Long-term protective effect of high-risk human papillomavirus testing in population-based cervical screening

We prospectively evaluated the 5-year predictive values of adding high-risk human papillomavirus (hrHPV) testing to cytology for the detection of ⩾cervical intraepithelial neoplasia (CIN)3 lesions in a population-based cohort of 2810 women. At baseline, nine (0.3%) women had prevalent lesions ⩾CIN3, all being hrHPV positive. After 5 years of follow-up, four (6.5%) of the 62 hrHPV-positive women with normal cytology developed lesions ⩾CIN3, vs only one (0.05%) of the 2175 hrHPV-negative women with normal cytology. High-risk human papillomavirus testing or combined screening revealed a much higher sensitivity, at the cost of a small decrease in specificity, and a higher negative predictive value for the detection of lesions ⩾CIN3 till the next screening round (5 years) than cytology alone.


MATERIAL AND METHODS
A cohort of 3170 women (mean age 45 years; range 29 -61 years) was enrolled from March 1995 till October 1998 by 60 general practitioners in the small district of Amstelveen, The Netherlands. In all, 194 women were excluded because of abnormal cervical cytology and/or histology during 2 years preceding the intake, 159 women because of a negative b-globin PCR test, and seven women because of inadequate cytology, leaving 2810 women for analysis. Cytological screening was performed according to the CISOE-A classification, routinely used in cervical screening in The Netherlands, and the referral policy was according to the nationwide guidelines (Bulk et al, 2004). High-risk human papillomavirus testing was performed by GP5 þ /6 þ PCR-EIA, using a cocktail probe of 14 hrHPV types, and independent of cytology results. The hrHPV test result was blinded. Informed consent was obtained and the study was approved by the Medical Ethics Committee of the VU University Medical Center. The primary end point of the study was the detection of histologically proven cervical intraepithelial neoplasia grade 3 or cervical carcinoma (XCIN3) up to and including the next screening round (after 5 years). Follow-up data were retrieved from the Dutch nationwide pathology registry (PALGA) in 2004.
Sensitivity, specificity, negative predictive value (NPV) and positive predictive value (PPV) of cytology, cytology and hrHPV testing combined, and hrHPV testing, for the detection of prevalent and incident lesions XCIN3 were computed using two by two tables. Results are presented as percentages with 95% confidence intervals (95% CI). To determine whether any gain in test performance from the addition of the second test (hrHPV testing) was greater than that expected if the second test offered no diagnostic information, expected performance characteristics for cytology combined with a random test having the similar prevalence as hrHPV were computed (Franco and Ferenczy, 1999).
The median follow-up time until histological or cytological diagnoses was 4.6 years (range 0.1 -8 years). The follow-up results of women with normal cytology at baseline are presented in Table 2. Of the 2687 women with normal cytology, 304 did not have follow-up data registered in PALGA. Another 146 women were excluded because they reached age 60 years or over and therefore were not called for further screening, leaving 2237 (83.3%) women with normal cytology for follow-up analysis. Among the 62 women of this cohort with an hrHPV-positive test at baseline, four (6.5%) cases had incident lesions XCIN3. Among the 2175 women with a negative hrHPV test, one (0.05%) lesion XCIN3 (an adenocarcinoma) was registered within this follow-up period. This woman with a symptomatic adenocarcinoma, who was tested hrHPV negative by GP5 þ /6 þ PCR EIA twice, revealed human papillomavirus (HPV) type 31 by E7 type-specific PCR in the baseline and follow-up smear. This suggests viral integration with disruption of the L1 region that is targeted by the GP5 þ and GP6 þ primers. For hrHPV-positive women with normal cytology, the relative risk to develop lesions XCIN3 in 5 years was 140.3 (95% CI; 15.9 -1237.3) compared to hrHPV-negative women with normal cytology. The overall detection rate after 5 years in this study of lesions XCIN3 by cytology was 0.5%.

DISCUSSION
Combined screening revealed a much higher sensitivity, at the cost of a small decrease of specificity, and a higher NPV at the next screening interval (after 5 years) than cytology alone. Cytology alone had reasonable performance characteristics, but on adding hrHPV testing we can achieve much better performance characteristics in population-based screening. The overall detection rate by cytology of 0.3% lesions XCIN3 at baseline and 0.5% after 5 years in this study is comparable to the detection rate of lesions XCIN3 in cervical cancer screening in The Netherlands (Bos et al, 2002;Anttila et al, 2004). The high NPV of the combination of a negative hrHPV test and a normal smear is in accordance with Clavel et al (2004), who reported an NPV of 99.9% in a partly hospital-based population, with a much shorter interval (median 2.8 years) for women of 15 -79 years of age. Sherman reported an NPV of 99.2% for hrHPV in cervical lavage specimens during annual screening for women of 16 -94 years of age (Sherman et al, 2003). The data are also in line with the data of the Manchester cohort (Peto et al, 2004).
Our data are the first that were prospectively obtained in population-based screening, with a 5-years screening interval in women 30 -60 years of age. Some methodological aspects of this study need to be discussed. The fact that histology was not obtained in all women might induce a verification bias in advantage of combined screening and hrHPV testing. However, women were followed according to current practice standards in nationwide screening with cytology after 5 years, and in this setting women with normal cytology are considered to be free of disease. As indicated by the overlapping 95% CIs, the gains in sensitivity and NPV by combined testing or hrHPV testing compared to cytology alone were not significant (Pp0.05). This might be due to our relatively small population with a low prevalence of lesions XCIN3, resulting in wide 95% CIs. However, with this low prevalence of lesions XCIN3, still a borderline significance was reached (Pp0.1). The increase in sensitivity by combined testing compared to cytology alone may be misleading because improvements in sensitivity would be expected by adding a second test, even if the second test performed randomly with respect to disease identification. For this reason, expected performance characteristics for cytology combined with a random test were computed to determine if any gain in test performance from the addition of the second test (hrHPV testing) was greater than that expected if the second test offered no diagnostic information (Franco and Ferenczy, 1999). By comparing the sensitivity and NPV of cytology and hrHPV combined screening to cytology with a random test, no different significance levels were revealed as when compared to cytology alone.
Cost-effectiveness studies and modelling studies show that cervical screening may become much more efficient in terms of decreasing numbers of false-negative and false-positive smears, if a test is used with a substantial higher sensitivity and long-term NPV than conventional cytology (Canfell et al, 2004). Negative test results in combined screening predicted that the future risk for lesions XCIN3 was very low. The higher sensitivity for lesions XCIN3 and the long-term NPV for lesions XCIN3 of hrHPV testing in combination with classical cytology and of sole hrHPV testing show that these could be such a test.
The use of hrHPV testing in cervical screening could lead to several different screening strategies, including combined cytology and hrHPV testing, or primary screening by hrHPV with cytology reading only of women tested hrHPV positive. In addition, the high sensitivity and NPV of hrHPV testing opens possibilities for longer screening intervals with still acceptable rates of incipient lesions. Modelling studies and confirmation of our results in larger studies will help to clarify this discussion and to devise more efficient cervical screening strategies.