ARTISTIC: a randomised trial of human papillomavirus (HPV) testing in primary cervical screening

Current design

The purpose of the National Health Service Cervical Screening Programme (NHSCSP) is to reduce cervical cancer incidence and mortality. Currently all women in England aged between 25 and 64 years are invited to attend for a cytology sample – a sampling of exfoliated cervical cells, formerly known as a ‘smear test’ – every 3 years between the ages of 25 and 49, and every 5 years between the ages of 50 and 64. The rationale of cervical screening is based on the detection of preinvasive lesions known as cervical intraepithelial neoplasia (CIN). These lesions precede the development of invasive disease, usually by many years, and offer the opportunity of detection and treatment before the development of cancer. Treatment of CIN involves excision or ablation of disease. Treatment is highly effective in preventing cancer1 and generally preserves reproductive potential. The National Screening Committee’s criteria for appraising the viability, effectiveness and appropriateness of a screening programme are included for reference (Appendix 1).

Two important developments in the screening programme have occurred in the last 3 years. The first was the introduction of liquid-based cytology (LBC), and the second was that the age at which women were initially invited for screening was increased from 20 to 25 and the screening interval was reduced from 5-yearly to 3-yearly for women aged 25 to 49. The reasons for these changes were:

• to avoid ineffective screening among 20- to 24-year-olds in whom over 400,000 samples per year were being read in an age range in which there were fewer than 50 cancers/year (this screening resulted in large numbers of low-grade abnormalities with the potential that overtreatment would do more harm than good)

• epidemiological data from the NHSCSP had demonstrated that in younger women, 3-yearly cytology was required to maintain the necessary level of protection. 2

Effectiveness

Although no randomised trials of cervical cytology as a means of secondary prevention have been performed, there is convincing evidence from disease-incidence rates in countries that have achieved systematic screening to demonstrate that cytology screening is effective. Wide population coverage is essential for a successful programme. In the UK, cervical screening was essentially opportunistic until 1988 with little impact on disease rates. Following the introduction of a systematic computerised call/recall system that issued invitations to every eligible woman at regular intervals, there has been an increase in coverage from 40% to over 80%,3 and a 50% reduction in disease incidence between 1988 and 2004. This has been accompanied by a fall in deaths UK-wide from 2000 per year to around 1000 per year when a continuing increase would have been expected from the increasing trend in mortality in younger women since the 1960s. 4 Furthermore, screening has led to a higher proportion of cancer being discovered sufficiently early that fertility can be preserved.

This success disguises the fact that a proportion of women who undergo screening develop interval cancers even if they comply with regular tests. The most common reason for such screening failure is inadequate sampling, but reading errors account for a proportion of cases. Internationally the sensitivity of cervical screening to detect CIN2 or greater has been estimated to be between 30% and 80%,5 but this is dependent on the quality assurance systems in place. The UK has one of the best programmes in the world with strict national quality assurance and accreditation processes in place for every step of the pathway from sample taking to colposcopic management.

Epidemiology

For 50 years cervical cancer has been considered to have an infectious aetiology, and it is now universally accepted that the necessary initial event is infection of the cervix by human papillomavirus (HPV). 6 There are over 100 types of HPV, of which a subset of around 20 have been associated with cervical cancer by virtue of the presence of HPV DNA being detectable in the cancer cells. Indeed, around 80% of cervical tumours worldwide are associated with types 16, 18, 31, 33 and 45, with type 16 being by far the most important, accounting for at least 50% of cases. 7

The accepted model of tumour development involves infection by HPV as a result of sexual exposure, following which most women clear the infection. In a proportion of cases, however, integration of the HPV genome into the cervical cells and expression of the oncogenes E6 and E7 result in dysregulation of the cell cycle and malignant transformation. HPV infection is a necessary event for cervical carcinogenesis, but other promoters such as cigarette smoking increase the risk. The majority of women will acquire an HPV infection at some time, but only a minority will develop cervical cancer, which can therefore be regarded as an uncommon complication of HPV infection. This necessity for HPV in cervical carcinogenesis is demonstrated by the fact that viral DNA can be identified in almost 100% of cervical cancers. 6 Type-specific HPV DNA detection which persists over 1 or 2 years or more confers a very high relative risk (> 400 in one study),8 when compared with women who were cytology negative (–ve) and HPV –ve. This is compelling evidence of an aetiological role requiring the persistent presence of HPV DNA before development of CIN3.

Potential uses of HPV testing

This scientific background suggests two important clinical applications. The first is a strategy of primary prevention through prophylactic HPV vaccination. This has become a reality with two recently published phase III trials9,10 demonstrating that vaccines against HPV16 and HPV18 can prevent the development of type-16-associated and type-18-associated high-grade CIN. A UK-wide vaccination programme was introduced for 12- to 13-year-old girls in 2008 with a catch-up to age 18 over a 3-year period. The bivalent vaccine directed against HPV types 16 and 18 (Cervarix^®, GlaxoSmithKline) was selected for the UK vaccine programme.

The second important role for HPV is as a biomarker of cervical neoplasia, so that HPV testing can be implemented as a test for screening or to aid clinical management. There are three obvious settings for HPV testing: (1) triage of mild cytological abnormalities to select for colposcopy, (2) as a ‘test of cure’ following treatment of CIN and (3), potentially most importantly, as a primary screening test. HPV testing for triage and test of cure are currently being evaluated in a Sentinel Site project being conducted by the NHSCSP. The effectiveness and cost-effectiveness of HPV testing in primary screening required rigorous testing in a controlled trial within the NHSCSP.

Liquid-based cytology

Nationwide conversion from so-called conventional cytology, which relied on ‘smearing’ exfoliated cells onto a glass slide before fixation, to LBC was completed in 2008. The LBC process involves the exfoliated cells being put into a liquid preservative suspension and either mechanically filtered onto a glass slide or collected by a cell-enrichment process. The principal advantages of LBC are a major reduction in inadequate samples for reading and more rapid throughput of samples in laboratories. 11

Clinical issues

The rationale for HPV testing in primary cervical screening is to increase the sensitivity to detect CIN3, which is generally accepted as the true cancer precursor lesion. By doing so, a drop in deaths could be expected among screened women who develop cancer despite being screened. The evidence for increased sensitivity comes from a number of studies which compare estimates of sensitivity for cytology and HPV testing in primary screening. In a meta-analysis of European studies12 the median sensitivity of cytology was about 50% although sensitivity in the only UK study in the analysis was almost 80%. By contrast, HPV sensitivity for detecting CIN2 or worse was estimated at over 95%. Furthermore, because HPV testing is an objective procedure there was far less variation in results than was the case for cytology. Other evidence for increased sensitivity comes from studies of HPV ‘triage’ to manage women with low-grade cytological changes. In the ALTS trial in the USA,13 HPV testing identified more disease than repeated cytology. Furthermore, this added sensitivity comes with a very high negative predictive value suggesting a role for HPV as an initial ‘stand-alone’ test without cytology. The problem with HPV testing is a lack of specificity, particularly in young women (< 30 years) in whom the high-risk HPV positive (+ve) rates are around 20%. Typing the HPV infection is needed not only to estimate true viral persistence, but also to determine which specific types are responsible for high-grade disease and which are less relevant in that respect. The Hybrid Capture (HC2, QIAGEN) test used in ARTISTIC (A Randomised Trial In Screening To Improve Cytology) uses a cocktail of probes to detect 13 high-risk (HR-) HPV types.

The purpose of the present study was to evaluate HPV testing in primary screening both as an adjunctive test with cytology and as a stand-alone test which would be backed up by cytology for HPV +ve women. At the time of planning the study it was not possible to undertake a trial that would involve HPV testing alone because cytology was, and remains, the international standard of screening and there was insufficient evidence regarding the role of HPV as a stand-alone test. The most rigorous acceptable design was considered to be a randomised trial that compared the current standard, i.e. cytology versus cytology plus HPV testing. To maximise the opportunity to evaluate HPV as a stand-alone test, it was decided to undertake HPV testing with cytology in all women but to conceal the HPV result in the standard arm. This would also permit controlled observation of the psychological impact of HPV testing.

Not only would a randomised trial be capable of robust comparisons of cytology versus cytology plus HPV testing as a primary screen, but the entire cohort data would provide valuable data on which could be modelled the outcomes of different screening strategies. This could include: (1) the current standard, (2) cytology with HPV triage and (3) HPV screening (including varying cut-offs) with cytology triage.

Other current randomised studies of HPV testing in primary cervical screening

Four other European randomised trials are evaluating HPV in primary screening in addition to ARTISTIC. These are being conducted in Sweden (Swedescreen), the Netherlands (POBASCAM), Finland (Finnish Public Health Trial), and Italy (NTCC). They began in 1997, 1999, 2002 and 2003 respectively. The characteristics of the trials are shown in Table 1. Results from Swedescreen14 and POBASCAM15 have been published – reporting data over two rounds of screening. The NTCC study16 has reported data from a prevalence round. All employed conventional cytology.

Implications of vaccination

Since the initiation of this study, prophylactic HPV vaccines have become licensed for the prevention of CIN2/3 for females aged 9–26 years. Currently available vaccines against types 16 and 18 will be capable of reducing the incidence of high-grade CIN (CIN2/3) by over 50% but low-grade cytological abnormality, most of which is either HPV –ve or associated with other HPV types, by perhaps only 20%. It will be important to determine how best to screen vaccinated females aged 25 years and over. If HPV testing were proposed as a primary determinant of risk followed by cytology if HPV +ve, then understanding the risk over time for cytology –ve/HPV +ve women, will be important. The data from the ARTISTIC trial will allow useful estimates of the impact of these vaccines on the prevention of abnormal cytology as well as CIN2/3, and will also provide a valuable contribution to the development of models of the cost-effectiveness of vaccination.

Economic issues

Cervical screening is generally accepted as a cost-effective intervention costing the NHS around £150–180 million17 with an estimated 800–1000 lives saved per year,18 although in the long term up to 5000 future deaths per year may be prevented by current screening. 4 Given that the average age at which cervical cancer deaths occur lies between 60 and 65 years,19 the number of life-years saved is well within the accepted range for a cost per quality-adjusted life-year (QALY) gained.

Any major change to the screening programme would need to be more effective and cost-effective, either in terms of saving more lives or in achieving greater programme efficiency. It could be expected that the use of HPV testing will save lives by more sensitive disease detection.

The systematic review by Cuzick et al. ,20 which informed the design of the ARTISTIC economic evaluation, undertook a modelling exercise to assess whether HPV testing could be cost-effective in primary screening. Three test combinations (cytology, cytology plus HPV testing, and HPV testing alone) were examined in two models in which the screening outcomes were adjusted favourably or less favourably towards the use of the HPV test. NHS-derived costs for cervical screening and the management or treatment of cervical dysplasia were applied, and the effective measures of screening were life-years gained and deaths prevented. Effects and costs were calculated for both 3-yearly and 5-yearly screening between 20 and 64 years of age. The authors concluded that additional HPV testing in primary screening would not be cost-effective unless the cost of HPV testing could be substantially reduced, or, alternatively, fewer tests were performed by lengthening the screening interval from 3 years or by lowering the age at which women are no longer invited for screening following a series of –ve HPV tests. As it was, the costs for HPV screening entered in the models omitted laboratory costs for analysing cervical samples. The authors acknowledged that the issues of cost were not clearly resolved; would the cost of the HPV test be substantially reduced if tests were used at a very high volume? Further modelling was needed to quantify uncertainties of the key parameters, preferably in a large-scale study with at least a 5-year follow-up.

The data capturing methods for ARTISTIC were designed to record cost-generating events (cervical screening, cytology, HPV testing, colposcopy, biopsy and treatment) for all women from the point of recruitment until their end point in the trial. Associated unit costs for these events would be estimated and attached to the individuals’ events. By performing HPV tests on cervical samples in both the control and intervention arms, the three test combinations of cytology, cytology plus HPV testing, and HPV testing alone could be evaluated through modelling. Economic evaluations of screening programmes often included travel and time costs incurred by those being screened on the basis that these may affect uptake and, ultimately, cost-effectiveness of a programme. Although women undergoing cervical screening incur personal expenses, mostly when attending a general practitioner (GP) surgery, women’s costs were not measured.

The systematic review20 was limited to modelling a cost per life-year gained because information on the value of health states associated with cervical cancer was not available. However, such information is required if a cost per QALY is to be estimated. As cervical screening impacts intermittently and temporarily on the lives of most women, the time trade-off (TTO) technique for valuing descriptions of cervical screening outcomes was selected for administration with women drawn from the general population. Previous research had shown that respondents more easily follow the TTO technique for eliciting valuations than an alternative method, the standard gamble technique. 21

In the protocol for the ARTISTIC trial, follow-up of women was limited to 3 years, hence modelling beyond the end point of the trial would be needed to determine whether the HPV test would have an impact on life-years gained. The model used in the systematic review20 was based on the natural history of the disease. Modelling had been applied previously to determine the effectiveness of cervical screening intervals22 and the cost-effectiveness of alternative technologies for Papanicolaou testing of the cervix. 23 For our purposes, a time-varying Markov model would be appropriate to estimate the lifetime costs and effects associated with different screening strategies. This assumed, however, that the trial results would show that the cytology and virology screening techniques performed differently in identifying women with precursor lesions of carcinoma.

Psychological/psychosocial issues

Cervical screening creates anxiety among a proportion of women who receive an abnormal result,24–26 although this tends to resolve following diagnosis and treatment. 27 Studies have indicated that testing +ve for HPV may be responsible for an adverse psychosocial impact, which is related to the sexually transmitted nature of HPV infection.

This effect has been shown in a quantitative study using psychometric measures, which compared women with negative or abnormal cytology who were identified as HPV +ve, with corresponding women who were identified as HPV –ve. 28 Qualitative research29 has identified a number of key concerns that women may have when told that they have tested HPV +ve. These include the stigma of a sexually transmitted infection, the link with cervical cancer and the fact that it may persist. The primary outcome measure of ARTISTIC is detection of CIN3 or worse (CIN3+) at 3-year follow-up, but these concerns reinforce the importance of measuring the psychosocial and psychosexual impact of HPV testing in the ARTISTIC trial.

The objective of the psychological study in ARTISTIC was to determine whether receiving an HPV +ve test result could be associated with increased psychological distress when compared with cytology alone and whether it could also exert a negative effect on psychosexual functioning. The randomised structure of the ARTISTIC trial offered an opportunity to compare the impact of HPV testing when combined with cytology in the revealed arm with a control group of women defined by randomisation in the concealed arm whose HPV status could be matched, but who were unaware of the HPV result.

It was also hypothesised that among women with –ve cytology, receiving an HPV +ve result could be associated with increased distress compared with women receiving an HPV –ve report. If we are to evaluate HPV testing in primary cervical screening we need to understand its psychological effects.

Reading of cytology slides and samples

Cytology results were reported according to the laboratory routine using the British Society for Clinical Cytology (BSCC) classification (described in Table 2). A cytoscreener read the slides with abnormal slides checked by a biomedical scientist or cytopathologist. Rapid review of every negative or inadequate slide was performed before reports were authorised. Cytoscreeners were unaware of HPV test results.

Biopsies were also reported according to routine practice using classification according to agreed guidelines. 30 All pathology results were checked by a consultant pathologist. There was no central review of cytopathology or histopathology, and pathology was reported blind to HPV results.

Concealment

Staff reading cytology slides and biopsies were unaware of the allocation of the sample and concealment was maintained throughout the trial. Women in the concealed arm could request their HPV result if they insisted, but in practice only three women requested their HPV results.

Management protocol

The protocol did not require significant deviation from national guidelines for the management of abnormal cytology samples. In the revealed arm women received their cytology and HPV test result and in certain cases management was dependent on the HPV result. Women were reminded about their next LBC test – either routine cytology or a follow-up HPV test – by letter, sent approximately 1 month before the due date. A reminder was then sent 2 weeks after the test was due. A final reminder was sent around 3 months later.

Women with inadequate cytology samples were retested and re-entered the trial with follow-up as appropriate. In both arms of the trial, women with moderate or severe dyskaryosis were referred for colposcopy and managed according to standard guidelines. Once a woman had been referred to colposcopy, the trial office ceased to send further recall letters to avoid discrepant management from that advised by the colposcopy clinic. Colposcopy clinics were asked to send copies of all correspondence relating to the appointment to the trial office and all related cytology, HPV, colposcopy and histology results were recorded in the trial database. Women in the concealed arm with mild or borderline cytology had a second LBC test at 6 months. If dyskaryosis persisted a colposcopy referral was made and any CIN was monitored or treated according to local policy. If the second sample was negative or borderline, women were recalled for a third LBC test at 12 months. If this third test was abnormal then a referral for colposcopy was made, and if it was negative, the women were recalled for a fourth test at 24 months. Women in the revealed arm who had two consecutive borderline or worse results which were HPV +ve were referred for colposcopy.

The management protocol is summarised in Figure 2. Women who were initially cytology and revealed HPV –ve were managed exactly as in the control arm. Women in the revealed arm who tested cytology –ve and HPV +ve at baseline were invited to be retested for HPV only at 12 months. Most HPV infections are transient and persistence over 12 months or longer was chosen to give greater specificity for detection of CIN than a single HPV +ve result. If this second HPV test was +ve then the women would receive a result letter offering a choice between a colposcopy examination at their local clinic or a repeat HPV test at 24 months. It was explained that if a 24-month HPV test were +ve then a referral would be made for a colposcopy examination. A postage-paid envelope was included for women to return the form indicating their preference. Patient choice was offered to improve compliance and provide valuable information on women’s reactions to being persistently HPV +ve. All women in the trial were recalled to have an LBC and HPV test 36 months after round 1 and were followed up by cytological and histological record linkage. At 36 months, women in both arms were managed according to NHSCSP guidelines, with the exception of those who had borderline cytology; those women were referred to colposcopy after only two consecutive borderline results (rather than the usual three) to ascertain histology within the timescale of the trial.

FIGURE 2.

ARTISTIC trial protocol for the management of women with negative and abnormal cytology and HPV +ve and –ve tests, in the revealed arm. Referral to colposcopy varied depending on cytology history, some women were referred to colposcopy only after three consecutive borderline results or after a mild result followed by a borderline result. Reproduced with permission of Cancer Research UK from Kitchener H, Almonte M, Wheeler P, Desai M, Gilham C, Bailey A, et al. HPV testing in routine cervical screening: cross sectional data from the ARTISTIC trial. Br J Cancer 2006;95:56–61. 31

Round 1

The round 1 (entry) sample was defined as the first cytologically adequate sample after randomisation that gave a satisfactory HPV result by HC2.

Round 2

The round 2 (first routine follow-up) sample was defined as the first cytologically adequate sample taken between 30 and 48 months after the date of the round 1 sample. We do not employ the term ‘exit round’ because women are continuing to be followed up in a third round, 6 years following recruitment to the study. Alternative analyses of CIN2 and CIN3+ (see Table 15) included women with round 2 samples 26–54 months after round 1, in order to reduce the number of excluded lesions.

Abnormal sample

An abnormal sample was defined as borderline or worse cytology, and/or HPV detected by HC2 for women in the revealed arm. Cytology is categorised in Table 2. Cytology +ve is consistently used to mean abnormal cytology of any degree.

Follow-up of abnormal round 1 samples

We considered periods of 6 months (i.e. < 6, 6–11.9, 12–17.9, etc.) to define time of second sample for the analysis of viral persistence in women who were recalled following a round 1 sample that was cytologically abnormal or HPV +ve.

HPV typing and HC2 cut-off point

The main analysis for ARTISTIC was carried out using the recommended cut-off value of ≥ 1 relative light unit/mean control (RLU/Co). In a separate analysis, we present the result of screening tests in round 1 using three additional cut-off points for the HC2 assay (2, 4 and 10 RLU/Co).

CIN1, CIN2 and CIN3

CIN1, CIN2 and CIN3 are worsening grades of CIN in terms of the likelihood of cancer developing and reduced likelihood of spontaneous regression. CIN3+ included CIN3, carcinoma in situ, cervical glandular intraepithelial neoplasia (CGIN), adenocarcinoma in situ, microinvasive carcinoma, invasive squamous carcinoma and adenocarcinoma. Incomplete excision or ablation is sometimes followed by recurrence. Women were therefore censored at CIN3+, so any later CIN2 or CIN3+ was excluded. In analyses of CIN2 women were censored at CIN2, so any later CIN2 was excluded. One CIN3 in round 2 in a woman with CIN2 in round 1 was excluded in the initial analysis but included in the alternative analysis in Table 15.

Disease outcomes

CIN2, CIN2+ (CIN2 or worse) and CIN3+ (CIN3 or worse) in round 1 were defined as the worst histology within 30 months of an abnormal round 1 sample, as there was sometimes a long interval between the first abnormal sample and eventual referral for colposcopy and subsequent histology. CIN2, CIN2+ and CIN3+ in round 2 were defined as worst histology within 30 months of a cytologically abnormal round 2 sample. CIN and cancer diagnosed as a result of histology following HPV detection in a cytologically –ve round 2 sample were ignored in the primary analyses to provide consistency in comparing the revealed and concealed arms.

HPV results

HPV detection (HC2 +ve) includes the 13 high-risk human papillomavirus (HR-HPV) types detected by the HC2 test. Results on HR-HPV types are presented for HPV16 alone, HPV16 and/or HPV18 (HPV16/18), and HPV16 and/or HPV18 and/or HPV31 and/or HPV33 and/or HPV35 (HPV 5 HR types).

Persistent HPV infection

A woman was considered to have a persistent HPV infection justifying referral for colposcopy in the revealed arm if she had:

• two consecutive HC2 +ve samples over a minimum of 12 months

• two HC2 +ve/line blot assay +ve specifically for HPV16, HPV16/18 or HPV 5 HR types (as defined above).

Cytological outcomes

Assuming 10% loss to follow-up by the next screening round, the study had a power of 96% (2p < 0.05) to detect a reduction from 8% to 2% in the prevalence of high-grade CIN at 36 months among women initially HPV +ve but cytologically –ve. The study would also have 90% power to detect an overall reduction of 40% between the two arms in the incidence of high-grade cytology and high-grade CIN at 36 months. This assumes a 1% rate in round 1 of moderate/severe dyskaryosis in the concealed arm. For the subgroup of patients who have an abnormal result and who are HPV +ve, a comparison will be made between subjects randomised to the revealed and concealed arms for high-grade cytological abnormalities at the 36 months rescreen although differences in management are unlikely to affect outcome, and this comparison is of less interest. The study had 80% power to show an overall reduction of 25% between the two arms in the incidence of low-grade cytology.

Practice and GP recruitment

In early 2001, comprehensive general practice lists were obtained from the patient data departments of the four Health Authorities. Information about the trial was sent to the senior partner at the surgeries along with an invitation to attend an evening seminar on how they could participate in the study. The seminar would also provide an update on cervical screening. An event was arranged in each of the areas. Contact was made with screening co-ordinators within Manchester where links already existed. They suggested that due to the nature of cervical screening it may be more appropriate to target practice nurses to gain support for the study.

A series of meetings was arranged which coincided with the decision to employ LBC as the technique for taking samples, so we were able to promote the meetings as an opportunity to be trained and gain experience in this new technique. The event was Continuing Professional Development accredited. Marie Curie Cancer Care supported the initial LBC/ARTISTIC trial training sessions which had over 100 attendees from general practices, FPCs and colposcopy units. Further training was hosted by Practice Nurse forums and later cascaded to new staff either by those who had been on the Marie Curie Cancer Care-supported course or by research nurses on the trial.

Participation in the trial by practices was straightforward. Detailed information about the trial and a consent form were included in recall invitations sent out to women by Health Authority screening staff. Additional copies of paperwork were also sent out to practices. Women were given information about HPV with their invitation for screening, backed up if necessary with information by the sample taker. Once signed consent to take part in the trial was obtained, a cervical sample was taken according to NHSCSP guidance33 and sent to either Stepping Hill Hospital cytology laboratory or MCC based at Central Manchester Health-care Trust (depending on where the woman was recruited), and then to the MRI Virology Department for HPV testing. A cytology request form and the completed patient consent form accompanied the cervical samples when they were sent for processing.

Regular meetings were held across all four Health Authorities throughout the 2-year recruitment phase keeping practices updated with recruitment figures and details of how the trial was progressing. GPs and FPCs received regular newsletters (3- to 6-monthly) with frequently asked questions, recruitment statistics, protocol reminders and trial updates.

Service support costs

The GPs and FPCs were given a reimbursement amounting to £10 for each woman recruited into the trial. Invoices were sent to the trial office on a 6-monthly basis. The target for the under-30 age group was reached more quickly than for the older age groups, therefore practices were asked to target recruitment of women aged 50–64 years. The recruitment reimbursement was increased from £10 to £15 in this age group and ceased for women under 30 years old.

Reimbursements also included postage and stationery (envelopes, paper, printing and photocopying) used to maintain contact with both women and sample takers.

Links with local screening co-ordinators

Regular meetings were held with screening co-ordinators in order to establish the trial. Local screening co-ordinators in Salford & Trafford and Manchester were involved in altering the recall dates for those women in the ARTISTIC study under 50 years of age and under routine recall from 5 years to 3 years, as the standard screening interval in these Health Authorities was still 5-yearly. This brought their recall in line with the trial protocol. Lists of women were sent to the co-ordinators who adjusted the date of recall to make it compatible with the study recall date so that women would receive their invitation letter from the trial office and the Health Authority at the same time, and subsequently be more likely to comply with the 3-year screening protocol. Permission was not given to bring forward the recall date for women over 50 years old, although the women did receive a 3-year invitation letter from the trial office.

Recruitment, consent and randomisation procedures

The study population was screened in over 127 general practices and FPCs which had agreed to participate in the trial. Of those women who were able to access the trial, and were offered the opportunity to participate, fewer than 5% refused.

All women consented to randomisation and separately, to storage of samples for further research. Women were assigned a unique trial number for identification purposes on the main Access 2000 trial and HPV databases. A consent form and a trial information sheet were produced for the study (see Appendices 4 and 5 respectively).

Allocation of women to HPV test revealed or HPV test concealed was in the ratio 3 : 1. Simple randomisation was used because blocking or minimisation is unnecessary to maintain balance in a large sample. For reasons of cost it was not feasible to use an independent randomisation service. Instead, randomisation was carried out by a research assistant independent of the study using a list prepared by one of the trial statisticians. Allocation was concealed from the practice nurse recruiting women into the study because allocation was made when the trial consent forms arrived at the trial centre. To achieve the desired age distribution the minimum age was increased from 20 to 30 when the recruitment target for women aged 20–29 had been reached, then to 40 when there were enough aged 30–39, and so on. Randomisation status of the women was concealed from the staff at the Cytology and Virology Laboratories.

Logistics of cytology and HPV testing

Figure 3 represents the pathway of the sample after its collection in primary care. At the outset of the trial there was no central transport or internal postal system that could deal with the sheer volume of written material involved in the study. Information sheets and consent forms had to be delivered by car by trial staff to practices and Health Authorities.

FIGURE 3.

Logistics of cytology and HPV testing.

The introduction of LBC also led to transport issues; conventional samples could be routinely posted, but existing systems did not support the transportation of LBC samples. HPV testing was undertaken centrally at the MRI Virology Department. Cervical samples taken were read at the MCC and Stepping Hill Cytology Laboratory in Stockport. LBC samples had to be couriered from GP surgeries to either MCC or Stepping Hill Hospital. The samples from Stepping Hill Hospital were couriered to MCC for the slides to be prepared, then back to Stepping Hill for reading. All vials were retained at MCC and taken in batches to the MRI virology laboratory for HPV testing because it was on the same site.

Protocol amendments

• Following a complaint on behalf of a trial participant’s partner midway through round 1, it was determined by the Ethics Committee that all women should be given at least 24 hours between receiving information regarding the trial and giving consent to participate, and the information leaflet was amended to be more explicit about HPV being sexually transmitted.

• GPs were asked to flag the records of women who defaulted HPV testing at 12 months to provide opportunistic encouragement to undergo repeat HPV testing.

• Women with borderline cytology were referred for colposcopy after two not three borderline results in round 2 of the trial to avoid undue delay in diagnosis.

• Ethical approval was granted to flag women participating in ARTISTIC in the NHS central register for automatic notification of incident cancers and death.

Since the ARTISTIC trial was implemented, there have been several national and local changes to the NHSCSP which have had an impact on the trial.

• At the time the trial began in 2001, NHSCSP guidelines recommended that women aged 20 to 64 should participate in cervical cancer screening every 3–5 years. In 2003, national guidelines were changed so that those women under 25 were no longer invited to attend cervical screening. Population-based data on the natural history of CIN in women screened in the English programme aged less than 25 years will therefore no longer be available. The ARTISTIC trial will provide data on long-term outcomes following HPV infection and CIN in 2575 women aged 20–24 years at recruitment.

• The roll-out of LBC to all General Practices and FPCs in the Greater Manchester region, and nationally, had two impacts on the trial. First, when the trial began the only practices and FPCs using LBC were those participating in the trial. This made it easy to identify which samples were in the trial and should therefore be HPV tested. LBC was more or less fully rolled-out in Greater Manchester by October 2005, which meant that the number of samples which could not be HPV tested as they had been mistakenly collected in ‘conventional’ cytology decreased dramatically, but it became more difficult to distinguish ‘ARTISTIC’ samples at the laboratories. To facilitate this process, letters were written to all surgeries to remind sample takers to flag the cytology request form as ‘ARTISTIC’, with lists of women recruited from their surgery. The roll-out of LBC to non-ARTISTIC practices also enabled samples from non-ARTISTIC practices to be HPV tested, for example, if a woman changed her GP or moved to a different Health Authority. In some cases we were able to perform HPV tests on samples from women who had moved out of the Greater Manchester area completely by asking the practice nurse to notify the local cytology laboratory to forward the sample to the Manchester Virology Laboratory. The roll-out of LBC also led to a number of samples being taken in SurePath medium, rather than ThinPrep.

• Referral to colposcopy after one mildly dyskaryotic cytology result instead of two began in Manchester, Salford & Trafford and Wigan & Leigh in January 2005. Stockport, however, continued to refer to colposcopy only after two results showing mild dyskaryosis.

Receipt of LBC samples in the virology laboratory

On receipt of samples in the virology laboratory, the sample vial and request form demographics were checked before labelling vials and request forms with a unique bar-coded identification number. All information was then manually added to an Access 2000 database.

Sample processing for the QIAGEN Hybrid Capture^® 2 test

High-risk HPV screening using the QIAGEN Hybrid Capture 2 test

High-risk HPV screening using the Roche AMPLICOR Microwell Plate Assay

During the ARTISTIC trial the opportunity arose to compare the HC2 test with the AMPLICOR test (Roche). There was a possibility that AMPLICOR would be more sensitive than the HC2 test and so there was a need to assess whether this might provide any clinical benefit. A comparison between HC2 and AMPLICOR was undertaken to determine whether one or the other test had greater utility in the setting of borderline cytology where the greatest need for effective triage exists. The AMPLICOR testing was performed on archived samples for which women had consented to further testing. None of the AMPLICOR results were acted on clinically. The original LBC samples had been spun down and resuspended in phosphate-buffered saline and stored at – 70°C. Thawed samples were tested for β-globin as a test of DNA integrity. The results of the comparison between the tests is shown in Appendix 6, Tables 64 and 65.

The AMPLICOR test was also assessed in a real-time comparison with HC2 on the residual material from a group of 5020 ARTISTIC women during round 2 whose LBC specimen had been used for cytology and HC2. The results of this are also shown in Appendix 6, Tables 66 and 67.

HPV genotyping by prototype line blot assay

Development of the database

Data were recorded using Microsoft Access 2000. 38 Participants were identified on the database by a unique trial number (1–25078). This was verified by a 10-digit NHS number. Participants’ demographic information (date of birth, address, first, last and previous names, comments, registered GP details, clinic/practice venue) and all cytology, HPV results and sample dates were recorded.

The Virology Laboratory kept a separate Microsoft Access 2000 database of samples that were flagged for HPV testing as part of the trial. The virology database detailed the woman’s trial number, laboratory identification number, name, date of birth, NHS number and HPV result. Any samples which came to the virology laboratory from the cytology laboratories in the ARTISTIC study for HPV testing that were not recognised as being part of the trial were included in a separate table of unidentified women, this helped to avoid samples being lost as a result of errors in information recorded on the cytology request form, such as the wrong date of birth, or a change of surname. This database was passed on to the trial office on a regular basis to link copies of flagged cytology reports with HPV results, and enter this information on the main trial database.

Recording data/colposcopy data

On receipt of a consent form, demographic details were entered on the trial database and women were assigned trial numbers consecutively as their details were entered on the database. Copies of flagged cytology results were collected on a regular basis and matched to the woman’s record on the database by performing a search on the ‘date of birth’ or ‘NHS number’ variable. Demographic details were updated if those on the cytology form differed from those on the database, after verification using NHS Open Exeter or by contacting the surgery.

The trial office received copies of correspondence from colposcopy clinics in Greater Manchester relating to colposcopy appointments, these were entered on the database in a subform, detailing the hospital, cytology and HPV result, biopsy result, treatment and the results of any follow-up tests as appropriate.

An audit of the colposcopy data was carried out to ensure that records were complete. The trial co-ordinator obtained an honorary research contract to access patient files at the colposcopy clinics, to confirm the number of attendances and events.

‘Missing’ data

By comparing data gathered in the central ARTISTIC trial office and at the LSHTM, some screening results were discovered to be missing in either database. The trial co-ordinator carried out a manual search using NHS Open Exeter to complete these missing screening records.

All women in the trial have been flagged on the NHS central cancer registry for cancer incidence and mortality, to obtain complete information in the long-term for missing data of which we are currently unaware.

Statistical plan

Statistical analysis of the primary outcome, CIN3+ detected in the screening round 2, was based on a test of proportions with confidence intervals of the difference. An intention-to-treat estimate of the effect of the intervention was determined using numbers randomised as the denominator. This assumes that all subjects without a follow-up screen are negative for CIN2/3+, and may be thought of as a measure of effectiveness of screening. An efficacy estimate was obtained by using the numbers screened in round 2 as the denominator. An efficacy estimate across both screening rounds was estimated by combining the estimate of the proportion of CIN3+ or CIN2+ for round 1 with the efficacy estimate for round 2.

Inclusion criteria: women were included if they:

• were aged 20–64 years attending the NHS screening programme in Greater Manchester (Manchester, Salford & Trafford, Stockport and Wigan)

• had an adequate round 1 sample defined as the first sample after randomisation that was cytologically adequate and gave a satisfactory HPV result by HC2.

Women were invited for their next routine LBC test 3 years after round 1 (round 2), but there was considerable variation in the actual interval. Women with no cytology result in the 30- to 48-month interval were initially excluded from analyses of results in round 2. A further analysis using a round 2 interval of between 26 and 54 months allowed fewer exclusions of second-screening round lesions.

The analysis of treatment policy was based on intention to treat. Women were therefore classified as HPV test revealed or HPV test concealed according to random allocation, irrespective of management. The primary outcome, i.e. the outcome of paramount importance was CIN3+ identified in round 2. Although CIN2+ represents the lesions that are treated, CIN3+ is widely accepted as the true cancer precursor and is the prime target for screening.

Introduction

The main objective of the economic evaluation alongside the ARTISTIC trial was to assess the cost-effectiveness of HPV testing in addition to LBC when compared with a cervical screening programme using LBC only. Secondary objectives included subgroup analyses to identify characteristics of screened women that render HPV testing more cost-effective than if HPV testing were applied to all women. Further scenarios were developed and modelled to explore the cost-effectiveness of alternative approaches to the use of HPV testing within the NHSCSP.

Measuring costs

The cost analysis was carried out from the NHS and personal social services perspective. All costs refer to 2006. The trial data capture methods recorded cost-generating events incurred by women from the point of recruitment to their end point in the study. The key cost-generating events according to the protocol were:

• cervical screening at recruitment and 36 months for all women, and, selectively, repeat screening at 6, 12 and 24 months

• colposcopic examinations, biopsies and treatments for CIN

• histopathology analysis of biopsied material

• gynaecological treatments for severe CIN or cervical cancer.

Unit costs were estimated for these cost-generating events and attributed to the women experiencing the events in order to estimate total costs. Unit costs were derived from observational studies, most undertaken specifically for ARTISTIC, and existing tariffs and contracts, and from published sources. 39 Staff costs reflected the new pay system for NHS staff. 40

The NHSCSP operates a comprehensive failsafe system to minimise the risk of women failing to be screened or managed appropriately. Within local cervical screening programmes (CSPs), the Exeter call/recall computer system generates letters reminding non-responders to attend for routine smears, and letters for women with abnormal smear results who have defaulted or who have had an inadequate smear result. Cytology laboratories have their own failsafe systems for checking that an appropriate referral has been made for women whose test result requires investigation by colposcopy (especially test results of severe dyskaryosis indicative of invasive cancer or glandular neoplasia that must be referred urgently). Failsafe arrangements in colposcopy clinics issue reminder letters to women who default from appointments and notification letters to their GP or other responsible clinician. However, costs were not prepared for the different components of this integrated system.

Measuring health benefits

The purpose of ARTISTIC’s TTO postal survey was to provide women’s valuations of health states following cervical screening involving HPV testing. The health states (i.e. scenarios) corresponded to the states in the Markov model intended to establish cost-effectiveness. The valuations would be used to generate QALYs for cost–utility analyses of cervical screening programmes. Research ethics approval was obtained. Questionnaires were sent to almost 1600 ARTISTIC women whose cytology and HPV results were –ve in both round 1 and round 2 of the trial, and more than half of the questionnaires were completed.

Utility scores were generated for five health states, but they were not incorporated in a Markov model as originally intended, because of the similarity in the clinical results for two arms of the trial. It is for this reason that an account of the TTO survey is not presented in this report, but more information is available at http://brunel.ac.uk/about/acad/herg.

Synthesis of costs and benefits

Modelling beyond the trial end point

Although there was an intention in the trial protocol, which was originally developed in 1998–99, to undertake Markov modelling beyond the end point of the trial to determine the impact that HPV testing could have on life-years gained, the feasibility of completing the modelling for this report depended upon the clinical results, and in particular, the performance of HPV testing in identifying additional cases of CIN2 or CIN3+. For the clinical analyses presented in this report, 54 months is the maximum duration of follow-up for women recruited in the trial, but many women had not yet been followed up for this length of time. Moreover, the emerging clinical results for round 2 indicated an unexpectedly low incidence of high-grade pathology. As the trial progresses through a third round of screening, that is 72 months after recruitment, the observed incidence rates of CIN2+ for the two screening techniques (LBC versus HPV testing) should be sufficiently reliable for modelling purposes.

A range of sensitivity analyses were carried out to explore the effects of key variables on the overall cost results, such as HPV test cut-off levels, unit costs for the colposcopy-related events and LBC inadequate rates for the NHSCSP.

Protocol amendments

Questionnaire measures

The General Health Questionnaire (GHQ-28)54 measures generalised psychological distress ‘over the past few weeks’. A cut-off score GHQ ≥ 4 was employed to estimate the numbers of probable cases of affective disorder in the sample.

The Spielberger State–Trait Anxiety Inventory (STAI)55 assesses two domains: state anxiety levels ‘right now, that is, at this moment’ (STATE) and trait anxiety ‘how you generally feel’ (TRAIT). GHQ and STAI higher mean scores indicate greater levels of general psychological distress and state and trait anxiety respectively.

The Sexual Rating Scale (SRS)56 determines sexual satisfaction with the woman’s current partner, and the participants were instructed to complete the SRS only ‘if they had a current partner’. These are rescaled as percentage scores (0–100%) generated from the SRS data, with higher scores indicating greater levels of sexual satisfaction in their current relationship.

Initially, the questionnaire data were collected in face-to-face interviews (n = 106) but postal delivery was subsequently adopted by the investigators because of time and economic costs. Non-responders were posted repeat questionnaires approximately 2 weeks after the initial mailing, and women who returned all four questionnaires completely blank were coded as non-responders. Those participants who completed at least one of the four questionnaires were classified as responders.

Statistical methods

Round 2 exclusions

Eleven subsequent CIN2 cases (11 revealed, 0 concealed) and six CIN3+ cases (five revealed, one concealed) were ignored because the woman’s round 2 sample was cytologically negative. This convention is necessary to give uniform round 2 follow-up between the arms of the trial, because HPV results were ignored in the concealed arm. In addition, 12 CIN2 cases (seven revealed, five concealed) and 14 CIN3+ cases (10 revealed, four concealed) were excluded because there was no preceding round 2 sample. Six of these 12 excluded CIN2 cases and all 14 excluded CIN3+ cases are included in the alternative analyses (Table 15) in which round 2 is defined as the first adequate cytology result between 26 and 54 months after round 1. Details of all excluded CIN2 and CIN3+ cases are given in Table 3. To minimise exclusion of CIN3+ cases in this alternative analysis an abnormal cytology result on the date of histology was assumed for three CIN3+ cases diagnosed 29, 31 and 35 months after round 1 with no abnormal smear record.

Accrual

Women entered the study between July 2001 and October 2003 as shown in Figure 6. Accrual was steady, slowing only a little after recruitment of women below 30 years old was stopped. Figure 7 shows the rate of return for round 2 cytology samples. Between July 2004 and April 2007, 14,639 women entered round 2 with adequate cytology, representing almost 60% of the original cohort. Of these women, 1325 did not have an HPV test. There was no difference between the proportions who attended round 2 from the two arms.

FIGURE 6.

Accrual to trial between July 2001 and October 2003.

FIGURE 7.

Accrual and follow-up curves of women returning for cervical samples in round 2 by original month of accrual.

The accrual from each Health Authority is shown in Table 4. The routine recall policy varied among Health Authorities before the 2005 national guidance on 3-year and 5-year follow-up, although invitations were sent from the trial office to all women at 3 years. Health Authority recall dates for women in the trial were altered in 2005 to bring them in line with the trial protocol (discussed in more detail in Chapter 2, Links with screening co-ordinators).

Cytology, HPV and histology data from round 1

Table 5 shows the characteristics in round 1 of those women who did and did not attend for round 2 screening. A higher proportion of older women attended for screening in round 2 (66% aged 40–64, 43% aged 20–29). This was reflected in the round 1 HPV +ve rates for those who did and did not attend for round 2 screening; 12.6% versus 19.7% respectively. There was however no difference in either baseline age or HPV rates between the arms in round 2 indicating that the pattern of adherence to round 2 did not introduce bias.

Of the 25,078 samples collected in round 1374 (1.5%) were inadequate for cytology and 141 (0.6%) were insufficient for HPV testing. The cytology results are tabulated against the HPV results in Table 6. Overall there was a cytology –ve rate of 87.2%, 7.3% of smears showed borderline changes, 3.6% mild dyskaryosis, 1.1% moderate dyskaryosis and 0.8% severe dyskaryosis. This latter proportion, 1.9% moderate and severe dyskaryosis combined is almost identical to the calculation (1.8% moderate/severe) made before the trial. It can be seen that the proportions of cytological abnormality are almost identical between the arms. The cytologically negative women who were HPV +ve represented the only real difference between the arms in terms of potential disease detection, as women in the revealed arm were offered colposcopy if the HPV tests were persistently positive over 12–24 months. There were 1675 such women accounting for 9.1% of the revealed arm.

Overall 3813 (15.6%) women were HPV +ve, while women with negative cytology had an HPV rate of 10.4%. Women who were HPV +ve had an abnormal cytology rate of 14.6% borderline; 16.1% mild dyskaryosis and 10.9% moderate or worse. HPV +ve rates rose as the grade of cytological abnormality increased; borderline, mild, moderate and severe dyskaryosis had HPV +ve rates of 31%, 70%, 86% and 96% respectively.

The striking relationship between HPV infection and age is shown in Figure 8. The HPV +ve rates fall from 40% in women aged 20–24 to less than half (18.5%) in women aged 30–34, down to just 6% in women aged over 55 years.

FIGURE 8.

Prevalence of high-risk HPV (HR-HPV) by Hybrid Capture 2 (HC2) according to age quinquennia. Reproduced with permission of Cancer Research UK from Kitchener H, Almonte M, Wheeler P, Desai M, Gilham C, Bailey A, et al. HPV testing in routine cervical screening: cross sectional data from the ARTISTIC trial. Br J Cancer 2006;95:56–61. 31

The relationship between cytology, HPV status and age is shown in Table 7. If the moderate and severe dyskaryosis rates are combined the values are 112/2575 (4.3%) and 103/2591 (3.9%) in women aged 20–24 and 25–29 years respectively, falling to 223/13,731 (1.62%) in women aged 30–49 and 25/5613 (0.45%) in women aged 50–64. This fall in abnormal cytology is largely the result of the falling rates of HPV infection, because among HPV +ve women the rates of high-grade cytology remained steady with advancing age; 203/1749 (11.6%) and 197/1697 (11.6%) for age 20–29 and 30–49 respectively. There was however a fall in HPV +ve women aged 50–64, 17/367 (4.3%).

Figure 9 depicts graphically the relationship between age and cytology and the rate of HPV infection. It is clear that rates of HPV infection in negative cytology, as well as in borderline and mild dyskaryosis are very age dependent. In moderate dyskaryosis the effect of age appears lessened and almost disappears with severe dyskaryosis. This reflects the very high association with CIN3 in severe dyskaryosis whatever the age. In younger women a high proportion of mild abnormalities merely reflects HPV infection.

FIGURE 9.

Prevalence of high-risk HPV (HR-HPV) by Hybrid Capture 2 (HC2) and cytology grade within different age bands. Reproduced with permission of Cancer Research UK from Kitchener H, Almonte M, Wheeler P, Desai M, Gilham C, Bailey A, et al. HPV testing in routine cervical screening: cross sectional data from the ARTISTIC trial. Br J Cancer 2006;95:56–61. 31

The histological data from round 1 are shown by age and grade of cytology for both HPV +ve and HPV –ve women in Table 8. In total there were 313 CIN3+ lesions and 586 CIN2+ (273 CIN2). Only nine CIN3+ lesions were detected in women aged 50 years or more and all of these occurred in association with HPV +ve/high-grade cytology. Only 10 CIN2 lesions occurred in women aged 50 years or more and only three of these occurred in association with HPV +ve/high-grade cytology.

Among the 313 women with CIN3+, 91 (29.1%) were detected in women with low-grade cytology and 212 (67.7%) in women with high-grade cytology. For CIN2+ the proportions were 225/586 (38.4%) and 329/586 (56.1%) respectively. When histological outcomes in women with abnormal cytology were analysed by initial HPV status (by HC2), 93.3% of CIN2+ and 97% of CIN3+ were detected in HPV +ve women. The proportion of women with moderate and severe dyskaryosis overall is 50 times greater in HPV +ve women (10.9%) compared with HPV–ve women (0.22%).

The same data by randomisation are shown in Appendix 6 (Table 58: Revealed arm and Table 59: Concealed arm), and corresponding data for round 2 in both arms together (Table 60).

A total of 313 cases of CIN3+ and 273 of CIN2 were diagnosed in round 1. There were no significant differences in proportions between the arms. Only 28 CIN2 and nine CIN3+ lesions were detected in the HPV –ve group, as against 245 CIN2 and 304 CIN3+ in the HPV +ve group. The latter included 22 CIN2 and 10 CIN3+ lesions in women in the revealed arm with negative cytology who were referred for colposcopy because they had persistent HPV infection.

The 12 cancers included in the CIN3+ category of lesions are summarised in Table 9. Eight of the nine cancers detected in round 1 were HPV +ve with moderate or worse cytology (severe dyskaryosis in six cases, moderate in one and glandular neoplasia in one). The remaining round 1 case, an adenocarcinoma, was HPV –ve with borderline cytology. Two of the three cancers detected in round 2 had negative cytology at both round 1 and round 2, and one was also HPV –ve in round 1. This was an adenocarcinoma that may have occurred high in the cervical canal and was perhaps inadequately sampled. The third round 2 cancer had borderline cytology in both round 1 and round 2. This woman was also HPV –ve in round 1, and had no HPV test in round 2.

Cytology, HPV and histology data in round 2

A total of 14,639 women had an adequate cytology result in round 2, representing 60% of the original cohort. HPV and cytology results for round 2 of screening are shown by randomisation in Table 10 and by age in round 1 in Table 11. Comparison against the corresponding round 1 results (Tables 6 and 7) shows a remarkable and unexpected reduction in cytological abnormality in both arms, the cytology –ve rate rising from 87.2% in round 1 to 95.1% in round 2. The borderline and mild dyskaryosis rates more than halved from round 1 to round 2, from 7.3% to 3.1% and from 3.6% to 1.5% respectively. The reductions in the moderate and severe dyskaryosis rates were even greater, from 1.1% to 0.2% and from 0.8% to 0.1% respectively.

As shown in Table 12, these low rates of cytological abnormality in round 2 led to a much lower number of colposcopies. Overall, 1925 women had one or more colposcopy clinic consultations. Eighty per cent of the colposcopies were performed in the revealed arm in both rounds 1 and 2 (a first colposcopy within 30 months of a round 1 sample was classified as being in round 1).

The numbers of CIN3+ and CIN2 lesions detected amongst all 24,510 women in round 1 and in 14,230 women in round 2 are shown by randomised arm and by cytology and HPV status in round 1 in Table 13a. The low prevalence of cytological abnormality is reflected in the low numbers of CIN2 and CIN3+ lesions identified in round 2 as shown in Table 13b. This unexpectedly low incidence rate of high-grade histology means that the trial has low power to detect the reduction in incidence of CIN3+ in the revealed arm, which was the primary outcome at the end point of the trial.

High-grade histology in round 1 and round 2: amended definition of round 2 sample to reduce exclusions

There was a need to define time limits for round 2 for the purpose of the data analysis for this report, hence the 30–48 months definition. As the trial progressed, it became apparent that this resulted in a number of excluded cases which fell outside the definitions of both round 1 and round 2. Because of the need to exclude as few cases as possible in a per protocol analysis, a further definition of round 2 was adopted covering months 26–54. The results are shown in Table 15 and there is a separate CONSORT diagram to accompany this (Figure 10). The differences between the arms in round 2 for CIN3+ (0.21% revealed, 0.41% concealed) and for CIN2+ (0.49% revealed, 0.80% concealed) both become statistically significant (p = 0.05 and p = 0.03 respectively), but there remains no difference between the arms where rounds 1 and 2 are summed. Round 2 cases of CIN2+ whose round 2 sample was cytology –ve were excluded to avoid bias between the arms (see round 2 exclusions in Table 3).

FIGURE 10.

CONSORT diagram of the ARTISTIC trial for round 2 definition of 26–54 months.

In addition to the more stringent per protocol analysis, an analysis on the basis of ‘intention to treat’ is presented in Table 16. This includes every initial CIN2+ and CIN3+ for every randomised woman over round 1 and 2.

Decline in cytological abnormality and histological disease from round 1 to round 2

Several factors independent of LBC may have contributed to the marked decline in disease rates from round 1 to round 2. These include:

• Women were about 3 years older in round 2, and disease rates drop sharply with age in younger women.

• Women screened in round 2, whose last routine smear was the round 1 sample taken approximately 3 years earlier, are at lower risk than the cross-section recruited in the trial, many of whom had not been screened within 3 years of recruitment.

• Histological follow-up of abnormal cytology in round 2 is still incomplete for some women.

Any change in the CIN2+ diagnosis rate in women with abnormal cytology can be assessed by calculating cumulative (Kaplan–Meier) CIN2+ rates. These are similar up to 12 months following abnormal cytology in round 1 and round 2: moderate or worse 68% [95% confidence interval (95% CI) 63% to 72%) in round 1, 60% (95% CI 46% to 74%) in round 2; borderline/mild 4.3% (95% CI 3.6% to 5.2%) in round 1, 5.4% (95% CI 3.9% to 7.5%) in round 2. Age-specific HPV prevalence rates in women with negative cytology were also similar in round 1 and in round 2 in women aged up to 40, although slightly higher above age 40 (respective HPV rates in round 1 in women with negative cytology at age 20–24, 25–29, 30–39, 40–49 and 50–64 were 27%, 18%, 10%, 6% and 6%, compared with 25%, 19%, 9%, 6%, 4% and 3% in round 2).

The contribution of the biases listed above to the extraordinary decline in disease rates during the ARTISTIC trial can be adjusted for by comparing cytological abnormality rates in round 1 and round 2, adjusting for current age in round 1 and at follow-up (round 2), and restricting the round 1 analysis to women whose previous smear was 30–48 months earlier and cytologically negative. This logistic multiple regression gives an adjusted odds ratio for round 2 against round 1 of 0.45 (95% CI 0.37 to 0.56) for borderline/mild and 0.21 (95% CI 0.10 to 0.43) for moderate or worse cytology.

A further explanation for this dramatic change is that a high proportion of cytological abnormality, particularly high-grade disease, which was detected by LBC in round 1 was missed by the preceding conventional smear test. This was unexpected because recent studies suggest similar sensitivity for these tests. 58

There are two factors which may have had an important influence on the performance of LBC. The first of these was the rigorous training which required medical and non-medical laboratory staff to complete a formal curriculum before being able to read slides independently. Non-medical staff had to read 400 unmarked and 20 test slides and medical staff 200 unmarked and 20 test slides. All staff had to achieve 95% sensitivity identifying high-grade slides and had to achieve an 80% pass mark for the test slides. In other countries, training was often provided by the manufacturer and consisted of 3–5 days of lectures and viewing slides.

The second factor may have been the high rate of low-grade abnormality in round 1 comprising 7.3% borderline and 3.5% mild dyskaryosis. In fact, the cytological abnormality rate was 16.8% in the first 6 months of recruitment and the age adjusted odds ratio (95% CI) in round 1 for any abnormality fell in successive 6-month periods from 1.0 (reference) to 0.75 (0.68 to 0.82), 0.71 (0.64 to 0.79), 0.60 (0.53 to 0.69) and 0.61 (0.49 to 0.76). This resulted in a colposcopy rate of 5.2% in the concealed arm and 6.8% in the revealed arm, the extra cases in the revealed arm being the result of cytology –ve/HPV +ve women. This relatively high colposcopy rate will have contributed to a high detection rate of CIN2+ and CIN3+. This will have had an impact on the incidence of disease in round 2.

Management preferences of women after two HPV +ve (cytology –ve) results

Of the 1675 women in the revealed arm who tested cytology –ve and HPV +ve, 1040 (62%) had returned for a first repeat HPV test within 30 months (Figure 11). Out of 1040 such women tested, 439 again tested HPV +ve (42.2%), of whom 427 responded to the letter offering either a colposcopy or a repeat HPV test at 24 months. Colposcopy was preferred by the majority of women (61.8%), all of whom attended. A further HPV test before round 2 was chosen by 163 women (39.2%). Only 50 of these 163 attended again for HPV testing before round 2, and a further 72 were retested at their next (round 2) routine recall. Twenty-seven (54%) of the 50 women who returned tested HPV +ve for a third time and were referred for colposcopy by the trial office.

FIGURE 11.

Flow of women who were cytology –ve/HPV +ve in the revealed arm in round 1. *Includes all women who returned for a second HPV test up to 30 months since the first one.

The effect of patient choice in the revealed arm

As a result of choosing colposcopy if persistently HPV +ve at 12 months and referral for colposcopy if still HPV +ve at 24 months, 10 CIN3+ and 32 CIN2+ were detected. A further one CIN3+ and eight CIN2+ were detected in round 2 amongst women who chose to be retested at 24 months, and a remaining 11 CIN3+ and 16 CIN2+ were diagnosed as a result of being rescreened in round 2.

Performance of HPV as a stand-alone test with cytology reserved for HPV +ve women

Because cytology combined with HPV testing did not appear more effective than cytology alone it is important to consider HPV testing as an initial sole test, cytology being reserved for HPV +ve women. The key data in this context are the lesions that would have been missed by HPV testing alone. For this purpose, all of the lesions detected in ARTISTIC have been mapped through rounds 1 and 2 according to initial HPV test results. As Figure 12 shows, in round 1 there were nine CIN3+ and 37 CIN2+ in the HPV –ve group of 20,697 women, compared with 304 and 549 respectively for the HPV +ve group of 3813 women. Therefore 97% of CIN3+ and 94% of CIN2+ were detected in the HPV +ve group.

FIGURE 12.

Number of CIN2 and CIN3 or worse histological lesions detected in rounds 1 and 2 by cytology and HPV status in round 1. Numbers in boxes refer to women who came back for screening in round 2 who were not previously diagnosed with CIN2 or CIN3+ in round 1 and 11 extra CIN3 cases (nine in the revealed arm and two in the concealed arm) listed in Table 3.

Figure 12 allows comparison of lesions missed by an HPV test and by initial cytology. Compared with combined testing, HPV initial testing would therefore have missed nine CIN3+ and 37 CIN2+ compared with 10 CIN3+ and 32 CIN2+, which would have been missed by initial cytology. These figures are almost identical.

The bottom half of Figure 12 and Figures 13 and 14 provide cytology and HPV data in round 1 on women with a diagnosis of CIN2+ and CIN3+ in round 2. In 12,441 women who had been HPV –ve in round 1 and were rescreened, there were only 10 CIN3+ (0.08%) and 22 CIN2+ (0.18%) cases in round 2. Rates were about twice as high among 12,666 women who had been cytology –ve in round 1, among whom there were 22 CIN3+ (0.17%) and 50 CIN2+ (0.39%) cases in round 2. In contrast, there were 21 CIN3+ (1.2%) and 51 CIN2+ (2.9%) cases in 1789 women screened in round 2 who had been HPV +ve in round 1.

FIGURE 13.

Number of CIN2 and CIN3 or worse histological lesions detected in round 2 in women HPV +ve in round 1. *Numbers in boxes refer to women who came back for screening in round 2 who were not previously diagnosed with CIN2 or CIN3+ in round 1. **Number of cases in the revealed arm: seven CIN2 and two CIN3+ in women with borderline/mild (B/M) cytology in round 2, and four CIN2 and six CIN3+ in women with moderate or worse (Mod+) cytology in round 2.

FIGURE 14.

Number of CIN2 and CIN3+ lesions in round 2 in women with HPV –ve test in round 1. * Numbers in boxes refer to women who came back for screening in round 2 who were not previously diagnosed with CIN2 or CIN3+ in round 1. B/M, borderline or mild; Mod+, moderate or worse.

To demonstrate screening outcomes following an initial HPV screen, the cytology and histology results in round 2 are also shown for women who had been HPV +ve (Figure 12) and HPV –ve (Figure 13) in round 1.

With respect to cytological outcomes, it is noteworthy that in women who were HPV +ve in round 1 the total rate of cytological abnormalities in round 2 was almost as high as it was in round 1. These were 1587/3813 (41.6%) in round 1 and 170/497 (35.5%) in round 2. With respect to high-grade cytology only, there were 417/3813 (10.9%) in round 1 but only 37/1789 (2%) in round 2 (Figure 12). This confirms the effect of sensitive screening in round 1 for high-grade disease; the impact was far less for low-grade abnormalities, many of which clearly represent little more than HPV infection.

Using all of the data from the revealed arm it was possible to calculate the relative sensitivity and specificity for CIN2+ detection under different screening policies based on 220 CIN2 and 233 CIN3+ lesions detected in the whole revealed arm in round 1. These are shown in Table 69 in Appendix 6.

Sensitivity of combined and separate cytology and HPV testing in the detection of CIN2/3 when backed up by routine colposcopy

A concurrent colposcopic study conducted locally, but outside the trial (Flynn M, et al. , unpublished data) involved 557 women (aged 20–64) who were routinely screened in a single primary care practice in Greater Manchester using the same tests (HC2 and ThinPrep) as in ARTISTIC but these women all consented to undergo colposcopy, and in the event of any colposcopic abnormality, a biopsy was performed. The standard use of colposcopy was to ascertain, within the limits of sensitivity of colposcopy/biopsy, the presence of any underlying disease which might not be detected, particularly in cytology –ve/HPV –ve women for whom no colposcopy was performed in ARTISTIC. Sixty-nine women (12.4%) underwent biopsy, and as can be seen in Table 17, 444 women (78%) were cytology –ve/HPV –ve. Of these women, none were found to have CIN, confirming the very high negative predictive value of this combination.

In a total of 490 HPV –ve women, 46 of whom had some cytology abnormality, none had CIN2+. Of 473 cytology –ve women, only one woman had CIN2. There were two cases of CIN3 and one of CIN2, all of whom were HPV +ve.

Optimal cut-off for a +ve Hybrid Capture 2 test

HC2 allows for a range of cut-off values because of the semiquantitative nature of the assay. We employed the manufacturer’s recommended threshold for a +ve HC2 test result of ≥ 1 RLU/Co in the trial, but the large size of the study allows analysis of different cut-offs and resultant cytological and histological outcomes in terms of sensitivity and specificity.

Explanation for CIN3 associated with an HC2 –ve result at baseline

There were nine women who were HC2 –ve at baseline and who subsequently developed CIN3. Typing data on residual material from these women found that three contained HPV16, one contained HPV6, four were LBA negative and one had insufficient sample for testing. Roche AMPLICOR results were available for eight of the samples (one had insufficient sample) of which three tested AMPLICOR positive (Table 20). Possible reasons for the failure of the HC2 assay to detect HPV in these samples may include sensitivity issues or the fact that this assay does not control for DNA integrity or sample adequacy.

HPV genotype profile in the screened population

Prevalence

The proportion of women with a single HR-HPV type who had moderate or worse cytology (right-hand side of Table 21) was 26% for HPV16, between 12% and 19% for HPV types 18, 31, 33 and 58, 7% to 9% for types 35, 45, 51 and 52, and less than 5% for types 39, 56, 59 and 68. The proportion with borderline or mild cytology ranged from 23% to 42% for the different high-risk types. Similar data for HPV16, HPV18 and other HR HPV types collectively are shown in Appendix 6, Table 63.

The proportion of different grades of cytology positive for HPV types 16, 18, 31, 45 and 52 is shown in Figure 15. This demonstrates graphically the increasing prevalence with cytology grade.

FIGURE 15.

Prevalence rates for four of the commonest five HPV types and HPV45 by cytological grade. Reproduced with permission of Cancer Research UK from Sargent A, Bailey A, Almonte M, Turner A, Thomson C, Peto J, et al. Prevalence of type-specific HPV infection by age and grade of cervical cytology: data from the ARTISTIC trial. Br J Cancer 2008;98:1704–9.

Prevalence rates for each HR-HPV type are shown in Table 22, both overall and by age group. HPV16 was the commonest genotype at all ages (overall prevalence 3.3%), followed by HPV52 (1.5%), HPV18 and HPV31 (both 1.3%), HPV51 (1.2%) and HPV39 (1.1%). There was a marked decline in the prevalence of HR-HPV with age, both overall (27.3% below age 30, 6.1% at age 30 or over) and for each HPV type, but less so for HC2 +ve samples in which no HR-HPV was detected (6.4% of women aged under 30, 4.5% at age 30–64).

TABLE 22 - Prevalence of high-risk HPVs overall and as a proportion of HR-HPV +ve women, by age group

Reproduced with permission of Cancer Research UK from Sargent A, Bailey A, Almonte M, Turner A, Thomson C, Peto J, et al. Prevalence of type-specific HPV infection by age and grade of cervical cytology: data from the ARTISTIC trial. Br J Cancer 2008;98:1704–9.

Type	20–29 years			30–39 years			40–49 years			50–64 years			All ages
	n	% of all women	% of HR-HPV+ women	n	% of all women	% of HR-HPV+ women	n	% of all women	% of HR-HPV+ women	n	% of all women	% of HR-HPV+ women	n	% of all women	% of HR-HPV+ women
16	499	(9.7)	35.5	207	(2.7)	26.6	62	(1.0)	23.9	37	(0.7)	26.1	805	(3.3)	31.2
18	191	(3.7)	13.6	89	(1.2)	11.4	20	(0.3)	7.7	19	(0.3)	13.4	319	(1.3)	12.3
31	185	(3.6)	13.2	102	(1.3)	13.1	27	(0.4)	10.4	12	(0.2)	8.5	326	(1.3)	12.6
33	130	(2.5)	9.3	41	(0.5)	5.3	10	(0.2)	3.9	2	(0.04)	1.4	183	(0.7)	7.1
35	49	(1.0)	3.5	39	(0.5)	5.0	17	(0.3)	6.6	3	(0.1)	2.1	108	(0.4)	4.2
39	159	(3.1)	11.3	75	(1.0)	9.6	17	(0.3)	6.6	15	(0.3)	10.6	266	(1.1)	10.3
45	94	(1.8)	6.7	58	(0.8)	7.4	20	(0.3)	7.7	18	(0.3)	12.7	190	(0.8)	7.4
51	189	(3.7)	13.5	73	(1.0)	9.4	26	(0.4)	10.0	17	(0.3)	12.0	305	(1.2)	11.8
52	211	(4.1)	15.0	106	(1.4)	13.6	39	(0.6)	15.1	11	(0.2)	7.8	367	(1.5)	14.2
56	98	(1.9)	7.0	58	(0.8)	7.4	13	(0.2)	5.0	13	(0.2)	9.2	182	(0.7)	7.0
58	97	(1.9)	6.9	46	(0.6)	5.9	18	(0.3)	7.0	7	(0.1)	4.9	168	(0.7)	6.5
59	125	(2.4)	8.9	40	(0.5)	5.1	23	(0.4)	8.9	11	(0.2)	7.8	199	(0.8)	7.7
68	50	(1.0)	3.6	30	(0.4)	3.9	8	(0.1)	3.1	6	(0.1)	4.2	94	(0.4)	3.6
16 and/or 18	654	(12.7)	46.6	289	(3.8)	37.1	80	(1.3)	30.9	54	(0.96)	38.0	1077	(4.4)	41.7
Any HR-HPV	1404	(27.3)	100	779	(10.3)	100	259	(4.2)	100	142	(2.5)	100	2584	(10.6)	100
HC2+ no HR-HPV	329	(6.4)	–	368	(4.8)	–	270	(4.4)	–	222	(4.0)	–	1189	(4.9)	–
HC2 –ve	3417	(66.4)	–	6452	(84.9)	–	5582	(91.3)	–	5246	(93.5)	–	20,697	(84.6)	–
All women	5150	(100)	–	7599	(100)	–	6111	(100)	–	5610	(100)	–	24,470	(100)	–

HPV prevalence rates by age group and cytology are shown in Table 23 for HPV16, HPV18 without HPV16, and for other HR-HPVs combined. Below age 30 a high proportion of infected women carried two or more different HR-HPV types (44% of women with HPV16, 50% with HPV18 but not HPV16, 24% of all women with other HR-HPVs). Multiple infection was less common at age 30–64 (23% of women with HPV16, 20% with HPV18, 14% of women with other HR-HPVs). The proportion with moderate dyskaryosis or worse was 15.3% (396/2584) in women with any HR-HPV infection, 1.2% for HC2 +ves with no HR-HPV, and 0.22% for HC2 –ve women. The risk of moderate or worse cytology was highest in women infected with HPV16 irrespective of the presence of other HPVs (26.2% for HPV16 alone, 25.3% together with other HR-HPVs).

TABLE 23 - Prevalence of single and multiple infections with HPV16, HPV18 and other high-risk HPV types by age group and cytology result; data are n (%)

–ve, negative; B/M, borderline or mild; Mod+, moderate or worse.

	No. of women	HPV16		HPV18, not HPV16		HPV16 and/or 18		HR-HPV, not 16/18	HC2 +ve		HC2 –ve
		Single HR-HPV	Multiple HR-HPV	Single HR-HPV	Multiple HR-HPV	No other HR	Other HR		HR-HPV	No HR HPV
Age
20–29	5150	280 (5.4)	219 (4.3)	77 (1.5)	78 (1.5)	374 (7.3)	280 (5.4)	750 (14.6)	1404 (27.3)	329 (6.4)	3417 (66.3)
30–39	7599	158 (2.1)	49 (0.6)	65 (0.9)	17 (0.2)	225 (3.0)	64 (0.8)	490 (6.5)	779 (10.3)	368 (4.8)	6452 (84.9)
40–49	6111	51 (0.8)	11 (0.2)	15 (0.2)	3 (0.1)	68 (1.1)	12 (0.2)	179 (2.9)	259 (4.2)	270 (4.4)	5582 (91.3)
50–64	5610	27 (0.5)	10 (0.2)	14 (0.2)	3 (0.05)	43 (0.8)	11 (0.2)	88 (1.6)	142 (2.5)	222 (4.0)	5246 (93.5)
Cytology
–ve	21,364	235 (1.1)	85 (0.4)	96 (0.5)	47 (0.2)	340 (1.6)	123 (0.6)	807 (3.8)	1270 (6.0)	940 (4.4)	19,154 (89.7)
B/M	2650	146 (5.5)	131 (4.9)	53 (2.0)	41 (1.6)	210 (7.9)	161 (6.1)	547 (20.6)	918 (34.6)	235 (8.9)	1497 (56.5)
Mod+	456	135 (29.6)	73 (16.0)	22 (4.8)	13 (2.9)	160 (35.1)	83 (18.2)	153 (33.6)	396 (86.8)	14 (3.1)	46 (10.1)
Total	24,470	516 (2.1)	289 (1.2)	171 (0.7)	101 (0.4)	710 (2.9)	367 (1.5)	1507 (6.2)	2584 (10.6)	1189 (4.9)	20,697 (84.6)

Of the CIN2+ lesions found before round 2, 108/329 (33%) and 83/225 (37%) respectively, were identified in high-grade and low-grade cytological abnormalities, which were HR-HPV +ve but types 16/18 –ve.

Viral persistence

Detection of persistent HPV infection is important because it confers a high risk of developing high grade CIN. True persistence implies type-specific persistence but in terms of clinical utility it is also relevant to assess the effect of a persistent HC2 +ve result.

The prognostic significance of persistent as opposed to cleared infection has already been referred to. ARTISTIC provided an opportunity to examine the effect of HPV persistence and clearance between rounds 1 and 2 in terms of screening outcomes. Persistence data are presented as HC2, HC2/LBA+ and type-specific persistence. HC2 persistence does not represent evidence of a true persistent infection because one type could have been replaced by another type, but in terms of clinical utility it represents persistence of the positive standard HPV screening test result as the means of risk assessment. HC2/LBA+ excludes possible false-positive HPV results because a type has been detected and type-specific persistence provides the clearest evidence of a truly persistent infection by a specific oncogenic type. HC2 alone would be the most sensitive measure of HPV persistence, and type-specific persistence the most specific.

Table 24 shows HPV persistence in women with all cytology grades in round 1 over a range of intervals up to 48 months after round 1. The results beyond 6 months are hardly affected by clinical intervention. Treatment of CIN clears HPV in the majority of cases, but these results are based on the next HPV result after round 1, which would precede treatment except in women with high-grade cytology in round 1, most of whom would be treated within 6 months. Persistence rates are similar irrespective of the measure of HPV positivity, including HC2 with or without confirmation by LBA and various type-restricted analyses including HPV16 detection, declining from over 80% within 6 months to about 40% at 18–24 months and remaining around 20–30% from 2 to 4 years.

Type-specific persistence rates are shown in Appendix 6, Table 68. These data are complicated by the fact that multiple infections are included but there were no major type-specific differences in persistence rates among the 13 high-risk types represented in the HC2 test.

Table 25 suggests similar HPV persistence rates in cytology –ve women in the concealed arm in whom there was no treatment intervention before round 2, but numbers are small for type-specific/restricted analyses.

The overall results (both arms combined) for HPV persistence in round 2 are shown in Table 26. This shows HPV persistence at 30–48 months of 28%, by HC2 but substantially lower rates for type-specific/restricted persistence of around 10–17%.

The impact of HPV persistence on cytology in round 2, in terms of odds ratios, is shown in Table 27 excluding those who had been treated in round 1. Abnormal cytology rates (borderline +) are consistently around 40% for women with persistence. The round 2 abnormality rate is only 4.8% in woman who had become HC2 –ve, but substantially higher (11–15%) in those who had become negative by the other more specific but less sensitive measures of HPV. It is therefore clear that women who have persistent HPV infection over 3 years are at considerable risk of developing an abnormality. When the results are compared between the arms, there are no striking differences (Table 28).

Cost analyses

For the cost analyses, resource-use events experienced by all 24,510 women in the trial were identified in the trial data sets and costed accordingly. The items of resources used by individual women covered the protocol-driven events and additional events of relevance, such as follow-up cervical cytology tests arranged by GPs for women who were negative in round 1, or colposcopy clinic follow-up visits for surveillance. The cost analyses generated a mean cost per woman, according to trial arm, for the trial itself and for other scenarios with alternative screening policies.

Trial costs

Cost comparisons between randomised arms were carried out for round 1 and for the ARTISTIC trial overall. The costed events conform with the trial protocol; that is, LBC screening alone for the concealed arm, and LBC screening and HPV testing for the revealed arm. The consolidated costs incorporated in the baseline cost analyses for the trial and the additional scenarios (2, 3a and 3b) are presented in Table 41. The table also identifies the numbers of resource use events in the concealed and revealed arms to which the costs were attributed. The time period covered by round 1 was the first 30 months after a woman’s valid round 1 sample, and the cost results for the full trial covered all resource events in both rounds 1 and 2 until 1 May 2007.

TABLE 41 - Summary of consolidated costs for resource-use items and resource usage in round 1 and the full trial according to randomisation

Resource item	Cost (£) 2006	Numbers of resource use events
		Round 1		Full trial
		Concealed (6124 women)	Revealed (18,386 women)	Concealed (6124 women)	Revealed (18,386 women)
Primary care
Sample taking consultation (adequate samples only)	14.21	7925	24,204	12,788	38,905
Cytology laboratory
Negative LBC sample	5.41	6513	19,842	11,030	33,511
Abnormal LBC sample	12.73	1412	4362	1758	5394
Adjustment for inadequate samples (2.5%)	0.48
Virology laboratory
HPV test on LBC sample	6.61		23,067		36,662
		NA		NA
Repeat HPV test (LBC sample taking consultation plus test)	17.98		963		1237
Colposcopy clinic
New attendance	300.51	320	1247	394	1531
Follow-up attendance	150.26	414	1607	460	1754
Inpatient treatment
Hysterectomy	4920	3	4	3	4

The mean costs per woman for round 1 and the full trial are presented in Table 42. These costs incorporate all cytology-, virology- and colposcopy-related events. In round 1, the mean (SD) cost per concealed woman was £55.97 (£177.87), (95% CI, £51.52 to £60.42); the mean (SD) cost for the revealed arm was significantly greater at £72.40 (£174.63), (95% CI, £69.88 to £74.92) (p < 0.001). The difference between the mean (SD) costs over the two screening rounds was equally significant: £77.10 (£186.99), (95% CI, £72.42 to £81.78) for the concealed arm compared with £99.96 (£187.41), (95% CI, £97.25 to £102.67) for the revealed arm (p < 0.001).

TABLE 42 - Mean cost per woman in the ARTISTIC trial covering screening and colposcopy-related events for round 1 and for the full trial

Trial arm		ARTISTIC trial mean cost per woman
		Round 1	Full trial
Concealed (n = 6124)	Mean cost (£)	55.97	77.10
	95% CI	51.52 to 60.42	72.42 to 81.78
	SD	177.87	186.99
Revealed (n = 18,386)	Mean cost (£)	72.40	99.96
	95% CI	69.88 to 74.92	97.25 to 102.67
	SD	174.63	187.41
Total (n = 24,510)	Mean cost (£)	68.30	94.25
	95% CI	66.1 to 70.5	91.9 to 96.6
	SD	175.58	187.56
p-value		p < 0.001	p < 0.001

Mean costs were calculated for age groups, banded in 5-year intervals, to observe differentials both across age ranges and between arms within age bands (Table 43).

TABLE 43 - Mean cost (£) per woman according to age group for round 1 and for the full trial

Age groups		ARTISTIC round 1		ARTISTIC full trial
		Concealed arm	Revealed arm	Concealed arm	Revealed arm
20–24	Mean cost	81.41	129.27	106.73	165.28
	95% CI	67.48 to 95.34	119.07 to 139.47	91.67 to 121.79	154.2 to 176.36
	n	647	1952	647	1952
25–29	Mean cost	84.16	99.56	110.50	129.44
	95% CI	69.21 to 99.11	90.93 to 108.19	94.19 to 126.81	120.01 to 138.87
	n	645	1945	645	1945
30–34	Mean cost	64.72	80.14	86.09	109.20
	95% CI	50.3 to 79.14	73.75 to 86.53	71.24 to 100.94	102.21 to 116.19
	n	927	2759	927	2759
35–39	Mean cost	47.78	69.26	70.92	95.81
	95% CI	39.9 to 55.66	66.77 to 71.75	62.11 to 79.73	88.25 to 103.37
	n	965	2982	965	2982
40–44	Mean cost	58.43	65.04	76.72	91.19
	95% CI	43.11 to 73.75	57 to 73.08	61.11 to 92.33	82.82 to 99.56
	n	828	2551	828	2551
45–49	Mean cost	43.30	54.66	62.65	81.86
	95% CI	35.76 to 50.84	49.5 to 59.82	54.36 to 70.94	76.2 to 87.52
	n	685	2032	685	2032
50–54	Mean cost	44.62	46.56	66.01	70.89
	95% CI	26.69 to 62.55	42.15 to 50.97	47.64 to 84.38	66.08 to 75.7
	n	623	1755	623	1755
55–59	Mean cost	30.89	43.46	46.40	66.74
	95% CI	24.39 to 37.39	38.77 to 48.15	39.62 to 53.18	61.68 to 71.8
	n	500	1473	500	1473
60–64	Mean cost	27.69	37.19	42.26	58.51
	95% CI	20.88 to 34.5	33.06 to 41.32	34.99 to 49.53	53.68 to 63.34
	n	304	937	304	937
Total	Mean cost	55.97	72.40	77.10	99.96
	95% CI	51.52 to 60.42	69.88 to 74.92	72.42 to 81.78	97.25 to 102.67
	n	6124	18,386	6124	18,386

In the revealed arm, women aged 20–24 years had the highest mean cost both in round 1 and in the full trial, a finding attributable to higher rates of HPV infection among the youngest women. In the concealed arm, the slightly older age group, 25–29 years, had the highest mean costs in both scenarios. Figure 16 demonstrates how, across the age spectrum, there was generally a systematic decline in mean costs from the youngest to the oldest age groups in the two arms. Observed deviations from this trend in the concealed arm were attributable to the high cost of a hysterectomy procedure.

FIGURE 16.

Mean costs for age groups: concealed arm vs revealed arm (in washed shades).

Costs for alternative screening policies

Based on the trial data, alternative scenarios for sequences of LBC and HPV tests and protocols for managing women were analysed. For ease of reference, the protocols for managing women according to the scenarios are described in Box 1, and illustrated in flow charts in Figures 17 and 18.

BOX 1 - Protocols for managing women following routine recall

Scenario 1

The ARTISTIC trial protocol for Round 1, and for the full trial (Figure 17).

Scenario 2

Primary screening with LBC; women with borderline or mild dyskaryotic results to be triaged by HPV testing on the LBC smear sample;

• those with a positive HPV result to be referred for colposcopy

• those with a negative HPV result would return to routine recall.

All women with moderate or worse cytology to be referred for colposcopy.

All women with normal cytology would return to routine recall.

Scenario 3a

Primary screening with HPV testing using QIAGEN HPV cervical samplers; women with a positive result to be advised to be resampled in primary care with an LBC cervical sampler for cytological examination;

• those with an abnormal LBC report to be referred for colposcopy

• those with a normal LBC report to have a second HPV test at 12 months; if the result is positive, they would be referred for colposcopy; if the result is negative, they would return to routine recall.

All women with a negative HPV result would return to routine recall.

Scenario 3b

Primary screening with HPV testing using LBC cervical samplers; women with a positive result to have the cytology examined promptly;

• those with an abnormal LBC report to be referred for colposcopy

• those with a normal LBC report to have a second HPV test at 12 months; if the result is positive, they would be referred for colposcopy; if the result is negative they would return to routine recall.

All women with a negative HPV result would return to routine recall.

FIGURE 17.

Flow chart for the ARTISTIC trial protocol. m, months.

FIGURE 18.

Flow charts for the protocols for alternative screening policies. m, months.

The trial protocol included a further permutation: revealed women with borderline or mild dyskaryosis in round 1 were to be recalled at 24 months even if their 12-month cytology and HPV results were negative (see Figure 2).

Before analysing the scenarios for alternative screening policies, the following adjustments were made to the cost inputs for the alternative screening scenarios, the resource use events having occurred within the first 30 months:

• For scenario 2, involving primary screening with LBC, and HPV testing to triage women with borderline or mild cytological changes. About 10% of all screened women would be affected. Hence the HPV test cost would be £10.11 based on a national workload of about 500,000 tests per year. As those women in the revealed arm with negative LBC reports would not be tested for HPV, they would have a £0 costs for colposcopy.

• For scenario 3, in which HPV testing is the method for primary screening, women who have a positive test are triaged with LBC. In the revealed arm of the trial, 15.6% of women had a positive HPV screening result. The triaged women with positive cytology would be managed according to the trial protocol.

There are two versions of scenario 3. In the first version (scenario 3a), the screening sample is taken with a cervical sampler designed by QIAGEN specifically for HPV testing. Women identified for triaging return to their GP surgery to have another sample taken with an LBC sampler. In the second version (scenario 3b), the screening sample is performed with an LBC cervical sampler, and so the sample’s cytology can be analysed without triaged women reattending. The costs of the HPV stand alone screening in scenario 3a was £20.08 (£5.87 for the test itself – see Table 33 – and £14.21 for the return visit by the triaged women for LBC sample taking). In scenario 3b, the costs for HPV testing were the same as in the trial, but the LBC laboratory costs would decrease overall as a result of the lower volume of samples identified for processing. Finally, we assumed that vials containing smear specimens identified for triage (either by LBC or HPV testing) would be transferred between laboratories using routine interlaboratory transport arrangements.

Cytological screening using LBC, followed promptly by an HPV test for women with borderline or mild dyskaryosis would be the most cost-saving strategy for a national screening programme. Table 44 shows that the mean cost for the LBC screening policy (scenario 2) of £43.98 per woman was significantly cheaper than the current practice of screening with LBC, as undertaken in the concealed arm of the trial (that is, a mean cost of £55.97 per woman) (p < 0.001).

TABLE 44 - Mean cost per woman for alternative screening policies

Trial arm and circumstances		Mean cost per woman for screening scenarios
		Scenario 2 LBC with HPV for borderline or mild triage	Scenario 3a HPV sampler screening followed by LBC sample triage	Scenario 3b HPV screening and LBC triage on same sample
Concealed as observed in the trial (n = 6124)	Mean cost (£)	55.97	55.97	55.97
	95% CI	51.52 to 60.42	51.52 to 60.42	51.52 to 60.42
	SD	177.87	177.87	177.87
Revealed according to proposed policy (n = 18,386)	Mean cost (£)	43.98	53.75	54.87
	95% CI	42.01 to 45.95	51.63 to 55.87	52.72 to 56.95
	SD	136.18	146.88	143.63
p-value		< 0.001	0.333	0.627

According to Table 45, for every age group, apart from the youngest of 20–24 years, a policy of LBC screening and HPV triage would be less costly than the current practice of LBC screening. It is worth noting, moreover, that the recently revised guidance on women’s eligibility for screening no longer extends to women under 25 years of age.

TABLE 45 - Mean cost per woman according to age group for alternative screening policies

The shaded cells indicate the least costly option for each age group.

Age groups		Screening policies
		Concealed arm round 1	Revealed arm round 1	Scenario 2 LBC with HPV for borderline or mild triage	Scenario 3a HPV sampler screening with LBC triage	Scenario 3b HPV screening with LBC sampler and triage
20–24	Mean cost	81.41	129.27	85.35	110.80	107.70
	95% CI	67.48 to 95.34	119.07 to 139.47	77.03 to 93.67	101.8 to 119.8	98.9 to 116.5
	n	647	1952	1952	1952	1952
25–29	Mean cost	84.16	99.56	65.19	83.06	82.00
	95% CI	69.21 to 99.11	90.93 to 108.19	58.16 to 72.22	75.38 to 90.74	74.52 to 89.48
	n	645	1945	1945	1945	1945
30–34	Mean cost	64.72	80.14	49.66	63.19	63.83
	95% CI	50.3 to 79.14	73.75 to 86.53	44.58 to 54.74	57.56 to 68.82	58.35 to 69.31
	n	927	2759	2759	2759	2759
35–39	Mean cost	47.78	69.26	41.27	50.03	51.73
	95% CI	39.9 to 55.66	66.77 to 71.75	34.98 to 47.56	43.43 to 56.63	45.22 to 58.24
	n	965	2982	2982	2982	2982
40–44	Mean cost	58.43	65.04	36.58	41.39	43.60
	95% CI	43.11 to 73.75	57 to 73.08	30.97 to 42.19	35.56 to 47.22	37.85 to 49.35
	n	828	2551	2551	2551	2551
45–49	Mean cost	43.30	54.66	29.10	33.86	36.21
	95% CI	35.76 to 50.84	49.5 to 59.82	25.76 to 32.44	30.29 to 37.43	32.76 to 39.66
	n	685	2032	2032	2032	2032
50–54	Mean cost	44.62	46.56	26.30	29.43	32.05
	95% CI	26.69 to 62.55	42.15 to 50.97	23.53 to 29.07	26.42 to 32.44	29.14 to 34.96
	n	623	1755	1755	1755	1755
55–59	Mean cost	30.89	43.46	23.91	25.02	27.92
	95% CI	24.39 to 37.39	38.77 to 48.15	22.69 to 25.13	23.04 to 27.22	26.49 to 29.35
	n	500	1473	1473	1473	1473
60–64	Mean cost	27.69	37.19	22.81	25.66	28.40
	95% CI	20.88 to 34.5	33.06 to 41.32	20.89 to 24.73	23.19 to 28.13	25.81 to 30.45
	n	304	937	937	937	937
Total	Mean cost	55.97	72.40	43.98	53.75	54.87
	95% CI	51.52 to 60.42	69.88 to 74.92	42.01 to 45.95	51.63 to 55.87	52.79 to 56.95
	n	6124	18,386	18,386	18,386	18,386
	SD	177.87	174.63	136.18	146.88	143.63