Organized screening using cervical cytology is an effective healthcare policy for the prevention of cervical cancer. Use of direct testing for the causative agent of cervical cancer, the human papillomavirus (HPV), has been investigated in numerous studies ever since the first HPV tests were launched 25 years ago. This review summarizes some of the important aspects of the current evidence base and discusses some of the major challenges that need to be considered in order to exploit HPV testing for more cost-effective cervical screening.
THE EVIDENCE BASE
The most reliable type of evidence comes from randomized controlled trials (RCTs). Several large RCTs, together encompassing more than 400 000 women have been conducted (Table 1).
The primary endpoint of these trials has been whether the HPV-based screening entails a reduction in high-grade cervical intraepithelial neoplasia (CIN) in a subsequent round of screening, except for the Indian trial that had reduction in invasive cancer as the endpoint. The trials have reached their primary endpoint, but are continuing to report data from the long-term follow-up. An exception is the Finnish trial that has not yet reached its primary endpoint but has reported interim data. The results have been summarized in at least four formal meta-analyses, published more than 2 years ago. The European Union guidelines project is currently performing an updated systematic review, in which an interim version has been published in collaboration with the Italian health technology assessment [1▪▪].
Direct evidence of greater effectiveness can be shown by lower occurrence of invasive cancer after HPV-based screening than after cytology-based screening. However, invasive cancer is a rare disease, particularly when effective cervical screening programs exist, and valuable evidence can also be obtained by evaluating the protection against the immediate cervical cancer precursors, CIN grade 3 (CIN3).
All four published RCTs that had CIN2 and CIN3 reduction as the endpoint showed a marked reduction of CIN2 and CIN3 at the subsequent screening round in the arm that used HPV or HPV/cytology cotesting compared with the arm that used cytology [1▪▪]. Two RCTs (POBASCAM and NTCC) found a significant reduction of invasive cancers at round 2 in the HPV compared with the cytology arm. NTCC also found a significant reduction of the cumulative incidence of invasive cancers over the first two screening rounds combined [1▪▪]. The Indian trial that reported on follow-up of invasive cancer incidence after HPV screening or cytology screening also found a significant reduction with HPV-based screening.
Another key parameter to evaluate is whether HPV testing causes unnecessary interventions by overdiagnosis of precursor lesions that would have spontaneously regressed or would not have progressed to cancer. Overdiagnosis of regressive lesions can be evaluated in randomized trials by comparing the overall detection of precursor lesions in the HPV and in the cytology arm over two (or more) screening rounds. More than two screening rounds are more informative for assessing overdiagnosis, as it is possible that an increased cumulative detection rate may reflect a very early detection of nonregressive lesions that may have been diagnosed on more extended follow-up. However, already after two rounds of screening, a greatly increased cumulative detection rate of precursor lesions does suggest overdiagnosis. For CIN2, both NTCC, Swedescreen and POBASCAM (but not ARTISTIC), found an increased cumulative detection rate over two rounds of screening. The increased detection rate was particularly strong in NTCC and in particular for women below 35 years of age [relative detection rate 2.81 (95% confidence interval 1.69–4.66) and 3.38 (2.11–5.43) in the two phases of NTCC, respectively]. The cumulative detection rate of CIN3 over two rounds tended to be less increased than for CIN2 and was significantly increased only in NTCC [1▪▪].
A cost-effective use of HPV testing will therefore need to consider age-stratified use of the test and the use of triaging tests among HPV-positive women.
Comparison of the gains of cotesting all women with HPV testing and cytology compared to only HPV-testing fail to find a significant difference in the protective effect [1▪▪]. Therefore, the much lower specificity and greatly increased cost of cotesting of all women implies that cotesting of all women is not an adequate policy.
Comparison of the effect of direct referral of HPV-positive women and of using cytology for triaging HPV-positive women can be made using available data in several of the trials. The most important outcome (reduction of CIN3+ in the next screening round) is quite similar [1▪▪], arguing that the more specific strategy (triaging of HPV-positive women with cytology) is preferable.
COST-EFFECTIVENESS OF HUMAN PAPILLOMAVIRUS-BASED SCREENING
A large number of studies have investigated the cost-effectiveness of HPV-based screening compared to cytology screening, the most recent one being the study of De Kok et al. [2▪▪]. That study found that in most tested scenarios, HPV-based screening was more cost-effective. However, the key information emerging from the cost-effectiveness studies is that the parameters that will most strongly influence cost-effectiveness are the price of the HPV test and the HPV prevalence. As HPV prevalences are strongly age-dependent, the cost-effectiveness can be maximized if invitations to HPV-screening avoid the youngest age groups. The price will of course not be known until a tender for purchasing has been issued. The experience from the HPV vaccination field has been that the cost of vaccines when nationwide tenders for organized vaccination are issued is only a fraction (about one fifth) compared to the cost when the vaccines are purchased for nonorganized use. The 2011 tender for primary HPV screening in the organized program in the greater Stockholm county (Sweden) had a similar outcome, resulting in that the switch to HPV testing will (at least in the case of our program) entail a substantially improved cost-effectiveness.
HETEROGENEITY OF HUMAN PAPILLOMAVIRUS TESTS
A systematic search identified that there exists no less than 148 different commercially available HPV tests. In addition, there were 44 additional commercially available HPV tests that were variants of some of the other 148 tests (M. Poljak et al., unpublished observation). These tests have varying levels of documentation and formal approvals, as well as different performance characteristics. As key performance characteristics such as predictive values are dependent on the HPV prevalence in the studied population, it is also not possible to perform direct comparative evaluations unless the same samples are tested with the different tests. However, for organized programs, the multitude of HPV tests available can instead be considered as an advantage, as there will be fierce competition when the purchasing tenders are issued.
BIOBANKING OF SCREENING SAMPLES
The problem with how to ensure that different HPV tests have adequate performance characteristics in the population served can be overcome if the program has a repository of cervical screening samples with defined outcomes [3▪] that can be used to verify whether tests considered for purchasing perform as expected in the population to be served by the program. In order to be optimally useful, such biobanks would need to be regularly linked to the cervical screening registries to identify samples taken before development of disease [4▪]. When evaluating whether an HPV-positive woman has a transient infection that does not require referral or a persistent infection that does require referral, it is also an advantage for the woman if this can be decided by testing an archival sample rather than by following up with repeated future testing. Finally, it is anticipated that there will be a rapid development of new cervical screening tests, and the development and evaluation of new tests will be facilitated if comprehensive archives of samples with accumulated longitudinal follow-up data on outcomes is available [3▪]. Sweden has recently launched a national biobanking system for cervical screening samples, based on joint operating procedures and nationwide tenders for standardized automation and storage systems.
EXTERNAL QUALITY ASSURANCE FOR HUMAN PAPILLOMAVIRUS TESTS
External quality assurance (EQA) is an essential component for accreditation of any test to be used for clinical routine. Although there are several national EQA programs that issue proficiency panels for HPV-based screening, there is as yet no international proficiency panels specifically designed for EQA of HPV screening. The WHO has developed an international EQA system with annually issued proficiency panels for EQA of HPV genotyping services [5▪], but the demands are somewhat different for an HPV screening test as for an HPV genotyping test. In particular, HPV screening tests do not need to distinguish between all the different HPV types and have different demands on analytic sensitivity. If international EQA were to become available also for HPV screening, this would greatly facilitate the quality assurance and open the possibility to compare results also between different programs.
CYTOLOGY FIRST, HUMAN PAPILLOMAVIRUS FIRST, OR ALTERNATING CYTOLOGY AND HUMAN PAPILLOMAVIRUS FIRST?
There are three different strategies to be considered. Primary cytology with HPV triaging for low-grade cytological abnormalities is currently a routine practice in many countries. This strategy will tend to be favored for populations with high HPV prevalences such as women under the age of 30 years. Primary HPV testing with cytological triage will be favored in populations with low HPV prevalences. As the protective effect of HPV-screening lasts longer than the protective effect of cytology, a population that has been identified as particularly interesting to consider for HPV-based screening is women in the age group in whom the screening program stops (typically at 60 years of age). A third possibility is to alternate whether to screen with cytology or with HPV first. This possibility has some distinct advantages. First, in populations that have previously been screened with HPV, the new HPV infections detected at the next screening round will represent the recent infections that will mostly be transient and will not have persisted long enough to motivate a referral. In other words, the performance of cytology relative to HPV testing will be increased in women who have recently been screened with HPV. Second, no test is perfect and for any test applied there will be some women who are missed. For the HPV test, this could, for example, be rare viral variants with mutations in sequences targeted by primers or probes used in the testing or rare HPV types that are not tested for but that can cause cancer in rare circumstances (such as HPV 26, HPV 51, or HPV 73). Re-designing the HPV tests to include these would cause considerable loss in specificity and a more realistic strategy to catch these would be to alternate with a cytology test. Third, implementation of a major technology switch will need to be evaluated with an intermediate endpoint to obtain an early feedback on whether the technology switch works as expected. To catch the CIN2/3 intermediate endpoints, there would need to be cytology screening performed in screening rounds after the HPV testing rounds.
As an example of strategy choice, the organized cervical screening program in Stockholm chose to pilot HPV screening by first offering it to women 56–60 years of age as an ‘exit test’ before the women leave the screening program. In phase II of the implementation, HPV screening is offered to women age 30 and above. In the age band 30–40 years of age, the next round of screening after the HPV test will be by cytology (alternating strategy), whereas among the women above 40 years of age (who have the lowest HPV prevalences), only HPV screening will be used (Fig. 1).
SCREENING INTERVALS AND FOLLOW-UP OF HUMAN PAPILLOMAVIRUS-POSITIVE, CYTOLOGY-NEGATIVE WOMEN
The follow-up of HPV-positive, cytology-negative women has been different in the different RCTs. As there was no detectable heterogeneity in the outcomes of the RCTs, this implies that the least ambitious follow-up strategy can be chosen. International guidelines are clear in stating that it is safe for cytology-negative women to defer screening for 3 years. This recommendation applies to all cytology-negative women (regardless of their HPV status) and a possible strategy is thus to not refer HPV-positive, cytology-negative women, but let them stay in the organized program if the next screening occurs within 3 years. With increasing use of HPV testing, the results of previous HPV screenings will be known in the screening databases enabling distinction between new HPV infections (that do not need referral) and persistent HPV infections (that need referral).
For HPV-negative women, the evidence base suggests that the protective effect of testing HPV-negative lasts twice as long as the protective effect of being cytology-negative. It seems safe to extend screening intervals for HPV-negative women from 3 to 5 years (and to 6 years if 5 years was the previously recommended interval).
RANDOMIZED HEALTHCARE POLICIES
The current European Union cervical screening guidelines from 2009 state that HPV-based primary screening should only be used within organized programs and only in controlled implementation projects with evaluation. We have recently reviewed the concept on how this can be done [6▪▪]. The only unbiased method for evaluation is if the policy switch is performed in a randomized manner, that is, the old policy and the new policy are used concomitantly and which policy to use is determined by randomization. Randomized healthcare policies (RHPs) is commonly used in a cluster-randomized design (in which the unit of randomization is, for example, a hospital). This was, for example, how the Swedish RHP on HPV-based triaging of low-grade cytological abnormalities was designed [7▪]. For primary screening, it has been possible to use RHPs with randomization at the sample level. The HPV screening policy depicted in Fig. 1 is thus only implemented for 50% of the population and only if the RHP evaluation in 2015 will be favorable, will the new policy be offered to the entire population.
PRIORITY SETTING IN RELATION TO EXPECTED HEALTH BENEFITS
As organized cytology-based screening programs already have a strong cancer-protective effect, the relative health gains of switching to an improved test are not expected to be large. In settings in which an organized program does not exist, the expected health benefits of implementing organization are much larger than the expected benefits of changing the test. Even among the organized programs, the extent of quality assurance used differs. Sweden and Finland have used nationwide audits in case–control format as a method to (using relative risks and proportion of affected cases) quantify which improvements of the program have the largest expected health benefits [8▪–10▪].
Evidence implies that primary HPV-based screening is at least as effective as cytology-based screening. From a public health perspective, the expected health gains are limited. However, primary HPV screening has the potential to offer a more cost-effective program. In order to exploit the benefits of HPV-based screening, there would need to be uniform use of a standard and quality-assured HPV test; a tendered pricing that ensures lower cost than cytology-based screening; a system ensuring that invitations for HPV testing are not issued to too young age groups with high HPV prevalences; a system to ensure that screening intervals are extended among HPV-negative women; a system to ensure that unnecessary referrals of low-risk women with recently acquired HPV do not occur; a screening organization that can perform adequate pilot implementations and evaluations, preferably using RHPs; and a quality assurance system in place that includes all processes of the program (not just the test). A number of European countries already have organized, invitational screening organizations and would have all these requirements in place. There are also many countries that are in the process of building organized programs. Thus, the use of HPV-based screening is likely to become widespread as an integral part of stringently quality-controlled, organized programs.
The author thanks the personnel of the Karolinska Hospital and the Regional Cancer Center Organised Screening Unit for continuous support and discussions.
Conflicts of interest
There are no conflicts of interest.
REFERENCES AND RECOMMENDED READING
Papers of particular interest, published within the annual period of review, have been highlighted as:
- ▪ of special interest
- ▪▪ of outstanding interest
Additional references related to this topic can also be found in the Current World Literature section in this issue (pp. 82–83).
1▪▪. Ronco G, Biggeri A, Confortini M, et al. Health technology assessment report: HPV DNA based primary screening for cervical cancer precursors. Epidemiol Prev
2012; 36 (3–4 Suppl. 1):e1–e72.
This health technology assessment report is, to our knowledge, the only systematic review of primary HPV screening that has been published within the last 18 months. The review is in Italian with English abstract, but does contain all the references to original publications and previously published work.
2▪▪. De Kok IM, van Rosmalen J, Dillner J, et al. Primary screening for human papillomavirus compared with cytology screening for cervical cancer in European settings: cost effectiveness analysis based on a Dutch microsimulation model. BMJ
2012; 344:e670doi: 10.1136/bmj.e670.
The study has compared the cost-effectiveness of HPV screening and cytology screening for a broad range of scenarios and found that for most scenarios, HPV-based screening is more cost-effective.
3▪. Arbyn M, Andersson K, Bergeron C, et al. Cervical cytology biobanks as a resource for molecular epidemiology. Methods Mol Biol
This study describes the theory and practice of biobanking of cervical screening samples.
4▪. Dillner J, Andersson K. Biobanks collected for routine healthcare purposes: build-up and use for epidemiologic research. Methods Mol Biol
This study describes the theory and practice of how to build clinical laboratory biobanks that are optimally useful for epidemiological research.
5▪. Eklund C, Forslund O, Wallin KL, et al. WHO Human Papillomavirus Laboratory NetworkThe 2010 global proficiency study of human papillomavirus genotyping in vaccinology. J Clin Microbiol
2012; 50:2289–2298.doi: 10.1128/JCM.00840-12.
The latest annual report from the WHO HPV LabNet global proficiency panel studies. Of interest, as a model for how similar global proficiency panels could be organized also for HPV-based screening.
6▪▪. Hakama M, Malila N, Dillner J. Randomised health services studies. Int J Cancer
2012; 131:2898–2902.doi: 10.1002/ijc.27561.
A commentary detailing the theory behind how to implement a policy change with reliable evaluation of whether the policy change works as expected.
7▪. Söderlund-Strand A, Eklund C, Kemetli L, et al. Genotyping of human papillomavirus in triaging of low-grade cervical cytology. Am J Obstet Gynecol
A follow-up study of the Swedish randomized healthcare policy of HPV-based triaging of low-grade cytological abnormalities, as an example of the use of cluster randomization in randomized healthcare policy studies.
8▪. Andrae B, Andersson TM, Lambert PC, et al. Screening and cervical cancer cure: population based cohort study. BMJ
2012; 344:e900doi: 10.1136/bmj.e900.
A report from the audit in case–control format of the Swedish organized screening program, as an example of how quality assurance of the entire screening process can be performed.
9▪. Lönnberg S, Nieminen P, Luostarinen T, Anttila A. Mortality audit of the Finnish cervical cancer screening program. Int J Cancer 2012. doi: 10.1002/ijc.27844 [Epub ahead of print].
A report from the audit in case–control format of the Finnish organized screening program, as an example of how quality assurance of the entire screening process can be performed.
10▪. Silfverdal L, Kemetli L, Sparén P, et al. Risk of invasive cervical cancer in relation to clinical investigation and treatment after abnormal cytology: a population-based case–control study. Int J Cancer
2011; 129:1450–1458.doi: 10.1002/ijc.25749.