OBJECTIVE: To evaluate the liquid-based ThinPrep Papanicolaou test. DESIGN: Prospective comparison of the ThinPrep test with the conventional Papanicolaou test. SETTING: Cervical smear specimens sent to a private practice, Hong Kong. PATIENTS: A total of 16,541 ThinPrep test specimens and 7258 conventional Papanicolaou smears from Hong Kong women who had been screened for cervical cancer between mid-July 1998 and mid-January 1999. MAIN OUTCOME MEASURES: Specimen adequacy, endocervical cell content, epithelial cell abnormalities, and micro-organisms present in both types of cervical smears; histological diagnosis of cervical biopsy specimens of women who had the ThinPrep test. RESULTS: Compared with the conventional Papanicolaou smear test, the ThinPrep test showed a reduction in the frequency of 'unsatisfactory' (0.56% versus 1.36%; P<0.01), 'satisfactory but limited' (1.67% versus 15.87%; P<0.01), and 'atypical squamous cells of undetermined significance' reports (1.72% versus 3.64%; P<0.01). The ThinPrep test was also more effective at detecting squamous intraepithelial lesions, showing a 58% increase for low-grade lesions (2.66% versus 1.68%; P<0.01) and 28% increase for high-grade lesions (1.71% versus 1.34%; P<0.01). The sensitivity and positive predictive value of the ThinPrep system were 97.5% and 94.2%, respectively. The liquid-based method yielded a higher percentage of samples that contained endocervical cells compared with conventional smear specimens (70.57% versus 51.23%; P<0.001). CONCLUSIONS: The ThinPrep test has a high sensitivity and positive predictive value. The ThinPrep test gives higher-quality specimens and has a higher detection rate of squamous intraepithelial lesions than the conventional Papanicolaou smear test. The drawbacks of the liquid-based system, however, pertain to cost and the additional procedures and training needed.