Apple replant disease (ARD) arises from repeated apple (Malus domestica) planting in the same area, disrupting physiological and morphological plant functions. Recent studies demonstrated that ARD occurs locally with low mobility in soil. The patchy distribution of ARD makes field identification of its severity difficult. Moreover, variability in soil properties can affect growth. Here, we aimed to identify drivers of small-scale growth variations with a pair-wise sampling approach at two ARD-affected orchards. We selected neighboring trees showing maximum differences in stem diameter growth. With this spatially explicit approach, large-scale heterogeneity in soil properties as a reason for differing growth was minimized. This design was applied to field plots differing in pre-cultures, i.e., comparing grass with Tagetes patula, the latter supposedly suppressing potential vectors of ARD. Various soil physical and chemical properties, the root phytoalexin content, and the bacterial and archaeal community composition were assessed. At one site, principal component analysis (PCA) separated neighbors with differing growth due to high particulate organic matter content, while no differences in soil physical properties, indicative of aeration differences or soil disturbances, could be detected. Elevated particulate organic matter content likely resulted from localized tree shredding. The worse-growing partners exhibited higher phytoalexin contents at the first site, which are general indicators of biotic stress and observed to increase in the presence of ARD. However, this was not associated with alterations of the rhizosphere bacterial and archaeal community composition as would be expected for ARD. At the second site, PCA showed no separation between tree groups, i.e., none of the measured variables could explain growth differences. Our work demonstrated that sampling neighboring trees with varying stem growth can identify co-occurring differences in related variables, some of which may reflect differences in ARD severity. Still, results were highly site-dependent and determined by the variables chosen for analysis.