According to the List of Hazardous Substances of the Agency of the Toxic Substances and Disease Registry (ATSDR 2017), some metallic elements such as Pb or Cd are still considered as the most polluting elements in the world. These elements accumulate in sediments, and there are various methods available that differentiate lithogenic sources from anthropogenic sources. For that, the natural geochemical background is required, however, its definition is far from unanimous: it can be a global or local value, single or statistically estimated value, depending on the studies. Our study is focused on the Subaé River in the state of Bahia, Northeastern Brazil, which was historically contaminated by lead metallurgy. The river sediments were sampled at 21 locations: one at the discharge point of the plant's rainwater basin, ten upstream, and ten downstream. The total contents of Fe, Ti, Mn, Pb, Zn, Cu, Cr, particle-size distribution, and organic matter were analyzed. The conventional geochemical indexes, i.e., the pollution degree (mCd), pollution load index, and enrichment factor indicated contamination by Zn, Pb, Cr, and Cu. Then, a new enrichment factor is introduced, assuming that the sediments come from erosion of soils in the watershed. After obtaining the natural concentrations of metals in the clay, silt, and sand fractions of uncontaminated reference soils, we corrected the pollution indexes by calculating a new individual background in each sediment sample, according to its grain size distribution. This new approach provided more precise indexes in the assessment of sediment pollution, by highlighting higher contamination of Zn and Pb (around 50%) and at the same time the absence of Cr and Cu contamination, two metals not involved with the metallurgical activity.