Fifteen to thirty percent of cases with histologically confirmed CIN2-3 in cervical biopsies regress spontaneously (ie, show CIN1 or less in the follow-up cervical cone). The balance between immune-reactive cells from the host and high-risk human papillomavirus (hrHPV) genotypes may provide a biological explanation for this phenomenon. We retrospectively studied 55 cases of CIN2-3 in a cervical biopsy with subsequent cervical cone to assess whether hrHPV genotypes (by AMPLICOR and Linear Array tests) CD4, CD8, CD25, CD138 and Foxp3 cells (by quantitative immunohistochemistry) in the cervical biopsies can predict regression (defined as CIN1 or less in the follow-up cone biopsy). Eighteen percent of the CIN2-3 cases regressed (median biopsy-cervical cone time interval: 12.0 weeks, range: 5.0-34.1 weeks). HPV-16 correlated with low CD8+ and high CD25+. None of the regressing CIN2-3 lesions contained HPV-16. The regressing CIN2-3 lesions had lower numbers of stromal CD138+ and higher numbers of stromal CD8+cells; higher stromal and intra-epithelial ratios of CD4+/CD25+ cells; higher ratios of CD8+/CD25+ cells and lower ratios of CD8+/CD4+, CD138+/Foxp3+ and CD25+/Foxp3+ cells in the stroma. With multivariate survival analysis, stromal CD8+ cell numbers, CD4+/CD25+ cell ratios and CD138+ cell numbers are found to be independent regression predictors. In conclusion, in non-HPV-16 CIN2-3 lesions, assessing stromal immune cells can be a useful prognostic indicator of regression or persistence.