Aug 5, 1996 - incidence of invasive cervical cancer in women who had not been screened during the preceding 0.5-5.5 years than in those who had been ...
Journal of Medical Screening 1996;3: 14{}--145
140
Relation between the incidence of invasive cervical cancer and the screening interval: is a five year interval too long? Amanda Herbert, K Stein, T N Bryant, Catherine Breen, P Old
Department of Histopathology, Southampton General Hospital, (Southampton University Hospitals NHS Trust), Southampton S016 6YD, United Kingdom A Herbert, consultant C Breen, research assistant Southampton and South West Hampshire Health Commission K Stein, senior registrar P Old, consultant Southampton University Hospitals NHSTrust TN Bryant, senior lecturer Correspondence to: Dr Herbert. Accepted for publication 5 August 1996
Abstract Objective-To examine the incidence of invasive cervical cancer per 100000 women years at risk and relative risk according to screening history among eligible women aged 25-69 in Southampton and South West Hampshire during the three years after completion of the first round of comprehensive screening. Results-There was a significantly higher incidence of invasive cervical cancer in women who had not been screened during the preceding 0.5-5.5 years than in those who had been screened (relative risk (RR) 2.6; 95% confidence interval (CI) 1.6 to 4.3). Among the latter group of women (with interval cancers) there was a significantly higher incidence in those with a long interval of 3.5-5.5 years since their most recent smear than in those with a short interval of 0.5-3.5 years (RR 2.2; 95% CI 1.3 to 3.8). Among women with non-interval cancers, there was a significantly higher incidence among those who had no cytology record than among those who had been screened but were overdue for a smear (RR 3.0; 95% CI 1.2 to 7.3). When screen detected cancers were excluded from the figures the relative risks for all the comparative groups described above were greater, though the 95% confidence limits were wider because the numbers were smaller. The most pronounced difference in incidence was between symptomatic cancers in women with a short screening interval (5.8 per 100000 women years at risk) and in women with no cytology record (71.3 per 100 000 years at risk). Most cancers were interval cancers (76%) because of the high screening coverage: 89.2% of eligible women aged 25-69 had been screened during the preceding 0.5-5.5 years. The overall incidence per 100 000 women years at risk approached that of interval cancers, and was nearer to that observed in the short than the long interval because 74.7% of women had been screened within 3.5 years. Conclusion-The results confirm the effectiveness of screening but suggest that a five year screening interval may be too long, at least during the early rounds of screening. (Journal of Medical Screening 1996;3: 14Q--145)
Key words: cervical cancer incidence; cervical cancer screening; intervals for cervical screening.
The effectiveness of cervical screening has been demonstrated in many countries in the world by case-control studies of women with and without cancer and by comparison of the incidence of invasive cervical cancer in screened and unscreened women. I 2 Many studies/ including two recent reports from the United Kingdom since the introduction of comprehensive screening: 5 have shown that protection from invasive disease decreases with the time since the latest negative smear test. Screening in Nordic countries has been highly successful with intervals varying between two and five years." Although the lowest incidence is recorded in Finland, which has a programme starting at age 30 with recall after five years, more than half the smears are taken outside the recommended recall interval, as is also said to be the case in other Nordic countries" Until the 1980s much of the screening in the United Kingdom was opportunistic and the incidence rate was one of the highest in Europe.' A comprehensive screening programme was implemented throughout England in 1988, which has resulted in a progressive fall in mortality from cervical cancer since 1991. 8 As yet there has been no reported fall in the incidence of invasive cancer; but national figures are not available for the 1990s. The National Health Service cervical screening programme recommends a maximum screening interval of five years" but in practice, as in the Nordic countries, there is considerable variation in the screening interval. In many districts, including Southampton and South West Hampshire, recall by the central family health service authority (FHSA) is at five years, though most general practitioners recall their patients earlier. In some districts, again including Southampton and South West Hampshire, three-yearly family health service authority recall is recommended by gynaecologists and the laboratories for women with a history of dyskaryosis or treatment of cervical intraepithelial neoplasia (CIN). In Southampton and South West Hampshire this accounts for about 10% of tests, in addition to the initial annual follow up smears (local unpublished data). Since invitations for screening and cytology records were centralised, data on screening coverage within 3.5 and 5.5 years have been available from FHSAs for individual districts
Incidenceof cervical cancer and the screening interval
141
Table 1 WVmen at risk each year in four screening groups, and screening coverage according to age. Results are given as No (%) Screened 0.5-5.5 years (interval cancers)
Not screened 0.5-5.5 years (non-interval cancers)
Age (years)
Eligible
Short
Overdue
No record
25-39 40-54 55-69
51 709 (100) 38214 (100) 26099 (100)
41 944 (81.1) 28 728 (75.2) 15 985 (61.2)
4957 (9.6) 6 698 (17.5) 5179 (19.8)
3562 (6.9) 2212 (5.8) 2552 (9.8)
1246 (2.4) 576 (1.5) 2383 (9.1)
116 022 (100)
86657 (74.7)
16834 (14.5)
8326 (7.2)
4205 (3.6)
Total 25-69
Long
throughout England from analysis of Korner returns collected nationally," which allows a direct comparison of the incidence of invasive cancer in women screened within those intervals. This study covers the first three years after completion of the first round of comprehensive screening, which was also the first time in which FHSA cervical cytology records were complete for Southampton and South West Hampshire. A previous study covering the three consecutive three year periods before, during, and after the introduction of call and recall showed a significant downstaging of invasive cancer in the screening age groups but no significant fall in incidence per 100000 women of all ages." The third period of the study, in which the incidence of invasive cancer was 15.1 per 100 000 women of all ages, with a high prevalence of screen detected and microinvasive cancer, forms the basis of this report. Population, subjects, and methods POPULATION
The population comprised the total female population of Southampton and South West Hampshire in 1992 (the midpoint of the study,) estimated as 221 026 by the Office of Population Censuses and Surveys. 13 STUDY GROUPS AND CALCULATION OF WOMEN
YEARS AT RISK
The study group of women at risk comprised 116022 women aged 25-69, registered with general practitioners in Southampton and South West Hampshire and eligible for screening-that is, excluding those who had had hysterectomies for unrelated disease. Subjects were divided into four screening groups on the basis of the interval since their most recent cervical smear, based on the average of Korner statistical (KC53) returns for the years 1991-92 and 1993-94. These returns provide yearly figures, in five-year age bands, of total women registered, removed from recall, with no cytology record, screened within 5.5 years, and screened within 3.5 years." The four study groups were defined as follows: (a) a short interval group, comprising women screened within 3.5 years as recorded on KC53; (b) a long interval group, consisting of women screened within an interval of 3.5-5.5 years, calculated by subtracting those screened within 3.5 years from those screened within 5.5 years: the long and short interval groups combined formed the baseline for women with interval cancers; (c) an overdue group, consisting of women who had a cytology record but had not been screened
within 5.5 years, calculated by subtracting the sum of the numbers screened within 5.5 years, those withdrawn from recall, and those with no cytology record from the total number of women registered; the overdue group includes women with unrepeated inadequate smears, estimated from laboratory follow up records as about 1%; (d) a group with no cytology record taken directly from the KC53 data. Cancers occurring in women in the overdue and no cytology record groups combined are referred to as non-interval cancers. FHSA registrations at that time were based on general practitioner rather than patient post codes and were slightly higher, in all age and sex groups, than Office of Population Censuses and Surveys estimates. Thus the figures for incidence are not identical with those based on Office of Population Censuses and Surveys estimates, and, additionally, are confined to women in the screening age group who were eligible for screening during the preceding five years. Because screening coverage is known to vary with age the screening groups were divided into three 15 year age bands of women aged 25-69 to allow the effect of age on differences in incidence in the screening groups to be assessed (table 1). Women years at risk were calculated for each age band and screening group by taking the average registrations for 1991-92 and 1993-94. CASES OF INVASIVE CERVICAL CANCER
The study cases of invasive cervical cancer comprised 83 cancers diagnosed during three years, 1991-93, in women aged 25-69 resident in Southampton and South West Hampshire (18 cases diagnosed during the same period of time in woman aged 70 and over and one in a woman aged less than 25 are excluded from the study). Cases were obtained from local cancer registry data, histology records at Southampton General Hospital, Wessex radiotherapy and oncology units, and records from adjacent district general hospitals and local private hospitals." STAGE AND HISTOLOGICAL TYPE OF CANCERS
The study group comprised 57 squamous cell carcinomas, 21 adenocarcinomas, and five adenosquamous carcinomas. Twenty nine cancers (21 squamous cell and eight adenocarcinomas) were diagnosed at a depth of invasion less than 3 mm and lateral extension less than 7 mm (35% of cancers). Sixty of the 83 cancers (72%) were diagnosed at stage I. MODE OF PRESENTATION
Cases of invasive cancer were categorised as screen detected if the cancer was found as a direct result of a routine screening test. Symptomatic cancers were found as a result of investigation of symptoms. Criteria for establishing mode of presentation have been described in a previous study," Forty three (52%) of the 83 cancers were screen detected.
Herbert, Stein, Bryant, Breen, Old
142
40 Symptomatic cancers
83 Cancers
120
100
80
/
/
/
/
1'.. . . . ......... ..................
.........
..................
/ 60 40
/
20
/
/
/
/
.........
~
/ : ............
/ /
/
/
/
/
/
/
............
......
................
/
_______________
~---------------o
0'-----------'----------'-----------'---------'---------' 25-39
40-54
55-69
40-54
25-39
55-69
Age at diagnosis
I Figure 1
0
Interval
A
Non-interval
Screening histories were defined using FHSA and laboratory records. Smears taken as part of the investigation of symptoms were excluded from the screening histories, as were abnormal smears recommending referral that resulted in diagnosis within six months (referral smears in screen detected cases). Women with cancer who had been screened within 0.5-5.5 years conform to the NHS cervical screening programme criteria for interval cancers and include 32 women with negative smears and 31 who had had abnormal smears, with reports either recommending early repeat or referral. 14 In the latter group there had either been a delay of more than six months between Table 2 Wfnnen with invasive cancer with routine smears within the preceding 0.5-5.5 years (interval cancer) ,. Interval since latest smear CytDwgy result
0.5-3.5 years (short)
3.5-5.5 years (long)
Negative, normal recall Negative, follow up after treatment of CIN Abnormal, repeat Abnormal, refer
13 2 12 17
16 (10) 1
Total
44 (15)
(7) (1) (6) (1)
1 1 (1)
19 (11)
* The number of symptomatic cancers is sbown in parentheses.
Table 3 Incidence per 100 000 women years at risk in women screened and not screened within 0.5-5.5 years Eligible
Total cancers 1991-93* 25-39 40-54 55-69 Total (25-69) Incidence per woman years at risk* 25-39 40-54 55-69 Total (25-69)
51709 38214 26099 116022
24 37 22 83
15.5 32.3 28.1 23.8
I
Incidence of invasive cancer per 100 000 women years at risk: interval and non-interval cancers.
SCREENING HISTORIES OF WOMEN WITH CANCER
Women at risk 25-39 40-54 55-69 Total (25-69)
- Both groups
(12) (15) (13) (40)
(7.7) (13.1) (16.6) (11.5)
Screened 0.5-5.5 years (interval cancers)
46901 35426 21 164 103491
23 28 12 63
16.3 26.3 18.9 20.3
(11) (9) (6) (26)
(7.8) (8.5) (9.5) (8.4)
* The number of symptomatic cancers is shown in parentheses.
Not screened 0.5-5.5 years (non-interval cancers) 4808 2788 4935 12531
1 9 10 20
6.9 107.6 67.5 53.2
the referral smear and diagnosis or cancer had developed after treatment of CIN. Three women with negative smears were being followed up after treatment of CIN carried out more than 5.5 years before (table 2). Seventy per cent of cancers (31 of 44) in the short interval group were in women with a history of abnormal smears or treatment of CIN and included more than half of the screen detected cancers in the study. Results INCIDENCE OF INVASIVE CANCER
Wbmen screened and not screened during the preceding 0.5-5.5 years The incidence per 100 000 women years at risk was 53.2 for women who had not been screened during the preceding round of screening (non-interval cancers) compared with 20.3 for those who had been screened (interval cancers), giving a relative risk of2.622 (95% confidence interval 1.586 to 4.334). When screen detected cancers were excluded the difference was 37.2 compared with 8.4, giving a relative risk of 4.447 (95% confidence interval 2.323 to 8.514). The difference in incidence of interval and non-interval cancers was reversed for women aged less than 40 and there was no difference when screen detected cancers were excluded. Because of the high screening coverage the overall incidence of cancer in women aged 25-69 was closer to the incidence of interval than non-interval cancers (table 3). There was a peak in the incidence of cancer in women aged 40-54 except for symptomatic interval cancers, which increased slightly with age (fig 1).
(I)
(6) (7) (14)
(6.9) (71.7) (47.3) (37.2)
Screened women (interval cancers) with long and short intervals since their latest smear The incidence per 100 000 women years at risk was 37.6 in the long interval group compared with 16.9 in the short, giving a relative risk of 2.223 (95% confidence interval 1.298 to 3.806). When screen detected cancers were excluded, the difference was 21.8 compared with 5.8, giving a relative risk of 3.775 (95%
Incidenceofcervical cancer and the screening interval
143
63 Cancers
50
26 Symptomatic cancers
.... -- ------
20
10
•
---- ..... - - - - - - •
o---------------~---------------~
40-54
55-69
25--39
40-54
Age at diagnosis o Short interval Figure Z
•
Long interval
WOmen overdue for screening compared with those with no cytology record Even though the numbers were small there was a significant difference between the incidence of cancer in women with no cytology record compared with those overdue for screening, giving a relative risk of 2.970 (95% confidence interval 1.215 to 7.260). The former group had an incidence of 95.1 per 100 000 women years at risk for all cancers and 71.3 when screen detected cancers were excluded. incidence in the overdue group was 32.0 for all cancers and 20.0 when screen detected cancers Incidence per 100 0000 women years at risk in the four screening groups Screened 0.5-5.5 years (interval cancers)
Nan-screened 0.5-5.5 years (nan-interval cancers)
Short
Lang
Overdue
No cytology record
Women at risk 25-39 40-54 55-69 Total (25-69)
41944 28728 15985 86657
4957 6698 5179 16834
3562 2212 2552 8326
1246 576 2383 4205
Total cancers 1991-93* 25-39 40-54 55-69 Total (25-69)
17 19 8 44
Incidence per woman years at risk" 25-39 40-54 55-69 Total (25-69)
13.5 22.0 16.7 16.9
(7) (5) (3) (15)
(5.6) (5.8) (6.3) (5.8)
All interval cancers
Incidence ofinterval cancers per 100 000 women years at nsk.
confidence interval 1.734 to 8.217). In both instances the difference was greatest in the youngest age band. Because most of the women had been screened within 3.5 years the overall incidence of interval cancers was closer to the incidence for the short than the long interval (table 4). Figure 2 shows the incidence of interval cancers in women with long and short intervals since their latest smear.
Table 4
-
6 9 4 19
(4) (4) (3) (11)
40.3 44.8 25.7 37.6
(26.9) (19.9) (19.3) (21.8)
1 4 3 8
(1) (2) (2) (5)
9.4 60.3 39.2 32.0
* The number of symptomatic cancers is shown in parentheses.
0 5 7 12
(9.4) (30.1) (26.1) (20.0)
(0) (4) (5) (9)
0.0 (0.0) 289.3 (231.5) 97.9 (69.9) 95.1 (71.3)
were excluded, which was similar to the long interval group (37.6 and 21.8 respectively). In all groups of women with non-interval cancers there was a peak incidence among women aged 40-54. There was only one non-interval cancer among women aged 25-39. Figure 3 shows the difference in incidence of invasive cancer in the four screening groups.
Discussion This three year study of a single district confirms the effectiveness of cervical screening and relates the incidence of invasive cancer to the screening interval. Although there was a significantly lower incidence per 100 000 women years at risk of interval compared with non-interval cancers, the greatest difference was seen between women screened within 3.5 years compared with those with no cytology record. The low incidence in women screened within 3.5 years is particularly interesting in that this group includes "high risk" women with a history of dyskaryosis or treatment of eIN. There was no difference in those screened after a long interval (3.5-5.5 years) and those who were overdue for screening. Our results are similar to those of previous studies, which showed that protection from invasive cancer was inversely related to the interval since the most recent smear'? and which reported a sharp increase in odds ratio for invasive cancer four to five years after a negative smear.' Although the official FHSA recall interval in Southampton and South West Hampshire is five years, most general practitioners chose to recall their patients earlier, as a result of which 74.7% of eligible women aged 25--69 had been screened within 3.5 years. Because such a high proportion of women had been recalled early, the overall incidence per 100 000 women years at risk in the study group was closer to that of the short than the long interval. National figures collected through Korner returns show that percentages of women screened within 3.5 years are similar for Southampton and England as a whole, sug-
144
Herbert, Stein, Bryant, Breen, Old
....
100
III
'':::
.
ftj III
IV
80
CD
>
c: CD
E 0
60
0 0
s
8
8...
40
0
Q,
CD
0
•
Qj
u
• ••
-1
Short long Overdue No cytology Interval Non-interval Both groups
- - ~ -
20
c
CD 'l:l
'u
.E Figure 3
0
All cancers (n
=83)
Symptomatic (n
=40)
Comparison of the incidence of invasive cancer in four screening groups in wome" aged 25-69.
gesting that the situation nationally may be similar." These findings support the British Society for Clinical Cytology statement that three year screening should be encouraged" and suggest that it would be unwise to restrict screening to a five year interval. The data have been calculated for three 15 year age bands within the screening age group to ensure that the differences between the long and short interval were not related to younger women being more likely to be screened at a shorter interval. There is no evidence to suggest that was the case: indeed the differences for all cancers and symptomatic cancers were greatest for the youngest age group, though numbers of cancers were too small for the trends to be significant. The data have been analysed with and without screen detected cancers, because the latter represent a successful outcome of the programme and include low stage cancers with a good prognosis, particularly those diagnosed at less than 3 mm depth of invasion. For all comparative groups: non-interval v interval cancers, long v short interval, no cytology record v overdue for recall, the relative risks were greater when screen detected cancers were excluded, though the confidence limits were naturally wider. Studies of cervical cancer in the Nordic countries have shown that incidence and mortality are reduced when screening is centrally organised and that screening coverage is probably more important than screening interval. Nevertheless, as in England, many smears are taken before the official screening interval.' Probably, the screening interval is more important in early rounds of screening as the incidence of cancer falls after three or more negative smears have been recorded. I,.,7 Even in the USA, where annual screening is recommended, it is recognised that a longer interval should be adequate after three negative smears.' In our study the difference in incidence between the long and short interval tended to fall with age, suggesting that it may be more critical when fewer smears have been taken. Although full screening histories were available for all women with cancer, there were no baseline data for numbers of smears recorded for women in the different screening groups.
Now that 89% of eligible women aged 25-69 have been screened, it is to be expected that most cancers in the screening age group will be interval cancers. The relative proportion of interval cancers increases in proportion to the screening coverage of women at risk ~ '" and would reach 100% if all women were screened. In this study 76% of cancers were interval cancers. Because of the high screening coverage the overall incidence of cancer in the screening age group was similar to that of interval cancers, making the latter the critical measure of the effectiveness of screening. It should be recognised that the relative proportion of interval cancers depends on the recommended screening interval and whether or not screen detected cancers, micro invasive cancers, and adenocarcinomas are included. In a New York State study of eligible women aged 20-69 46% of all histological types of cancer were in women who had had one smear within three years. I. In Aberdeen 29% of women had been screened within five years, but adenocarcinomas were excluded and women of all ages included.' Sasieni et al report 70% and 58% of micro invasive and fully invasive cancers, respectively, occurring in women screened between six months and five years before diagnosis, using the same age group and definitions as our study, whose cases were included in the former study.' We believe that the high proportion of interval cancers in our study relates to the fact that Southampton and South West Hampshire were one of the last districts to introduce call and recall, towards the end of September 1987. A high prevalence of undiagnosed cancer and CIN is to be expected in early rounds of comprehensive screening. The highest incidence of cancer was seen in the 3.6% of women who had no record of a smear. Most of these cancers were symptomatic, and therefore more likely to be diagnosed at a later stage and to carry a worse prognosis than screen detected ones." The relatively low incidence in women who were overdue for a smear, which lowered the overall incidence of non-interval cancers, may be explained by the inclusion in that group of women who had been treated for CIN in the past.
!
145
Incidence of cervical cancerand the screening interval
This study confirms the effectiveness of cervical screening but shows that the overall incidence of invasive cancer in the screening age group has now approximated that of interval cancers because screening coverage is so high. At least in early rounds of screening, the incidence of interval cancers depends on the length of the screening interval, and might have been significantly higher if screening had been restricted to the five year FHSA recall interval.
7 8 9 10 11
We would like to thank Cansearch (the campaigning arm of the British Society for Clinical Cytology) and the Wessex Cancer Trust for the grants which enabled this study to be carried out. We would also like to thank Karen Stone for typing the paper and preparing the tables.
2 3
4
5
6
!ARC Working Group on Cervical Cancer Screening. Summary chapter. In: Hakama M, Miller AB, Day NE, eds. Screening for canar of the ..urine cervix. Lyons: International Agency for Research on Cancer, 1986. (!ARC scientific publications No 76.) Cannistra SA, Niloff JM. Cancer of the uterine cervix. N Engl] Med 1996;334: 1030-8. !ARC Working Group on Evaluation of Cervical Cancer Screening Programmes. Screening for cervical cancer: duration of low risk after negative results of cervival cytology and its implications for screening programmes. BM] 1986;293:659-64. Macgregor JE, Campbell MK, Mann EMF, Swanson KY. Screening for cervical intraepithelial neoplasia in north east Scotland shows fall in incidence and mortality from invasive cancer with concomitant rise in preinvasive disease. BM] 1994;308: 1407 -11. Sasieni PD, Cusick J, Lynch-Farmery E and the National Co-ordinating Network for Cervical Screening Working Group. Estimating the efficacy of screening by auditing smear histories of women with and without cervical cancer. Br] Cancer 1996;73:1001-5. Hakama M, Magnus K, Pettersson, et al. The effect of organised screening on the risk of cervical cancer in the Nordic countries. In: Miller AB, Chamberlain J, Day NE,
12
13 14 15 16
17
18 19
Hakama M, Prorok PC, eds. Cancer screening. Cambridge: Cambridge University Press, 1991. (Published on behalf of the International Union Against Cancer.) Cancer Research Campaign. Cancer of the cervix ..uri. London: CRC, 1994. (Fact sheet No 12.) Sasieni P, Cusick J, Farmery E. Accelerated decline in cervical cancer mortality in England and Wales. Lancet 1995;346:1566-7. Office of Population Censuses and Surveys. Cancer statistics: registrations, England and WfUes 1989. London: HMSO, 1994. (Series MB1 :21.) National Health Service. NHS management exec..tive health. servia guidelines. National cervical screening programme. (HSG(93)41.) Department of Health. Cervical cytology 1991/2; 1992/3. London: Government Statistical Service, DH Statistics Division 2B. (Summary information from form KC 53, England.) Herbert A, Breen C, Bryant TN, et al. Invasive cervical cancer in Southampton and South West Hampshire: effect of introducing a comprehensive screening programme.] Med Screen 1996;3:23-8. Wessex Regional Health Authority. OPCS mid-year population estimates (1992 based). 1992. Herbert A, Johnson J, Patnick J, et al. Achievable standards, benchmarks for reporting, criteria for evaluating cervical cytopathology. Cytopathology 1995;6:suppl 2. British Society for Clinical Cytology. Statement on recall internal for routine cervical screening. Newsletter March 1992. Lynge E, Poll P. Risk of cervical cancer following negative smears in Maribo County, Denmark. In: Hakama M, Miller AB, Day NE, eds. Screening for cancer of the .. urine cervix. Lyons: International Agency for Research on Cancer, 1991. (!ARC scientific publications No 76.) Boon ME, de Graff Guilloud JC, Rietveld WI, WijsmanGrootendorst A. Effect of regular 3 yearly screening on the incidence of cervical smears: the Leiden experience. Cytopathology 1990;1:201-10. Mitchell H, Medley G, Giles G. Cervical cancers diagnosed after negative results on cervical cytology: perspective in the 1980s. BM] 1990;300:1622-6. Nasca PC, Ellish N, Caputo TA, Saboda K, Metzger B. An epidemiological study of Pap screening histories in women with invasive carcinomas of the uterine cervix. NY Stase ] Med 1991;91:152-6.
