It has been argued that in vitro toxicity testing of engineered nanoparticles (NPs) should consider delivered dose (i.e., NP mass settled per suspension volume) rather than relying exclusively on administered dose (initial NP mass concentration). Delivered dose calculations require quantification of NP sedimentation in tissue cell culture media, taking into consideration fundamental suspension properties. In this article, we calculate delivered dose using a first-principles "particles in a box" sedimentation model, which accounts for the particle size distribution, fractal dimension, and permeability of agglomerated NPs. The sedimentation model was evaluated against external and our own experimental sedimentation data for metal oxide NPs. We then utilized the model to construct delivered dose-response analysis for a library of metal oxide NPs (previously used for hazard ranking and prediction making) in different cell culture media. Hierarchical hazard ranking of the seven (out of 24) toxic metal oxide NPs in our library, using EC50 calculated on the basis of delivered dose, did not measurably differ from our ranking based on administered dose. In contrast, simplified sedimentation calculations based on the assumption of impermeable NP agglomerates of a single average size significantly underestimated the settled NPs' mass, resulting in misinterpretation of toxicity ranking. It is acknowledged that in vitro dose-response outcomes are likely to be shaped by complex toxicodynamics, which include NP/cellular association, triggering of dynamic cell response pathways involved in NP uptake, and multiple physicochemical parameters that influence NP sedimentation and internalization.