A weighted kernel machine regression approach to environmental pollutants and infertility

In epidemiological studies of environmental pollutants in relation to human infertility, it is common that concentrations of a large number of exposures are collected in both male and female partners. Such a couple‐based study poses some new challenges in statistical analysis, especially when the effect of the totality of these chemical mixtures is of interest, because these exposures may have complex nonlinear and nonadditive relationships with the infertility outcome. Kernel machine regression, as a nonparametric regression method, can be applied to model such effects, while accounting for the highly correlated structure within and across exposures. However, it does not consider the partner‐specific structure in these study data, which may lead to suboptimal estimation for the effects of environmental exposures. To overcome this limitation, we developed a weighted kernel machine regression method (wKRM) to model the joint effect of partner‐specific exposures, in which a linear weight procedure is used to combine the female and male partners' exposure concentrations. The proposed wKRM is not only able to reduce the number of analyzed exposures but also provide an overall importance index of female and male partners' exposures in the risk of infertility. Simulation studies demonstrate good performance of the wKRM in both estimating the joint effects of exposures and fitting the infertility outcome. Application of the proposed method to a prospective infertility study suggests that the male partner's exposure to polychlorinated biphenyls might contribute more toward infertility than the female partner's.