We developed an in vitro model for studying the cytotoxicity of pharmacologic agents on corneal epithelium employing 3H-thymidine incorporation. Primary rabbit corneal epithelial cell cultures were established, and the cells plated prior to each experiment. 3H-thymidine incorporation was measured after the addition of drug or vehicle to these confluent cells, and dose-response curves were generated. Marked inhibition of 3H-thymidine incorporation was reached at chemotherapeutic concentrations achieved clinically for cytosine arabinoside (10(-7) M), methotrexate (10(-3) M), and 5-fluorouracil (10(-6) M). A 10(-4) M concentration of 2-deoxycytidine, a naturally occurring competitive inhibitor of cytosine arabinoside, protected cells up to a concentration of 10(-5) M. We utilized these data to undertake an in vivo prophylaxis study in 13 leukemia patients receiving high-dose iv cytosine arabinoside. Topical deoxycytidine 10(-4) M and 1% prednisolone phosphate, given 12 hours prior to the start of antileukemic therapy, were effective in reducing symptoms and signs of keratitis; both were better than historical placebo-treated eyes. Ophthalmic preservatives were studied in vitro at concentrations used clinically: benzalkonium chloride (BAC) (0.004-0.02%) was the most toxic, thimerosal (TMS) (0.001-0.004%) intermediate, and chlorobutanol (CHB) (0.2-0.5%) the least toxic. Antiviral agents (final concentration) included: trifluridine (TFT) (1.0%), ethyldeoxuridine (EDU) (2.0%), and idoxuridine (IDU) (0.1%). Dose but not time-dependent concentrations of these 3 agents were noted to cause toxicity; however, (E)-5(2-bromovinyl)-2'-deoxyuridine (BVDU) (0.1%) was non-toxic. Similarly, tobramycin and amikacin were significantly less toxic than gentamicin and neomycin in this system. These in vitro cytotoxicity data correlate well with previous in vivo and pre-clinical corneal epithelial toxicity studies. Our model may be useful in the toxicologic study of future topical ophthalmic agents.