This report investigates different options for using human papillomavirus (HPV) testing in cervical cancer prevention. These options are evaluated by a stochastic model of the progression of pre-malignancy and its relationship to HPV infection. Three screening policies are compared: 2 based on cytological screening, with or without HPV testing, and I in which HPV testing is the primary screening method. A policy of HPV testing for women with mildly abnormal smears would have little effect on the overall incidence of invasive cancer when compared with a policy of repeat cytology, provided follow-up is efficient. Moreover, the potential value of HPV testing as a primary screening method is strongly dependent on the proportion of neoplasias that are HPV-negative. Important factors in assessing the future role of HPV testing would be cost-effectiveness and benefits from improved compliance.Cervical screening is effective in reducing cervical cancer mortality but expensive, complex to organisc and often considered inefficient (Benedet et al., 1992; Smith et al., 1989;Herbst et al., 1992). Detection of human papillomavirus (HPV) DNA in exfoliated cervical cells offers a potentially automatic and cheap diagnostic test without the subjectivity and sampling problems of cervical cytology or the expense of colposcopy. Clinical series and case-control studies have shown a very strong association between HPV types 16, 18, 31 and 33 and cervical cancer, supported by a dose-response relationship for HPV 16 and an increased rate of progression to more advanced pre-cancer among HPV 16/18-positive women (Mufioz and Bosch, 1992). The common genital papillomaviruses have been divided into risk groups for progressive pre-cancer and invasive cancer, and a set of HPV types detected by PCR has been associated with at least 93% of, and possibly all, cervical cancer (Bosch et al., 1995;Meijer et al., 1992).HPV testing may have future application either in the management of women with mild cytological abnormalities, improving the sensitivity of cytology for high-grade pre-cancer (Cuzick et al., 1995;Cox et al., 1995), or as an alternative screening test (Meijer et al., 1992). An overview of the likely impact of HPV testing is important to direct research on HPV detection technology and its application in practice. However, to determine appropriate investigations is not simple since cytology, colposcopy and HPV tests can be combined sequentially in many complex ways as part of a screening programme. There is still uncertainty about the prevalence and natural history of HPV infections and their relation to cervical pre-cancer, as wcll as the accuracy and sequential use of HPV tests. There are also many variations in possible screening intervals, coverage and follow-ups.We have previously described a mathematical model of the development and progression of pre-cancer (Sherlaw-Johnson et al., 1994). This was used to compare colposcopic with cytological management of mildly abnormal smears under a wide range of variations in natural history...