Particular types of human papillomavirus (HPV) infection may preferentially progress from high-grade squamous intraepithelial lesions (HSIL) to squamous cell carcinoma of the cervix (SCC). We performed a meta-analysis of published data to compare HPV type distribution in HSIL and SCC. HPV16, 18 and 45 were each more prevalent in SCC than HSIL, whereas the reverse was true for other oncogenic types including HPV31, 33, 52 and 58. These data suggest that HSILs infected with HPV16, 18 and 45 preferentially progress to SCC. This may have implications for follow-up protocols of future HPV-based cervical cancer screening programmes and for HPV vaccine trials. (Walboomers et al, 1999). However, only a fraction of precancerous lesions progress to ICC. A strong candidate factor for differential progression is HPV type (Lorincz et al, 1992).Identifying HPV types that preferentially progress from highgrade squamous intraepithelial lesions (HSIL) to ICC has implications not only for follow-up protocols in ICC screening programmes, but also for prophylactic type-specific HPV vaccine trials. For ethical reasons, final outcome measures in such trials will be the prevention of HSIL. However, it is important to know whether the HPV type distribution in HSIL is representative of those that go on to cause cancer.Articles presenting HPV type-specific prevalence data were identified from Medline. Studies had to include at least 20 cases of squamous cell or histologically unspecified cervical cancer (Clifford et al, 2002) and/or 20 histologically verified cases of HSIL. In this study, HSIL refers both to lesions classified by the Bethesda system, that is, CIN2/3, and those classified separately as CIN2 and CIN3. Studies had to use polymerase chain reaction (PCR)-based assays to identify HPV, and to present prevalence of at least one type other than HPV6, 11, 16 or 18 (Clifford et al, 2002).This report includes 8594 squamous cell carcinoma of the cervix (SCC) cases (including 2725 of unspecified histology), as previously reported in Clifford et al (2002), and 4338 HSIL cases (1733 reported as CIN2/3, 1824 as CIN3, 729 as CIN2 and 52 as cervical carcinoma in situ)(detailed information on the HSIL studies is reported in the Appendix). Compared to SCC, cases of HSIL were more likely to be from (i) Europe and South/Central America rather than other regions, (ii) studies that detected HPV from exfoliated cells rather than biopsy specimens and (iii) studies that used 'broad'-spectrum (MY09/11, GP5 þ /6 þ and SPF10) rather than other PCR primers (Table 1).Type-specific prevalence is presented for the 14 most common high-risk (HR) types identified in SCC ( Table 2). As not all studies tested for all 14 types, sample size varies between type-specific analyses. Type-specific prevalence is expressed as a percentage of all cases tested for HPV, and thus represents the prevalence in either single or multiple infections.Overall, HPV prevalence was slightly higher in SCC cases (87.6%) than HSIL (84.2%) (SCC : HSIL ratio 1.04, 95% CI 1.03 -1.06) ( T...