HPV type-specific risks of high-grade CIN during 4 years of follow-up: A population-based prospective study

Abstract We followed a population-based cohort of 5696 women, 32–38 years of age, by registry linkage with cytology and pathology registries during a mean follow-up time of 4.1 years to assess the importance for CIN2+ development of type-specific HPV DNA positivity at baseline. HPV 16, 31 and 33 conveyed the highest risks and were responsible for 33.1, 18.3 and 7.7% of CIN2+ cases, respectively. Women infected with HPV 18, 35, 39, 45, 51, 52, 56, 58, 59 and 66 had significantly lower risks of CIN2+ than women infected with HPV 16. After adjustment for infection with other HPV types, HPV types 35, 45, 59 and 66 had no detectable association with CIN2+. In summary, the different HPV types found in cervical cancer show distinctly different CIN2+ risks, with high risks being restricted to HPV 16 and its close relatives HPV 31 and HPV 33. Main Infection with ‘high-risk’ types of HPV is the major cause of cervical cancer, and the distribution of different HPV types in cancer tissue has been extensively analysed (Walboomers et al, 1999; Bosch et al, 2002; Munoz et al, 2003). Fifteen HPV types that infect the genital mucosa have been proposed as ‘high-risk’ HPV types as they have been found more often in cervical cancers than among healthy subjects (Munoz et al, 2003). Data on the HPV type-specific risk of cervical neoplasia has hitherto mostly been based on cross-sectional case–control studies (Bosch et al, 1995; Clifford et al, 2003; Munoz et al, 2003). Case–control studies are sensitive to several biases, notably selection bias, differential sampling bias and other reverse causality biases. Absolute risks and population attributable proportions for each HPV type form the basis for decisions regarding which HPV types should be included in HPV screening tests as well as in vaccines and it is therefore important to estimate the type-specific risks using population-based prospective studies, a study design that minimises major sources of bias. Only two previous prospective studies have assessed the risk associated with several individual HPV types (Schiffman et al, 2005; Berkhof et al, 2006), while several prospective studies have assessed certain clusters of HPV types (Koutsky et al, 1992; Liaw et al, 1999; Sherman et al, 2003; Szoke et al, 2003; Peto et al, 2004; Winer et al, 2005) or only investigated HPV types 16 and 18 (Khan et al, 2005). Therefore, we HPV-tested a population-based cohort of women and used a comprehensive registry-based follow-up to identify the risks for development of histopathologically verified high-grade cervical intraepithelial neoplasia (CIN2+) associated with infections with 14 different so-called ‘high-risk’ HPV types. Materials and methods Cohort definition A population-based multicentre study was started in Sweden in May 1997 with the main purpose to evaluate the effect of HPV testing in primary cervical cancer screening. Women aged between 32 and 38 years (mean age: 35.1 years) in five regions in Sweden (Gothenburg, Malmö, Stockholm, Umeå and Uppsala) who took part in organised cervical screening were invited to take part in the study. Following informed consent, 12 527 women were enrolled and randomised either to action on HPV tests (6257 women) or to no action on HPV tests (6270 women), as described in detail elsewhere (Elfgren et al, 2005). All women had a cervical brush sample taken at baseline that was used for routine cytological screening and then frozen in 1 ml of 0.9% NaCl for future HPV DNA analysis. Referral to colposcopy was based on routine clinical management. Furthermore, HPV-positive women in the intervention arm that did not have an abnormal enrolment smear in cytology and pathology registries were invited for a second HPV test and cytology on average 19 months later and if persistently positive invited to colposcopy. A matched number of women from the control arm were also invited for HPV test, cytology and colposcopy (Elfgren et al, 2005). For the present study, a population-based cohort was formed from 6257 women in the intervention arm as well as 409 women randomly selected from the control arm that had HPV tests performed on their baseline samples. All women were followed by registry linkages using unique personal identification numbers with both the regional cytology and pathology registries in the enrolling regions, as well as with the national cervical screening registry, to detect development of CIN2+. All women with an abnormal histopathological diagnosis as well as all women invited for colposcopy within the study protocol had their specimens re-evaluated by a single expert pathologist (WR) who was masked to the HPV status of the women. For 22 specimens that could not be located in the pathology archives, the original diagnosis was retained. Overall, 126/148 (85%) of the CIN2+ diagnoses in the present study have been confirmed by expert pathologist review. Women who had an inadequate (β-globin-negative) HPV test at baseline (172 women) or who had no cytological or histological samples registered during follow-up (796 women) were excluded from the analysis. The final population-based cohort thus consisted of 5696 women with a mean follow-up time of 4.1 years. HPV DNA testing The virus laboratory that performed the HPV analyses was masked to cytological and histological diagnoses of the women. Cervical brush samples were analysed using a general HPV primer GP5+/6+-mediated PCR-enzyme immunoassay consisting of a pool of digoxigenin-labelled HPV type-specific oligonucleotide probes of 14 high-risk HPV types (types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 and 68) (de Roda Husman et al, 1995; Jacobs et al, 1997). Human β-globin was amplified simultaneously in the PCR-EIA assay to test for sample DNA quality. All HPV-positive samples were typed by reverse dot blot hybridisation (RDBH) using HPV type-specific plasmids corresponding to the different HPV types tested for in the PCR-EIA (Forslund et al, 2002). PCR-EIA-positive samples negative in RDBH were cloned and sequenced. Samples were considered HPV positive only if successfully typed by RDBH or by DNA sequencing. Statistical analysis Statistical analyses were performed using STATA 9.0. Absolute cumulative risks of future CIN2+ and CIN3+ with binominal exact 95% confidence intervals (CI) were calculated for each HPV type. Relative rates for CIN2+ and their 95% CI were calculated using Poisson regression. Women were censored at their last testing date, except for women who developed CIN2+ who were censored at the date when the diagnostic biopsy was taken. Since lesions that were detected early during follow-up might have been present already at baseline we split the data according to length of follow-up of more or less than 6 months. A likelihood ratio test was then performed to assess if there was interaction between HPV infection and time of CIN2+ diagnosis. There was weak evidence for interaction (P-value 0.10) and thus we adjusted for time of di (...truncated)