Cumulative 5-year diagnoses of CIN2, CIN3 or cervical cancer after concurrent high-risk HPV and cytology testing in a primary screening setting
Article first published online: 8 MAR 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 116, Issue 1, pages 136–143, 10 August 2005
How to Cite
Hoyer, H., Scheungraber, C., Kuehne-Heid, R., Teller, K., Greinke, C., Leistritz, S., Ludwig, B., Dürst, M. and Schneider, A. (2005), Cumulative 5-year diagnoses of CIN2, CIN3 or cervical cancer after concurrent high-risk HPV and cytology testing in a primary screening setting. Int. J. Cancer, 116: 136–143. doi: 10.1002/ijc.20955
- Issue published online: 12 MAY 2005
- Article first published online: 8 MAR 2005
- Manuscript Accepted: 21 DEC 2004
- Manuscript Received: 13 OCT 2004
- Deutsche Forschungsgemeinschaft. Grant Number: SCHN 294/6-4
- cervical cancer;
- high-risk HPV;
- predictive value
The aim of our study was to assess the cumulative 5-year diagnoses of CIN2, CIN3 or invasive cervical cancer (CIN2+) after concurrent screening by high-risk HPV test and Pap smear in a primary screening setting. Four thousand thirty-four women from Eastern Thuringia/Germany were recruited from 1996 to 1998 for baseline screening that included routine cytology, high-risk HPV testing by consensus primer PCR GP5+/6+ and routine colposcopy. After a median of 59 months 3,153 women participated in final screening using identical methods. Women with abnormal cytology or colposcopy or a positive high-risk HPV test at any time during the study period were recalled for expert colposcopy and histologic verification. CIN2+ was detected in 160 women resulting in a cumulative 5-year proportion of 4.4% (95% CI: 3.7–5.0%). Of 3,702 women who were high-risk HPV negative at baseline, 34 (1.1–95% CI: 0.7–1.4%) had either prevalent CIN2+ or developed CIN2+ within the observation period. HPV/cytology double negatives at baseline were at lowest risk for CIN2+ (1.0–95% CI: 0.7–1.4%) compared to screening positives (16.8–100% depending on combined test results). The 5-year negative predictive value in HPV−/Cyto− women was 99.0% (95% CI: 98.6–99.3%). This suggests that a prolongation of the screening intervals in this group is feasible. However, it should be noted that 1 woman developed a microinvasive carcinoma within the observation period. Moreover, 2 women with prevalent cancer were missed by both tests. The prognostic relevance of concurrent high-risk HPV/cytology screening needs to be verified further by randomized trials. © 2005 Wiley-Liss, Inc.
Changes within the current primary screening setting for prevention of cervical cancer are under active discussion. It has been shown that infections with high-risk human papillomaviruses are carcinogenic.1 However, before a HPV test can be recommended for screening the efficiency and cost-effectiveness of the new system has to be demonstrated. Whereas higher sensitivity and negative predictive value of HPV testing has been shown in several, mainly cross-sectional studies, the long-term impact on incidence and mortality has been evaluated in models only.2, 3 Thus, the current recommendations of various scientific societies and expert groups are controversial. The IARC working group on efficacy of cervical cancer screening concluded that there is sufficient evidence that HPV testing can reduce mortality from cervical cancer. However, further work has to be done to have an affordable, simple and reliable test.4 So far, no updated, comprehensive model calculations for the cost-effectiveness of HPV testing have been published for Europe since important data are still lacking. Our study funded exclusively by the government may add additional knowledge to this important issue of women's health.
We followed a cohort of women originally recruited in the scope of a HPV diagnostic study5 over a period of about 5 years to prove the hypothesis that risk of high grade lesions remains low in women testing HPV negative. Our prospective part of the study aimed at estimating the cumulative proportion of high-grade cervical intraepithelial lesions (CIN2, CIN3 or invasive cervical cancer) after a negative high-risk HPV test. In addition, depending on the results of the concurrent high-risk HPV/cytology screening at baseline, the study population was stratified into 4 groups to compare the 5-year predictive values for histologically confirmed high-grade lesions.
We previously reported the detection of cervical intraepithelial neoplasia grade II (CIN2) or III (CIN3) or invasive cervical cancer by routine cytology and high-risk HPV testing in 4,761 women within the scope of a cross-sectional diagnostic study.5 We followed the participants of our study to evaluate CIN cumulative over a period of about 5 years. Women who attended routine screening were recruited for initial study screening in 7 gynecologic practices in Eastern Thuringia/Germany between 1996 and 1998. Final screening and final diagnostic verification was done between 2001 and 2003. Written or oral informed consent was given by the participants. The primary inclusion criteria were normal cytology within 1 year prior to the study and age between 18 and 70 years. Women with prior CIN, hysterectomy, conization or current pregnancy were excluded. Additionally, for the evaluation of prognosis women without a second study follow-up examination or with undefined HPV or cytology at baseline screening were excluded as well. Both parts of our study were approved by the local ethics committee.
Screening and verification of diagnoses
All women entering the study had an initial screening in the gynecologic practices that included conventional cytology by cytobrush, detection of high-risk HPV and screening colposcopy. Cytology was read in 4 cytology laboratories on a routine basis according to the Second Munich Cytological Classification. Pap class III or worse were scored as test positives. This grading corresponds to the Bethesda classification ASC-H, LSIL, HSIL or cancer. High-risk HPV-DNA testing was performed in the research laboratory at the Department of Gynecology of the Friedrich-Schiller-University Jena. The HPV test used was based on the general primer GP5+/GP6+ mediated PCR as described previously.5, 6 Patients with positive findings in either one of these tests (Pap class III or worse and high-risk HPV positive, colposcopically suspected CIN1 or worse) were referred immediately to the colposcopy clinic at the Department of Gynecology for expert colposcopy and histologic verification. In order to identify screening failures women negative on all 3 tests had to be rescreened after 6 (± 2) months and were referred to verification in case of a positive finding. All histologic diagnoses within 12 months of entry were the subject of the first histology review in the scope of the cross-sectional study.5
Routine interim examinations.
Over the 5-year period, routine screening (annual as recommendation) was done at the discretion of the woman and the gynecologist mainly by cytology. If indicated by abnormal findings at entry women were monitored additionally by HPV testing and screening colposcopy. Any suspicious finding from these examinations had to be documented on the case report form (CRF).
Women were invited 4 months before ending the 5-year observational period by the study office to visit their gynecologist for final study screening. The same 3 screening tests as those for the baseline examination were used. Women, completely negative in all tests and without history of dysplasia during the whole study period completed study participation and went back to regular screening. In order to make predictions about the prognostic status, the data collected on the CRF was confirmed by the study gynecologists. A positive result from either routine or final screening was a given reason to recall the woman for final diagnostic verification as described below. Histological diagnoses from routine pathology during the interim period were reviewed and, if indicated, used as primary endpoint.
Final diagnostic verification.
If at least 1 positive screening test or history of histologically confirmed CIN1 was documented regardless of the date during the study (inclusive baseline screening), women were referred to the colposcopy clinic at the Department of Gynecology of the Friedrich-Schiller-University Jena for final expert colposcopy and biopsy. Colposcopic assessment followed international terminology.7
For the diagnostic “gold standard” at least 1 colposcopy-guided punch biopsy was taken at the point where the most severe lesion was suspected. In case of an invisible squamocolumnar junction or a colposcopically visible endocervical lesion, an endocervical curettage (ECC) was performed. Additionally, a Pap smear and a HPV smear were taken. All women with histologically-confirmed CIN3 were committed for conization. However, patients with histologically confirmed CIN1 or CIN2 were referred to colposcopic follow-up after 6 or 3 months, respectively.
Different measures were applied to assure the quality of the final diagnoses. Owing to the known error rate of cytology a second smear was taken at the initial and final screening. Whereas a sample (28%) was reread initially to assess quality,5 all final 2nd smears were read by the reference laboratory of the Institute for Pathology of the Friedrich-Schiller-University Jena. A woman was recalled for final verification if 1 of both smears were classified as Pap III or worse. Agreement (either classified as negative, positive or not evaluable) was 97% (kappa: 11%). All of the smears taken for HPV testing were suspended in a Tris buffer (10 mM Tris, pH 8.0, and 50 mM KCl) according to the protocol of Jacobs and colleagues.6 After performance of the test all samples were stored at −80°C. Of these, a random sample (n = 113) was sent to the laboratory for Pathology and Medical Microbiology in Eindhoven, The Netherlands, to be reevaluated for the presence of high-risk HPV DNA. The test was slightly more sensitive at our study laboratory (observed agreement 95%, kappa: 88%). Discrepant results were obtained for 6 cases. Five of these had photometric extinction values which were just above the cut-off, i.e., contained only low amounts of HPV. We cannot exclude that repeated thawing and freezing of the samples may have led to degradation of the viral DNA, thereby reducing sensitivity upon reevaluation. Routine pathology diagnosis used to guide clinical management and histologic specimens taken during the study period were reviewed by 2 expert pathologists. Blinding to the screening tests was necessary to enable estimation of test accuracy. A majority diagnosis (2 of 3 including the original diagnosis) was accepted as final. A consensus opinion was achieved by joint review for cases with initially 3 different diagnoses. Specimens from 626 women were selected for review, whereas 591 were reevaluated. The results from routine diagnoses were used for 35 (5.6%) women since the histologic specimens were not available. For most cases several histologic sections were available for review. The study progress and quality control was regularly reported to the scientific study board and to the collaborating gynecologists.
CIN2 or worse (CIN2+) lesions detected within 5 years after initial screening were defined as the primary endpoint. A priori, we estimated that the negative predictive value of the HPV test could be estimated at a confidence level of 95% with a precision of 0.2% to 0.6% for negative predictive values of 99.5% to 97.5%, respectively if 80% of the cross-sectional study population were available for final analysis. Secondarily, proportions of CIN2+ and CIN3+ were calculated for 4 risk groups defined by the combination of the HPV/cytology screening results: HPV negatives/cytology negatives (<PapIII), HPV negatives/cytology positives (=PapIII) and HPV positives/cytology negatives (<PapIII), HPV positives/cytology positives (=PapIII). We adapted the lifetable method to handle the problems of censored data and observational imbalance in a more appropriate way. Censoring also occurred at the date of hysterectomy or conization. Women who were otherwise missed at final screening were censored at date of last histology, if available, or at date of last study screening. Women who passed the study without any suspicious finding were assumed to have no disease. Follow-up was aggregated into 6 intervals: <12, 12 to <24, 24 to <36, 36 to <48, 48 to <56 and ≥56 months. To avoid biased final estimates due to diagnostic delay, all cases observed later than 56 months were assumed to be prevalent at 5 years. Predictive values and 95% confidence intervals, using standard error after Greenwood's formula, were reported for the 1- and 5-year estimates. Sensitivity analyses (SA) were done to evaluate the impact of delayed examination or incomplete information about participants who dropped out early or refused diagnostic verification. Three different approaches cover a range of risks from the lowest (probably an underestimate) to a reasonable upper limit. Supposing that no case remained undetected, predictive values were calculated as the proportion of detected cases per risk group at study entry (SA1), which would equate to the lower limit of estimates. In contrast, upper risk limits could be determined if women, who dropped out early, refused expert colposcopy, had undefined final screening or histologic results, were censored at the date of last histologic diagnosis if available or were excluded (SA2). In a third calculation, the analysis was restricted to those women who attended final screening and if required expert colposcopy within 6 years after entry (SA3).
Statistical comparisons were done by the Gehan and McNemar test. All data were managed in an ACCESS database and were analyzed with SPSS 10.1 statistical software.
The study cohort was recruited from 4,761 women participating in the cross-sectional diagnostic study described above. Some of these women were excluded for the following reasons: One gynecologic practice with 326 participants refused further participation. Additionally, 69 women could not be allocated to 1 of the 4 risk groups due to an undefined initial HPV or cytology screening test and 332 women attended the initial screening, with no follow-up. A total of 4,034 women fulfilled the inclusion criteria for the follow-up study and could be included in the cohort. They had at least 1 further study examination (screening or expert colposcopy/biopsy) during median follow-up period of 59 months.
The median age of the cohort was 35 years (range 18 to 70) at entry. Figure 1 describes the flow of the participants during follow-up. After a median of 59 months (range 38 to 85) 3153 (78%) women attended the final screening. Delayed observation after the prespecified tolerance interval of 72 months occurred in 109 participants. Women were not eligible for final screening if they had already been identified as cases (118 due to initial screening and 25 interim cases), underwent hysterectomy (82) or conization (21) or died (10). Another 625 women dropped out early since they refused further participation (66), could no longer be contacted (530) or were not eligible otherwise (29). A final histologic verification was recommended to 386 women with at least 1 positive screening test or a histologically confirmed CIN1 at baseline or anytime during follow-up. Despite repeated invitations by letter and telephone only 261 (68%) women underwent expert colposcopy/biopsy at the end. Out of the 125 women who did not have expert colposcopy/biopsy, 92 had negative results at the final screening visit. Observations of women without recommended verification were censored at 56 months. Since 338 out of 386 invited women underwent at least 1 histologic verification at any time during follow-up, we included these visits as date of last information in the sensitivity analysis.
At study entry, the prevalence of positive results for routine cytology was 1% (n = 40). High-risk HPV was detected in 332 (8.2%) women. By combining both tests, 4 risk groups for the estimation of prognosis were defined: 3,685 (91.3%) HPV negatives/cytology negatives (median age 35, range 18 to 70), 17 (0.4%) HPV negatives/cytology positives (median age 38, range 22 to 52), 309 (7.7%) HPV positive/cytology negative (median age 30, range 17 to 70) and 23 (0.6%) HPV positive/cytology positive (median age 35, range 25 to 57).
In the cohort eligible for the final screening, the prevalence of positive cytology was 0.2% (n = 6), and 131 (4.3%) women were found to be infected with high-risk HPV. In Table I results of the final screening by risk groups are shown. If we compared prevalence profiles in the subgroup available for both screening rounds, similar distributions with slightly lower HPV prevalence at the end could be observed (p = 0.3). After about 5 years, 29 women (0.9%) remained HPV positive, 117 (3.7%) became HPV negative and 102 women (3.2%) had newly detected infections.
|Screening||At study entry||Total|
|HPV−/Cyto−||2,875 (91.7%)||6 (0.2%)||117 (3.7%)||0||2,998 (95.7%)|
|HPV−/Cyto+||5 (0.2%)||0||0||0||5 (0.2%)|
|HPV+/Cyto−||99 (3.2%)||2 (0.1%)||28 (0.9%)||1 (0.0%)||130 (4.1%)|
|HPV+/Cyto+||1 (0.0%)||0||0||0||1 (0.0%)|
|Subtotal1||2,980 (95.1%)||8 (0.3%)||145 (4.6%)||1 (0.0%)||3,134 (100.0%)|
|HPV n.e. /Cyto−||11||0||1||0||12|
|Total||3,685 (91.3%)||17 (0.4%)||309 (7.7%)||23 (0.6%)||4,034 (100.0%)|
The histologically confirmed diagnosis of CIN2 or CIN3 or invasive cervical cancer was defined as the primary endpoint of our prognostic study. As CIN1 lesions have a potential for progression, they were reported as well. Overall, 77 women with CIN1, 50 with CIN2 and 101 with CIN3 were detected. Of these 68, 45 and 95 CIN diagnoses were reviewed, respectively, by an expert panel. The majority of CIN1 (60%) and CIN2 (62%) was detected in women younger than 30 years compared to CIN3 or invasive cancer, which was identified more frequently in women 30 years and older (Table II). Invasive cancer was confirmed by reviewers in 9 women, and 3 of them belong to the lowest risk group, i.e., initially negative in both tests. Two of these 3 women were diagnosed with prevalent cancer in the first year of follow-up, both at the age of 34 years. In 1 of these patients, squamous cell carcinoma FIGO stage Ib1 was diagnosed after a colposcopic finding of CIN3. This case was missed by HPV-DNA detection due to a partial deletion of the L1 gene, which had occurred during integration of the HPV68 genome into the host DNA. In the other patient, squamous cell carcinoma FIGO stage IIb positive for HPV 16 was diagnosed after the control screening (HPV positive and Pap IIK) 5 months after baseline. The third patient with invasive cancer was high-risk HPV negative at baseline but was positive for HPV 33 after 63 months when the final screening sample was taken. Cytology was normal for both smears. This patient underwent biopsy followed by conization 14 months later. A squamous cell carcinoma stage Ia2 was diagnosed at the age of 63.
|< 30 years||46 (60%)||31 (62%)||37 (37%)||1 (11%)||115 (48%)||69 (43%)||38 (34%)|
|≥ 30 years||31 (40%)||19 (38%)||64 (63%)||8 (89%)||122 (51%)||91 (57%)||72 (66%)|
The distribution of CIN diagnoses over time is shown in Table III for the risk groups defined by screening results at study entry. Detection within the first and last interval was strongly related to diagnostic follow-up of the 2 study screenings. Detection in between was based on findings from routine screening or surveillance of women with positive screening findings or CIN1 diagnoses at the beginning. From the total of 160 cases, 118 were prevalent during the cross-sectional study. Nineteen cases were probably incident during follow-up since they had a negative consensus diagnosis within the first interval (14 without dysplasia and 5 with CIN1).
|Screening at study entry||Histology||Time interval of first diagnosis (months)||Total|
|Population at risk1||3,685||3,240||3,093||3,057||2,989||2,904||3,685|
|Population at risk1||17||13||10||10||9||8||17|
|Population at risk1||309||190||172||156||149||148||309|
|Population at risk1||23||2||1||1||1||1||23|
|Population at risk1||4,034||3,445||3,276||3,224||3,182||3,146||4,034|
Table III further displays the number of women at risk and censored observations due to reasons described above. About half of censoring occurred during the first interval if women who were initially negative by screening dropped out after having a control exam (details in reference 5). The consequences for estimation of predictive values are discussed below as part of the sensitivity analysis.
Based on data from Table III proportions of cumulative CIN2 or worse (CIN2+) and CIN3 or worse (CIN3+) are reported in Table IV. Overall, the proportion of CIN2+ after about 5 years was 4.4%. In the 4 risk groups, proportions were significantly different (p < 0.001) and increased from 1% in HPV−/Cyto−, 16.8% in HPV−/Cyto+, 37.7% in HPV+/Cyto− up to 100% in HPV+/Cyto+ groups. Cases related to HPV infection at first screening were mostly prevalent. After merging of risk groups according to HPV or cytology, the cumulative proportion of CIN2+ was 1.1% in HPV negatives compared to 42% in HPV positives, and 3.8% in cytological negatives compared to 66.7% in cytological positives. Comparisons between the 2 screening tests should be done with caution since the groups are not independent. Similar figures could be obtained for the more severe diagnosis of CIN3+, which was detected in 110 women (3%). Lower risks were estimated for the double negative group (0.7%) and the HPV+/Cyto− group (26.6%). The 2 cases in the HPV−/Cyto+ group had CIN3 diagnoses, therefore risk remained unchanged.
|Screening at study entry1||Populationat risk atstudy entry||Numberof cases||Cumulativeproportionafter 1 year (%)||95% CI (%)||Cumulativeproportionafter 5 years(%)||95% CI (%)|
The same analyses were performed for a restricted population of 2,814 women aged 30 years or older at baseline. Cumulative 5-year proportions and 95% confidence intervals of CIN2+ were 0.6% (0.3%, 0.9%) in 2,623 HPV−/Cyto−, 17.5% (0%, 40%) in 15 HPV−/Cyto+, 38.7% (30.6%, 46.7%) in 156 HPV+/Cyto−, 100% (84.7%, 100%) in 20 HPV+/Cyto+ and 3.5% (2.8%, 4.2%) in the total subgroup. Detailed figures are available from the authors on request.
The 5-year negative predictive value for CIN2+ in HPV−/Cyto− women was 99.0% (98.6%, 99.3%) in the total study population and 97.9% (97.0%, 98.9%) in women screened below age of 30 compared to 99.4% (99.0%, 99.7%) in those screened at age 30 or older.
Three different assumptions were subjected to sensitivity analyses (SA). In SA1 lower limit estimates were calculated supposing that no case remained undetected (Fig. 2). The same population was covered as in the main analysis. In total, 4% cumulative CIN2+ and 2.7% cumulative CIN3+ were estimated by SA1. In contrast, upper risk limits (SA2) could be determined if observations were excluded or were censored more restrictively in women with protocol deviations. For this analysis, 3,420 women (85% of the total population) were included. The age distribution was comparable to that of the total population. The cumulative 5-year proportion of CIN2+ and CIN3+ were 4.9% and 3.4%, respectively. In a third calculation (SA3), the population analysed was restricted to 3913 women (97% of the total population) who attended final screening and if indicated had expert colposcopy within 6 years after entry. Compared to the main analysis, SA3 yielded nearly the same results (cumulative 4.3% CIN2+ and 3.0% CIN3+).
The results of our study show a low detection rate of 0.3% for moderate or severe dysplasia or presymptomatic cancer after a negative HPV test if controlled by repeated screening and if indicated by expert colposcopy and biopsy within the first 12 months. The proportion remains low for the total follow-up period and was cumulatively 1.1% over 5 years. In cytologically negative women 2.6% were already detected with CIN2+ during the first year, most of them being infected by high-risk HPV. By the end of our study, this percentage increased to 3.8%. Statistical comparisons between single tests would require independent groups, which were not implemented in our study. Valid comparisons can be made between the 4 risk groups established by combination of HPV and cytology testing. The cumulative percentage of CIN2+ in double negatives after 5 years was much lower (1%) than that observed among women with a positive screening test after 1 (6.4–86.4%) and 5 years (16.8–100%). Out of 32 women with CIN2+ belonging to the initially double-negative group, 24 underwent verification due to a HPV positive finding during follow-up, 5 due to Pap III or worse during follow-up and 3 women had suspicious findings at routine colposcopy.
Questions may arise about the validity of our data. More women with CIN2+ would probably be identified by adding an HPV test in cervical cancer screening, due to a longer detection interval at the molecular level than at the cytologic level and by the higher sensitivity compared to cytology screening.8 But the detection of regressive lesions, which are not distinguishable from progressive ones, would almost certainly increase as well. The 5-year proportion of CIN2+ was 4.4% in the total study population and 6.7% in women below age 30 compared to 3.5% in the older subgroup. A meta-analysis, which included HPV screening studies with a sufficient number of CIN2+ cases, covered a range in detection rate from 0.64 to 4.3 for cross-sectional designs.9 The highest referred to an unscreened Chinese population where all participants underwent multiple biopsies to circumvent verification bias.10 Kulasingam et al.11 estimated a 3.2% prevalence of CIN3+ in a younger population (mean age 25) where 25% underwent a biopsy, and Wright et al.12 reported a prevalence of 3.4% ín previously unscreened black South African women. The lowest detection rate was reported by Petry et al.13 from a Western German population aged 30 and older, where 5.4% of the women underwent biopsy. In our study, where half of participants were aged 35 or younger, histologic specimens were taken from 9.3% during the diagnostic part and from 15.5% altogether. For 1989, which reflects the last date of reliable data for an Eastern German population, cervical cancer incidence was reported to be about twice that of Western Germany. This may still hold true for the present and could explain in part the variation within Germany. In contrast to the diagnostic studies, a large longitudinal study14 of HPV natural history reported a crude proportion of 0.8% CIN3 over 122 months. HPV was assessed retrospectively and detection was based on the Pap test only.
Thus, the overall detection rate is a complex figure. It reflects the underlying cervical cancer risk of the population examined, the efficacy of current Pap smear screening, the age distribution of participants as an indicator for numerous prevalent CIN lesions that will probably never progress and the extent and quality of histologic verification.
Some results have been published from the follow-up of HPV/cytology negative cohorts. Bory et al.15 recovered 2 CIN2+ cases out of 2,432 women (0.08%) after 27 months. Five cases (4 HSIL and 1 micro-invasive carcinoma) were found in 4,401 women (0.1%) after 34 months by Clavel et al.16 Sherman et al.14 reported that during the first 45 months the cumulative incidence of CIN3 was 0.16% in HPV and cytology double negatives. It is not surprising that no CIN2+ was detected by colposcopy/biopsy in random samples of about 200 to 250 women.11, 13, 15 The chance of detecting at least 1 case is low taking into account the baseline risk of CIN2+ in that group. In our study, the women selected for histologic verification were not comparable to the randomly selected subgroups mentioned above. The 32 women with CIN2+ were selected for colposcopy/biopsy due to suspicious screening findings at a control examination after initial screening (n=22), during routine screening (n=5) or at the final examination (n=5).
In order to identify women with CIN2+ we followed our study cohort more actively. Therefore, our CIN2+ figures might be higher compared to those from other studies. Additionally, the life-time risk of cervical cancer in Eastern Germany before the introduction of Pap smear screening was one of the highest in Europe. It was calculated from data of the Eastern German cancer registry to be 3.8%.17 After 30 years of Pap smear screening, Germany has still a higher incidence rate compared to other European countries.18 The percentage of CIN2+ detected by concurrent high-risk HPV/cytology screening in our study within 5 years is higher than the underlying lifetime cervical cancer risk. Moreover, our figures support the assumption that HPV related CIN even at younger age might be of minor predictive value for the development of cervical cancer.3 In particular, in German-like settings with minimum age of 20 and annual screening, substantial modifications of the program will be necessary if HPV test is considered for primary screening. Preliminary cost-effectiveness models have shown that screening intervals for HPV and cytology negatives not only could but should be extended to avoid increasing diagnostic workup, surveillance and probably over-treatment of regressive lesions. Mandelblatt et al.2 estimated the greatest reduction in incidence and mortality by a biannual HPV and Pap test at the expense of the negative side effects mentioned above. Physicians may argue that longer intervals will compromise the overall participation and therefore the success of the program. It can be suspected from a comparison of European screening programs that poor screening rates will be more likely to occur in settings without an invitational system.19 Authors of a current review20 about new directions in cervical cancer control recommended a minimum screening age of 25 and extended intervals of 8 to 10 years but conceded the case for 3 to 5 years in settings that currently support annual or biannual screening.
The results of our study confirm the prior hypothesis from several HPV studies that HPV and cytology double negative women constitute a large group with low risk for prevalent or incident dysplasia over a longer period.21 To determine the length of the protective interval an independent cohort with only cytological screening would be required for comparison, which goes beyond the scope of our study. One may be concerned about the 2 prevalent cases of invasive cancer that were missed even by a combination of cytology and high-risk HPV testing. Whether HPV testing could have detected these cases at a precancerous stage, which is the main target of cervical cancer screening, remains speculative.
Our results may be biased by some sources. One of them is the percentage of censored observations (22%) especially due to early dropout and withdrawal from verification. Women who dropped out early had a slightly higher prevalence of positive screening findings at entry. Thus, the total number of CIN2+ cases might be underestimated on this basis. Withdrawal from diagnostic verification was observed in 32% of 386 women invited after final screening according to comprehensive recall criteria. From those with positive screening findings at the end 78% had final expert colposcopy/biopsy. Owing to the complexity of reasons and different time points of censoring, sensitivity analyses were used to study the outcomes based on varying assumptions. In all cases, the major messages remained valid. Discontinuation of the diagnostic follow-up after screening is a known phenomenon. Reduced compliance was also reported in the HART study (72%),22 in a German study (75%)13 and in a Canadian study (80%).23 Participation, coverage of the target population and the compliance with early diagnostic and treatment might be of higher impact on the efficiency of a screening program than more sensitive screening tests.
All histologic specimens taken within the first year were reviewed in the scope of our diagnostic study. Thirty-five (5.6%) of the histologic specimens that were collected subsequently and were possibly related to CIN or cancer were not reviewed. According to our experience with expert review, slightly more down-grading than up-grading would be expected in the nonreviewed cases.
Despite limitations, our study yields interesting data about concurrent HPV/cytology screening in a population that is likely to participate in screening. Attention should be paid to over-diagnosis and reduced compliance with diagnostic follow-up and treatment. Both should be considered in further modeling of cost-effectiveness. Since HPV infection is very common evaluation of potential benefit and harm will need to be included in further studies. As in other European countries, implementation of HPV testing into cervical cancer screening is a matter of current discussions in Germany. A rigorous evaluation of long-term effectiveness is recommended by the IARC if current Pap screening is to be modified.4 As long as the proof of efficiency and cost-effectiveness, either by modeling or more credible by randomized controlled trials, is still lacking, several other measures could probably improve effectiveness of current cytology screening. One such measure would be to improve participation of women overall and in particular of lower social/economic status in current screening programs.24, 25, 26
The excellent support by the pathologists C. Bergeron, C. Crum and V. Schneider is gratefully acknowledged. A. van den Brule served as reference virologist and kindly reevaluated a random sample of smears for HPV-DNA. Our study would not have been possible without the continuos support of U. Baumbach, G. Götz, D. Menzel, R. Möller, A. Mönch-Hering, K. Pfützenreuter, M. Riefenstahl and G. Wetzel, the gynecologists who performed the screening of all women though not reimbursed. We thank M. Schiffman for critical reading of the manuscript.
- 4International Agency for Research On Cancer a Part of the WHO. IARC confirms efficacy of cervix cancer screening for women 25–65 in reducing mortality. Press release No 151. 3-5-2004. Available from URL: http://www.iarc.fr/pageroot/PRELEASES/pr151a.html
- 15Recurrent human papillomavirus infection detected with the hybrid capture II assay selects women with normal cervical smears at risk for developing high grade cervical lesions: a longitudinal study of 3,091 women. Int J Cancer 2002; 102: 519–25., , , , , , .
- 17Cancer incidence in five continents. 1970. Springer-Verlag Berlin, Heidelberg, New York., , .
- 18GLOBOCAN 2000: Cancer incidence, mortality and prevalence worldwide. 2001;Version 1.0. Lyon, IARCPress., , , .
- 26Utilization of health screening studies and measures for health promotion]. Gesundheitswesen 1999; 61 Spec No: S163–S168., , . [