Recently, LD Score regression1 has been proposed as a computationally fast method to contrast confounding biases with polygenicity and to quantify their contribution to the inflation of test statistics in GWAS. In this communication, we extend the LD Score regression approach by applying the generalized estimation equations (GEE) framework, which is capable of incorporating more external information from reference panels about the correlation structure of test statistics. We apply our GEE approach and LD Score regression to simulated and real data to compare their performance. We show that our proposed methodology obtains more efficient estimates while preserving the robustness and desired properties of LD Score regression.