Several microbicides, including nonoxynol-9 (N-9) and cellulose sulfate (CS), looked promising during early trials but failed in efficacy trials. We aimed to identify Phase I mucosal safety endpoints that might explain that failure. In a blinded, randomized, parallel trial, 60 healthy premenopausal sexually abstinent women applied Universal HEC placebo, 6% CS or 4% N-9 gel twice daily for 13½ days. Endpoints included immune biomarkers in cervicovaginal lavage (CVL) and endocervical cytobrushes, inflammatory infiltrates in vaginal biopsies, epithelial integrity by naked eye, colposcopy, and histology, CVL anti-HIV activity, vaginal microflora, pH, and adverse events. Twenty women enrolled per group. Soluble/cellular markers were similar with CS and placebo, except secretory leukocyte protease inhibitor (SLPI) levels decreased in CVL, and CD3(+) and CD45(+) cells increased in biopsies after CS use. Increases in interleukin (IL)-8, IL-1, IL-1RA, and myeloperoxidase (MPO) and decreases in SLPI were significant with N-9. CVL anti-HIV activity was significantly higher during CS use compared to N-9 or placebo. CS users tended to have a higher prevalence of intermediate Nugent score, Escherichia coli, and Enterococcus and fewer gram-negative rods. Most Nugent scores diagnostic for bacterial vaginosis were in N-9 users. All cases of histological inflammation or deep epithelial disruption occurred in N-9 users. While the surfactant N-9 showed obvious biochemical and histological signs of inflammation, more subtle changes, including depression of SLPI, tissue influx of CD45(+) and CD3(+) cells, and subclinical microflora shifts were associated with CS use and may help to explain the clinical failure of nonsurfactant microbicides.