There has been significant progress in the development of in vitro ocular models over the past decade (see Table 1). Originally monolayers of nonocular and ocular cells were used for in vitro toxicity tests. Stratified and differentiated corneal epithelial cells and organotypic models incorporating the three corneal cell layers were then developed and found to better replicate the in vivo tissue responses. Recently, assays using neural cells to evaluate the neurogenic pain component have been developed (Lilja & Forsby, 2004; Lilja, et al., 2007). Progress in identifying the corneal epithelial stem cells (Ahmad, et al., 2007; Shortt, et al., 2007; Stepp & Zieske, 2005) suggests that a future approach may involve the incorporation of stem cells into in vitro corneal models in an attempt to get them to recover from injury much like the in vivo cornea.
Researchers have found that much more than having an appropriate cell type is involved in creating a useful in vitro model that can replicate essential in vivo behaviors. The substrate on which the cells are grown, as well as the composition of the culture media, has long been known to exert an influence on the growth, phenotype, differentiation, and repair of cultured cells. Research continues to be conducted in developing cell scaffolds for in vitro models and tissues for clinical use (Doillon, et al., 2003; Duan, et al., 2007), and in studying the effect of cell basement membrane and cell mediators on corneal wound healing (Stramer, et al., 2003). For continued progress, models and assay endpoints must be carefully selected and optimized for an in vitro model to perform well in a validation study. The model’s biological relevance and the mechanistic relationship of the assay endpoint to human eye irritation need to be evident and described as part of the test method validation.
A widely recognized problem in evaluating and validating new ocular test methods is the use of Draize rabbit eye test data as the standard against which new methods are judged. The results from a human cell-based test method cannot be expected to perfectly correlate with Draize data, but the variability in the Draize test itself makes obtaining a good correlation even less likely. In addition, only limited amounts of Draize data are available, and many people consider conducting new Draize tests just for the purpose of a validation study to be unethical. The available data are usually very old, making it difficult to trace the exact composition and manufacturing specifications of the substance that was tested (S. Ward, personal observation). Available Draize data are of variable quality and generated using a variety of protocols, and they generally overpredict the human ocular response; even data collected with good laboratory practice (GLP) standards have shown large intralaboratory and interlaboratory variability (Buehler & Newman, 1964; Freeberg, et al., 1984; Gad & Chengelis, 1991; Weil & Scala, 1971). Furthermore, the rabbit ocular response to some chemicals is mechanistically different than the human response (Grant & Schuman, 1993). Additionally, Draize data provide information on limited endpoints of ocular injury and are unlikely to provide useful comparative data for validating methods based on new human mechanistic endpoints (Bagley, et al., 2006; Nussenblatt, et al., 1998). To accurately predict human eye irritation potential, a new toxicity test method must be relevant to the biological and toxicological processes that occur when a chemical or chemical mixture makes contact with the human ocular surface. For reasons noted above, this type of method (a mechanism-based assay in a human tissue model) may not be appropriately assessed using current validation study approaches. Therefore, the usefulness of developing new Draize data to validate in vitro methods has had little support.
A general conclusion from an ocular symposium highlighting in vitro ocular methods that were sufficiently developed to the point of having some correlative data to the Draize rabbit eye test was that it was difficult to assess the performance of the in vitro methods because the only comparative data were the Draize rabbit data (Salem & Katz, 2003). Current momentum is toward the development and use of human data for validating alternative ocular test methods (Bagley, et al., 2006; Nussenblett, et al., 1998) and/or other innovative approaches.
In May 2005, ICCVAM held a scientific symposium entitled Mechanisms of Chemically-Induced Ocular Injury and Recovery to assess research needs for obtaining validated methods of ocular toxicity testing for regulatory purposes. A formal report from this workshop is still pending; however, the discussions revealed that only modest progress has been made over the past decade in elucidating the mechanisms of chemical eye injury and in the development and validation of in vitro ocular toxicity test methods. At this workshop, a panel including ophthalmologists and ophthalmic researchers concluded that although some indicators of injury exist, because of the lack of quality human data they still cannot be linked to specific levels of human chemical eye injury.
The European Cosmetic, Toiletry and Perfumery Association (COLIPA) eye irritation research program has been funding research aimed at understanding the mechanisms of chemical eye injury for the identification of endpoints for use in in vitro assays that will be predictive of the in vivo response (de Silva, 2002; Eskes, et al., 2005). COLIPA is also supporting research on the Draize low-volume eye test for understanding mechanisms of eye injury and for developing comparative data for the evaluation of in vitro assays.
Additional topics on emerging research and policies relevant to eye irritation/corrosion testing will be provided on AltTox.org in the future; please check back again.