BACKGROUND: The ATHENA study was designed to assess the performance of carcinogenic human papillomavirus (HPV) testing and HPV16 or HPV18 genotyping compared with liquid-based cytology for cervical cancer screening in a large US population aged 21 years and older. We did a subanalysis of this population to compare the screening performance of the cobas HPV test versus liquid-based cytology in women aged 25 years and older, and assess management strategies for HPV-positive women. METHODS: Women aged 25 years or older who were attending routine cervical screening were enrolled from 61 clinical centres in 23 US states. Cervical specimens were obtained for liquid-based cytology and HPV DNA testing with two first-generation assays (Amplicor HPV test and Linear Array HPV genotyping test) and the second-generation cobas HPV test (with individual HPV16 and HPV18 detection). Colposcopy and diagnostic biopsies were done on women with atypical squamous cells of undetermined significance (ASC-US) or worse cytology, those who tested positive with either first-generation HPV test, and a random sample of women who tested negative for HPV and cytology. All women not selected for colposcopy received their results and exited the study. Participants and colposcopists were masked to cytology and HPV test results until the colposcopy visit was completed. The primary endpoint for this substudy was histologically confirmed cervical intraepithelial neoplasia grade 3 (CIN3) or worse. This study is registered with ClinicalTrials.gov, number NCT00709891; the study is in the follow-up phase, which is due to be completed in December, 2012. FINDINGS: From May 27, 2008, to Aug 27, 2009, 47,208 women were enrolled, of whom 41,955 met our eligibility criteria. Valid cobas HPV and liquid-based cytology test results were available for 40,901 women (97%), who were included in this analysis. Of these, 4275 women (10%) tested cobas HPV positive and 2617 (6%) had abnormal cytology. 431 women were diagnosed with CIN2 or worse and 274 with CIN3 or worse. In women who had colposcopy, the cobas HPV test was more sensitive than liquid-based cytology for detection of CIN3 or worse (252/274 [92·0%, 95% CI 88·1-94·6] vs 146/274 [53·3%, 95% CI 47·4-59·1]; difference 38·7%, 95% CI 31·9-45·5; p<0·0001). Addition of liquid-based cytology to HPV testing increased sensitivity for CIN3 or worse to 96·7% (265/274, 95% CI 93·9-98·3), but increased the number of screen positives by 35·2% (5783/40,901 vs 4275/40,901) compared with HPV testing alone. As a triage test to identify CIN3 or worse in HPV-positive women, detection of HPV16, HPV18, or both alone was equivalent to detection of ASC-US or worse alone in terms of sensitivity (150/252 [59·5%] vs 133/252 [52·8%]; p=0·11) and positive predictive value (PPV) (150/966 [15·5%] vs 133/940 [14·1%]; p=0·20). Among HPV-positive women, detection of HPV16, HPV18, or both or low-grade squamous intraepithelial lesion or worse cytology had better sensitivity (182/252 [72·2%]; p<0·0001) and similar PPV (182/1314 [13·9%]; p=0·70) for detection of CIN3 or worse than ASC-US or worse cytology alone. Furthermore, detection of HPV16, HPV18, or both or high-grade squamous intraepithelial lesion or worse cytology had higher sensitivity (165/252 [65·5%]; p=0·0011) and PPV (165/1013 [16·3%]; p=0·031) for detection of CIN3 or worse than ASC-US or worse cytology alone. INTERPRETATION: HPV testing with separate HPV16 and HPV18 detection could provide an alternative, more sensitive, and efficient strategy for cervical cancer screening than do methods based solely on cytology. FUNDING: Roche Molecular Systems.