Dose-response by in vitro testing is only valid if the fraction of the particle dose that deposits onto adherent cells is known. Modeling tools such as the 'distorted grid' (DG) code are common practices to predict that fraction. As another challenge, workflow efficiency depends on parallelized sample preparation, for which freeze-thaw protocols have been explored earlier, but not their implications on dosimetry. Here we assess the sensitivity of the DG code toward freeze-thaw protocols and variations in user-defined parameters, including the estimation of particle-cell affinity and determination of agglomerate size, which we measure by DLS or AUC. We challenge the sensitivity by materials of varying composition, surface functionalization, and size (TiO2, CeO2, BaSO4, 2x Ag, 3x SiO2). We found that the average effective density is robust, but the dose predictions by different approaches varied typically 2-fold and up to 10-fold; this uncertainty translates directly into the uncertainty of no-effect-concentrations. The use of standardized dispersion protocols increases the uncertainty in doses. The choice of a measurement method and minor details of the particle size distribution strongly influence the modeled dosimetry. Uncertainty is high for very well dispersed nanomaterials; since then, the assumed affinity of particles to cells has a decisive influence. Against this background, the modulation of deposited dose by freeze-thaw protocols is a minor factor that can be controlled by aligning the protocols of sample preparation. However, even then, the uncertainty of deposited doses must be considered when comparing the in vitro toxicity of different nanomaterials.
Keywords: affinity; deposited dose; dosimetry; effective density; nanomaterials.