Dr A Herbert, Department of Histopathology, 2nd Floor, North Wing, St Thomas’ Hospital, London SE1 7EH, UK. Email email@example.com
Objective To examine screening histories of women with invasive cervical cancer in a 12-year period during the introduction of organised screening to assess its effectiveness.
Study Design Screening histories were classified into six categories: no cytology record (1), previous cytology more than 5 years before diagnosis (2) and previous cytology within 5 years of diagnosis (3–6). Categories 3–6 were described as interval cancers and comprised previous negative cytology (3), previous repeats for low-grade cytology (4), previous recommendations for investigation (5) and previous treatment (6).
Setting Southampton and South West Hampshire (SSWH).
Population Women resident in SSWH registered with invasive cervical carcinoma between 1985 and 1996.
Methods Data were held for analysis on an anonymous spreadsheet.
Main outcome measures Association of screening history categories with 3-year time period, age group, type and stage of cancer and route to diagnosis; incidence per 100 000 women aged 25–64 years screened and not screened within 5 years in 1991–93 and 1994–96.
Results Interval cancers increased as a proportion of all cancers from 31.5% in 1985–87 to 48.6% in 1994–96 (P= 0.002) and showed a peak in 1991–93. Their incidence decreased from 20.1 to 10.9 per 100 000 eligible women aged 25–64 years screened within 5 years (P= 0.008) between 1991–93 and 1994–96, while incidence in women not screened within 5 years was unchanged (44.2 and 40.4). Factors other than previous negative smears were recorded in 50.9% of interval cancers. Interval cancers were more likely to be low stage, screen detected and were diagnosed in younger women (P < 0.00001).
Conclusions Interval cancers should be assessed as a proportion of eligible women screened within the same period of time and not as a proportion of all cancers. This audit demonstrates the importance of accurate cytology, appropriate follow up, prompt investigation and effective treatment of high-grade precancerous lesions.
In a previous study of this data set on women resident in Southampton and South West Hampshire (SSWH), we showed that screen-detected cancers became more frequent as screening coverage increased before, during and after the introduction of organised screening in 1988.1 During the study period (1985–87 to 1994–96), incidence fell from 16.8 to 10.4 per 100 000 women, which was similar to England as a whole.2 However, incidence (and mortality) alone does not give the complete picture of the programme’s potential effectiveness.3 Cervical cancer continues to occur in previously screened women, often leading such women to wonder why this happened despite participating in the programme. Audit of their personal history can yield such information and can identify areas of the programme in which procedures or standards could be improved. Such an audit is now recommended by the NHS Cervical Screening Programme (NHSCSP) as a routine component of quality assurance.3
However, it has to be recognised that the greater the screening coverage the higher the proportion of women with cancer who would have previously been screened,4 and this proportion would reach 100% if all women were screened. An audit carried out in Leicestershire showed that 80.3% of women with cancer had previously been screened, but the details of age range, stage of cancer or route to diagnosis were not provided.5 In a previous study of one 3-year period (1991–93) of the current data set, we found that 75.9% of women aged 25–69 years had been screened in the preceding 5.5 years.6
This article is the culmination of a study carried out in SSWH during the 1990s. We now analyse the screening histories of all the cases and believe that our findings provide a useful baseline for the current audits, which the NHSCSP now requires.3
We have already shown that symptomatic, screen-detected fully invasive and screen-detected microinvasive cancers were progressively more likely to be interval cancers.1 We now present the detailed screening histories to determine reasons why cancers were not prevented in a screened population, particularly in the context of the numbers of women screened and not screened within 5 years.
Population, subjects and methods
The study population comprised the women resident in SSWH estimated by the Office of Population Census and Surveys to have an average total female population of 214 731 in 1986, 217 631 in 1989, 221 026 in 1992 and 224 249 in 1995, the midpoints of the four 3-year periods of the study.7
Screening coverage and incidence of cancer
At the time of study, women aged between 20–64 years were eligible for cervical screening. Centralised call and recall was introduced in 1988 and fully implemented by 1990. There are no data in SSWH on the number of women screened and unscreened in the first two periods of the study (1985–87 and 1988–90) because records had not been transferred to the central computer. Therefore, incidence in women previously screened and not screened within 5 years could not be calculated for the first two periods of our study. However, the overall incidence of cancer per 100 000 total female population was known for all periods of the study and has been published in a previous article on the same data set.2 Screening coverage in the third and fourth periods of our study was published nationally in slightly different age groups. Since 1994–95 (national records run from 1 April to 31 March each year), Department of Health’s Annual Statistical Bulletins records the number of women in each health district eligible for screening, which excludes those known to have had a hysterectomy, and the percentage of women aged 25–64 years screened within 5.0 years.8 Between 1991 and 1993, national statistical summaries recorded coverage within 5.5 years in women aged 20–64 years.9 For the present study, coverage in 1991–93 has been calculated from local Wessex Regional Health Authority datasheets for 1991–92 and 1992–93, which records coverage in 5-year age bands, allowing us to exclude women aged 20–24 years.10 Thus, although the interval for screening coverage changed by 6 months, it was possible to calculate the number of women aged 25–64 years screened and not screened in the third and fourth periods of the study. Women invited but not attending for screening are not separated on regional or national records into those never screened and ever screened, so it is not possible to distinguish between women with no previous cytology and those screened more than 5 (or 5.5) years before diagnosis.
Data collection and study groups
Invasive cervical cancers in women resident in SSWH diagnosed between 1985 and 1996 were retrieved and recorded as previously described.1 The findings have been analysed as four age bands: 20–34, 35–49, 50–64 and 65 years and older; and four periods of time: 1985–87, 1988–90, 1991–93 and 1994–96. Cancers were classified as symptomatic, screen-detected fully invasive or screen-detected microinvasive. The decision as to whether the cancers were screen detected was made on clinical grounds as described in the previous paper.1
Cytology tests performed within 6 months were assumed either to be tests leading to the diagnosis of cancer or tests taken during the investigation of symptoms; these are excluded from the screening histories. Screening histories classified on the basis of routine screening tests carried out at least 6 months before diagnosis are as follows:
1No cytology record.
2Previous cytology record but not within 5.5 years of diagnosis.
3Negative cytology reported within 5.5 years of diagnosis.
4Early repeat(s) recommended for low-grade or inadequate cytology within 5.5 years of diagnosis (with no recommendation for colposcopy).
5Recommended for investigation more than 6 months before diagnosis, but no record of treatment.
6Previous treatment of cervical intraepithelial neoplasia (CIN) before diagnosis of cancer.
Categories 3–6 are defined in our study as ‘interval cancers’ and all had cytology tests within 5.5 years. The histories were ascertained from local laboratory records, medical notes and correspondence with GPs at the time of diagnosis.
Cytology slide review
Ninety-four slides from 85 women with negative smears reported within 5.5 years of cancer diagnosis (category 3) were reviewed internally at the time of the original audit. Similarly, 53 slides from 37 women recommended for early repeat smears for low-grade or inadequate cytology (category 4), and 20 negative smears from eight women on follow up after treatment (category 6) were reviewed at the time of the audit. A summary of the original slide review is presented with the results.
Chi-square test was used as a test of inference to compare difference between proportions. Chi-square for trend was used to test significance of trend of proportions across study years. Statcalc calculator in the EPI INFO 6 software was used for calculation of these tests.
Screening histories and demographic details
Of 382 women with invasive cervical cancer, 148 (38.7%) had no previous cytology record (category 1), 61 (15.9%) had been screened more than 5.5 years before diagnosis (category 2) and 173 (45.2%) had been screened within the preceding 5.5 years (interval cancers: categories 3–6) (Table 1). Cancers have been analysed according to time period, age band, route to diagnosis, histological type and International Federation of Obstetrics and Gynecology (FIGO) stage of cancer for categories 1, 2 and 3–6 (Table 1). A similar analysis of four categories of interval cancers is shown in Table 2.
Table 1. Analysis of cancers in the main screening history categories (1, 2 and 3–6)
Table 2. Analysis of interval cancers in four categories (3–6)
Total interval cancers
Negative cytology <5.5 years (category 3)
Low-grade repeats <5.5 years (category 4)
Delay in referral (category 5)
Previous treatment for CIN (category 6)
Age bands (years)
Route to diagnosis
Screen detected, fully invasive
Screen detected, microinvasive
Squamous cell carcinoma/NOS
The largest number of cancers was seen in women aged 35–49 years, which age band held 127 (47.2%) of 269 cancers in women eligible for screening (20–64 years). Screen-detected cancers comprised 126 (33.0%) of all cancers and were separated into microinvasive (stage IA) and fully invasive (stage IB+) cancers. The latter group was shown in a previous study to be stage IB in 83.3% of cases.1 Histological type was known except for eight cases that were either undifferentiated or type unknown. Adenocarcinomas or adenosquamous carcinomas comprised 92 (24.1%) of all cancers, and there was no significant difference in that percentage in the three main screening group categories. FIGO stage was known except for two symptomatic and one screen-detected squamous cell carcinomas, which were only known to be fully invasive, not otherwise specified. These three cases have been included with stage IB cancers in Tables 1 and 2 as that was the largest group.
Women not screened within 5.5 years (categories 1 and 2)
Category 1: no cytology record (n = 148)
The numbers of cancers in women who had no previous cytology record showed a significant decrease across the four periods of the study (chi-square for trend = 5.889, P= 0.015). The main fall was between 1988–90 and 1991–93, which corresponds to the introduction of organised call and recall (Table 1 and Figure 1). Invasive cancers in this category were more frequent with increasing age and stage (P < 0.00001). Of these 148 cancers, 132 (89.1%) were symptomatic, while 13 (8.8%) were screen-detected fully invasive and 3 (2.0%) were screen-detected microinvasive. Of 113 women aged 65 years or older, 85 (75.2%) had no previous screening record.
Category 2: previous cytology more than 5.5 years (n = 61)
A downward trend was also observed in the number of women in category 2 in proportion to cancers as a whole, but this trend was not significant (P= 0.08) (Table 1 and Figure 1). Of the 61 women who had been screened more than 5.5 years, 18 (29.5%) were known from their medical or laboratory records to have had previous abnormal cytology: 14 had repeats recommended for atypical/ borderline changes and 1 for mild-to-moderate dyskaryosis, 1 had refused treatment for high-grade dyskaryosis and 2 had defaulted after treatment of high-grade CIN.
Interval cancers (categories 3–6)
Demographic and clinical details of the four categories of interval cancers are shown in Tables 2 and 3. Of 173 women with interval cancers, negative cytology had been reported in the 5.5 years before cancer diagnosis (category 3) in 85 (49.1%) women. Low-grade or inadequate smears had been reported and early repeat recommended (category 4) in 37 (21.3%) women. Eighteen (10.4%) women had been referred for colposcopy more than 6 months before cancer diagnosis but had not had any treatment (category 5), while 33 (19%) women had previously been treated for CIN (category 6).
Table 3. Flow chart of screening histories of 382 cancers
No cytology record
Cytology or treatment
More than 5.5 years before diagnosis
6 months–5.5 years before diagnosis
Less than 6 months screen detected, n (%)
n/k, not known, data not recorded.
The next test was the referral smear or symptoms developed before it was due.
Includes referral smears more than 6 months before diagnosis.
There was a significant trend towards interval cancers across the four periods of the study (Table 1 and Figure 1). Interval cancers comprised 34 (31.5%) of 108 cancers in 1985–87 and 34 (48.6%) of 70 cancers in 1994–96 (chi-square = 13.339, P= 0.002) but showed a striking peak in 1991–93 when they represented 63.7% of cancers (Figure 1). This was seen in all categories of interval cancers (Table 2). Cancers were more likely to be interval cancers with decreasing age (P < 0.00001) and stage (P < 0.00001) (Figure 2). Interval cancers were equally likely to be symptomatic (49.7%) as screen detected (51.3%), but this trend was different among the four categories of interval cancers (Table 3): there was a highly significant trend towards the likelihood of cancers being screen detected from category 3 to category 4 and to categories 5 and 6 (P= 0.00001).
Category 3: negative cytology within preceding 5.5 years (n = 85)
Category 3 comprised 85 (49.1%) of 173 interval cancers and represented about half of all interval cancers in all periods of time (47.1–55.0%) and age bands (46.8–53.8%) (Table 2). Of these 85 cancers, 25 (29.4%) were screen detected on investigation of subsequent abnormal smears. Among interval cancers, category 3 comprises 30 (61.2%) of 49 adenocarcinomas compared with 55 (44.4%) of 124 squamous carcinomas (P= 0.067). There were 94 negative smears reported from these 85 women. Reviews of 73 slides were recorded, of which 41.1% were confirmed as negative, 11.0% reviewed as inadequate, 17% as low-grade dyskaryosis and 31.0% as high-grade dyskaryosis.
Category 4: low-grade repeats within previous 5.5 years (n = 37)
There had been 53 abnormal or inadequate smears from these 37 women, of which three had been reported as mild-to-moderate dyskaryosis and 13 as mild dyskaryosis. Borderline/atypical reports had been given in 30, inadequate/bloodstained in five and the result was unknown in two. In 14 of these 37 women, repeats were late or not carried out, while four had subsequent negative smears. The remainder either developed symptoms or had a smear recommending colposcopy before the repeat test was due: 21 (56.7%) women were later referred for abnormal cytology on which screen-detected cancers were diagnosed. Slide reviews of 39 of the 53 slides were recorded, of which 19 (48.7%) were confirmed as low-grade or inadequate dyskaryosis and 20 (51.2%) were reviewed as high-grade dyskaryosis.
Category 5: delay in referral (n = 18)
Abnormal smears for which colposcopy was recommended had been reported more than 6 months before diagnosis in 18 women. Of these cancers, 15 (83.3%) were screen detected, even though diagnosis was delayed, while three (16.7%) presented with symptoms (Table 3). The delay in diagnosis varied between 8 and 23 months and was most often caused by waiting times for initial biopsy and/or subsequent treatment (11/18; 61.1%). The cytology report was delayed in one case, and initial colposcopy was negative in two cases. Two women defaulted from colposcopy appointments and two had moved since the cytology report was issued. Six of these women also had negative or mildly abnormal smears within 5.5 years, which were not reviewed.
Category 6: previous treatment for CIN (n = 33)
These 33 cases, of which 26 (78.8%) were screen detected, were described in detail in a separate study11 and comprised 25 women whose treatment was carried out within 5.5 years and eight who had recent follow-up smears after earlier treatment (Table 3). Two had treatment of cervical intraepithelial glandular neoplasia (CGIN), 19 of CIN3, nine of CIN2 and 1 of CIN1. There had been 20 negative follow-up smears from 15 of these women. Thirteen of those smears were reviewed, of which nine were confirmed as negative, one was reviewed as inadequate, two as mild/borderline and one as high-grade dyskaryosis.
Incidence of cancers in screened and unscreened women
As explained above, incidence in screened and unscreened women could not be calculated for 1985–87 or 1988–90 as data on screening coverage were not available.
Incidence of cancer in women aged 25–64 years screened and not screened within 5.5 years in 1991–93 and within 5.0 years in 1994–95 is shown in Table 4 and Figure 3. Annual incidence of total cancers in previously screened women declined significantly (chi-square = 7.04, P= 0.008) from 20.1 to 10.9 per 100 000 women between 1991–93 and 1994–96. Similarly, incidence of symptomatic cancers declined from 7.8 to 4.0 and screen-detected cancers from 12.2 to 7.8. In women who had not been screened, there was no significant change (chi-square = 0.0, P= 0.950) in the incidence of total cancers (44.0–42.2), symptomatic (28.3–30.5) or screen-detected cancers (15.7–11.7).
Table 4. Incidence of cancer in women aged 25–64 years screened and not screened in 1991–93 and 1994–96
Eligible population aged 25–64 years (total minus ceased)
Screening coverage aged 25–64 years
Number screened within 5.5 years
Number not screened within 5.5 years
Cancers in 1991–93 in women aged 25–64 years, n= rate per 100 000 women per year (%)
Eligible population aged 25–64 years (total minus ceased)
Cancers in 1994–95 in women aged 25–64 years, n= rate per 100 000 women per year (%)
Interval cancers, defined in our study as occurring in women screened within 5.5 years, were significantly more likely to be associated both with early age and early stage at diagnosis. Thus, 86.5% of cancers in women aged 20–34 years were interval cancers as were 83.3% of stage IA cancers, those percentages decreasing with age and stage, respectively (Figure 2). The first of those observations is in accordance with that of Sasieni et al.,12 among whose cases ours were included. Those authors concluded that screening was less effective in younger women while we believe that cancer is more difficult to prevent when it develops early in life because there are less opportunities to detect precancerous changes before invasion develops. Furthermore, screen-detected microinvasive cancers are more likely to occur in previously screened women just as are cases of CIN3 and CGIN, most of which are diagnosed in young women, which we have shown rarely to be detected on the first screening test13—demonstrating the known poor sensitivity of a single test.
Our study also shows that numbers and to a lesser extent percentages of cancers in women who had not been screened within 5.5 years decreased as screening coverage increased reaching steady levels in the final two periods of the study (Figure 1). However, that proportion will decrease further with time as more previously screened (and treated) women enter the older age groups; and there will be progressively fewer women in the population who have never been screened. The Leicester audit5 showed that less than 20% of cancers were in women who had never been screened, which was taken as an indication that screening had been ineffective. Cancers in women who have never been screened should be taken in the context of the number of women in the same population who have never been screened and not as a proportion of cancers. Unfortunately, records do not accurately separate women who have never been screened from those who have been screened more than 5 years before diagnosis, although an attempt to separate these groups was made in a previous study.6 In the third and fourth periods of our present study, when screening coverage was relatively constant, incidence was unchanged at 44.0 and 42.2 per 100 000 women aged 25–64 years who had not been screened within 5 years. Cancers in such women comprised 19.2% (14 of 73) and 37.5% (18 of 48) of cancers in women aged 25–64 years in 1991–93 and 1994–96, respectively (Table 4). The first figure is similar to the Leicester study, whose age range was not presented.
More importantly, interval cancers should be related to the population of women screened within 5 years, which is relatively easy to calculate from Department of Health’s Annual Statistics Bulletins.8 In our study, incidence in previously screened eligible women decreased significantly during the third and fourth periods of the study from 20.1 to 10.9 per 100 000 women. Incidence in women screened within 5 years is an overestimate for the majority of well-screened women. We showed in a previous study that incidence was significantly lower in women screened within 3.5 years, who comprised the majority (75%) compared with 3.5–5.5 years.6 Nevertheless, it is important to question why interval cancers were so frequent in the third period of our study.
We believe that there were several reasons. Those women were previously screened during a decade when workload increased steadily, but sometimes in sudden bursts, almost doubling during the 1980s, resulting in recurring backlogs in screening and waiting lists for colposcopy (unpubl obs.). Although resources eventually followed those demands, there was often a delay in appointing and training new staff leading to considerable pressure on the workforce, while many aspects of quality control had yet to be introduced. Although there was a certificate of competence for cytology screeners, the test was not mandatory for qualified biomedical scientists; there was no rapid review or preview of negative smears; there was no external quality control; there were no standards for cytology reporting rates. All these measures were introduced in Southampton early in the 1990s. The interval cancers detected in 1991–93, many of which were screen-detected microinvasive cancers, resulted from improved quality control as much if not more than from increased screening coverage, which had increased steadily during the whole of the previous decade. The time period 1991–93 also coincided with the introduction of excisional biopsies instead of local ablation for high-grade CIN, so it is possible that some microinvasive cancers might not have been recognised in previous years.14 However, the 65 interval cancers during 1991–93 should be placed in the context of high-grade CIN and CGIN successfully detected. The figures for CIN2 are not recorded, but there were 700 women with CIN3 and 14 with CGIN (3 of the latter were mixed CIN3/CGIN) reported in SSWH 1991–93.13
Another reason for the high rate of interval cancers was the risk of cancer in women aged 35–49 years, who comprised 83/173 (48.0%) of all women with interval cancers. Women in that age band in 1991–93 were born between 1942 and 1958, representing a birth cohort that has been shown across Europe to be at increased risk of cervical cancer15 almost certainly because of the sexual freedom allowed by the availability of reliable contraception.16 Our study suggests that screening in the 1980s was failing to control that increased risk but that increased screening coverage and improved quality control in the 1990s rapidly reversed the trend. Overall incidence had fallen to 10.4 per 100 000 women in 1994–96, at which time incidence of interval cancers in eligible women aged 25–64 years was 10.9 per 100 000 women. The latter figure could provide a useful baseline for the increasing proportion of cancers that are interval cancers in young women, now that cancers in older women have so successfully been prevented and would undoubtedly be more helpful than relating the number of interval cancers to the overall number of cancers, which largely reflects the number of older unscreened women in the population.
Cancers may develop in screened populations for a number of reasons, which were described by Chamberlain17 and supported by later observations.18 These reasons have been used in our study to categorise screening histories as recommended in the first edition of ‘Achievable standards, benchmarks for reporting and criteria for evaluating cervical cytopathology’.19 The most frequent and clinically important category among interval cancers comprised the 49.1% that occurred in women who had negative smears within the previous screening round. Most (70.6%) of those cancers were symptomatic and only 41.1% of negative smears were confirmed as such on review, which is consistent with other studies.20–22 Cancers also occurred in women who had repeat tests recommended for low-grade or inadequate cytology, some of whom had not been followed up as recommended. That category represented 21.3% of interval cancers, and about half of the slides were considered to be high-grade on review. There had been a delay of 8–23 months, for a variety of reasons, since a recommendation for colposcopy in 11.8% of women with interval cancers. Most of those cancers were screen detected (83.3%). Previous treatment for CIN had been carried out in 19.0% of women with interval cancers, which were described in more detail in a separate publication.11 All except one had high-grade CIN, of which most was CIN3 or CGIN but nine of 32 (28.1%) had CIN2.11 In the present study, two of 61 women screened more than 5.5 years before diagnosis were also known to have received treatment for CIN and defaulted from follow up. Overall, 35 of 382 (9.1%) cancers were in women previously treated for CIN. These should be placed in the context of the total number of women treated for high-grade CIN. If the percentage of cases in our study is representative of the country as a whole, 202 (9.1%) of 2221 cancers in 2004 might have been in women who had previous treatment. In that year, 17 345 cases of CIN3 alone were registered and probably a similar number of cases of CIN2, for which there is no national figure. Thus, 1.2% of women with treated CIN3 and probably about half that percentage of those treated for CIN2 and CIN3 combined may have developed invasive cancer. As with cancers in unscreened women and interval cancers, it is the denominator that counts.
Audit of all aspects of screening draws attention to reasons why many of these cancers were not prevented—and demonstrates the importance of accurate cytology, appropriate follow up (and compliance with recommendations), prompt investigation and effective treatment of high-grade CIN. We believe that our study will provide a useful baseline for later audits in the UK and also for countries introducing organised screening on a background of opportunistic screening.
Disclosure of interest
None of the authors have interests that conflict with the study.
Contribution to authorship
A.H. conceived and carried out the initial 12-year audit, while employed as Lead Cytopathologist at Southampton University Hospitals Trust (SUHT). Anshu, while a Commonwealth Fellow at Guy’ and St Thomas’ (GSTT) NHS Foundation Trust helped write the paper and prepared the tables and graphs from the spreadsheet. S.S.G. carried out the statistical analysis. M.G. led the Southampton Cytology Laboratory throughout the time of the study and collaborated in collecting the information about screening at that time. N.S. is now the Lead Cytopathologist at SUHT and helped with this study and was co-author of the previous study of the same data set when it was transferred to an anonymous spreadsheet.
Details of ethics approval
Approval for this audit was gained at the time the project was started in January 1992 when a 3-year grant was successfully applied for and Catherine Breen appointed as a research assistant.
Wessex Cancer Trust and Cansearch, which was the campaigning arm of the British Society for Clinical Cytology, jointly provided the grant that enabled the original project to be set up and carried out. GSTT Charity provided a grant for A.H. to carry out the final analysis of the data set as a baseline for a GSTT cervical cancer audit.
We thank Catherine Breen, the research assistant who set up the original audit database at Southampton General Hospital and Jennifer Smith, Director at that time of Wessex Cancer Intelligence Unit who provided population and coverage data for a previous study of the full data set referred to in the text.