A new fuzzy document clustering algorithm based on topic homogeneity is introduced. In detail, a novel dissimilarity measure is proposed, derived from the p-value of a hypothesis test that assesses the homogeneity of topic distributions between two documents. First, the topic distributions are derived through Latent Dirichlet Allocation, and then a bootstrap procedure is applied to obtain the p-value. Finally, the resulting dissimilarity matrix is integrated into the fuzzy relational clustering procedure. The performance of the proposal is evaluated using a benchmark dataset.