A report on the current status of European research on the use of human papillomavirus testing for primary cervical cancer screening

Authors


Abstract

Cervical cancer remains a significant public health concern, both at a global and a European level. A number of new technologies such as diagnostic tests for human papillomavirus (HPV) have a potential to assist with the reduction of this disease. However, both the efficacy and the cost-effectiveness of these new technologies must be established in properly designed trials before they can be implemented within national public health programs. Our study reviews the randomized controlled trials that are currently being conducted in Europe to establish the performance of HPV testing as a primary cervical cancer screening test. © 2005 Wiley-Liss, Inc.

Background

Cervical cancer is the second most common cancer in women worldwide and is the most common female cancer in the Caribbean, East-, Central and Southern Africa and in South–Central Asia. Globally, an estimated 471,000 women develop cervical cancer and 233,000 die from it every year (estimate for 2000).1 The disease primarily affects younger women, with the majority of cases appearing between the ages of 30 and 50,2 an age when many women are actively involved in their careers, caring for their families or both. Consequently, the total years-of-life lost is proportionately higher than that for most other cancers with a later onset, and the impact on society as a whole is greatly increased.

Cervical cancer still remains an important public health issue in Europe, where it is the 10th most common cause of cancer deaths in women1 and while it is currently not among the most frequent cancers, a very important distinction with cervical cancer is that it is the only cancer that is almost completely preventable through regular screening. Each year in Western Europe, 13,000 women develop cervical cancer and 6,000 women die from this disease, while the situation in Eastern Europe is worse with ∼31,000 women developing cervical cancer and about 17,000 dying every year.3 This difference is largely due to the absence of effective cervical screening in Eastern Europe and the implementation of properly organized screening programs would inevitably decrease the burden of this disease in these countries.

Cervical cancer screening

Squamous cervical cancer is particularly amenable to screening as it has a long preclinical phase and identifiable precursor lesions that can be treated with high efficacy using simple outpatient procedures if they are detected early.

Screening based on the Pap smear has substantially reduced the incidence of cervical cancer in those countries where it is practiced effectively.3, 4 However, recent studies indicate that it has limitations. Data available from organized screening programs show that the initial declines in disease incidence seen following the establishment of screening have now levelled-off, indicating that the maximum effect of Pap smear-based screening may have been reached in these countries.5, 6 Further, a meta-analysis of studies unaffected by verification bias has shown that the pooled sensitivity of the Pap smear using low-grade squamous intra-epithelial lesions (LSIL) as the threshold to detect histologically confirmed cervical intraepithelial neoplasia of grade 2 or worse (≥CIN2) was 77% (95% CI: 58% to 97%).7

Given these data, an important question for the public health community is whether the introduction of new technologies can yield further progress in the battle against cervical cancer. It is therefore important to ensure that all available screening tools are properly evaluated, both in terms of their performance as well as their cost-effectiveness within the context of a screening program, and that they are fully implemented when found to be appropriate.

The human papillomavirus and cervical cancer

There is now an overwhelming body of evidence demonstrating that infection with certain types of the human papillomavirus (HPV) is the primary risk factor for the development of cervical cancer and its precursor lesions.8, 9 More than 130 different HPV types have been identified with ∼40 of these infecting the anogenital epithelium. These have been classified as either low-risk (LR) or high-risk (HR) for the development of cervical cancer based upon their identification in cervical tumor samples.10 A recent analysis of 11 studies has designated 15 anogenital HPV types as HR (16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 73 and 82), with a further 3 types designated as probably HR (26, 53 and 66)11 although some of these designations have been disputed by others.12

This body of evidence is strong, consistent across different populations and demonstrates that HPV is a necessary (although not a sufficient) cause of cervical cancer.9, 13 On the basis of these data, it is logical to conclude that HPV testing could be a useful cervical cancer screening tool and its use has been proposed for primary screening, for the triage of equivocal Pap smears14, 15, 16 and for the follow-up of patients after treatment for CIN.17, 18, 19

HPV testing for cervical cancer screening

With regard to its role as a primary screening test, a number of research studies have demonstrated that HPV testing has a higher sensitivity and negative predictive value (NPV) for the detection of prevalent cervical cancer precursors than either conventional or liquid-based cervical cytology methods, albeit with a lower specificity and positive predictive value (PPV).20 Some of these studies are presented in Table I.

Table I. Sensitivity and Specificity of HPV Testing Compared to Cervical Cytology (Using ≥ASC-US as the Referral Threshold) for the Detection of ≥CIN2
StudyDescriptionHPVPap (≥ASC-US)HPV/Pap (≥ASC-US)
SensSpecSensSpecSensSpec
  • 1

    Hybrid Capture 2 HPV test, positive threshold ≥ 1.0 pg unless otherwise indicated.

  • 2

    Hybrid Capture 1 test (Digene Corp. Gaithersburg, MD) (HC2 positive threshold ≥ 2.0 pg).

  • 3

    Sensitivity of HC1 was suboptimal and would have contributed to the difference seen between HPV testing alone and the combination of HPV testing with Pap.

Cuzick et al. (1999)21United Kingdom: n = 1,703 conventional Pap, HC21, ♀ ≥ 35 yrs95957999NANA
Schiffman et al. (2000)22Costa Rica: n = 1,119 conventional Pap, HC2, ♀ ≥ 18 yrs88897894NANA
Ratnam et al. (2000)23Canada: n = 2,098, 69% HC12/31% HC2, ♀ 18–69 yrs, (adjusted for verification bias)853 (68)58 (91)56 (40)62 (92)973 (76)39 (86)
Clavel et al. (2001)24France: n = 5,651 liquid cytology, HC2, ♀ ≥ 15 yrs100868893NANA
Petry et al. (2003)25Germany: n = 8,468 conventional Pap, HC2, ♀ ≥ 30 yrs9896449810094
Cuzick et al. (2003)26United Kingdom: n = 10,358 conventional Pap, HC2, ♀ ≥ 30 yrs97 (96)93 (94)7796(100)(94)

These studies21, 22, 23, 24, 25, 26 and others show that the sensitivity of a single HPV test for the detection of prevalent CIN2+ is 10–54% higher than that of a single Pap smear, although the specificity is 1–7% lower. They also demonstrate that the NPV of HPV testing approaches 100% in all studies using well validated techniques for the detection of HPV, and more recent data show that this unprecedented NPV remains high prospectively with 4-year27 and 5-year28 disease rates in HPV negative women being equivalent to those of cytology-negative women at 2 years. This observation indicates that HPV testing could be cost-effective in many European countries by allowing the screening interval to be safely lengthened.

When interpreting these data, it is important to note that the studies have methodological limitations that restrict their use for public health policy formulation.29, 30 These studies essentially used a cross-sectional design with double testing of all women and short-term follow-up by colposcopy with biopsy for those having one or more positive screening tests. Such a design is appropriate to assess the relative sensitivity of each test to detect prevalent high-grade cervical intra-epithelial (CIN2+), but cannot be used to study the impact of different screening strategies on the incidence of invasive cancer. Indeed, a considerable number of the CIN2+ lesions detected would have regressed or be misclassified lower-grade disease and therefore would not have contributed to the cancer burden. In addition, this design does not account for the long duration of precancerous stages and the consequent increase in sensitivity that accompanies repeated Pap smear testing (i.e. program sensitivity).

These limitations can be addressed to some extent by comparing groups of women who are screened and followed-up with different strategies for at least 1 screening round and randomized controlled trials (RCTs) are the obvious design for such comparisons. Alterations to large-scale public health programs require substantial investments of public funds and can have enormous repercussions (either positive or negative) for the populations being served. Therefore, the burden of proof required to make changes to existing, successful programs is justifiably high and by today's standards for evidence-based medicine, RCTs are required as they are the best way to obtain unbiased assessments of the technologies being investigated.

Notwithstanding the above concerns, the studies that have been published show that HPV testing has the potential to offer significant benefits to cervical cancer screening programs and they do provide valuable data required for the design of more comprehensive trials that meet the requirements needed for public health policy formulation.

Ongoing HPV screening trials

At present, there are 5 population-based RCTs that are underway in Europe to evaluate the use of HPV testing as a primary screening tool. A summary of these studies is presented in Table II.

Table II. Key Parameters of the European RCTs
StudyCountryTotal recruitmentAge range (yrs)HPV testCytologyDesignComparisonMain study outcomesInitiatedAnticipated completion
Finnish Randomised Public Health TrialFinland200,00025–65HC2Conventional Pap smearRandomised intervention evaluationHPV alone vs. Pap aloneCumulative incidence of CIN2,CIN3 and cancer after initial screening20032009
SwedescreenSweden12,52732–38PCR (GP5+/6+)Conventional Pap smearRandomised controlled trialHPV + Pap vs. Pap aloneComparative prevalence of histologically confirmed ≥CIN2 at the exit screen20012006
POBASCAMThe Netherlands44,10230–60PCR (GP5+/6+)Conventional Pap smearRandomised controlled trialHPV + Pap vs. Pap aloneProportion of histologically confirmed ≥CIN3 found at any time during the trial from recruitment to exit screen19992007
ARTISTICUnited Kingdom25,00020–64HC2ThinPrep Pap Test (Cytyc)Randomised controlled trialHPV + cytology vs. cytology aloneComparative prevalence of histologically confirmed ≥CIN3 at the exit screen20012007
NTCCItaly95,00025–60HC2ThinPrep Pap Test (Cytyc) or conventional Pap smearRandomised controlled trialPhase 1: HPV + LBC vs. Pap Phase 2: HPV alone vs. Pap aloneComparative detection of histologically confirmed ≥CIN2 from the recruitment screen up to and including the exit screen20022007

The Finnish randomized public health trial

In Finland, a randomized public health trial to evaluate HPV testing as a stand-alone primary screening test was initiated in 2003, by the Finnish Cancer Registry and the Cancer Society of Finland. The ultimate objective of this trial is to compare the effectiveness and efficiency of screening with HPV to screening using either computer assisted cytology review (Papnet®) or the traditional Pap smear.

The Finnish cervical cancer screening program is organized and nationwide. Women are invited every 5 years, starting from age 25–30 and continuing to age 60–65. The annual number of personal invitations issued is in the region of 250,000 and 72% of these women attend. Since 1999, about two-thirds of the target population (as specified at the municipal level) have been individually randomized 1:2 to computer assisted screening using the Papnet® system or screening by conventional Pap smear. Since 2003, 20,000 women per year from the conventional Pap smear arm have been randomly allocated to either HPV testing or the conventional Pap smear. Within 5 years (i.e. 1 screening interval), this will include 100,000 women in each arm, or 200,000 in total. The participation rate in this trial has been in the region of 70% and has been identical between the arms. Interim data show that the number of screen positive women is substantially higher in the HPV arm, as is the detection rate for cytological abnormalities. However, the difference in the detection of high-grade lesions between the arms is smaller. Further details on the implementation of the new screening technologies within the organized screening program can be found in Nieminen et al. 200331 and Kotaniemi-Talonen et al.32

Swedescreen

The Swedescreen trial is being conducted by Prof. Joakim Dillner and colleagues in 5 counties in Sweden. In this study, 12,527 women aged 32–38 who attended the organized population-based cervical screening program have been randomized 1:1 to HPV testing for high-risk types undertaken with the GP5+/6+ PCR-enzyme immunoassay, or the control arm (HPV sample taken, but frozen). Enrolment started in May 1997 and was completed in 2001. Women with an abnormal Pap smear at baseline were referred to colposcopy as part of the screening program. HPV-positive women with normal smears together with an equal number of women randomly selected from the control arm were called back for a second HPV test ≥12 months later. Women who had persistent HPV infection with the same type were referred to colposcopy together with an equal number of women randomly selected from the control arm. The primary outcome of the study is the incidence of histopathologically confirmed ≥CIN2 at the next screening round, 3 years later. Baseline data from the trial have been reported,33 together with determinants of HPV persistence.34 Among women who were cytologically negative at baseline but had persistent high-risk HPV infection, 28% were found to have histologically confirmed ≥CIN2 at follow-up.35

Population based screening study Amsterdam

The population based screening study Amsterdam (POBASCAM) trial is being conducted within the national screening program in the Netherlands by Prof Chris Meijer and colleagues. This study evaluates 2 questions: (i) whether the use of HPV testing in conjunction with conventional cervical cytology will allow the screening interval to be lengthened from the current standard of 5 years, and (ii) whether HPV testing can be used for the triage of women with borderline or mild dyskaryosis (BMD – equivalent to ASC-US and LSIL) to provide a reduction in the number of women referred to colposcopy.

In this trial, 44,102 women attending for routine cervical cancer screening, aged 30–60 years, were recruited from January 1999 to September 2002, and were randomized 1:1 to having their HPV test revealed (intervention arm) or concealed (control arm). Cervical cytology was performed with the conventional Pap smear and HPV testing for high-risk types was undertaken with the GP5+/6+ PCR-enzyme immunoassay. The participants will now be followed for a total of 5 years, at which time both arms will be rescreened by cytology and HPV testing with all histologically confirmed lesions ≥CIN 3 up to and including those detected at the exit screening round being included in the final analyses.

In the control arm, women with normal cytology will be recalled at the next screening round in 5 years. Women with BMD will have repeat cytology at 6 and 18 months. Those still having BMD after either 6 or 18 months will be referred to colposcopy while those regressing to normal cytology after 18 months will not be recalled until the next screening round. All women with >BMD at any point in the trial will be referred to colposcopy.

In the intervention arm, women who are cytology and HPV negative at baseline will be recalled at the next screening round in 5 years, while those who are cytology negative but with a positive HPV test were advised to repeat both tests at 6 and 18 months. Participants with persistent HPV infection will be referred to colposcopy while those who have cleared their HPV infection will not be recalled until the next screening round. Women with BMD will have both tests repeated at 6 and 18 months. Those having persistent HPV infection at 6 or 18 months will be referred to colposcopy while those who have cleared their HPV infection at 18 months will not be recalled until the next screening round. All women with >BMD at any point in the trial will be referred to colposcopy independent of their HPV test.

The primary outcome measure is the proportion of histologically confirmed ≥CIN 3 lesions found at any time during the trial from recruitment up to and including the exit screen. Since women with normal cytology at the exit screening round will not be referred for colposcopy, and therefore, will not have a histological endpoint, it will be assumed that no CIN lesions are present. The potential for lengthening the screening interval will be evaluated by assessing whether the proportion of histologically confirmed ≥CIN 3 lesions among women with normal cytology and a negative HPV test at baseline is less than that among women with normal cytology in the control arm. The use of HPV testing for the triage of women with BMD will be evaluated by assessing whether the cumulative proportion of histologically confirmed ≥CIN3 in women with BMD/HPV− at baseline or BMD/HPV+ at baseline and HPV clearance at 6 months is less than that found in women with BMD at baseline and normal cytology at 6 months, as screened by conventional cytology.

Recruitment for POBASCAM has been completed; the baseline data published36 and the final results from the 5-year follow-up should be published at the beginning of 2007.

A randomized trial of HPV testing in primary cervical screening

The randomized trial of HPV testing in primary cervical screening (ARTISTIC) trial is funded by the UK Health Technology Assessment Program and is being conducted in Manchester by Prof Henry Kitchener and colleagues. The study is designed to evaluate the costs, medical effects and psychosocial impact of using HPV testing in conjunction with routine cervical cytology within an organized cervical screening program. The study will also provide an estimate of the efficacy and costs of using HPV testing as a stand-alone screening test and investigate the methodological issues of using HPV testing within a population-based screening program.

In this trial, 25,000 women aged 20–64 who are attending their general practitioners for routine cervical screening are being randomized 3:1 to having their HPV result revealed (intervention arm) or concealed (control arm). Cervical cytology is performed using the ThinPrep Pap test (Cytyc Corp.) and all cytologies in both arms of the trial are accompanied by HPV testing for high-risk types performed using Hybrid Capture 2 (HC2) (Digene Corp.). All women in the trial will have exit cytology with a HPV test at 36 months after entry, and all will be followed cytologically and histologically by record linkage for at least 6 years. The primary outcome measure is the prevalence of ≥CIN3 at 36 months.

In the control arm, women with normal cytology will be recalled for screening by cytology and HPV testing at 36 months, while women with moderate or severe dyskaryosis will be referred to colposcopy and managed according to standard practice. Women with mild dyskaryosis or borderline cytology will have repeat cytology at 6 months or will be referred to colposcopy if dyskaryosis persists. If this cytology is normal or borderline, a third smear will be taken at 18 months if normal, or at 12 months if borderline, and the women referred to colposcopy if any abnormality is detected. If it is normal, no further action will be taken until 36 months.

In the intervention arm, management also depends on the HPV result. Women with normal cytology who are HPV negative will be recalled for screening by cytology and HPV testing at 36 months, while those who are HPV+ will be retested at 12 months. Those who are still HPV+ at 12 months may choose between colposcopy and a repeat HPV test at 24 months. If this test is still positive, the women will be referred for colposcopy; if not, the women will go on to the exit cytology and HPV test at 36 months. Women with mild dyskaryosis or borderline cytology who are HPV− will be managed as the control arm. Those who are HPV+ will be managed in a similar fashion except for women whose third smear is normal but HPV+ who will be referred to colposcopy. Women with moderate or severe dyskaryosis will be referred for colposcopy and managed by standard guidelines.

The ARTISTIC trial commenced recruitment in July 2001 and by September 2003 had reached its recruitment target. Baseline data are expected to be published in early 2006, with the full trial results to be published late 2007.

New technologies for cervical cancer screening

The new technologies for cervical cancer screening (NTCC) trial is being conducted in Italy by Dr. Guglielmo Ronco and colleagues to evaluate: (i) the performance of liquid-based cytology and HPV testing as stand-alone tests for cervical cancer screening compared to that of the established standard of the conventional Pap smear, (ii) the potential cost-effectiveness ratio of each new technology, (iii) the appropriate age to start and to stop screening.

In this trial, a total of 95,000 women aged 25–60 years were recruited from the routinely organized cervical cancer screening programs, in 6 regions of Italy. Recruitment was conducted in 2 phases of ∼50,000 women each; the first comparing liquid cytology combined with HPV testing using Hybrid Capture 2 [Digene Inc.] to the conventional Pap smear, and the second comparing HPV testing alone to the conventional Pap smear.

In phase 1 of the trial, women were randomized 1:1 to receive either conventional cervical cytology (the control arm) or liquid cytology with HPV testing (the intervention arm). In the control arm, women with normal cytology were scheduled for recall at the next screening round in 3 years. Women with ≥ASCUS were referred to colposcopy. In the intervention arm, women with normal cytology and a negative HPV test were scheduled for recall at the next screening round in 3 years. Women with normal cytology who were HPV+ were either recalled in 1 year to repeat both tests if <35 years of age, or they were referred to colposcopy if ≥35 years of age. In phase 2 of the trial, the women recruited were randomized 1:1 to receive either conventional cytology (the control arm) or HPV testing as a stand-alone screening test (intervention arm). In the control arm, women were managed as in phase 1. In the intervention arm, women who were HPV- were scheduled for recall at the next screening round in 3 years. All women who were HPV+ were referred to colposcopy.

In both phases, HPV positive women who had no lesion detected at colposcopy were scheduled to repeat both tests at yearly intervals as long as they remain HPV positive, and were referred to colposcopy if cytology became ≥ASCUS.

The main outcome measures for the trial are the comparative detection rates for histologically confirmed ≥CIN2 found at recruitment, during the follow-up or at the exit screening round at 3 years. A reduction in the detection of lesions after recruitment in the intervention arm compared to that of the control arm will indicate that the higher sensitivity of the new technology should allow the screening interval to be extended. Other outcome measures include: colposcopy referral rate, duration of follow-up and number of repeat tests conducted.

Recruitment started in early 2002 and was completed in summer 2003 for phase 1, and at the end of 2004 for phase 2. Publication of data on test performance at recruitment is expected at the end of 2005 for phase 1 and in 2006 for phase 2. The rescreening phase will continue until the end of 2007.

Summary

A large number of studies examining the performance of HPV testing as a primary screening test have been published in the scientific and medical literature. However, most of these have either had methodological shortcomings or did not completely address the issues required to make changes to public health interventions such as the cervical cancer screening programs in Europe. This issue was recognized by members of the European research community who moved to implement several large-scale randomized controlled trials that will provide a higher standard of evidence upon which to decide whether HPV testing should be used within cervical cancer screening programs and whether it should complement or replace cervical cytology. The final results of these studies should be published in 2006 and 2007, and will provide a much stronger foundation upon which to make decisions about the ongoing development of cervical cancer screening programs in Europe.

On this basis, it is now important to decide whether the studies that are being conducted in 5 European countries will provide data sufficient for other European countries to make decisions about their cervical screening programs or whether yet more studies will be required. Large scale RCTs are complex and expensive undertakings that require many years to produce results and it is therefore necessary to carefully consider if additional studies would be an effective use of scarce public health/research resources. An alternative or complementary approach could be the use of computer modelling, and several well-validated models of cervical cancer screening now exist.37 These could be updated with pooled data from the ongoing RCTs and adapted to the healthcare systems/costs of other European states to provide estimates of the cost and effects arising from the implementation of new technologies. However, these results would only be estimates rather than observed end-results from RCTs. Continued post-trial surveillance including linkage to population-based cancer registries and the randomized or controlled application of HPV screening as public health policy can complete the experimental trials and provide the final level of evidence although these results would not be available for many years.38

The use of new technologies for cancer screening was recently addressed by the European Commission (EC) in its recommendation that was published in December of 2003.39, 40 This recommendation states: “No screening test other than those listed in the Annex [the Pap smear] is scientifically justified to be offered to people with no symptoms in an organized population-based program before it has been shown in randomized controlled trials to decrease disease specific mortality”. This statement was qualified in the case of cervical cancer where the EC accepted that “a reduction in cervical cancer incidence can be considered a very helpful indicator”.

Here, it is important to note that the relatively low incidence of cervical cancer in Europe means that trials capable of detecting statistically significant reductions in its incidence would require the recruitment of a very large number of women who would need to be followed for many years at a very high cost. This issue was acknowledged in the design of the trials summarized earlier, which use intermediate end-points such as prospective reductions in ≥CIN3 together with subsequent modelling to estimate the impact on cervical cancer incidence. Even with this end-point, the trials still required the recruitment of many of thousands of women and yet they still will not meet the requirements established by the EC to measure reductions in cancer mortality or incidence as the trial outcomes.

The question of the conclusive endpoint of these screening trials remains the subject of intense debate, dividing opponents and advocates of the early introduction of HPV-based primary screening. Should HPV testing be demonstrated to produce a significant reduction in incident ≥CIN3 at subsequent screening rounds, accepting this outcome as sufficient proof has the advantage that reduced research funding would be required and a more effective screening policy could be implemented earlier, which would result in more life-years saved. The disadvantage is that the ultimate outcome of a reduction of cancer incidence is not directly observed and the modelled effects might, therefore, be partially spurious. Answering this fundamental question of the level of evidence that will be acceptable to make changes to large-scale public health programs will not be simple as it has both scientific and political dimensions. However, it needs to be resolved as soon as possible so that both the research and the public health communities can make plans for the appropriate development of cervical cancer screening in Europe.

Ancillary