HPV testing compared with routine cytology in cervical screening: long-term follow-up of ARTISTIC RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 98/04/501. The contractual start date was in September 2013. The draft report began editorial review in April 2017 and was accepted for publication in November 2017. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Henry Kitchener reports grants from the University of Manchester during the conduct of the study. He is chairperson of the Advisory Committee for Cervical Screening (Public Health England) as well as of the National HPV Pilot Steering Group. Any views expressed in this report are those of the authors and not of Public Health England.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Gilham et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Data collection during the trial (2001–9)

ARTISTIC compared cytology with and without HPV testing among 24,510 women attending routine cervical screening in 2001–3. Over 60,000 HPV tests [hybrid capture 2 (HC2) with full HPV typing of those testing positive] were performed on routinely collected liquid-based cytology (LBC) cervical samples until September 2009. Methods are described elsewhere3,5 reporting that women were randomly allocated in a ratio of 1 : 3 between having their HPV results concealed (concealed arm, n = 6124) or revealed (revealed arm, n = 18,386). All women with abnormal cytology were recalled for retesting or referred to colposcopy according to national guidelines. 3 After March 2008, the laboratory became part of the Sentinel Sites project in which low-grade cytological abnormalities were triaged using HPV. Women were referred with borderline or mild cytology only if they tested positive for HPV; otherwise, they were returned to routine recall. In addition, women in the revealed arm with normal cytology who were hybrid capture 2 positive (HC2+) were recalled for repeat HPV testing, and those whose HPV infection persisted were referred to colposcopy.

Liquid-based cytology was carried out using ThinPrep^® (Hologic, Crawley, UK) and HPV testing using HC2 (Qiagen, Crawley, UK). A cut-off point of 1 relative light unit/mean control (RLU/Co) pg/µl was used to identify positive HPV samples, which were genotyped using at least one of three HPV typing assays. All HC2+ samples during rounds 1 and 2 were genotyped using the prototype Roche Line Blot Assay (Roche Molecular Diagnostics, Pleasanton, CA, USA). 3 This was replaced by the commercially available Linear Array assay and the PapilloCheck^® assay (Greiner Bio-one GmbH, Frickenhausen, Germany) during round 3. In addition, approximately two-thirds of archived HC2+ round 1 samples were also tested by the PapilloCheck assay and one-third of archived HC2+ round 2 samples were tested by the Roche Linear Array^®. In all three rounds, any of the 13 high-risk human papillomavirus (HRHPV) types detected by HC2 (HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68) that was detected by any of the assays was included in the analysis. All other HPV types were ignored in all analyses.

After the trial ended in 2009, all women returned to routine cytology every 3 years if they were aged < 50 years, and every 5 years if they were aged 50–64 years. We have published reports on baseline data,8,9 follow-up through two further screening rounds3–5,10 and various other HPV-related issues. 6,11–13 These analyses were based on follow-up to 2009 through two cytology laboratories in the Manchester area. Any women moving out of the area were lost, and only 60% of the ARTISTIC cohort had a cytology record between 2006 and 2009.

Further data collection (2010–16)

Definition of round 2 follow-up

In previous reports, round 2 was defined as the first adequate cytology between 26 and 54 months after entry, which included 10% with no adequate HPV test on their round 2 sample. In this report, round 2 is defined as the first adequate HPV test between 30 and 48 months after entry. Under this definition, 13,591 women had a round 2 test, on average, 3 years and 7 months after entry. In analyses beginning at round 2, 185 women with CIN3+ registration/histology before this date and six women whose flagging data indicated that they had cancelled from the NHSCR before this date were excluded.

Human papillomavirus clearance

To estimate and compare clearance rates of recently acquired and long-standing HPV infections (i.e. those already present at entry), clearance from the second HPV test in ARTISTIC (regardless of time since entry) was assessed. Women with any histology, regardless of result, between the second and third HPV tests were excluded in case the action of taking a biopsy affected the clearance rate. In addition, women with CIN3+ histology before the second HPV test were excluded. A total of 1026 women with at least three HPV tests whose second test was positive for HRHPV were included in the analysis. Each infection with a HPV type was considered independently, so women with more than one HPV type were counted more than once in the analysis. Clearance of each infection was calculated by time between the second and third HPV test and stratified by whether or not the same HPV type was present at entry (i.e. whether the infection developed since entry or was prevalent at entry).

Human papillomavirus recurrence

The recurrence of type-specific HPV infections was counted in women who had HRHPV infections at round 1 and subsequently tested HC2–. The time to ‘clearance’ (defined as testing HC2–) was stratified into ‘short term’, defined as clearance of < 24 months after round 1, and ‘long term’, defined as clearance 24–54 months after round 1. Recurrence in these two groups of women was defined as any subsequent HRHPV+ result with one or more genotype identified that was also present at entry. Under the protocol, genotyping was not carried out on HC2– samples, which implies that ‘clearance’ means that the HPV infection had become undetectable by HC2.

Type-specific human papillomavirus persistence

The HPV types found in round 2 (30–48 months after entry) were compared with those identified in round 1 so that round 2 infections could be classified as new or persistent. Women with a type-specific persistent infection with or without a new infection of a different genotype were classified as persistent.

Human papillomavirus status

In most other analyses, women were classified hierarchically into mutually exclusive groups: HPV 16 or HPV 18, any of HPV types 31, 33, 45, 52 or 58 (without 16 or 18), any other HRHPV, HC2+ with no HRHPV, or HC2–.

Calculation of cumulative risk for cervical intraepithelial neoplasia or cervical cancer

Cumulative CIN3+ risks were estimated by Kaplan–Meier methods. Subgroups were compared using the log-rank test. Cumulative risks were calculated from entry (round 1) and also from round 2 when incident and persistent infections could be distinguished. Women were censored at date of last cytology prior to hysterectomy according to call–recall data. All analyses were censored on 30 April 2015 to allow for late registration of CIN3 and cancer.

Kaplan–Meier analysis of interval-censored data

The aim of the cumulative risk analyses is to estimate the probability that CIN3+ would be detected by a test at a given time after entry to follow-up, beginning either at entry to the trial or at round 2. This raises several issues.

1. We ignore the possibility that CIN3 occasionally regresses, so cumulative risks may overestimate the prevalence that would be observed if the next test were 5 or 10 years later.

2. CIN3+ histology is backdated to the beginning of follow-up if the histology/registration occurred within a year of the beginning of follow-up. These cases are shown at time zero giving an initial step in the Kaplan–Meier curve. Some of these cases were not present at the time of the initial abnormal cytology or positive HPV test but developed during the period of repeat testing.

3. Later CIN3+ histology/registration dates are backdated to the first test in the preceding year, then further backdated to the mid-point of the interval between that test and the preceding test. For three cancers and a single CIN3 lesion with no recorded cytology within a year, the histology date is taken as the end of the interval. One further cancer was censored on ceasing screening and thus excluded.

This modified Kaplan–Meier treatment of interval-censored data, in which lesions are assigned to the mid-point of the screening interval where they became detectable, gives results similar to a parametric analysis (Professor Peter Sasieni, Queen Mary University of London, 2017, personal communication). The second-order error is that the interval mid-point is not the expected value of the time when the CIN3 would be detectable, which may be slightly earlier or later depending on the assumed model of HPV acquisition and CIN3 development.

Section 1: description of data

Eligible data

The cohort that participated in ARTISTIC consisted of 24,510 women aged between 20 and 64 years and provided an adequate cytology and HPV test at enrolment. Fourteen women did not have a valid NHS number and so 24,496 women were flagged with NHS Digital for cancer registration and death. These women were then linked to the National Screening Programme call–recall system, which yielded a final cohort of 23,888 after excluding three non-matches and 605 type-2 opt-outs (those opting out of their data being accessed for medical research) (Figure 1). The call–recall data were linked by NHS number to ensure correct linkage. At least one cytology date and result from the laboratory database matched the call–recall data in all but 53 women (99.8%). These 53 women were individually examined and most had only one or two tests identified through the laboratory. We assumed that these cytology dates had not been recorded correctly on the call–recall system. For those whose call–recall screening records matched the Manchester laboratory data during the trial (2001–9), it was not uncommon for some records to be inconsistent between the systems.

FIGURE 1.

The CONSORT (Consolidated Standards of Reporting Trials) flow of diagram showing women followed up for cancer registration and mortality through national registration, and for cytology through call–recall cervical screening records.

Follow-up via local histology laboratories to 2009 when the trial ended and via cancer registrations to April 2015 identified 509 cases of CIN3+, including 23 invasive cancers. Three cases of CIN3 were excluded owing to censoring at last cytology as no link was made to call–recall data. In addition, one invasive cancer was excluded as the woman’s screening record indicated that she had ceased screening after reaching the age of 65 years. Therefore, 505 cases of CIN3+ were included in the analysis (Table 1).

TABLE 1 - The CIN3 and invasive cancer cases

Three CIN3 were excluded owing to censoring at last cytology (2007, n = 1; 2010, n = 2).

One invasive cancer excluded owing to censoring at ceasing screening at age 64 years (after 2010).

	Histology only (2001–9), n (%)	Registration only (2001–15), n (%)	Both histology and registration, n (%)	Total number
CIN3	99 (20.4)	94a (19.3)	293 (60.3)	486
Invasive cancer		11b (47.8)	12 (52.2)	23
Year of diagnosis
2001–3	61 (21.2)	7 (2.4)	220 (76.4)	288
2004–6	21 (22.8)	11 (12.0)	60 (65.2)	92
2007–9	17 (23.6)	30 (41.7)	25 (34.7)	72
2010–15	0 (0.0)	57 (100.0)	0 (0.0)	57
Total	99 (19.4)	105 (20.6)	305 (59.9)	509

Among women who had not ceased screening due to age by 2016 (n = 17,602), 642 women aged 30–64 years were recorded as ceased on the call–recall system due to hysterectomy: 0.3% aged 30–39 years, 2.5% aged 40–49 years, 5.3% aged 50–59 years and 6.9% aged 60–64 years.

Natural history of human papillomavirus infection

Table 2 shows the prevalence of HPV infection at rounds 1 and 2 by age. HC2 was the primary assay used and only HC2+ samples were further genotyped. Overall, 4% of samples tested positive by HC2 but no HRHPV was detected. These have been assumed to be HRHPV– in the analysis shown in Table 3, which shows prevalence of HRHPV by age at round 2 among women who were HPV– at round 1, and prevalence rates of type-specific persistence and new infection among those who were HPV+ at round 1. The prevalence of new infection falls steeply with age in those women who were HPV– at round 1, from 19.7% at 20–24 years to 1.0% at 55–64 years. In contrast, among women who were HPV+ at round 1, the proportion with persistence at round 2 (17.9%) shows little or no age trend. Round 2 HPV prevalence in women who were HPV– at entry in ARTISTIC and in the POBASCAM (the Netherlands)14 and NTCC (Italy) trials14 are shown by age in Figure 2 and by HPV type in Figure 3, in which ARTISTIC data are restricted to the age range of the NTCC trial. The screening interval was 3 years in ARTISTIC and NTCC and 5 years in the POBASCAM trial. Figure 2 shows remarkably similar HRHPV rates in England, Italy and the Netherlands above the age of 33 years, but higher rates in ARTISTIC for younger women. Figure 3 shows the data by HPV genotype.

TABLE 2 - Age-specific HRHPV prevalence at round 1 (entry) and round 2 (30–48 months later)

Age (years)	Prevalence, n (%)					Total HRHPV+, n (%)
	HPV 16/18	HPV 31/33/45/52/58	Other HRHPV	No HRHPV detected (HC2+)	HC2–
Round 1 (entry)
20–24	408 (15.7)	254 (9.8)	204 (7.9)	167 (6.4)	1560 (60.2)	2593 (33.4)
25–29	264 (10.2)	200 (7.7)	145 (5.6)	106 (4.1)	1874 (72.4)	2589 (23.5)
30–34	190 (5.2)	200 (5.4)	131 (3.6)	164 (4.5)	2998 (81.4)	3683 (14.2)
35–39	103 (2.6)	104 (2.6)	120 (3.1)	150 (3.8)	3462 (87.9)	3939 (8.3)
40–44	58 (1.7)	56 (1.7)	70 (2.1)	128 (3.8)	3069 (90.8)	3381 (5.4)
45–49	26 (1.0)	42 (1.6)	43 (1.6)	110 (4.1)	2496 (91.9)	2717 (4.1)
50–54	25 (1.1)	29 (1.2)	28 (1.2)	90 (3.8)	2210 (92.8)	2382 (3.4)
55–59	19 (1.0)	10 (0.5)	22 (1.1)	67 (3.4)	1853 (94.0)	1971 (2.6)
60–64	11 (0.9)	7 (0.6)	11 (0.9)	47 (3.8)	1165 (93.9)	1241 (2.3)
Total	1104 (4.5)	902 (3.7)	774 (3.2)	1029 (4.2)	20,687 (84.5)	24,496 (11.4)
Round 2 (30–48 months)
20–24	36 (9.4)	37 (9.6)	22 (5.7)	29 (7.6)	260 (67.7)	384 (24.7)
25–29	73 (8.1)	57 (6.3)	33 (3.7)	69 (7.6)	673 (74.4)	905 (18.0)
30–34	66 (4.7)	45 (3.2)	30 (2.2)	65 (4.7)	1187 (85.2)	1393 (10.1)
35–39	25 (1.3)	51 (2.5)	40 (2.0)	73 (3.6)	1819 (90.6)	2008 (5.8)
40–44	14 (0.6)	39 (1.7)	19 (0.8)	72 (3.2)	2129 (93.7)	2273 (3.2)
45–49	19 (1.0)	23 (1.2)	11 (0.6)	48 (2.5)	1846 (94.8)	1947 (2.7)
50–54	8 (0.5)	14 (0.8)	13 (0.8)	48 (2.9)	1574 (95.0)	1657 (2.1)
55–59	9 (0.6)	6 (0.4)	8 (0.5)	32 (2.1)	1464 (96.4)	1519 (1.5)
60–64	6 (0.6)	5 (0.5)	7 (0.6)	29 (2.6)	1053 (95.7)	1100 (1.6)
65–69	0 (0.0)	0 (0.0)	3 (0.7)	8 (2.0)	394 (97.3)	405 (0.7)
Total	256 (1.9)	277 (2.0)	186 (1.4)	473 (3.5)	12,399 (91.2)	13,591 (5.3)

TABLE 3 - Women who were HRHPV+ at round 2 by age at round 2. Results are shown separately for those HRHPV– at entry and those HRHPV+ at entry

Persistence indicates the same HRHPV type at both rounds.

Age (years) at round 2	Infection status at entry
	HRHPV–		HRHPV+
	Number of women	New HRHPV at round 2, n (%)	Number of women	New HRHPV at round 2, n (%)	Persisting HRHPV type at round 2, n (%)
20–24	239	47 (19.7)	145	20 (13.8)	28 (19.3)
25–29	648	77 (11.9)	257	37 (14.4)	49 (19.1)
30–34	1119	60 (5.4)	274	21 (7.7)	60 (21.9)
35–39	1814	68 (3.8)	194	16 (8.2)	32 (16.5)
40–44	2138	48 (2.3)	135	7 (5.2)	17 (12.6)
45–49	1869	37 (2.0)	78	8 (10.3)	8 (10.3)
50–54	1602	24 (1.5)	55	0 (0.0)	11 (20.0)
55–59	1482	14 (0.9)	37	3 (8.1)	6 (16.2)
60–64	1079	12 (1.1)	21	3 (14.3)	3 (14.3)
65–69	396	1 (0.3)	9	0 (0)	2 (22.2)
All women	12,386	388 (3.1)	1205	115 (9.5)	216 (17.9)

FIGURE 2.

New infections: prevalence of HRHPV at round 2 in the ARTISTIC, NTCC (Italy) and POBASCAM (the Netherlands) trials in women who were HRHPV– at entry.

FIGURE 3.

New infections: prevalence (%) at round 2 of each HRHPV type in the ARTISTIC, NTCC and POBASCAM trials in women who were HRHPV– at entry.

The proportion of HPV infections clearing increased by time interval to testing (Table 4). Clearance of new HPV infections (i.e. infections that had developed in the preceding screening interval) was higher than persisting HPV infections (i.e. those that were present at the previous screening round). Clearance of HPV 16 infection was lower than other infections and approximately 40% of new HPV 16 infections were still persisting in women tested over 3 years.

TABLE 4 - Clearance of new and persistent HRHPV infections

Next HPV test following first detection of 634 HRHPV types in 504 women who did not have that HRHPV type at entry.

Second HPV test for 609 HRHPV types detected at entry in 522 women.

Infections	Proportion of infections clearing by time to next HPV test, n/N (%)
	< 1 year	1–2 years	2–3 years	≥ 3 years
New HPV 16 infections	20/51 (39.2)	18/26 (69.2)	7/13 (53.9)	7/12 (58.3)
Other new infections	142/249 (57.0)	107/144 (74.3)	53/65 (81.5)	66/74 (89.2)
All new infectionsa	162/300 (54.0)	125/170 (73.5)	60/78 (76.9)	73/86 (84.9)
HPV 16: present at entry	20/94 (21.3)	12/22 (54.6)	9/21 (42.9)	7/12 (58.3)
Other infections: present at entry	91/243 (37.5)	52/104 (50.0)	44/65 (67.7)	36/48 (75.0)
All infections present at entryb	111/337 (32.9)	64/126 (50.8)	53/86 (61.6)	43/60 (71.7)

Short-term recurrence was assessed among 405 women who were HRHPV+ at entry and who tested HC2– less than 2 years after entry (Table 5). Of these, 27 (6.7%) women had a recurrent infection of the same HPV type present at entry. Long-term recurrence was assessed among 415 women who were HRHPV+ at entry and tested HC2– at least 2 years later. Of these, only five (1.2%) women had a later recurrence or reinfection of the HPV type present at entry.

TABLE 5 - The HPV type-specific recurrence following a negative HC2 test of HPV infections that were detected at entry

Recurrence	Months from HC2– test to next HPV test
	0	6	12	24	36	≥ 48
Short-term recurrence (following a HC2– test < 24 months after entry)
Number of women	26	81	122	118	29	29
Cumulative number of women	26	107	229	347	376	405
Cumulative recurrent infections (%)	2 (7.69)	7 (6.54)	14 (6.11)	21 (6.05)	22 (5.85)	27 (6.67)
Long-term recurrence (following a HC2– test 24–54 months after entry)
Number of women	11	54	81	77	175	17
Cumulative number of women	11	65	146	223	398	415
Cumulative recurrent infections (%)	0 (0.0)	0 (0.0)	1 (0.68)	1 (0.45)	5 (1.26)	5 (1.20)

Section 2: the long-term protection of a negative human papillomavirus test and, hence, the safe screening interval at different ages

Table 9 shows cumulative CIN3+ risks in women testing HC2– at rounds 1 and 2 by age. The 10-year risks decrease from 1.10% [95% confidence interval (CI) 0.69% to 1.77%] in women aged 20 –25 years at entry to 0.08% (95% CI 0.03% to 0.20%) in women aged ≥ 50 years at entry. The 10-year CIN3+ risk in all HPV– women aged 25–39 years was 0.40% (95% CI 0.28% to 0.56%) and significantly higher than the risk in women aged ≥ 40 years at entry (0.11%, 95% CI 0.06% to 0.20%; p < 0.0001). Table 10 shows cumulative risks from round 2 in those HC2– at both time points compared with those women who had cleared an infection present at entry. Risks were based on very small numbers, but were approximately three times higher among those women who cleared an infection at entry. Presumably this is confounded with sexual activity and, hence, risk of acquiring a new HPV infection. Figure 4 and the bottom panel of Table 9 show a similar CIN3+ risk 10 years after a negative HPV test (0.31%, 95% CI 0.18% to 0.49% in the revealed arm) and 3 years after a negative cytology test (0.30%, 95% CI 0.23% to 0.41% in the concealed arm). These data support a much longer screening interval after a negative HPV test than after a negative cytology test.

FIGURE 4.

Cumulative CIN3+ risk (%) following a negative HPV test or normal cytology at entry.

Section 3: the role of human papillomavirus typing

The increasing availability of partial genotyping allows risk-based stratification in an organised screening programme. Four of the six sites are utilising partial typing HPV tests (Roche COBAS and Abbott Realtime assays) in the English pilot study, which both allow identification of HPV 16 and HPV 18 infections. Table 11 and Figure 5 show that the cumulative risk for women with HPV 16 is much higher than for any other genotype. Women with HPV 16 constituted 22% of all HC2+ infections, but 57% of CIN3+ occurred among this group of women, giving a 10-year cumulative CIN3+ risk of 29.8% (95% CI 26.8% to 33.0%). HPV 18 was a much less common type among ARTISTIC women, constituting 9% of all HC2+ infections (including 2% who also had HPV 16) and the 10-year cumulative risk did not differ from the group of types including 31, 33, 45, 52 and 58. It is clear that assays that could identify these five types in addition to HPV 16 and 18 would be beneficial for a risk-based referral strategy.

FIGURE 5.

Cumulative CIN3+ risks by HPV detection from entry (483 CIN3, 23 invasive cervical cancers in 24,500 women).

Table 12 and Figure 6 show the risks following two tests approximately 3 years apart. The 10-year cumulative CIN3+ risk following a new HPV infection is low at 3.4% (95% CI 2.1% to 5.4%). The 10-year CIN3 risk from round 2 was no higher in those women who were HPV+ with another type at round 1 (2.7%, 95% CI 0.9% to 8.0%) than in those women who were HPV– at round 1 (3.6%, 95% CI 2.2% to 6.0%).

FIGURE 6.

Cumulative CIN3+ risks following round 2 for persistent and newly acquired HPV infection (82 CIN3+ in 13,591 women).

The highest risks were associated with type-specific persistent infections that, overall, conferred a 10-year cumulative risk of 20.4% (95% CI 15.65% to 26.45%). HPV 16 again showed the highest risks and the 10-year risk following a new infection reached 7.3% (95% CI 3.7% to 14.1%), suggesting that any women with HPV 16 detected could be referred immediately. Of the 331 women with double-positive HRHPV tests (see Table 3), 115 (35%) were positive with a new HPV type at the second test and 216 (65%) was type-specific persistent. The proportion of new infections was highest in younger women (43% in women in their 20s) and decreased to just 23% of double-positive infections in women aged ≥ 50 years.

Section 4: triage of human papillomavirus-positive women, particularly the interval to retesting for human papillomavirus-positive women

Table 13 shows cumulative CIN3+ risks by HPV type and cytology at entry. Referring moderate or worse cytology to immediate colposcopy is clearly a highly effective strategy for identifying those women at highest risk of disease (52% by 5 years). Among HC2+ women, the 10-year cumulative CIN3+ risks in those with negative and borderline/mild cytology were estimated to be 4.1% (95% CI 3.3% to 5.0%) and 10.3% (95% CI 8.7% to 12.2%), respectively. These risks vary by HPV genotype, and, hence, a partial genotyping assay could identify those at highest risk (e.g. HPV 16 or 18). Table 14 is restricted to the 1116 women who were HC2+ and cytology negative and had a follow-up sample taken 30–48 months later (round 2). The current UK protocol for HC2+ cytology-negative women is to repeat HPV and cytology testing at 12 and 24 months and refer all those still HPV+ at 24 months. Table 14 shows these women under a protocol of repeat testing after a 3-year recall interval. This excludes 10 CIN3+ that were diagnosed in cytology-negative HPV+ women in the revealed arm who were diagnosed before the round 2 test, but under a 3-year recall protocol would probably have been diagnosed at round 2. Analysis restricted to the concealed arm would avoid this bias. Only 277 out of the 1120 women in Table 14 were in the concealed arm and the results were consistent between the arms; therefore, they were pooled to maximise numbers. The majority of women cleared their infection by round 2, with 72% testing HC2– and a further 7% who remained HC2+ in whom no HRHPV was detected. A small proportion (1.3%) of women developed moderate or worse cytology by round 2, most (9 out of 15) of whom had CIN3+ immediately diagnosed. Out of 220 women who remained HPV+ with normal, borderline or mild cytology, 38% were infected with a new HPV type and 62% were type-specific persistent. Among the new infections, the 10-year cumulative CIN3+ risk remained low (3.7%, 95% CI 1.2% to 11.0%). The risks were slightly higher for those women with borderline or mild cytology, but there was insufficient power to detect a significant difference. Further genotyping of these samples to determine type-specific persistence would refer only those women at highest risk of developing CIN3+ to colposcopy.

Tables 15 and 16 show cumulative CIN3+ risks following new and persisting HPV infections at round 2 in women with negative and borderline/mild cytology, respectively. Similar patterns are seen, with the highest risks in women with persisting infections and in those with HPV 16 infections.

Section 5: optimal ages at starting and stopping screening

Table 17 and Figure 7 show that the cumulative CIN3+ risk 5 years after round 2 in women with type-specific persisting infection is similar regardless of age before the age of 40 years (23.8%, 95% CI 18.2% to 30.9%), but significantly lower in women aged ≥ 40 years (6.8%, 95% CI 2.3% to 19.7%; p = 0.02). Persistent HPV 16 infections account for 44% of the type-specific infections in those women aged < 30 years, but only 11% among women aged ≥ 40 years.

FIGURE 7.

Cumulative CIN3+ risks following round 2 by age in women with type-specific persistence of any HRHPV.

Risk-based screening and cancer prevention

The UK National Screening Programme will adopt HPV screening and triage protocols that strike an acceptable balance between cancer risk and the cost and patient inconvenience of excessive screening and referral rates and unnecessary treatment. Acknowledging that there is no such thing as zero risk, Castle et al. 18 proposed a risk-based strategy for cervical screening based on widely accepted practice, and suggested that women should be returned to routine screening if they have a < 2% CIN3+ risk, recalled earlier than the routine screening interval if they have a 2–10% risk, and referred to immediate colposcopy if they have a > 10% risk. These specific cut-off points may be considered arbitrary and do not consider the relationship between CIN3 and cancer risk. The aim of screening is to prevent invasive cancer and not CIN3; therefore, the acceptable cumulative CIN3+ risk when a woman is next screened depends on the ratio of cancer risk to CIN3 risk at the next screen. This ratio depends mainly on the delay between time of development of CIN3 during the screening interval and its treatment when it is detected at the next HPV test or after subsequent triage processes.

Many CIN3 cases develop soon after HPV acquisition and most develop within a few years. 19 Therefore, the CIN3 incidence rate in women who were HPV– at their last test is approximately proportional to their age-specific incidence rate of new HPV infection (with a lag of a year or two), and, hence, proportional to the overall population HPV prevalence (Figure 8), as most prevalent infections are newly acquired and transient even in older women (see Table 3).

FIGURE 8.

Age-specific HRHPV prevalence at round 2 by age at round 2 overall and in women who were HRHPV– at round 1, and cumulative 10-year CIN3+ risk by age at round 1 in women who were HC2– at round 1. (HRHPV prevalence rates at round 2 by age at round 1 in women who were HRHPV– at round 1 in 5-year age groups from 20–24 years to 55–59 years were 14.7%, 8.1%, 3.6%, 2.9%, 2.2%, 1.5%, 1.5% and 0.8%.)

The cumulative CIN3 risk at next screen is also roughly proportional to the screening interval (Figure 9). The cervical cancer incidence rate in women with untreated CIN3 appears to be fairly constant at about 1% per year, independent of age,1 so the cancer risk at next screen should be roughly proportional to the square of the screening interval. In unscreened populations with fairly constant HPV infection rates up to middle age, cervical cancer incidence should therefore rise linearly from age at first intercourse. From 10 to 30 years after first intercourse, the observed quadratic relationship corresponds closely to a linear increase with a diagnostic lag of 7.5 years. 20 In primary screening, women who test HPV– acquire new infections and develop resulting CIN3 throughout the interval to next test. In contrast, in the further interval before retesting when HPV+ women are being triaged, most CIN3 lesions will have developed long before or soon after their positive HPV test. Therefore, some cancers will be present at the screening visit when HPV is detected, and the cancer risk will further increase at a roughly constant rate throughout the triage process. This simplified and speculative model ignores several relevant factors, including the lag from HPV acquisition to CIN3 development and the rate of progression to detectable cancer, but it illustrates the weakness of screening and triage protocols based only on CIN3+ risk. An undefined model of cancer risk underlies the focus on CIN3 that informs cervical screening and triage protocols. An explicit model based on epidemiological knowledge should perhaps be developed and refined as an alternative basis for cost–benefit evaluation of the details of HPV screening and triage protocols, as it is unlikely that their effects on cancer incidence can be observed directly. Even the simple comparison of HPV testing against cytology required the pooled results of trials with a total of 176,000 randomised women to detect the effect on cancer incidence. 6

FIGURE 9.

Age-standardised prevalence (%) of CIN2+ during ARTISTIC by time since previous normal cytology to entry, and by time since cytologically normal entry to next test. Women whose previous cytology was abnormal or < 2 years before entry are excluded.

The long-term protection of a negative human papillomavirus test and, hence, the safe screening interval at different ages

The 10-year CIN3+ rate among women who were HPV– at entry to ARTISTIC is only 0.1% in those aged ≥ 40 years, suggesting that screening intervals for HPV– women aged 40–49 years might be extended even further than the 6 years previously suggested. 4 Our data are consistent with the long-term risks reported from women participating in the POBASCAM trial in the Netherlands that has led to the recommendation for the Dutch HPV screening programme to adopt a 10-year screening interval in women aged ≥ 40 years. 21 Many HPV– women have had an earlier infection, so this very low long-term risk after a single negative HPV test suggests that a HPV infection that has become undetectable rarely recurs and then progresses to CIN3. Frequent HPV testing, therefore, greatly increases the number of transient infections that are detected without increasing sensitivity for CIN3 detection.

As noted above, if most CIN3 cases (and a large proportion of CIN2 cases) persist, then their prevalence will increase almost linearly with increasing screening interval. This was seen in routinely screened women in Manchester in the 1990s, in whom CIN3 prevalence increased almost linearly with longer intervals since last normal cytology. 22 Age-standardised CIN2+ prevalence in ARTISTIC shows this pattern at round 2 (Figure 9), with an annual increase of about 0.2% per year since entry in women with a cytologically normal entry test. Figure 9 shows that CIN2+ prevalence at entry increased at a similar rate with increasing interval from the previous normal cytology to entry, with the addition of about 1.2% reflecting the CIN2+ prevalence missed by previous screening rounds that was detected at entry to ARTISTIC.

The role of human papillomavirus typing

The HPV+ women with borderline/mild cytology as well as those with negative cytology might be further triaged by HPV type to identify those who can be recalled later. A HPV test that identifies HPV types 16, 18 and 31/33/45/52/58 would enable this moderate-risk subgroup, which constitutes 58.4% of HPV+ women with borderline/mild cytology, to be referred immediately or followed up within (say) 1 year. See Table 13, which shows the 5-year CIN3+ risks of 10.9%, 4.1% and 2.9% with negative cytology, and 20.9%, 10.8% and 9.4% with borderline/mild cytology. Borderline/mild cytology with other HRHPV types (35, 39, 51, 59 and 68) has a 5-year CIN3+ risk of 3.5% and might be recalled 3 years later for a repeat HPV test. Immediate referral or early retesting of women with HPV 16 was recently recommended in the Netherlands23 and proposed more than 10 years ago in the USA,24 but the cost–benefit balance of more complex triage protocols based on full HPV typing following cytology triage deserves further consideration, particularly for vaccinated women. Table 13 shows that most of the CIN3+ risk in HPV+ women with abnormal cytology is present at, or soon after, HPV detection and increases by about only 2% from 2.5 to 10 years irrespective of cytology.

Triage and referral policy in relation to primary screening interval

Figure 9 confirms that the marked reduction from round 1 to round 2 in the proportion of HC2+ women aged ≥ 25 years who had moderate+ cytology (11.2% and 3.2%, respectively; see Table 6) reflects the elimination of long-standing high-grade disease that had been missed at previous routine cytology. The round 1 results, thus, represent what might be seen in primary HPV testing with a very long screening interval, whereas round 2 shows results after an interval of about 3 years. However, the overall CIN3 risks in HRHPV+ women were almost the same after stratifying by cytology, with 5-year CIN3 risks for normal, borderline/mild and moderate+ cytology of 3.1%, 9.5% and 52.4%, respectively, in round 1, and 2.5%, 8.4% and 44.7%, respectively, in round 2. This suggests that recall policy following cytology triage of HPV+ women need not depend on the primary screening interval.

Triage and referral policy in relation to type-specific human papillomavirus persistence

Table 12 shows that the 10-year CIN3 risk for women with a new HPV type at round 2 was only 3.4% (95% CI 2.1% to 5.4%). The risk was no higher in those whose HPV infection at round 1 had cleared (2.7%, 95% CI 0.9% to 8.0%) than in those who were HPV– at round 1 (3.6%, 95% CI 2.2% to 6.0%). In contrast, the 10-year CIN3 risk in women with type-specific persistence at round 2 is six times greater than for a new infection of the same type. Among 331 women with HRHPV at both rounds, 58% had the same type (10-year CIN3 risk 20.4%, 95% CI 15.6% to 26.4%). The 10-year CIN3 risk in the remaining 42% who had cleared the original type and acquired a new type was only 2.7%, so these women could safely be recalled 3 years later when most of their infections will have cleared. This large difference in risk (p < 0.0001), and, hence, the potential efficiency of further triage, is diluted with HPV tests that do not identify all 13 high-risk types. The price of different HPV tests and the disadvantages of more complex triage and referral protocols will determine whether or not the advantages of full HPV typing outweigh the costs.

A triage protocol in which women are recalled repeatedly until their CIN3 risk exceeds (say) 10% implies a cycle of repeated testing that ends only in HPV clearance or referral, and some limit must be set to detect and treat occult CIN3 that has regressed cytologically. A HPV infection that has persisted for 5 years should perhaps be referred for random cervical biopsy. In a prospective study in the Netherlands, six out of the eight women with untreated CIN who regressed to normal cytology and colposcopy but remained HRHPV+ over several years had occult CIN3 detected by random biopsy. 25

We excluded women with any recorded histology in calculating HPV clearance rates, as biopsy can induce clearance. 26 This might lead to an overestimate of clearance but, as the clearance rates for new and persistent HRHPV infections shown in Table 4 are similar to those reported by Nobbenhuis et al. 27 in women followed up without intervention, any bias seems to be small.

Triage of human papillomavirus-positive women

If the screening interval is increased, the time between appearance and diagnosis of CIN3 and, hence, the risk of progression to cancer will also increase, so the triage protocol must minimise the delay in referring those at highest risk to colposcopy. Nonetheless, the evidence from round 1 of ARTISTIC (see Table 13) indicates that the triage protocol being piloted in England could be too conservative, at least for HPV+ women with negative cytology. The risk of developing CIN3 within 5 years in HPV+ women is 52.4% for moderate/severe cytology, 9.6% for borderline/mild cytology and only 3.1% for normal cytology (see Table 13). The suggestion that a CIN3+ risk of about 10% is a reasonable threshold for immediate colposcopy18,28 would, therefore, imply rapid referral for HPV+ women with any abnormal cytology except perhaps the minority of women with borderline/mild cytology who have genotypes with the lowest associated risks. The minimal increase in the risk of developing cancer by retesting women with borderline/mild cytology rather than referring them directly to colposcopy has always been accepted, and it is not clear that this risk/benefit balance should be abandoned when they are found to have HPV. In round 1 of ARTISTIC, the proportion of HPV+ women with moderate + cytology was 12% below age 50 years and at age 50 years or older. By round 2 of ARTISTIC, only 5% of HPV+ women aged < 50 years and 1% of women aged ≥ 50 years had moderate/severe cytology. The majority of HPV+ women in a population regularly screened by HPV testing will, thus, have negative or borderline/mild cytology.

Triage: women with normal cytology

Of the women who were HC2+ with normal cytology at entry to ARTISTIC, 72% had cleared their HPV infection within 3 years (see Table 14). Only 1.3% (15 out of 1116) of women had moderate or worse cytology by round 2 (of whom, nine were diagnosed with CIN3). Fourteen per cent of women remained HPV+ with normal cytology, but approximately 60% of these (97 out of 158) had type persistent HPV and approximately 40% (61 out of 158) had new HPV types. The subsequent elevated cumulative risk was among those with type-specific infections. This also applied to those with borderline/mild cytology at the follow-up, but this was based on small numbers. This implies that the current policy in the pilot sites (repeat testing after 1 year and referral to colposcopy after 2 years if women are still HPV+ with normal cytology) is referring a large proportion of women to unnecessary testing and colposcopy. Recall after 2 or 3 years then referring only those developing moderate+ cytology and those with type-specific persistence may be a better strategy.

Triage: women with borderline/mild cytology

Among HPV+ 24- to 64-year-old women presenting with borderline/mild cytology at entry, 29% had cleared their HPV infection within 1 year and an additional 19% remained HPV+ but became cytologically normal. Those with HPV 16 with 5-year CIN3+ risk of 20.5% should clearly be referred immediately, but inviting all HPV+ women with borderline or mild cytology to return for a repeat test at 1 year would substantially reduce immediate referrals and could be a better strategy.

Ages at starting and stopping screening

Study limitations

Further research

The National Institute for Health Research has funded a further extension, under which follow-up for CIN2 diagnoses (which are managed clinically in the same way as CIN3 but are not recorded centrally) will be updated through local histology laboratories. The sample bank, augmented by further samples from women in the cohort who are screened between 2018 and 2022, will be used to retest samples from women who develop CIN3 or cancer and matched controls. Interim results for CIN2+ as well as CIN3+ will be provided to the NSC in order to inform its recommendations on screening intervals for HPV– women and triage protocols following HPV detection. The same data would be available in the future from national routine screening records if these were linked to histology records to capture CIN2 diagnoses as well as to national cancer registration. The establishment of a larger national bank of samples from routinely screened women would provide much larger numbers of samples from women who will develop cervical cancer for retesting by more sensitive PCR and with alternative screening or triage assays such as DNA methylation.

Conclusions

A screening interval of up to 10 years following a negative HPV test might be considered for women aged ≥ 40 years and perhaps for younger women following a negative HPV test. Immediate referral of all HPV+ women with mildly abnormal cytology, as adopted in the English pilot study of primary HPV screening, may not be essential and recall intervals of 1 year for high-risk subgroups and 3 years for the remainder could be considered. The risk is largely confined to women with a persisting type-specific HPV and a 10-year interval will identify them efficiently. The possibility of a small increase in risk of progression to invasive cancer between screening rounds and perhaps some increase in the proportion of women who fail to attend regularly must be weighed against the advantages for patients and reduction of NHS costs with longer screening intervals.

Routine HPV testing and follow-up to HPV clearance will divide the population into the low-risk HPV– majority and those with type-specific persistent HPV infection among whom virtually all cervical cancers will arise. Those whose infections do not disappear after repeated testing may consider curative treatment irrespective of cytological or colposcopic findings, particularly when they reach the age at which routine screening ends.

The ARTISTIC trial study group

Contributions of authors

Ms Clare Gilham (Medical Statistician) carried out the data management and statistical analysis and drafted the report.

Dr Alexandra Sargent (Clinical Scientist) conducted the virology testing for ARTISTIC.

Professor Henry C Kitchener (Professor of Gynaecological Oncology) was the clinical principal investigator for ARTISTIC and made critical comments on the report.

Professor Julian Peto (Professor of Epidemiology) contributed to the design of the study and the analysis and drafted the report.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.