellitus (GDM), but have focused on associations with individual chemicals. As methods have emerged to better assess simultaneous exposure to multiple chemicals, the association between phenol mixtures and GDM remains undetermined. In the present study, we utilized Bayesian kernel machine regression (BKMR) to investigate the relationship between phenol mixtures and GDM.Methods: A case-control study of 64 GDM cases and 237 controls was conducted among obstetric patients at the University of Oklahoma Medical Center, August 2009 to May 2010. Concentrations of bisphenol A (BPA), benzophenone-3, triclosan, 2,4-dichlorophenol, 2,5-dichlorophenol, butylparaben, methylparaben, and propylparaben were quantified in mid-pregnancy spot urine samples. Multivariable logistic regression was performed to assess associations between continuous measures of individual specific-gravity adjusted phenol concentrations and GDM while controlling for age, race and pre-pregnancy body mass index. Probit implementation of BKMR with hierarchical variable selection was used to evaluate the adjusted mean difference in GDM probability for each component of the phenol mixtures, while taking the correlation among the mixture components into account.Results: We observed that benzophenone-3 was positively associated with GDM [adjusted odds ratio (aOR per interquartile range (IQR) = 1.48 (95% Confidence Interval (CI) 1.11, 1.98)] when analyzing individual phenols using logistic regression. In contrast, BPA was negatively associated with GDM (aOR 0.63 (95% CI 0.41, 0.98)). In Probit-BKMR analysis, an IQR increase in Z score-transformed benzophenone-3 was associated with an increase in the probability of GDM (0.19, 95% Credible Interval: -0.09, 0.48), holding other phenols fixed at their medians, but the credible interval crossed zero. The BKRM analyses did not identify associations with the other phenol concentrations while accounting for correlation among the mixture components.Conclusion: Our results highlight the importance of addressing chemical mixtures in studies of perinatal environmental exposures.