A meta-analysis framework to assess the role of units in describing nanoparticle toxicity.

Despite ample research on nanoparticles, their environmental toxicity is still debatable. The lack of consensus is due in part to the challenge of comparing studies because of variability in parameters like test organism, test medium, and duration of experiment. However, the unit used to compare the toxicology of nanoparticles is one variable that experimentalists can control. Traditionally, mass per volume is the most common unit used to make comparisons, but there is growing evidence that alternative units such as surface area per volume or particles per volume may provide a better and more mechanistic measure of toxicity. Herein, we propose and test a meta-analytic framework to study the effect of units on nanotoxicology using data from the NanoE-Tox database, a freely available database containing 1518 toxicology values from 224 published articles of which 42 records met our basic inclusion criteria. These data were augmented with more recent data published over the past five years as archived by the Web of Science citation index. An additional 27 records from 1676 papers met the inclusion criteria and were also included in the analysis. The meta-analysis framework measures the degree of heterogeneity for each of three units (grams/L, particles/L, surface area/L) grouped by the type of test organism, particle chemistry, and manner in which a nanoparticle's size was measured (e.g., nominal particle size reported by the manufacturer vs. measurement of size for particles suspended in the liquid medium used in a subsequent toxicity experiment). The result of the meta-analysis reveals that surface area per volume reduces the heterogeneity in the Ag crustacean subgroup when nanoparticle size was measured in the test medium, and the ZnO crustacean subgroup when nanoparticle size was measured out the test medium and may therefore be a more appropriate estimate of the toxicity of soluble nanoparticles. No subgroups in our analysis showed a reduction in heterogeneity for particles per volume in either soluble or insoluble nanoparticles. The lack of conclusion on insoluble nanoparticles was not due to a limitation of our meta-analysis but rather highlights a critical deficiency in the primary literature. The majority of published studies fail to report common measures of error that are essential for further analysis (i.e. error of the measured nanoparticle size and/or interoperable error of the measured half-maximal concentration of the toxic endpoint). If future nanotoxicity studies report such error, as they should, then the framework of our meta-analysis could be used more broadly to provide a simple, statistically rigorous way to assess the role of units on the toxicity of nanoparticles.