Cytological screening for cervical cancer is hampered by high false negative rates. Inter-observer reproducibility needs optimizing. The potential of p16(INK4a) as a biomarker for cervical lesions was examined in a study of liquid-based cytology (LBC), HPV DNA testing by MY09/MY11 consensus PCR and type-specific PCRs and p16(INK4a) immunocytochemistry on a series of 291 patients selected from routine screening. Comparison of the number of p16(INK4a) immunoreactive cells/1,000 cells exhibited a significantly higher mean count in HSIL (8.80 +/- 1.13) than other cytological groups. The mean count of LSIL (1.09 +/- 0.18) was significantly higher than that of the negative group (0.82 +/- 0.40). ASC-H and HSIL combined showed a significantly higher mean count (6.46 +/- 1.17) than negative, ASC, ASC-US and LSIL. The mean count of immunoreactive cells/1,000 cells was significantly higher in HPV16 positive samples (3.22 +/- 0.72) than in samples containing infections with types of unknown malignant potential (0.83 +/- 0.26) or HPV negative samples (1.17 +/- 0.41). The mean count in infections with other high-risk HPV types (2.55 +/- 0.52) was significantly higher than that in HPV negative samples. Receiver-operating characteristic curves yielded a test accuracy (area under curve) of 0.76, 0.79, 0.88 and 0.95 for ASCUS, LSIL, ASC-H/HSIL and HSIL, respectively. Thresholds for 95% sensitivity were at 0.005, 0.007, 0.098 and 0.445 immunopositive cells/1,000 cells for ASCUS, LSIL, ASC-H/HSIL and HSIL, respectively. The 95% specificity threshold for the detection of HSIL was at 1.87 immunopositive cells/1,000 cells. P16(INK4a) immunocytochemistry can be used as an adjunct to LBC in cervical screening, because it has a good diagnostic accuracy to discriminate HSIL and ASC-H from other lesions. It could be used as a surrogate marker of high-risk HPV infections.