This dataset is a comprehensive resource characterizing mitochondrial DNA (mtDNA) variation in the Brazilian population, providing a valuable context for pharmacogenomic studies. The data was generated from the SABE cohort, a census-based sample of 1,171 elderly individuals from São Paulo, Brazil, and processed using whole-genome sequencing with GATK best practices. This dataset is complemented by reference data from the 1000 Genomes Project, which was used for comparative analyses, such as phylogenetic trees and networks. The dataset contains 431 distinct homoplasmic mtDNA variants, with a majority of them located in protein-coding genes involved in energy metabolism. It also includes eight variants with pharmacogenomic (PGx) relevance, categorized according to PharmGKB's levels of evidence. These variants are linked to drug responses, including susceptibility to aminoglycoside-induced ototoxicity and side effects of antiretroviral therapies. Mitochondrial haplogroup analysis revealed that the Brazilian cohort is highly diverse, reflecting its complex admixture history with African, European, Native American, and Asian ancestries. The data exposes a frequent mismatch between mitochondrial (maternal) and autosomal ancestry, highlighting the limitations of relying solely on autosomal data for PGx research in admixed populations. By providing this detailed record of mtDNA variants and their ancestral context, this dataset aims to support ancestry-informed approaches in drug prescription and public health in Brazil. It fills a critical gap by contributing data from a severely underrepresented population, which is crucial for advancing precision medicine on a global scale.