Keywords Cytology-screening, statistics-epidemiological surveys.

DOI: 10.1111/j.1471-0528.2007.01467.x www.blackwellpublishing.com/bjog Epidemiology An examination of the role of opportunistic smear taking in the NHS cervical screening programme using data from the CSEU cervical screening cohort study RG Blanks, SM Moss, DA Coleman, AJ Swerdlow Cancer Screening Evaluation Unit, Institute of Cancer Research, Sutton, Surrey, UK Correspondence: Dr RG Blanks, Cancer Screening Evaluation Unit, Sir Richard Doll Building, Institute of Cancer Research, Cotswold Road, Sutton, Surrey SM2 5NG, UK. Email roger.blanks@icr.ac.uk Accepted 19 June 2007. Published OnlineEarly 5 September 2007. Objective The objective of this study was to study the prevalence of opportunistic smear taking in the NHS cervical screening programme between 1999 and 2003 and the relationship of this to screening interval policy. Design A cohort study of nearly 2millionwomen,with data on screening at ages 20 64 years from 1988 to 2003 has been constructed. Data from 1999 to 2003 have been used in this analysis. Screening episodes have been divided into those where the primary smear was initiated by the national call/recall system (invitational), normally at3-or5-yearlyintervals,and those initiated by the GP or woman (opportunistic). Opportunistic smears were further classified as routine (occurring within 6 months of 3 or 5 years) or sporadic (occurring at other times). Setting NHS cervical screening programme. Population Four Health Authorities in England (now Primary Care Trusts) with supplementary studies on national data. Methods Screening episodes have been defined. All episodes start with a primary smear defined as being invitational or opportunistic in origin. Main outcome measure Proportion of primary smear that were invitational or opportunistic. Results In total, 72% of incident screen primary smears were invitational and 28% were opportunistic. The proportion of opportunistic primary smears was 17 and 43% in 3- or 5-yearly screening policy areas, respectively, resulting in a considerably reduced average screening interval for women aged 20 64 years in 5-year policy areas. Conclusion The NHS cervical screening programme is strongly influenced by opportunistic smear taking. In particular, nominally 5-year policy areas experienced much higher levels of opportunistic smear taking than those with a 3-year policy, causing the average interval in the 5-year areas to be much shorter than the policy would suggest. In future, with the change in national policy for inviting women aged 25 49 years every 3 years and those aged 50 64 years every 5 years, the level of opportunistic smear taking, particularly in the older group of women, needs to be carefully monitored. A lack of compliance may result in greater than predicted costs with little or no additional cancer prevention. Keywords Cytology-screening, statistics-epidemiological surveys. Please cite this paper as: Blanks R, Moss S, Coleman D, Swerdlow A. An examination of the role of opportunistic smear taking in the NHS cervical screening programme using data from the CSEU cervical screening cohort study. BJOG 2007;114:1408 1413. Introduction Background Cervical screening started in the UK in the mid-1960s. Although many women were having regular smear tests by the mid-1980s, there was concern that those at greatest risk were often not being tested and that not all those who had positive results were being followed up and treated effectively. As a result, the NHS cervical screening programme was set up in 1988 when the Department of Health instructed all Health Authorities to introduce computerised call/recall systems and to meet certain quality standards. The NHS call/recall system (the Exeter system ) invites women who have registered with a GP for screening and also keeps track of any follow-up investigation and when appropriate recalls women for screening at 3- or 5-year 1408 ª 2007 The Authors Journal compilation ª RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology

Opportunistic smears in NHS cervical screening intervals. Until 2003, the programme invited women between the ages of 20 and 64 years to be screened at a routine interval of either 3 or 5 years, depending on the Health Authority (now Primary Care Trust [PCT]) policy, but as a result of evidence from the UK programme, 1 the policy has changed to 3-yearly screening between the ages of 25 and 49 years and 5-yearly screening between the ages of 50 and 64 years. National screening programmes can be categorised as being organised (invitational) or opportunistic. Opportunistic smears are smears taken outside a call/recall invitation system and initiated by the woman herself or by her GP. A complexity of the UK programme is the role of these opportunistic smears taking place within an invitational screening programme, and as a result, the UK and some other national programmes have been called partially invitational. 2 High levels of opportunistic smear taking could reduce the average screening interval in practice compared with the local policy. We are undertaking a large cohort study to investigate further the effects of different screening intervals, the need for screening beyond the age of 50 years in women with a history of negative smears and the role of opportunistic smear taking. This paper outlines the design and construction of the cohort and presents initial results relating to the role of opportunistic smear taking. Methods Construction of cohort Study data were collected from four Health Authorities (now PCTs) in England. Two had a policy of 5-yearly screens, and two had a policy of 3-yearly screening. The population characteristics of the Health Authorities were broadly similar (Table 1) so that differences in outcome measures are most likely to reflect screening policy rather than population differences. The four Health Authorities agreeing to take part Table 1. Characteristics of Health Authorities included in the study cohort (2000/01 data) ONS area classification 1999 Townsend deprivation score* Interval policy ESI** formed two pairs of adjacent Health Authorities in different parts of England, one with a 3-year policy and one with a 5-year policy. Data from 1999 showed that approximately half of all Health Authorities in England had a 3-year policy and half a 5-yearly (or mixed 3 and 5 year) policy. We obtained a download of the records of all women who had been invited to screening from each of the four Health Authorities since the start of computerisation of records. The final cohort contains records of women screened between 1 January 1988 and 31 December 2003. In this paper, the present analysis has been restricted to smears taken between 1 January 1999 and 31 December 2003 inclusive because of problems with data quality and variations in coding between regions in the data before 1999 and also to get the best and most current estimates of opportunistic smear taking. Definitions To analyse the extent of opportunistic screening, it is necessary to define several parameters to differentiate types of smear within the study dataset. To obtain a consistent and pragmatic analysis, we have recognised the episodic nature of smear taking, and the analyses are based around smears that initiate new screening episodes rather than including all individual smears within an episode. Episodes and primary or secondary smears We have defined an episode as a series of smears starting with a primary smear that is initiated either by the call/recall system or by the GP or by the woman herself and ending with a closing smear that results in the woman being returned to routine recall (Figure 1). Any subsequent smears carried out as a consequence of the primary smear are termed secondary smears. When a woman has a nonnegative smear result, she may end up being referred to colposcopy (and suspended in the call/recall system). The smear that results in a referral to colposcopy is defined as the referral smear, and the woman may be treated and undergo a further series of smears before the episode is closed by a negative smear returning her to routine recall. The time between the primary smear and the closing smear is the length of the episode, and most episodes (where the primary smear is negative) are therefore of length Somerset Rural 22.3 5 4.66 Berkshire Prosperous 21.8 5 4.43 Dorset Coast and Services 22.2 3 3.60 Oxfordshire Prosperous 22.3 3 3.77 Primary smear Referral smear Closing smear ONS, Office for National Statistics. *Negative scores are more affluent, and positive scores are more deprived. **Estimated using 2000/01 KC53 returns. Additional secondary smears Figure 1. An example of a screening episode indicating primary, referral and closing smears. For most women, the primary smear will be negative and will therefore also be the closing smear. ª 2007 The Authors Journal compilation ª RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology 1409

Blanks et al. zero because the primary smear is also the closing smear. For some women, an episode can last for many years. Episodes can be further divided into prevalent screen episodes (a woman s first smear episode) and incident screen episodes (a subsequent smear episode). In this paper, we use only those incident episodes where a woman has a previous episode in the cohort database, for which the interval since the closing smear of the previous episode (the screening interval) can therefore be calculated. For a minority of women, the grouping of smears into episodes can be difficult either due to nonattendance for repeat smears or due to conflicting data for two smears; detailed rules [obtainable from the cancer screening evaluation unit (CSEU)] were therefore established to allow all smears to be considered as part of an episode. Primary smears as invitational or opportunistic Primary smears can be considered invitational where the woman is invited by the call/recall system (known as the Exeter system) itself or opportunistic where the smear is initiated by a GP or by the woman herself and not as the result of a system invitation. For a small proportion of primary smears, insufficient information exists to classify the primary smear, and these primary smears are classed as undefined. An episode can therefore also be defined as invitational or opportunistic depending on the origin of its primary smear. Opportunistic primary smears can be further divided into routine and sporadic. A routine opportunistic primary smear is defined as one taken at a normal routine interval; between 2.5 and 3.5 years for a 3-year policy area or between 2.5 and 3.5 years or 4.5 and 5.5 years for a 5-year policy area. Routine opportunistic smears will be used to measure routine 3-yearly screening occurring in 5-year policy areas. Opportunistic primary smears taken outside these intervals are defined as sporadic. A glossary of terms is included in the Appendix. Supplementary studies Data from the 2000/01 annual Health Authority returns to the Department of Health (KC53 returns) were used to check whether the screening interval in practice for the four Health Authorities was in line with the stated policy (Table 1) and to provide additional information for all of the 99 Health Authorities in England. The screening interval can be crudely estimated by dividing the proportion of women screened within the past 3 years by the proportion screened within the past 5 years (P). This ratio would be 0.6 if all women were rescreened at 5 years, and 1.0 if all women were rescreened at 3 years; an estimated screening interval (ESI) can therefore be estimated as 8 5P. This supplementary information has been used to determine how representative the four Health Authorities are of English Health Authorities overall and to help make inferences to all of England from the cohort. Results During 1 January 1999 to 31 December 2003, there were 733 814 primary smears initiating incident screen episodes (originating from a total of 577 109 women) of which 677 580 (92%) were episodes where the primary smear could be categorised as invitational or opportunistic. There were 56 234 undefined episodes (8%) where missing or contradictory information did not enable the episode to be thus categorised. Of the incident screen episodes, 87% were single smear episodes, where the primary smear was also the closing smear, and 13% were multi-smear episodes. In total, 25% of multi-smear episodes and 3.6% of all episodes included a referral smear resulting in the woman being referred to colposcopy. Of the 677 580 categorised primary smears, 486 074 (72%) were invitational and 191 506 (28%) were opportunistic. Of the opportunistic smears, 83 956 (44%) were routine opportunistic and 107 550 (56%) were sporadic opportunistic. Table 2 shows the 677 580 primary smears analysed by local interval policy and age. For all women (aged 20 64 years), the invitational proportion was far higher in 3-year policy (83%) than in 5-year (57%) policy areas, with 17 and 43% of primary smears, respectively, being opportunistic. Invitational smears were marginally more common for older women than younger women regardless of policy. Routine opportunistic smears were particularly common in 5-year policy areas (21%) where they represent routine 3-yearly screening in a 5-yearly policy area. In 3-yearly areas, only Table 2. Number and percentage of incident screen primary smears by type, age group and interval policy Type of primary smear and age group at smear Three-year policy (%) Five-year policy (%) Within age group, Invitational 20 24 years 14 898 (71.3) 6768 (45.8) 25 49 years 206 473 (82.1) 110 528 (55.3) 50 64 years 99 559 (87.7) 47 848 (62.2) 20 64 years 320 930 (83.2) 165 144 (56.6) Routine opportunistic 20 24 years 1983 (9.5) 3294 (22.3) 25 49 years 15 073 (6.0) 42 468 (21.2) 50 64 years 5275 (4.6) 15 863 (20.6) 20 64 years 22 331 (5.8) 61 625 (21.1) Sporadic opportunistic 20 24 years 4004 (19.2) 4721 (31.9) 25 49 years 29 883 (11.9) 46 963 (23.4) 50 64 years 8745 (7.7) 13 234 (17.2) 20 64 years 42 632 (11.0) 64 918 (22.3) 1410 ª 2007 The Authors Journal compilation ª RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology

Opportunistic smears in NHS cervical screening 6% of smears were classified as routine opportunistic. Sporadic opportunistic smears were more common in 5-year policy (22%) than in 3-year policy areas (11%) and also in younger women irrespective of policy. Figure 2A, B show the percentage of incident episode primary smears that were invitational or opportunistic for 3- and 5-year policy areas. Routine opportunistic smears in 3-year policy areas tended to occur mostly between 30 and 36 months and probably represent opportunistic smear taking in women due to an invitation in the near future, perhaps because the GP felt that they might not attend the invitation or because it was more convenient to do the smear there and then. Figure 2A shows that sporadic opportunistic smear taking in 3-yearly policy areas mostly relates to women screened between 24 and 30 months, with a small additional peak at 48 months, the latter presumably being smears taken in women who did not attend their 3-yearly invitation. In 5-yearly policy areas, where 21% of smears were sporadic, these occurred mostly between 42 and 54 months, representing early screening in a 5-yearly policy area. (Figure 2B). Overall, in 5-yearly areas, 84.7% of opportunistic smears occurred at less than 54 months (which could be considered A Percentage of primary smears B Percentage of primary smears 20 18 16 14 12 10 8 6 4 2 0 0 12 24 36 48 60 72 Screening interval (months) 20 18 16 14 12 10 8 6 4 2 0 0 12 24 36 48 60 72 Screening interval (months) Figure 2. Percentage of incident screen primary episodes that are invitational (solid line) or opportunistic smears (dotted line) by screening interval (time since previous closing smear) for all areas over 1999 2003 for women aged 20 64 years by (A) 3-year policy areas and by (B) 5-year policy areas. an early screen) compared with 3-yearly areas where only 43.1% of opportunistic smears were taken at less than 30 months. Details of percentages of opportunistic smears occurring in 3- and 5-yearly areas by age group and interval are given in Table 3, and for all smears, the details are given in Table 4. For brevity, these tables only include age groups 25 49 and 50 64 years, which are the age groups currently invited to screening following recent policy changes. Opportunistic smears that are not early may be smears taken at a more convenient time just before an invitation was due or a smear taken after an invitation when the woman was a nonresponder. Overall, 9.4% of all primary smears in women aged 25 49 years in 3-yearly areas were early, increasing to 50.6% in 5-year areas. For women aged 50 64 years, these figures were 5.4 and 38.1%, respectively. External validity of cohort results We investigated whether the above results can be applied to the whole of England. In the four Health Authorities included in the cohort using data from the 2000/01 annual returns, the ESI ranged from 4.66 in Somerset, which has a 5-year policy, to 3.60 in Dorset with a 3-year policy. Berkshire had a higher proportion of opportunistic smears than Somerset and also had a shorter ESI of 4.33 versus 4.66 (Table 1). Overall, for all 99 Health Authorities in 2000/01, the ESI varied from 3.56 to 4.82 with a mean of 4.01. Those Health Authorities with 3-year policies had a mean ESI of 3.76 (range 3.56 4.13), while those with a 5-year policy had a mean of 4.41 (range 4.10 4.82). Figure 3 shows the percentage of all smears taken opportunistically as reported in the annual returns against the ESI in 2000/01 from all Health Authorities with a 5-year screening policy. Increased use of opportunistic smear taking is associated with decreasing screening interval. Only one Health Table 3. Number of opportunistic smears (percentage) occurring within specified time periods in 3- and 5-year policy areas for age groups 25 49 and 50 64 years Policy Interval (months) Ages 25 49 years, Ages 50 64 years, 3 year 0 29 19 550 (43.5)* 5579 (39.8)* 30 41 15 073 (33.5)** 5275 (37.6)** 42 53 5641 (12.5) 1705 (12.2) 54 65 1840 (4.1) 487 (3.5) 66 2852 (6.3) 974 (6.9) 5 year 0 29 22 220 (24.8)* 5960 (20.5)* 30 41 31 730 (35.5)* 11 769 (40.5)* 42 53 21 783 (24.4)* 6338 (21.8)* 54 65 10 738 (12.0)** 4094 (14.1)** 66 2960 (3.3) 936 (3.2) *Early. **Correct interval. ª 2007 The Authors Journal compilation ª RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology 1411

Blanks et al. 80 60 40 20 4.00 4.20 4.40 4.60 4.80 Estimated Screening interval % opportunistic Fitted line Authority (Avon (ESI = 4.82) has evidence of a true 5-year interval in practice, as well as policy; the mean ESI in Figure 3 is 4.41, suggesting that opportunistic smear taking results in an average screening interval less than the 5-year policy. We estimate that the level of opportunistic smear taking in all 5-year policy areas is around 50%, rather than the 43% from our two areas combined in the cohort study and that Berkshire is more representative of the average 5-year policy area than Somerset. Our results may therefore be marginally conservative when applied to the whole of England. Discussion Berkshire Somerset Figure 3. Percentage of smears that are opportunistic versus ESI for Health Authorities in England with a 5-year screening policy (data from 2000/01 annual returns). A number of studies have examined the effectiveness of organised versus opportunistic smear taking 3,4 but have not come to consistent conclusions. Miller 5 has argued that when Table 4. Number of all smears (percentage) occurring within specified time periods in 3- and 5-year policy areas for age groups 25 49 and 50 64 years Policy Interval (months) Ages 25 49 years, Ages 50 64 years, 3 year 0 29 23 529 (9.4)* 6174 (5.4)* 30 41 167 780 (66.7)** 88 155 (77.6)** 42 53 38 082 (15.1) 12 010 (10.6) 54 65 9265 (3.7) 2981 (2.6) 66 12 773 (5.1) 4259 (3.8) 5 year 0 29 26 452 (13.2)* 6494 (8.4)* 30 41 47 763 (23.9)* 15 707 (20.4)* 42 53 26 906 (13.5)* 7113 (9.2)* 54 65 81 066 (40.5)** 42 219 (54.9)** 66 17 772 (8.9) 5412 (7.0) *Early. **Correct interval. invasive cancer is used as an endpoint, organised screening is more effective because opportunistic smears tend to be taken frequently in young women and have less impact on invasive cancer incidence. The use of opportunistic smear taking in England is in some ways different because of the use until recently of a dual screening policy where women could be invited at 3- or 5-yearly intervals. The NHS cervical screening programme is strongly influenced by opportunistic smear taking. In the extreme, one Health Authority with a 5-yearly policy actually had an ESI shorter than another Health Authority with a 3-yearly policy. Clearly, screening interval policy and practice may be very different as a result of opportunistic smear taking. With the recent change in screening programme policy in England (such that all women aged 25 49 years will, in future, be invited every 3 years and women aged 50 64 years every 5 years), careful monitoring will be required. The new policy will require GPs in the previous 3-year policy areas to screen women aged 50 64 years at a longer, 5-year, interval. If, as we have shown, many GPs in the previous 5-yearly policy areas screen women aged 50 64 years opportunistically at shorter intervals than 5 years, then GPs in the previous 3-year policies areas may be reluctant to adopt the new policy of screening at 5-yearly intervals for this age range. Hence, opportunistic smear taking could result in a national programme with screening intervals in practice that are markedly different to policy. Screening women aged 50 64 years, 3-yearly, is likely to result in higher costs with little gain in additional cancer protection. 1 In our cohort, 31% of smears in women aged 50 64 years in 5-yearly screening areas were early opportunistic smears occurring more than 6 months before the recommended screening interval; this is similar to the percentage (at all ages) of negative smears found to be interval smears in a study in Bristol, 6 which concluded that discouraging such smears could reduce laboratory workload by 30%. The routine measurement of the ESI for women aged 25 49 and 50 64 years from PCT returns (KC53 returns) may be a simple method of tracking screening interval, and hence the likely use of opportunistic smears, as the national programme changes to the new screening policy. In the future, improvements could be made to the information available from standard returns to enable a more precise measure of screening interval to be developed. Conclusions These results suggest that GPs (and women) often do not adhere to a local 5-year policy and in practice may screen women much earlier. The effect is that the real screening interval is much less between 3- and 5-year policy areas than would be anticipated and that opportunistic smears may limit the effect of the new English policy of 5-yearly screening at ages 50 64 years. 1412 ª 2007 The Authors Journal compilation ª RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology

Opportunistic smears in NHS cervical screening Contribution to authorship R.G.B. helped design the study, analyse the data and write the paper. S.M.M. helped design the study and write the paper. D.A.C. helped design the study and analyse the data. A.J.S. helped design the study and write the paper. Acknowledgements We thank Aida Sanchez for her work in the data management of the cohort and to the Health Authorities for supplying the data. The Cancer Screening Evaluation Unit receives funding from the Department of Health Policy Research Programme; the views expressed in this publication are those of the authors and not necessarily those of the NHS Executive or Department of Health. The study has Patient Information Advisory Group (PIAG) approval and was deemed by the Multi-Centre Research Ethics Committee (MREC) not to require their approval, as it did not include the collection of any other data that was already routinely collected. The study is funded by the Department of Health. j References 1 Sasieni P, Adams J, Cuzick J. Benefit of cervical screening at different ages: evidence from the UK audit of screening histories. Br J Cancer 2003;89:88 93. 2 Anttila A, Ronco G, Clifford G, Bray F, Hakama M, Arbyn M, et al. Cervical cancer screening programmes and policies in 18 European countries. Br J Cancer 2004;91:935 41. 3 Gustafsson L, Sparen P, Gustafsson M, Wilander E, Bergstrom R, Adami HO. Efficiency of organised and opportunistic cytological screening for cancer in situ of the cervix. Br J Cancer 1995;72:498 505. 4 Nieminen P, Kallio M, Anttila A, Hakama M. Organised vs. spontaneous Pap-smear screening for cervical cancer: a case-control study. Int J Cancer 1999;83:55 8. 5 Miller AB. The (in)efficiency of cervical screening in Europe. Eur J Cancer 2002;38:321 6. 6 Channer JL, Mackenzie EF. A study of the effect of introducing a restrictive cervical screening policy on laboratory workload and cervical cancer detection rates. Cytopathology 1990;1:13 17. Appendix Glossary Primary smear Referral smear Closing smear Episode Prevalent episode Incident episode Exeter (call/recall) primary smear Opportunistic primary smear/episode Routine opportunistic smear/episode Sporadic opportunistic smear/episode The first smear in an episode The smear that leads a woman being referred to colposcopy The last smear in an episode. A primary smear which is negative will also be the closing smear. Closing smears are always negative A single primary smear or a sequence of smears initiated by the primary smear that ends with a closing smear, which is a negative smear test that returns the woman back to routine recall A woman s first episode A subsequent screening episode An episode started by a primary smear, which has resulted from the woman being invited to be screened by the national call/recall system An episode started by a primary smear that resulted from a decision to do a smear test from the GP or the woman but not initiated by the call/recall system An episode started by an opportunistic smear taken at a standard interval between 2.5 and 3.5 years or between 4.5 and 5.5 years, usually the former. These smears are mostly the result of GPs undertaking 3-yearly screening in a 5-year policy area* An episode started by an opportunistic smear taken between 0 and 2.5 years, 3.5 and 4.4 years or more than 5.5 years after a previous closing smear and not initiated by the national call/recall system *Note that the small number of prevalent screen opportunistic smears would be considered routine opportunistic. ª 2007 The Authors Journal compilation ª RCOG 2007 BJOG An International Journal of Obstetrics and Gynaecology 1413