Prevalence of Human Papillomavirus Types in High-Grade Cervical Intraepithelial Neoplasia and Cancer in Italy

Two new approaches for the prevention of cervical cancer have emerged over the last few years: vaccination against human papillomavirus (HPV) in adolescent girls and HPV test as primary screening test (1-8). In Italy, the vaccine is currently recommended for all 11-year-old girls and the vaccination campaign aims at covering 95% of the target population within 5 years since it began in 2008 (9).

Population-based studies of HPV genotype prevalence are needed to predict how these two approaches might influence cervical cancer prevention. To monitor whether widespread vaccination might induce changes in the distribution of HPV types in high-grade lesions or cancer, we need solid baseline data on HPV epidemiology not only in the healthy population but also in women with cervical neoplasia. Moreover, the ability to develop the most appropriate screening algorithm and protocols for vaccinated cohorts depends on our understanding of the absolute risk of non-HPV16/18 cervical cancer by age and history of screening.

A meta-analysis of worldwide studies showed that the proportion of HPV16 and HPV18 among invasive cancers is quite stable and that they are present in the vast majority of invasive cases without any correlation with the geographic location (10, 11). On the other hand, the distribution of other HPV types shows some geographic variation (12-15). Useful information in public health decision making is the proportion of non-HPV16/18 in invasive cancers from younger women. Knowing this can help to determine the best age at which screening for vaccinated women should start (16).

Several studies about HPV types in precancer and cancer have been conducted in Italy, but in most of them, sample size was small. Table 1 summarizes results and sample sizes from previous Italian studies.

The aim of this multicentric study was to describe the distribution of HPV types in cervical neoplasia, by age, grade of lesion, and histologic type, with an appropriate sample size, from relatively recent samples collected by eight laboratories in six different regions from central and southern Italy. Our results might have an effect on decisions about the age at which HPV screening for vaccinated women should start.

This multicentric study involved eight centers from six different Italian regions: Abruzzo (Atri), Lazio (S. Giovanni Hospital in Rome and Belcolle Hospital in Viterbo), Campania (Naples: National Cancer Institute Fond Pascale), Sicily (Catania), Sardinia (Cagliari), and Tuscany (Florence: ISPO and S. Maria Annunziata Hospital). Specimens were retrieved from the archives of all centers, but molecular analyses were centralized in two laboratories (ISPO in Florence and National Cancer Institute Fond Pascale in Naples).

The coordinating laboratory collected the lists of patients with cervical neoplasia from the databases of the participating centers' pathology units (list of ICD-O codes, SNOWMED, text). Information collected included date and place of birth, date of diagnosis, diagnosis code, or text. The calendar period we considered coincided with the availability of electronic archives (1999-2008).

Only histologically confirmed diagnoses were included in the study. All invasive squamous cancers, adenocarcinoma (in situ or invasive), and all cases from women born abroad were included in the study, whereas a random sample of cervical intraepithelial neoplasia 2 (CIN2) and CIN3 was drawn to reach a minimum of 1,200 cases. The lists of sampled women were then given back to each center to confirm the original diagnosis checking all the medical records. Finally, the pathology units retrieved the samples from their specimen bank, selecting the specimen with the most severe diagnosis. When multiple paraffin-embedded tissues were present for a single case, the tissue sample used for this analysis was selected in the following order of priority: the original cervical punch biopsy sample, followed by the loop excision specimen, followed by the hysterectomy specimen. Histologic samples were not available for some screen-detected lesions (157 CIN2/3 and 1 invasive cervical carcinoma), so HPV typing was done on the original cervical samples.

Paraffin-embedded tissue samples were subjected to a sandwich technique for which we obtained an initial 10-μm tissue section for H&E staining, followed by three/five 10-μm sections that were collected in two separate tubes for PCR, and immediately sent to the reference facility.

PCR-safe precautions were taken to minimize the risk of contamination during tissue processing. Sections were cut from the tissue blocks with a standard microtome, carefully cleaning the microtome and using a new part of the blade for each block, changing gloves between specimens, and using disposable toothpicks (sticks) to transfer tissue ribbons to their storage tubes. Finally, tissue processing and HPV testing were done in different facilities.

For each sample, the following information was recorded: date of sampling, type of material (biopsy or surgery sample), number of sections in each tube, and date sent.

H&E slides immediately adjacent to the tissue ribbons used for HPV DNA typing had to be reviewed by the local pathologist for the presence of lesions to check that neoplastic tissue had been included. In one center (Naples), this morphologic check showed that the mean percentage of tumor cell staining was 64% (range, 5-100%), with a relevant proportion of sample slices containing a limited quantity of neoplastic tissue within a healthy tissue context.

Cervical scrape samples were collected in ThinPrep vial containing PreservCyt transport medium (Cytyc Corp.) or in Specimen Transport Medium (Digene, Qiagen). At the time of diagnosis, 400 μL of STM samples were stored at −80°C, whereas 1.5 mL of the ThinPrep vial were centrifuged (10 minutes at 3,000 rpm) and washed twice with 10 mL of 1× PBS. Pellets were resuspended in 2 mL PBS and stored at −80°C until DNA extraction.

DNA extraction and HPV genotyping were centralized in two laboratories (Florence and Naples) using harmonized protocols. ISPO in Florence analyzed samples from Abruzzo, Lazio, Sicily, Sardinia, and Tuscany, whereas the Pascale National Cancer Institute analyzed samples from Campania.

DNA was extracted from two sections (10 μm) of paraffin-embedded tissue using QIAamp DNA FFPE Tissue kit (Qiagen) according to the manufacturer's instruction with slight modifications. Paraffin was dissolved in 1 mL of xylene. Shaking at room temperature for 15 to 30 minutes followed. Sections were pelleted by centrifugation (14,000 rpm for 5 minutes) and washed twice in 100% ethanol. Pellets were dried at room temperature (for 30 minutes) and then at 37°C (for 15 minutes). Dried samples were resuspended in 180 μL ATL Buffer and 20 μL proteinase K, mixed, and then incubated at 56°C overnight (until complete lysis). Samples were then incubated for 1 hour at 90°C to partially reverse formaldehyde modification of nucleic acids. After this step, the manufacturer's protocol was strictly followed and DNA was eluted in 100 μL ATE Buffer.

DNA from cervical specimens was extracted using QIAamp DNA Mini kit (Qiagen) according to the manufacturer's instructions.

A sample known to be negative for HPV was included in each batch of extraction to exclude any contamination during this phase.

For each sample, DNA concentration was evaluated by spectrophotometry and expressed in ng/μL.

HPV genotyping was done using different PCR-based strategies in a well-defined algorithm. For all samples, we initially used a PCR assay based (30) on GP5+/GP6+ primers (Consensus High Risk HPV Genotyping kit, Digene, Qiagen), according to the manufacturer's instruction with slight modifications. An HPV-positive control and two negative PCR controls (a purified DNA sample negative for HPV and a DNA-free sample) were included in each PCR run. PCR products were loaded onto 2% agarose gel, stained with ethidium bromide, and visualized under UV illumination. GP5+/GP6+ primers amplify a broad spectrum of HPV genotypes by targeting a 150-bp fragment within the L1 open reading frame of the HPV genome (31). All samples and controls, independent of gel results, were subjected to the reverse line blot (RLB) for detection of 12 high-risk HPV types (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 59), 1 probably carcinogenic HPV type (68), and 5 possibly carcinogenic HPV types (26, 53, 66, 73, and 82; ref. 32). RLB strips were analyzed visually using an interpretation grid supplied with the kit. A biotinylated poly(dT) control for conjugate reaction is present in each strip to ensure good performance of the test and proper alignment of the strips on the interpretation sheet.

GP5+/GP6+ PCR-negative and RLB-negative samples were amplified for the β-globin gene sequence using GH20-PC04 primers (268-bp amplicon length) to assess DNA integrity (33).

To overcome any false negatives due to inhibitors commonly present in formalin-fixed, paraffin-embedded tissue, all samples negative for HPV at this first step were retested on 1:10, 1:50, or 1:100 dilution.

To overcome misclassification of the HPV genotype resulting from potentially degraded DNA in aging archival paraffin-embedded tissues, samples that still remained negative for HPV on diluted DNA were reamplified using INNO-LiPA HPV Genotyping Extra Amp kit (Innogenetics), targeting a 65-bp fragment within the virus L1 open reading frame (34-36). INNO-LiPA genotyping kit was also used on the GP5+/GP6+ PCR-positive but RLB-negative samples because it allows for detection of 28 HPV types (16, 18, 26 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82, 6, 11, 40, 43, 44, 54, 70, 69, 71, and 74). This step was done only in Florence. All clearly visible lines should be scored by using the INNO-LiPA HPV Genotyping Extra Reading Card. The remaining untyped samples (GP5+/GP6+ PCR positive and RLB negative or INNO-LiPA negative) underwent direct Sanger sequencing of the GP5+/GP6+ PCR product using BigDye 1.1 terminator chemistry (Applied Biosystems) and ABI Prism 310 Genetic Analyzer (Applied Biosystems) to identify the specific HPV types probably not included in the Digene genotyping kit or in the INNO-LiPA kit. PCR products were sequenced in both directions, and the resulting sequences were compared with known HPV sequences in public databases using BLAST. A sequence was considered a match if it had >90% nucleotide similarity to an HPV sequence in GenBank.

To overcome negative results due to low copy number of HPV DNA, samples that remained HPV negative (while β-globin positive) were further analyzed by nested PCR using the PGMY09/11 and GP5+/GP6+ primer sets (37), followed in Florence by RLB hybridization (Consensus High Risk HPV Genotyping kit), as previously described, whereas in Naples samples underwent direct sequencing of the nested amplification products.

In the last step, the remaining HPV-negative samples (β-globin positive) were amplified using HPV-specific primers for the E6 to E7 regions of four high-risk HPV types (18, 16, 45, and 31; ref. 33) to overcome HPV-negative results due to integration events affecting the HPV L1 gene, which is the target of all the PCR-based tests we used. The PCR products were loaded onto 2% agarose gel, stained with ethidium bromide, and visualized under UV illumination. Samples that presented a specific PCR fragment were considered positive for the specific HPV type.

Our goal in combining test results was to maximize the detection of HPV genotypes, not to evaluate the performance of any one test.

We searched Medline using the terms (“HPV” OR (“human” AND “Papillomavirus”)) AND (“type” OR “typing”) AND (“Italy” OR “Italian”) AND (“cervix” OR “cervical”) AND (“cancer” OR “neoplasia”). The search was updated on December 15, 2009. We excluded all studies targeted to the healthy general population and HIV-positive women, and not regarding cervical neoplasia. We included only studies of pathologic populations. We extracted the total HPV positivity, the proportion of HPV16 among isolated HPV, the proportion of HPV18, and the proportion of cases with HPV16 or HPV18 by grade of pathologic finding: histologic classification (if reported) or cytologic classification (if histology was not reported).

The following analyses were planned a priori in the study design: proportion of HPV16/18 by age group and grade of lesion (CIN2, CIN3 including adenocarcinoma in situ, invasive squamous carcinoma or adenocarcinoma, test for trend), proportion of HPV16/18 in adenocarcinomas (in situ and invasive), proportion of HPV16/18 in women born abroad, and distribution of HPV types by grade of lesion. Furthermore, we present the distribution of types by center and the distribution of adenocarcinomas (in situ and invasive).

It was impossible to calculate a priori the precision (power) of this study because we did not know how many invasive or adenocarcinomas would be retrieved.

Our sample included 1,162 cases. For 278 of them, it was impossible to obtain a biological sample mainly because the paraffin block had not been correctly stored or because the histologic sample had been completely sliced and used for the original diagnosis. One hundred and sixty-two additional cases were excluded because the diagnosis was not confirmed (131 of 884 = 14.8%, mostly cases reported as “Uterus NAS,” which resulted to be endometrial adenocarcinomas at the examination of medical records) or because there was no human DNA in the sample (31 of 753 = 4.1%; Table 2).

The laboratory in Florence analyzed 598 samples (158 cytologic samples and 440 tissue specimens) and the one in Naples analyzed 155 tissue specimens.

In Florence, of 598 samples, 4 samples were negative for HPV DNA, 537 were genotyped directly with GP5+/6+ RLB, 25 were typed on diluted DNA with GP5+/GP6+ RLB, 10 by INNO-LiPA, 14 by nested PCR, 8 samples were sequenced and for 3 of them a type was identified, and 2 more samples were genotyped with type-specific E6/E7 HPV primers. Finally, 3 samples remained untyped.

In Naples, 31 of 155 tissue biopsies (20%) were negative to human DNA determined by amplification of the β-globin gene. These were not considered in the following analyses; 47 of 124 samples were genotyped with GP5+/GP6+ RLB, directly or diluted, 13 were typed by nested PCR, 13 were typed by HPV16 E6/E7-specific primers, 5 samples were sequenced but remained undetermined, and 46 samples remained HPV negative.

Overall, 50 (6.9%) cases were negative for HPV DNA (13.2%, 4.7%, and 6.7% for CIN2, CIN3, and invasive cancer, respectively). Forty-six of these were clustered in one center, which analyzed paraffin blocks from an archive already used for several investigations, and all of them contained <20% of tumor cells as evaluated by H&E staining. Excluding this center, 99.3% of samples resulted positive for HPV DNA. Eight cases were positive for HPV, but it was not possible to determine the type.

The flowchart in Fig. 1 summarizes the results of HPV typing in the two laboratories, whereas Table 3 shows the HPV strains found by type of lesion.

The most frequent type was HPV16 in all classes. The rank of the other HPV types differs according to the grade of lesion: in CIN2/3, HPV16 is followed by HPV31, HPV18, HPV58, and HPV33; in invasive cancers, HPV16 is followed by HPV18, HPV45, HPV31, and HPV58.

The percentage of cases with HPV16 or HPV18, alone or in coinfection with other types, was 63.6% [95% confidence interval (95% CI), 54-72] in CIN2, 76.5% (95% CI, 72-81) in CIN3, and 80.2% (95% CI, 73-85) in invasive cancers. The observed trend is statistically significant (P = 0.002). The figures slightly decrease if we consider only cases with HPV16 and/or HPV18 not associated with other high-risk HPV types: 51.2% in CIN2, 66.7% in CIN3, and 75.1% in invasive cancers.

The distribution of cases among adenocarcinomas is similar to the types observed in squamous invasive cancers (16, 45, 18, 31, 58, and 35), and the only difference is a slightly lower proportion of HPV16 (P = 0.5, difference is not significant); the overall proportion of HPV16/18 is 71.4% (95% CI, 54-85), including coinfections with other high-risk HPV, whereas the percentage decreases to 60.0% if we exclude coinfections.

The proportion of HPV16/18 in invasive cancers shows a statistically significant decreasing trend with age: from 92% in women under 34 to 73% in women ages 55 and older (P = 0.036, test for linear trend). In women <35, only HPV16, HPV18, and HPV45 were found, whereas in older ages, other and rare types were also present. The trend is not significant for CIN2 (P = 0.4), and there is no trend for CIN3 (Fig. 2). The percentage of cases with HPV16 or HPV18 varied significantly among geographic regions in precancer [χ²(5) = 24.0, P < 0.0005 for CIN2; χ²(5) = 14.9, P < 0.01 for CIN3] but not in invasive cancers [χ²(5) = 6.7, P = 0.25 for invasive cancers; Table 4]. There was no difference in the proportion of HPV16/18 in CIN2 or CIN3 in liquid-based cytology (LBC) specimens compared with histologic specimens (63% in LBC and 68% in histologic samples for CIN2, P = 0.53; 69% in LBC and 71% in histologic samples for CIN3, P = 0.62).

Considering only CIN2 and CIN3, coinfections were more frequent, even if not significantly, in cytologic (18.8%) than in tissue samples (14.8%; P = 0.3). No differences in coinfections were found by age neither in CIN2/3 nor in invasive cancer. In 10.2% of invasive cancers, we observed a coinfection, whereas in CIN3, the rate was 15.6% and in CIN2 it was 17.4%; the decreasing trend is not significant (P = 0.15, adjusted for type of specimen). Coinfections varied also among regions [χ²(5) = 10.6, P = 0.06], with a higher proportion in Campania (25.0%) and Tuscany (20.5%). HPV16 was involved in 77% of coinfections. The frequency of combinations seems to reflect the frequency of single virus strains in each type of lesion, even if sample size does not allow for detection of medium or small differences. Matrices of all the observed combinations of types by grade of lesion are available online (Supplementary Table S1A-D).

There were 106 women who were born abroad. The proportion of HPV16/18 in women born abroad was 70.8% (95% CI, 56-83) in CIN2, 64.6% (95% CI, 49-78) in CIN3, and 90.0% (95% CI, 55-99) in invasive cancers. These figures do not differ from the ones observed in women born in Italy.

We found an invasive cancer, positive for HPV DNA, negative to both typing methods, in which we obtained a DNA sequence of HPV67.

One hundred and ninety-three articles were found in PubMed according to the searching criteria; 14 of which fulfilled the inclusion criteria. Five articles reported results for histologically confirmed invasive cancers (18-20, 22, and 27), six reported results for preinvasive lesions according to histologic findings (17-24), six reported results only according to cytologic results (25-29), and one by both classifications (17). Results are summarized in Table 1. In all studies, the percentage of HPV16 increased with the grade of pathologic findings. The percentage of HPV16 among CIN2 ranged from 29% to 53% [the high-grade squamous intraepithelial lesions (H-SIL) have a similar range], whereas the proportion of CIN3 ranged from 42% to 56% and in invasive cancers from 40% to 74%.

As expected, given the natural progression of the disease, the overall HPV positivity was high; the observed prevalence agrees with results from epidemiologic studies co-coordinated by the IARC (overall HPV/DNA prevalence: 92.5%; ref. 38) and with more recently published data in which HPV positivity ranges from 86% to 94%, depending on the continent (39). It is generally accepted that differences in the percentage of HPV-negative samples across studies can be explained by differences in the methods used to assess HPV DNA positivity, in the histopathologic quality of samples, and in the type of specimens analyzed (biopsies, surgical specimens, or fresh tissues). Evaluation of the present study has to take into account that it was not possible to check the histology of the biological sample collected at some centers, and that the vast majority of negative samples originated from one of these centers: If we ignore data collected at this facility, only 1 of 147 cancers resulted to be HPV negative. Similarly, only few positive samples were not typed.

The observed distribution of HPV types in CIN2 and CIN3 slightly differs from the distribution observed by the large IARC meta-analysis for H-SIL in Europe (10): In our cases, both HPV16 and HPV58 show a higher frequency. The same happens in invasive carcinomas, where we find a higher proportion of HPV16, HPV45, and HPV58 and a lower frequency of HPV18, HPV31, and HPV33 compared with the IARC analysis.

Table 1 summarizes the results of the 11 Italian studies found in our systematic review. The limited sample size does not permit an analysis for each type. Only a comparison for the proportion of HPV16 can be done, whereas HPV18 shows large random fluctuations and few studies clearly report the proportion of cases with HPV16 or HPV18 infection. In general, except for Sideri et al. (22) and Tornesello et al. (27), all the reviewed studies have a few invasive cancers. Our estimates for the proportion of HPV16 in CIN2, CIN3, and invasive cancer are, however, close to or little over the upper level of the ranges found in previous Italian literature.

The proportion of HPV16/18 increases with the grade of neoplasia in our data set as well as in all literature reports (14, 15). Furthermore, the variability among regions is higher for precancer than for cancer, as observed in previous meta-analyses (10). The observed proportion is slightly higher than what is observed in previous Italian studies for each of the three classes: CIN2, CIN3, and invasive cancers. In particular, we observed a rate of 80% in cancer samples, similar to the results reported before by Tornesello and colleagues (27). The difference between our study and the previous one is largely due to random fluctuation.

On the contrary, the finding of a higher proportion of HPV16/18 in Italian studies compared with that observed in the large IARC meta-analysis for Europe may not be random. From cancer registry data, we know that there is a time trend for higher proportion of less advanced cancers (40) but also for more screen-detected and interval cancers (41, 42).

We can envisage three possible explanations: (a) a real change has occurred due to the diffusion of cytologic screening; (b) laboratory techniques have improved specifically for the most frequent HPV types (we do not know whether bias has been introduced or corrected); and (c) laboratory contamination.

In Italy, public screening programs should actively invite all women ages 25 to 64 every 3 years for Pap testing. The Pap test, the eventual colposcopy, and any further analysis and treatment are free. Opportunistic screening is also widespread, and currently, most women do Pap test yearly (43). Screening programs are well established in central Italy, whereas in the south most Pap tests are done in opportunistic screening. The 3-year Pap test coverage is high in central Italy [∼80%, according to National Health Interview (43)] but much lower in southern regions (slightly more than 50%). In our samples, only 24% of informative cases are from southern regions, whereas 76% of them are from central regions. Furthermore, about half of the cases from the south are from Abruzzo, where Pap test coverage is ∼70%. It is possible that Pap test screening results in better prevention for non-HPV16/18 cancers due to a longer transformation time and to a lower proportion of adenocarcinoma in this group (44). Consequently, in a context of high Pap test coverage, lesions undetected by screening controls are more likely to be HPV16 or HPV18. This interpretation is consistent with the observation that the proportion of HPV16/18 in cancers is higher in high-income countries [Western Europe (45) and North America] with large diffusion of the Pap test and lower in low/middle- and low-income countries (South America, Asia, and Africa; ref. 13). This hypothesis is in contrast with the reported findings and with the interpretation given by Wheeler and collaborators (46). Another plausible explanation might ascribe this differences to changes in laboratory procedures that result in a higher sensitivity for the most common types and in particular for the two vaccine-targeted types (e.g., amplification with the E6/E7-specific primers of HPV16, HPV18, HPV31, and HPV45).

Finally, contaminations might have occurred during laboratory analysis, artificially increasing the positivity for HPV16, even if all the controls have been done. This hypothesis cannot explain the similarity of the results obtained in different laboratories (Florence and Naples).

The proportion of HPV16/18 in adenocarcinoma, in our data set, is lower than the proportion in squamous cancers even if the difference is not statistically significant. This finding is not consistent with previous results; in fact, the proportion of HPV18 in adenocarcinomas is higher than 35%, and the role of HPV16 + HPV18 is >80% in the large IARC meta-analysis (47). Conversely, we observed that HPV45 was present with greater frequency than HPV18 in adenocarcinoma, together accounting for 33% of cases. In our data set, we do not have any adenocarcinoma in situ and we cannot exclude an overstaging of our cases (i.e., some in situ lesions could have been classified as invasive). Overstaging could also explain partially why we have a lower proportion of HPV16/18 in adenocarcinoma.

The proportion of CIN3 due to HPV16/18 does not change with age. On the contrary, the proportion of invasive cancers due to HPV16/18 shows a decreasing trend with age. This trend is significant and consistent with published data (22, 48). This could be the consequence of the higher probability of progression in the HPV16 infection: Higher probability of progression in a multistep process should produce a higher proportion of fast developing cancers. If this is true, only relatively fast developing cancer would occur, in younger women, at ages closer to the start of sexual activity.

The proportion of non-HPV16/18 among women under 35 is 8% (1 of 12) in our study. If we apply this proportion to the very low incidence of cancer in this age group (10-year cumulative risk, 50/100,000), the absolute 10-year cumulative risk of cervical cancer for an Italian woman, ages 25 and vaccinated before the onset of sexual activity, will be ∼4/100,000. Similar conclusions were suggested by Wheeler and collaborators (46).

In our data set, the percentage of coinfection decreases in higher grades; this effect is independent of age. This trend is not significant if we adjust for type of specimen. In fact, there is a higher proportion of coinfections in cytologic specimens, which in our study were only used for CIN2 and CIN3. On the other hand, the trend is significant if we include data on coinfections in healthy women (27.7% in the general population, 16.8% in CIN2, 15.5% in CIN3, and 10.0% in invasive cancers; ref. 49). In the literature, there are inconsistent conclusions about this point (50, 51) even if more recent studies seem to be consistent with our data (16, 20, 24).

Three possible mechanisms may explain this finding: (a) coinfections by types other than HPV16 could be protective for progression; (b) the progression could cause a clonal expansion of the infection, reducing the probability to detect a coinfection; and (c) if the percentage of pathologic tissue in samples of invasive cancers is higher than the surrounding healthy tissue, the probability of identifying noncancerous infections in other parts of the cervix could be reduced.

If the first explanation is true, the effect of vaccination will be difficult to predict. On the other hand, it is difficult for retrospective studies like ours to predict the effect of vaccination in preventing CIN2 and CIN3 sustained by coinfection because it is impossible to know which HPV type causes the lesion.

The proportion of invasive cancer with HPV16/18 in central and southern Italy is 78%, and although this finding is consistent with previous results, it should be emphasized that it represents the highest value.

The proportion of invasive cancers due to HPV16/18 decreases with age at diagnosis. The absolute risk of an invasive cancer due to non-HPV16/18 in women under 35 is extremely low and probably does not justify any screening among younger women if they have been vaccinated before the onset of sexual activity.

F.M. Carozzi is an occasional advisor to Gen-Probe, Abbott, Sanofi, and GlaxoSmithKline.

Abruzzo: Angeloni Claudio, Lattanzi Amedeo, Maccallini Vincenzo, Caraceni Donatella, Fortunato C. Cagliari: Macis Rosalba, Pilia Massimo, Caredda Valeria. Campania: Carillo Giuseppe, Di Iasi Angela, Santarsiere Aldo, Casto Loredana, Manno Maria, Santangelo Claudia, Pini Maria Teresa, Gallicchio Giuseppina, Scherillo Isabella, Barretta Elena, De Santis Vincenzo, Ercole Filomena. Catania: Scalisi Aurora, Spampinato G., Cantarella M.A., Miano M.A. Lazio: Giorgi Rossi Paolo, Chini Francesco, Capparucci Paola, Marsili Laila Maria, Tufi Maria Concetta, Gomez Vito, Verrico Giovanna, Schiboni Maria Luisa, Pellegrini Antonella, Bove Emilia, D'Addetta Albina, Placidi Antonio. National Cancer Institute Fond Pascale: Buonaguro Franco Maria, Tornesello Maria Lina, Loquercio Giovanna, Losito Simona, Botti Gerardo, Vecchione Aldo. Tuscany: Carozzi Francesca M., Confortini Massimo, Bisanzi Simonetta, Sani Cristina, Venturini Giulia, Burroni Elena, Tinacci Galliano. Ministry of Health: Antonio Federici.

We thank all the pathology unit staff involved in the retrieving of specimens and Giovanna Patrone for the English editing.

Grant Support: This study has been financed by the Italian Ministry of Health and by the participating centers.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.