- Top of page
Background About 50 000 women are referred annually to colposcopy in England because of a low-grade smear. About 35% of these women have no colposcopic abnormality but are followed up in the colposcopy clinic because of uncertainty about the risk of significant pathology.
Objective This study determined the 5-year rate of disease when initial colposcopy was normal and smear was non-dyskaryotic.
Design Retrospective study.
Setting Colposcopy clinic of an inner city postgraduate teaching hospital.
Population Two thousand one hundred and fifty seven women referred between January 1990 and December 2001 with mild dyskaryosis (Low Grade Squamous Intraepithelial Lesion [LSIL]) or borderline nuclear changes (Abnormal Squamous Changes of Uncertain Significance [ASCUS]).
Methods Information was obtained from the colposcopy clinic database and Open-Exeter. Time plots of the disease-free rates were generated using the Kaplan–Meier method, and statistical comparisons were performed using Cox regression.
Main outcome measures Cumulative rates of cytological and histological abnormalities.
Results High-grade or invasive disease was diagnosed histologically in 12.8% of 805 women referred with borderline nuclear changes and in 35.8% of 1352 referred with mild dyskaryosis. Among 620 women with normal colposcopy and a negative or borderline repeat smear, high-grade disease was found after 5 years of follow up in 1.3% of women originally referred with a borderline smear and in 8.5% referred because of mild dyskaryosis.
Conclusion Women referred to colposcopy with borderline nuclear changes or mild dyskaryosis whose colposcopy findings are normal and whose repeat smear in the clinic is non-dyskaryotic may be discharged for routine 3-yearly screening in the community because the risk of high-grade disease in the next 5 years is small.
- Top of page
The National Health Service Cervical Screening Programme (NHSCSP) has been established since the mid-1980s. With the introduction of the national call and recall system and increased public awareness, there has been improved uptake of cervical screening and a fall in cervical cancer incidence. Overall, it is estimated that the programme prevents between 1100 and 3900 cases of cervical cancer each year.1 In 2003–04, 3.6 million women of all ages were screened. In March 2004, 80.6% of eligible women aged 25–64 years and resident in England had been screened at least once in the previous 5 years.2
The risk of high-grade lesions (cervical intraepithelial neoplasia [CIN] 2–3) is well established for women with smears showing moderate or severe dyskaryosis. As a result, these women are referred immediately to colposcopy for further evaluation and treatment if required. However, the majority of abnormal smears show only low-grade cytological changes. In 2003–04, of the 3.46 million adequate smears taken, 129 814 (3.7%) had borderline nuclear changes and 76 314 (2.2%) mild dyskaryosis, compared with the 48 188 (1.4%) with moderate or severe dyskaryosis or abnormal glandular cells combined.2 Overall, women with low-grade smears account for around 50% of colposcopy referrals for abnormal cytology in England.2
High-grade CIN is less likely in women with low-grade cytology than in those with severe or moderate dyskaryosis. High-grade lesions are found in 49–69%3–5 of women referred to colposcopy with mildly dyskaryotic smears and in 9–30%6–10 in those with borderline nuclear changes. The risk of developing invasive disease is increased when these women are followed up by cytology rather than referred to colposcopy.11 Two prospective trials concluded that immediate referral to colposcopy was the safer and more effective option.12,13 Indeed, 70% of women with mild dyskaryosis end up being referred for colposcopy even when cytological follow up is employed initially.14 Current national guidelines recommend that best practice is referral to colposcopy after one mildly dyskaryotic smear or three borderline smears.15
Approximately one-third of women referred with borderline nuclear changes or mild dyskaryosis will have normal colposcopy.3,7–9 Some of these women will have been referred with false-positive cytology, but some lesions detected by cytology may have regressed naturally. However, colposcopy does not identify all lesions, particularly those that are low grade.16 A study of 221 women referred with low-grade smears and normal colposcopy reported that 36–53% had abnormal smears and approximately 15% had histologically confirmed CIN or cancer after 5 years of follow up.17 They concluded that women with normal colposcopy should continue to be followed up colposcopically regardless of the severity of the referral smear. However, the results of smears taken at the first visit were not considered.
A second study of 295 women referred to colposcopy with a mildly dyskaryotic or borderline smear and found to have normal colposcopy and a non-dyskaryotic repeat smear reported that only 23 (7.8%) required treatment during follow up and only 6 had high-grade CIN.18 They suggested that such women could be discharged safely to cytological screening in the community.
However, the ASCUS-LSIL Triage Study (ALTS) study indicated that 11.3–13% of 1587 women referred with low-grade abnormalities had high-grade CIN diagnosed in a 2-year follow-up period even after negative colposcopy and biopsy or CIN1 on colposcopic biopsy diagnosis.19
The results of the study by Teale et al.18 led to the NHSCSP Guidelines in 2004 suggesting that women referred with mild dyskaryosis or less with a satisfactory, normal colposcopic examination and a normal repeat smear 6 months later might be returned to routine recall.20 However, Teale et al. recognised the limited nature of their data and called for further studies to be performed in other regions with larger numbers of women and longer follow up.
The objective of the present retrospective study was to identify women referred to the colposcopy service at Hammersmith Hospital with mild dyskaryosis or borderline nuclear changes and then to determine the rate of dyskaryosis, abnormal histology and treatment in those women whose initial colposcopy was normal and whose first colposcopy clinic smear was non-dyskaryotic.
- Top of page
Hammersmith hospital is a regional tertiary referral centre for gynaecological oncology and has a large colposcopy caseload. Over the study period, referrals for colposcopy came from GPs in the local area, genitourinary medicine clinics and private and NHS gynaecology departments. Colposcopy was performed either by experienced colposcopists or trainee colposcopists under supervision. Colposcopy was defined as normal if there was no evidence of changes consistent with human papillomavirus (HPV) infection, CIN or cancer and if the transformation zone was fully visible. A cervical smear was taken at the first visit. A biopsy was taken from any area suggestive of CIN unless the colposcopic appearances suggested that excisional treatment was indicated. Biopsies were also taken from most women with low-grade changes to confirm the absence of high-grade CIN requiring treatment.
Since 1989, the colposcopy clinic has employed a comprehensive database for storing all information relating to women seen.21 At a woman’s first appointment, identifying and demographic information are entered, as well as the reason for referral, the referral smear result, presenting symptoms, pertinent medical history, clinical and colposcopic findings, smear and biopsy results and the planned management. Thereafter, relevant clinical information and test results are added at follow up and treatment visits.
For the purpose of this study, a separate database was created in Lotus Approach (IBM™) that included all women from the clinic database who had been referred to colposcopy between January 1990 and December 2001 with a smear showing borderline nuclear changes or mild dyskaryosis. The database was contained in two files; one containing static data including the woman’s name, date of birth, hospital number, referral smear and clinical details relating to the first colposcopy visit and a second file of variable data with multiple records for each patient to include details of colposcopy follow up and results of smears taken in the community. For patient confidentiality, this database was protected by a password.
Each record in Lotus Approach was reviewed manually, and women were excluded if they had been pregnant at the first visit or had a history of treatment to the cervix. The main study population contained two broad groups of women; one containing women who had received treatment as a result of colposcopy, histology or cytology findings at their first visit and another consisting of women who were managed conservatively. For this second group of women, information on cytology reports or subsequent treatment once they had been discharged or had defaulted from the colposcopy clinic was acquired using Open-Exeter. This browser-based program allows access to local data on the national screening database and provides patient-specific details relating to further screening tests. While this is a national database, access was limited to information held by health authorities local to the hospital. Any results subsequent to those taken at or between visits to the colposcopy clinic were added to existing data in Lotus Approach. Where data records were incomplete or inconclusive, hospital case notes and the main colposcopy database were examined.
Having obtained all available follow-up data from the clinic and community, the grade of the worst dyskaryotic smear or date of the last non-dyskaryotic smear were recorded. Where a report of ‘abnormal glandular cells’ was made, this was taken as the worst result. The time interval between the first appointment and the worst dyskaryotic smear or the last nondyskaryotic smear was then calculated. Second, the date and result of the worst histology report was recorded. This may have been from a punch biopsy or a treatment biopsy. Adenocarcinoma in situ (AIS) was considered more severe than CIN3. In addition, the date of the first relevant treatment was recorded. Treatments affecting the cervix for reasons other than CIN or cervical cancer were not considered, except as a reason for censoring.
In the management algorithm proposed by Teale et al.,18 women whose colposcopy findings were consistent with HPV infection alone were included in the group with abnormal colposcopy and would therefore be reviewed in colposcopy at 6 months. However, such women have no evidence of CIN, and it might be possible to discharge these women safely back into the community in the same way as women who have normal colposcopy. To consider this hypothesis, the worst cytology and histology were reviewed and survival analysis was carried out for conservatively managed women with colposcopy findings showing HPV infection.
The statistical analysis was carried out using StatsDirect (www.statsdirect.com). Kaplan–Meier curves were plotted for the time-dependent variables. The first analysis used an endpoint of dyskaryosis of any degree. Women were censored at the time of their last smear result if no dyskaryosis had been found or at their last smear prior to hysterectomy for reasons other than gynaecological malignancy. A second survival analysis was performed with the endpoint of a biopsy showing high-grade CIN or worse. Censored women were as above, plus those women whose worst biopsy result had been CIN1. In order to consider possible modifications to the management algorithm proposed by Teale et al., the analysis was repeated with the addition of women whose colposcopy examination at the first visit showed colposcopic changes consistent with HPV infection. Cox regression analysis was carried out to determine whether there was a significant difference in the disease-free curves between the women who were referred with a borderline smear compared with those with mild dyskaryosis.
- Top of page
A total of 2380 women were referred to the colposcopy clinic between January 1990 and December 2001 with borderline nuclear changes or mild dyskaryosis on a cervical smear. Table 1 shows the population characteristics of the total group according to the referral smear. From the outset, 49 women were excluded because they had been pregnant at the first visit and 174 had undergone previous cervical surgery (Figure 1). A group of 2157 women made up the main study population. Of these, 563 (26%) women received treatment as a result of findings at their first colposcopy visit and the remaining 1596 were managed conservatively (Figure 1).
Table 1. Population characteristics for all women referred to colposcopy between January 1991 and December 2001 with a borderline or mildly dyskaryotic smear
|Age at first visit||36.1 (28.1, 45.9)||29.6 (24.6, 37.3)|| |
|Previous cervical surgery||75||99||174|
|Pregnant at first visit||12||37||49|
|Treated as result of first visit||119||444||563|
The disease burden for the 2157 women who made up the main study population is shown in Table 2. It is based on the most severe histology from treatment or diagnostic biopsy by the end of colposcopic follow up. Because 563 women were treated soon after the first visit, the median length of colposcopic follow up of these 2157 women was only 26 weeks (lower and upper quartiles 7and 69 weeks). The cumulative rate of high-grade or invasive disease diagnosed histologically by 5 years was 12.8% in the 805 women referred with borderline nuclear changes and 35.8% in the 1352 referred with mild dyskaryosis.
Table 2. The most severe histology results reported at any time during the study in 2157 women referred to colposcopy with borderline nuclear changes or mildly dyskaryosis, comparing those treated (n= 563) with those managed conservatively (n= 1592) as a result of findings at the first visit
|Worst histology||Referral smear||Total|
Invasive cervical cancer was found in 2 (0.09%) of the 2157 women referred with borderline nuclear changes or mild dyskaryosis. One woman referred with mild dyskaryosis was diagnosed with stage Ib1 invasive carcinoma of the cervix after her initial colposcopic examination; smear and punch biopsy had suggested only CIN1. She received treatment 6 months later for what appeared to be persistent CIN1, and it was only the treatment histology that correctly identified invasive disease. A woman referred with borderline changes was found to have a small adenocarcinoma of the cervix after 2.5 years of follow up. Her smears had never been truly normal although they had not been severe enough to warrant biopsy up until that point. A total of six women had a diagnosis of AIS. All of these women received treatment due to findings at the first colposcopy appointment. Three of the six women had normal or only CIN1 colposcopic findings at their first visit, highlighting the inability of colposcopy to detect glandular lesions and the importance of biopsy in diagnosis.22
Of the 561 women treated as a result of colposcopy findings at the first visit, 362 (64.5%) had a histological diagnosis of high-grade CIN and 161 (28.7%) had CIN1 (Table 2). Only 14.5% of those referred with borderline smears were treated at this time compared with 32.8% of the mild group (Table 2: Yates corrected χ2= 86.91, P < 0.0001). Abnormal smears at the first visit contributed to the decision to treat 22 of the 51 women with normal colposcopy in this group of 561 women.
Among the group of 1596 women who were managed conservatively, there were 670 women who had normal colposcopy findings and a nondyskaryotic repeat smear at the first visit (Figure 1). The remainder consisted of 129 women with normal colposcopy and an abnormal or inadequate smear and 797 with abnormal colposcopy. Abnormal smears in the 129 women with normal colposcopy led to the diagnosis of 12 women with high-grade CIN.
Of the 670 women with normal colposcopy and a nondyskaryotic smear, 48 were lost to follow up because they left the country or moved to a part of England where we did not have access to the Exeter database. Of those lost to follow up, 21 had been referred with borderline smears and 27 with mild dyskaryosis. The remaining 622 women were followed for a median of 159 weeks (lower and upper quartiles 78 and 270) and a total of 2292 women years. In this group, 354 women had been referred with borderline nuclear changes and 268 with mild dyskaryosis (Table 3). None of the women with normal colposcopy and a nondyskaryotic smear were found to have invasive disease.
Table 3. Most severe cytology and histology after colposcopy and community follow up for 620 women who had normal colposcopy at the first visit
|Referral smear||Borderline (n= 352)||Mild dyskaryosis (n= 268)||Total (n= 620)|
|Normal or borderline||316||229||545|
Only 14 (4.0%) of the 354 women referred with borderline nuclear changes received treatment, while 32 (11.9%) of the 268 women referred with mild dyskaryosis required treatment (P= 0.002). The median times to treatment were 95 and 91 weeks, respectively. Four women had a smear showing abnormal glandular cells during their follow up (Table 3). Two of these women were treated, but the histology from treatment was negative. The remaining two women were fully investigated but no further positive results obtained. Glandular abnormalities are more difficult to interpret and therefore false-positive cytology is more likely to occur.22
Kaplan–Meier analysis was carried out on the 620 women with normal colposcopy findings and a non-dyskaryotic smear at the first visit. This showed that 86.0% (standard error 2.3) of the 352 women referred with borderline nuclear changes and 79.8% (SE 3.1) of the 268 women referred with mild dyskaryosis had no dyskaryotic cytology reports after 5 years of cumulative follow up (Figure 2). The difference between the two groups was not statistically significant (P= 0.19). After only 12 months of follow up, 4.3% (SE 1.1) of women referred with borderline nuclear changes and 5.2% (SE 1.4) with mild dyskaryosis had a dyskaryotic smear. This is the proportion of women who would have been referred back to colposcopy if a smear had been taken at 12 months as recommended by Teale et al.18
Figure 2. Kaplan–Meier survival curves for women with normal colposcopy findings and negative or borderline smear at the first visit. Endpoint is any smear showing mild dyskaryosis or worse during follow up.
Download figure to PowerPoint
The cumulative rate of high-grade disease diagnosed histologically by 5 years in those with normal colposcopy findings and a non-dyskaryotic smear at the first visit was 1.3% in those women referred to colposcopy with a borderline smear and 8.5% in those referred because of mild dyskaryosis (P= 0.006, Figure 3). Age had no effect on these results (Cox proportional hazards z=−1.409441, P= 0.1587).
Figure 3. Kaplan–Meier survival curves for women with normal colposcopy findings and negative or borderline smear at the first visit. The endpoint is a biopsy showing CIN2 or worse during follow up.
Download figure to PowerPoint
Among the 155 women found to have HPV colposcopy and a nondyskaryotic smear at the first visit, a smear showing moderate dyskaryosis or worse was reported in 4 (5.1%) of the 78 women referred with a borderline smear and in 4 (5.2%) of the 77 women referred with mild dyskaryosis (Table 4). High-grade CIN was found in eight (10.3%) of those referred with a borderline smear and in seven (9.1%) of those whose referral smear was mild dyskaryosis. In all, 17 (11.0%) of these women were treated. Thus, the grade of referral smear made no difference to the rate of disease in women with HPV changes on colposcopy. The colposcopic finding of HPV changes increased the crude rate of high-grade CIN in women referred with borderline smears from 1.4 to 10.3% (P= 0.0002) and in those referred with mild dyskaryosis from 5.6 to 9.1% (P= 0.2, not significant). Colposcopy was a stronger predictor of high-grade disease than was the referral smear (Cox proportional hazards, z=−3.63468, P= 0.0003 versus z=−1.939908, P= 0.0524). The numbers of women with HPV colposcopy and a non-dyskaryotic smear at colposcopy were too small to estimate the rates of disease beyond 4 years, but by that time, the cumulative rate of high-grade disease diagnosed histologically was 17.1%. This compares with 1.5% and 7.0% at 4 years among those with normal colposcopy and referred with borderline or mildly dyskaryotic smear, respectively.
Table 4. Worst cytology and histology after colposcopy and community follow up for the 155 women who had HPV changes on colposcopy and a non-dyskaryotic smear at the first visit
|Referral smear||Borderline (n= 78)||Mild dyskaryosis (n= 77)||Total|
|Normal or borderline||68||66||134|
- Top of page
The 12.8% rate of high-grade CIN, AIS or invasion in the 805 women referred with borderline nuclear changes and the 35.8% rate in the 1352 referred with mild dyskaryosis are similar to previous reports.4–11 Invasive cervical cancer was found in 92.7 per 100 000 women, approximately ten times the background incidence of cervical cancer of 8.5 per 100 000 in the year 2002.23 This rate is similar to that reported in other colposcopy clinics in the UK.2 These results emphasise the disease burden found in women with apparently mildly abnormal smears.
However, one-third of these women had normal colposcopy and a negative or borderline repeat smear at their first appointment. With 2292 years of follow up in a population of 622 women, this is the largest published study of such women in the UK. After 5 years, high-grade disease was found in only 1.3% in those women referred to colposcopy with a borderline smear and in 8.5% in those referred because of mild dyskaryosis. This reinforces the current guidelines based upon the Birmingham data.18 It further suggests that a smear taken at the first colposcopy appointment should be used in this protocol so that women could be returned to routine 3-yearly screening after the first appointment instead of returning for another smear.
Women with HPV colposcopy and a nondyskaryotic smear had a 17.1% rate of high-grade CIN by 4 years. This figure is based on a relatively small group of women and so should be interpreted with caution. Managing these women in the same way as those with normal colposcopy would increase the number of women discharged by approximately 8%. However, they had almost twice the rate of high-grade disease by 4 years. Thus, further observation in the colposcopy clinic for this small group of women is appropriate.
HPV testing at 6 or 12 months after colposcopy suggesting no more than CIN1 is said to have better than 90% sensitivity in detecting women with high-grade disease.24 However, the specificity of this approach is about 48%, resulting in the referral of 55–64% of the women for further investigation. This is unacceptable.
A limitation of this study was the retrospective nature of the design. However, all the information was collected prospectively, and every effort was made to encourage women to comply with follow up. A further limitation was the restriction placed on access to Open-Exeter. Some 13.9% of the women on follow up had moved out of West London to another part of the UK, and no information was available about them after they had moved despite knowing where they had moved to. Providing national access to this important resource would benefit audit and research and help improve health care provision.
In summary, these data reinforce the national guidelines for women who have normal colposcopy after referral with low-grade smear results and show that these women can return safely to routine cytological surveillance. Taking a smear at the first colposcopy visit would make the process more efficient. Not only would this benefit colposcopy services allowing more time and resources for women with higher grades of dyskaryosis but also it would reduce inconvenience and anxiety for patients. The debate over best management of low-grade smear results is largely resolved, and most authors agree on referral to colposcopy.12–14 Nevertheless, it is important to recognise that one-third of these women could be discharged after only one visit. This would help avoid unnecessary intervention and over-treatment of low-risk women.
As this study was a retrospective audit of patients of W.P.S, referred up to 31 December 2001, ethical approval was not required.
This study was not funded.