Multiple metal‐impacted soils are often realistic scenarios for risk assessments, but tools to address these are currently lacking. The objective of this work was to evaluate whether assuming concentration addition (CA) of metal mixture effects was conservative for prospective risk assessment of soils that were elevated mainly in Ni and Cu and somewhat with Co, Pb, or As. Observed whole mixture toxicity for field soils with aged metal mixtures was compared to the expected whole mixture toxicity, assuming additivity of prospective single‐metal thresholds (“toxic units”) for the mixture components. Bioavailability‐adjusted single‐metal toxicity thresholds expected for those field soils were the median hazard concentration affecting 5% of species (HC5‐50) from the predicted no‐effect concentration (PNEC) calculator and calculated from the species‐specific dose–response multiple linear relationships (MLRs), all from the European Union Registration, Evaluation, Authorisation and Restriction of Chemicals (EU REACH) dossiers for metals. Generic single‐metal toxicity thresholds were based on Canadian Council of Ministers of the Environment soil quality guidelines (CCME SQGs) for agricultural soils. Observed toxicity thresholds were from the community‐based risk assessments conducted for Port Colborne and Sudbury, Ontario, Canada. Mostly, prospective single‐metal toxicity thresholds were protective relative to the observed toxicity, although that was species or ecological process dependent. The bioavailability‐adjusted single‐metal thresholds were less conservative than the CCME SQG method, even though the former is based on site‐specific EC10 values, and the latter is based on generic EC25 values. When within‐site variability in soil properties was used to calculate the 5th and 95th CI for the HC5 sum of toxic units (∑TUs), CA was conservative for far fewer endpoints. In addition, the prospective ∑TUs were more conservative predictions of the observed whole mixture toxicities for Port Colborne soils than for Sudbury soils. The most appropriate balance of accuracy and conservatism for identifying low‐level risk of the whole mixtures in these soils appeared to be the bioavailability‐adjusted HC5‐50, which was applicable to many endpoints and 2 quite different exposure concentration ratios. Integr Environ Assess Manag 2021;17:753–766. © 2020 SETAC