The efficacy of a microbicide depends on the balance between its specific activity and its safety. The experience with nonoxynol-9, the first microbicide to be tested in clinical trials, provides a good example of a compound with high in-vitro activity and poor clinical performance, possibly because of underestimated local safety issues. In order to identify compounds that may induce epithelial toxicity and inflammation early in drug development, we have established a preclinical evaluation system based on the assessment of cytotoxicity, cytokine secretion, and tissue inflammatory reaction using a human vaginal cell line (VK-2/E6E7) and a refined rabbit vaginal irritation model. Evaluated through this system, nonoxynol-9 displayed high cytotoxicity and proinflammatory activity. VK-2 cells incubated for 6 h with nonoxynol-9 released significant amounts of IL-1, IL-6 and IL-8. Gels containing 2 and 4% nonoxynol-9, administered to female rabbits intravaginally for 3 days, induced a potent inflammatory reaction evidenced by high levels of IL-lb and CD3+ T cells in cervicovaginal lavages and a significant influx of CD4 and nuclear factor kappa B cells into mucosal and submucosal tissues. Interestingly, cervicovaginal epithelium exposed to nonoxynol-9 also showed high levels of active nuclear factor kappa B immunoreactivity. This combined, sequential, preclinical evaluation system based on VK-2 cells in culture and a refined rabbit vaginal irritation model represents a valuable tool to assess the local safety profile of anti-HIV microbicide candidates.