Despite widespread cervical cancer-screening programmes, 30 000 of the 190 million women living in Europe still die every year from this disease (Ferlay et al, 1988). The overall decline in cervical cancer incidence observed since the introduction of the Papanicolaou (PAP) smear has levelled off and even started to increase in countries such as England, Wales and Finland (Vizcaino et al, 2000), where organised screening programmes with extensive quality control have been in place for many years. In Germany, an opportunistic cervical cancer-screening programme was established in 1971, which made free annual cervical cancer screening available to all women 20 years of age or older. The average annual participation rate in 1997 was roughly 50% (Kahl et al, 1999; Klug et al, 2000; Klug and Blettner, 2003). When looked at over a 3-year period, which is the screening interval used in many European countries, the population coverage reaches 80% (Schenck and von Karsa, 2000) or higher in the 25–50-year-old age group. Despite this considerable effort, Germany has one of the highest cervical cancer incidence rates (13.28 per 100.000) and standardised mortality rates (4.4 per 100.000 for 1993) in Western Europe, which is higher than in neighbouring countries with similar socio-economic structures such as France, Italy and the Netherlands. Infection with high-risk human papillomavirus (HR-HPV) is the major risk factor for the development of cervical cancer (Bosch et al, 1995; Walboomers et al, 1999) and it has been suggested that testing for HR-HPV should be included in existing cervical cancer-screening programmes (Vizcaino et al, 2000). Large-scale screening studies by Schiffman et al (2000), Clavel et al (2001) and Schneider et al (1996),(2000) demonstrated that HPV testing is more sensitive for the detection of high-grade cervical lesions than cytology although the lower specificity of HPV testing made its use in primary cervical cancer screening questionable. However, the majority of these studies did not compare HPV testing to routine cytology, but rather used a team of expert cytologists thereby not allowing to draw conclusions about the performance of HPV testing under routine screening conditions. In addition, these studies included a high proportion of younger women aged 18 years and above, although it is known that young, sexually active women have a high prevalence of both HR-HPV and high-grade cervical intraepithelial neoplasia (CIN), the overwhelming majority of which resolves spontaneously (Evander et al, 1995; Ho et al, 1995). It is possible that the higher sensitivity of HR-HPV testing in these young populations in comparison with cytology may partially be because of the identification of transient disease with a high tendency for regression. To address this particular issue, Cuzick et al (1999) examined the performance of HPV testing in women 35 years of age or older and again found that HR-HPV testing had a higher sensitivity than cytology, but a lower specificity and positive predictive value (PPV) leading to a higher referral rate than cytology alone. Importantly, all of the former studies are characterised by the lack of an HPV−/Pap− control group examined by colposcopy, thereby not allowing an accurate estimate of the false-negative rate of the screening regimen (Ratnam et al, 2000). Our aim was to determine the value of HPV testing in the routine primary cervical cancer-screening programme in Germany for the detection of high-grade cervical cancer precursors (⩾CIN2). To avoid confounding factors, we implemented different levels of quality controls (see study protocol in Figure 1 Figure 1 Study Protocol. ) to evaluate the use of HPV testing. All the 8083 study participants finally included were aged 30 years or older attending for routine cancer screening and represented a random unselected population that was highly representative of Western Germany as a whole. All positive and 5% of the negative HC2 samples were retested by a highly sensitive, broad-range, consensus-primer PCR with direct sequencing to evaluate HC2 performance. Digital images of the cervix were taken at the colposcopic examination and reviewed by independent experts blinded to previous findings. Additional expert reviews were performed on all histology samples and all cytology samples with any degree of dyskaryosis, all cytology samples from HR-HPV+ women and 5% of randomly selected cyto-/HPV-cytology samples. Most importantly, a colposcopic examination of a random sample of 250 women who were cytology and HPV− was performed in order to estimate the bias that would be introduced by disease missed in women who were falsely negative on both tests (Ratnam et al, 2000). Cytology labs undertaking the primary cytology were not informed about which Pap smears were included in the trial and the samples were therefore screened under routine conditions. The statistical analysis of the raw data was performed by an independent group (Erasmus University, Rotterdam) that was not involved in the screening trial nor in the collection of the data. MATERIALS AND METHODS Study population Between December 1998 and December 2000, 8466 women attending routine cervical cancer screening were recruited from 28 urban, suburban or rural, office-based gynaecological practices from Hannover and Tuebingen and the surrounding areas. Women were eligible for inclusion if they were attending for routine annual cervical cancer screening, were 30 years of age or older, had not undergone a hysterectomy, had no history of atypical cytology, CIN, or treatment for cervical disease in the preceding year and were currently not pregnant. Of the 8466 women enrolled in the study, 8101 women met the inclusion criteria. Of the 4566 participants from Tübingen, 233 (5.1%) were excluded, leaving 4333. Of the 3900 women enrolled from Hannover, 132 women (3.4%) were excluded leaving 3768. Reasons for exclusion (365 women) were: 43 women with genital warts; 13 women with a history of conisation or hysterectomy; 11 women who were pregnant; eight women with abnormal cytology within 1 year of study entry; 167 women who were under the age of 30 years and 123 women who had not given written consent. Written informed consent was obtained from the patients by the participating gynaecologists. The study was approved by the local ethics committees at the Universities of Hannover and Tuebingen. Screening examinations At the first gynaecological examination, the cervix was visualised and a sample was taken for routine cervical cytology following the procedures normally used in each gynaecological practice (most, but not all, samples were taken with a cotton-tipped swab, rolled onto a microscope slide and spray fixed). This practice of sample taking is the standard recommended procedure in Germany (Link M and Link H, 2001). A second sample was then obtained with a cervical sample device (Digene Inc. Gaithersburg, MD, USA) and suspended in 1 ml of specimen transport medium (STM/ Digene Inc, Gaithersburg, MD, USA) for HPV DNA testing. Women were allocated to follow-up according to the protocol illustrated (Figure 1). Cytological diagnoses All cervical smears were analysed at one of the cytology laboratories normally used by each participating office-based gynaecology practice and reported in accordance with the Second Munich Cytological Classification (Table 1 Table 1 Correlation between second Munich classification and the Bethesda system Munich classification Bethesda correlate PapI/II Normal/inflammatory PapIII ASC-H and AGUS–cannot exclude high-grade disease or cancer PapIIId CIN1–2 PapIva CIN3–carcinoma in situ PapIvb Carcinoma in situ–microinvasive carcinoma PapV Microinvasive carcinoma–invasive carcinoma Unofficial classification Bethesda correlate PapIIw Includes inadequate and ASC-US ). The ‘Pap II w’ (W=wiederholen=repeat) is a widely used category although it is not an official cytological classification. It is used by cytologists to describe inadequate specimens, minimal changes and koilocytes in the absence of further abnormalities or atypical squamous cells of undetermined significance (ASCUS). In this study, smears were considered positive if any degree of cytological abnormality was observed (⩾PapIIw, ≈borderline/ASCUS) in order to maximise detection of disease. The cytology laboratories had not been informed about the study and were therefore analysing the smears under routine screening conditions. All glass slides and accompanying forms lacked information on the patient's participation in a trial to prevent bias in reading. HPV DNA testing by the HC2 assay Samples for HPV testing were stored at 4°C for a maximum of 4 weeks prior to testing. All primary HPV testing was undertaken using the Hybrid Capture 2 test (HC2/Digene Inc, Gaithersburg, MD, USA) at the clinical diagnostic laboratory of the Medizinische Hochschule Hannover or at the Department of Experimental Virology at the University of Tuebingen. All samples were analysed for the presence of 13 HR-HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68) following the manufacturer's instructions. Therefore, a positive HPV result refers to a patient who is positive for one or more of the 13 HR-HPV types included in the high-risk probe mix. Infection with low-risk HPV types was not evaluated. Samples were considered positive if they attained or exceeded the FDA-approved threshold of 1.0 pg HPV DNA ml−1, which corresponds to 1.0 relative light units (RLU). All samples with RLU values in the range of 0.7–2.0 RLU were retested in duplicate, although subsequent statistical analysis was based only on the primary measurement. HPV testing by PCR DNA extraction, sample preparation and sample analysis procedures were carried out in separate rooms. All samples were tested for DNA integrity by PCR using primers for the human β-globin gene. Three different primer combinations were used (PPF1/PPR2, PPF1/CP5, CP4/CP5) producing amplicons ranging in size from 220 to 450 bp, all located in the highly conserved helicase region of the E1 gene. The sensitivity of this PCR system for different HPV types was established by testing serial dilutions of HPV plasmids for the following types: HPV1–8, 10–19, 21–26, 30–38, 40, 45–47 and 60, both in the presence and absence of human placenta DNA (1 μg). Primers PPF1/PPR2, which are not degenerate, reach the highest sensiti-vity of 10 genome copies per sample and result in a single 220 bp amplification product when applied to DNA extracted from clinical samples. PPF1/CP5 and CP4/CP5 have a detection limit of 100 copies, which is by far sufficient for the testing of clinical samples of the cervix. Primers PPF1/CP5, where the downstream primer is degenerate in three positions, detect at least 64 different HPV types and produce a 280 bp amplicon, while primers CP4/CP5 with five degenerate nucleotide positions generate a fragment of 450 bp that has been described elsewhere (Tieben et al, 1994). The nucleotide sequence of the primers is as follows: PPF1 5′-(nt 2082)-AAC-AAT-GTG-TAG-ACA-TTA-TAA-ACG-AGC-(nt 2108)-3′ PPR2 5′-(nt2336)-ATT-AAA-CTC-ATT-CCA-AAA-TAT-GA-(nt2314)-3′ CP4 5′-(nt1942)-ATG-GTA-CAR-TGG-GCA-TWT-GA-(nt1961)-3′ CP5 5′-(nt 2400)-GAG-GYT-GCA-ACC-AAA-AMT-GRC-T-(nt 2378)-3′ Numbers are according to the sequence of HPV16W12E (Genbank IDNr. AF125673). Direct automated sequencing of the PCR products was carried out using the Big Dye Terminator Cycle Sequencing Ready Reaction Kit (PE Biosystems, Foster City, CA, USA). Sequencing was performed in 47 cm capillaries with the use of an ABI 310 sequencer (PE Biosystems). Sequences with 5 years), those women with low-risk sexual behaviour who are negative both on cytology and on HPV test could be grouped safely on a recall interval of 3–5 years or more, which would generate savings that could be applied to a better surveillance for those women who are at increased risk.