Given the number of components scored in the Draize test and the diver
se mechanisms leading to distinct clinical signs (opacity, inflammatio
n, necrosis, etc.), no single-endpoint in vitro test is likely to suff
ice to predict ocular irritation potential. We have sought to develop
a test battery to measure these clinical signs to evaluate chemicals f
or worker safety. The bovine corneal opacity-permeability assay, which
measures two endpoints, corneal opacity and loss of integrity of the
epithelial cell barrier, has proven to be quite useful. With this comb
ination of endpoints, we find better than an 80% correlation with in v
ivo ocular irritation data. Cytotoxicity assays, on the other hand, ha
ve been of limited use. We have evaluated neutral red uptake, 3-(4,5-d
imethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide dye reduction, a
nd [H-3]leucine incorporation and have found that these endpoints show
good correlation with in vivo scores only when rankings are made with
in a single chemical class. In many applications, however, one needs t
o test across many chemical classes, and in that case the cytotoxicity
correlation is relatively poor. Measurement of inflammation is techni
cally difficult, although assessment of release of chemotactic factors
or eicosanoids shows some promise. The best overall approach to findi
ng alternatives to in vivo ocular irritation testing is to define a ba
ttery of tests in which the endpoints being measured correlate with sp
ecific in vivo mechanisms of irritation and which complement one anoth
er (i.e., multiple mechanisms measured). Coupling the in vitro results
with data generated from historical databases, structure-activity rel
ationships, physicochemical analysis, and other toxicity measurements
should give a solid base of information from which ocular irritation p
otential can be predicted with confidence.