This study investigates the harmonization of multi-site diffusion tensor imaging (DTI) data to address site-specific variability in fractional anisotropy (FA) and mean diffusivity (MD) maps. The authors demonstrate that DTI measurements are highly site-specific and that combining data without harmonization can lead to inaccurate inferences. They compare several harmonization methods, including global scaling, functional normalization, RAVEL, Surrogate Variable Analysis (SVA), and ComBat, and find that ComBat performs best at removing unwanted inter-site variability while preserving biological variability. ComBat, originally developed for genomics, is shown to be effective in DTI data by modeling site effects as additive and multiplicative factors. The study uses age as a biological variable to assess the effectiveness of harmonization methods, showing that ComBat preserves biological variation and removes site-induced noise. The authors also evaluate the robustness of these methods in the presence of confounding between site and age, finding that ComBat consistently outperforms other methods in replicating age-related associations across independent datasets. The results indicate that ComBat is a robust and effective method for harmonizing multi-site DTI data, particularly in the presence of confounding variables. The study highlights the importance of harmonization in multi-site neuroimaging studies to ensure accurate and reliable statistical analyses.This study investigates the harmonization of multi-site diffusion tensor imaging (DTI) data to address site-specific variability in fractional anisotropy (FA) and mean diffusivity (MD) maps. The authors demonstrate that DTI measurements are highly site-specific and that combining data without harmonization can lead to inaccurate inferences. They compare several harmonization methods, including global scaling, functional normalization, RAVEL, Surrogate Variable Analysis (SVA), and ComBat, and find that ComBat performs best at removing unwanted inter-site variability while preserving biological variability. ComBat, originally developed for genomics, is shown to be effective in DTI data by modeling site effects as additive and multiplicative factors. The study uses age as a biological variable to assess the effectiveness of harmonization methods, showing that ComBat preserves biological variation and removes site-induced noise. The authors also evaluate the robustness of these methods in the presence of confounding between site and age, finding that ComBat consistently outperforms other methods in replicating age-related associations across independent datasets. The results indicate that ComBat is a robust and effective method for harmonizing multi-site DTI data, particularly in the presence of confounding variables. The study highlights the importance of harmonization in multi-site neuroimaging studies to ensure accurate and reliable statistical analyses.