BJOG: An International Journal of Obstetrics & Gynaecology

A clinical review of borderline glandular cells reported on liquid-based cervical cytology

Authors


Mr Amit Patel, Northern Gynaecological Oncology Centre, Queen Elizabeth Hospital, Sheriff Hill, Gateshead, NE9 6SX, UK. Email amit.patel@ghnt.nhs.uk

Abstract

Please cite this paper as: Patel A, Thampy N, Hemming D, Naik R. A clinical review of borderline glandular cells reported on liquid-based cervical cytology. BJOG 2010;117:1051–1059.

Objective  To review the clinical outcome of women presenting with borderline glandular cells on liquid-based cervical cytology (LBC).

Design  Retrospective review.

Population and setting  Women seen at a colposcopy clinic over a 26-month period referred with borderline glandular cells on LBC.

Methods  Review of the case notes and cytology, pathology and colposcopy databases of all women referred with borderline glandular cells on LBC between June 2006 and August 2008.

Main outcome measures  Final histological diagnosis.

Results  Sixty-nine women were identified (0.19% of all smears). Twenty-seven women (39.1%) had premalignant or malignant lesions, five (7.2%) had cancers and 22 (31.9%) had intraepithelial neoplasia, 19 (27.5%) of which were cervical squamous intraepithelial neoplasia (CIN) and three (4.3%) of which were cervical glandular intraepithelial neoplasia (CGIN). No women under 35 years of age with normal and satisfactory colposcopy had premalignant or malignant lesions. Despite normal and satisfactory colposcopy, three women over 35 years had significant lesions: one high-grade CIN, one CGIN and one squamous cell carcinoma.

Conclusions  On the basis of our results, it would be considered acceptable to manage women under 35 years of age with normal and satisfactory colposcopy conservatively. In women above 35 years of age, we would recommend a diagnostic ‘large loop excision of the transformation zone’ procedure, irrespective of the colposcopic findings.

Introduction

Squamous abnormalities in the current cervical screening programme are relatively common and cytologically well defined. However, glandular abnormalities are comparatively rare and, although the cytological criteria for the diagnosis of endocervical dyskaryosis are well defined, their recognition in cervical screening can be difficult, as the features may be subtle and subject to the variability of individual laboratory and cytologist practice.1 The category of borderline glandular abnormality can be used (albeit infrequently) to identify the presence of abnormal glandular cells when the nuclear or cytological abnormalities are even less well defined. The cells should be clearly glandular in type and may show nuclear or architectural abnormalities (or both), but fall short of those recognised for endocervical dyskaryosis or glandular neoplasia (including endometrial neoplasia). The architectural features may include cell crowding, loss of cell polarity and multilayering in cell strips. The cytological features may include irregular or coarse nuclear chromatin, hyperchromasia, cytoplasmic vacuolation and variation in cell size.2 This group provides a diagnostic challenge to both the cytologist and colposcopist as the clinical significance is currently unknown.

The term ‘borderline nuclear abnormality’ was introduced in the British Society of Cervical Cytology (BSCC) working party report in 1986.3 This category was to be used in women in whom there was genuine doubt as to whether the cell changes were neoplastic.3 The 1994 National Coordinating Network (NCN)/BSCC/Royal College of Pathologists (RCPath) Guidelines defined the likely situations and appearances of borderline nuclear changes.2 The borderline glandular cell category is defined in the BSCC/NHS Cervical Screening Programme (NHSCSP) national guidelines criteria,2 NHSCSP ‘ABC’ publication,4 NHSCSP Cervical Pathology Atlas (2006)5 and the recently published revised BSCC terminology for abnormal cervical cytology,6 which is yet to be accepted by NHSCSP. This category is equivalent to atypical glandular cells of uncertain significance (AGUS) of the Bethesda system, which was introduced in 1988 soon after the introduction of borderline nuclear abnormality in 1986. In the 2001 Bethesda classification, AGUS was replaced by atypical glandular cells (AGC). The Bethesda system, however, also provides the option of further subclassification on the basis of the origin of these cells and severity of the changes.

The NHSCSP Colposcopy and Programme Management publication recommends that women with borderline glandular cells should be referred promptly after one report of borderline glandular cells for investigation by colposcopy, appropriate cervical biopsy and the selective use of endometrial biopsy.7

Most reports in the literature utilise the Bethesda system for the reporting of borderline glandular cells, that is AGUS. A meta-analysis of AGUS by Schnatz et al.8 reported 4.2% low-grade cervical squamous intraepithelial neoplasia (CIN 1), 6.4% high-grade cervical squamous intraepithelial neoplasia (CIN 2/3), 3.1% adenocarcinoma in situ (cervical glandular intraepithelial neoplasia, CGIN), 0.2% endometrial hyperplasia and 4.0% malignancy. A previously published review from our department by Mohammed et al.9 on borderline glandular cells on conventional cytology using the BSCC classification showed a significant risk of high-grade CIN (27.9%), high-grade CGIN (7.0%) and malignancy (16.2%). However, there are currently no reports of outcomes following borderline glandular cells on liquid-based cervical cytology (LBC) from the UK.

The English pilot study report on LBC found it to be at least as good as conventional smears in the identification of glandular abnormalities.10 The study reported a reduction in the rate of smears showing glandular neoplasia. However, it was unable to clarify whether such lesions were being reported as negative or as high-grade dyskaryosis.10 A meta-analysis of 14 studies comparing the sensitivity of LBC with conventional cytology demonstrated that the sensitivity may be up to 12% better with LBC. However, a meta-analysis of six studies reporting specificity found no difference between the specificity of LBC and that of conventional cytology.11

This paper is the first report on the histological outcomes after colposcopic referral of women with borderline glandular cells using the BSCC classification following the implementation of LBC into the UK cervical screening programme.

Methods

The Cervical Cytology Screening Department at the Queen Elizabeth Hospital in Gateshead provides the cervical screening services for the Gateshead and South Tyneside areas. The department uses the ‘borderline glandular cells’ category for women in whom the cytological changes are thought to be glandular in origin and are not sufficiently severe to be classed as dyskaryosis/?glandular neoplasia. Within this ‘borderline glandular cells’ category, there is a separate ‘borderline endometrial cells’ category in existence for the inclusion of women in whom the atypical cells may be of endometrial origin. However, we have not used this category since the introduction of LBC in the department, and endometrial cells, when present, are classified as either ‘normal endometrial cells’ or atypical and included in the ‘?glandular neoplasia’ category.

The departmental policy at the Northern Gynaecological Oncology Centre, Gateshead, during this period was that all cases of borderline glandular cells were discussed after the colposcopy visit at the weekly colposcopy correlation meeting held between the colposcopists and cyto-/histopathologists. The overall departmental policy for the assessment of these cases since the publication of previous conventional cytology series9 was to use punch biopsy for low-grade colposcopic changes and large loop excision of the transformation zone (LLETZ) for high-grade changes. Random punch biopsies were not used when colposcopy was satisfactory and normal. Instead, LLETZ in most cases with suspicious symptoms and repeat cytology with colposcopic assessment in most asymptomatic cases were carried out. Persistent abnormality on repeat cytology was subjected to LLETZ if there was continued normal and satisfactory colposcopy. Endometrial biopsy with pipelle was performed on women over 35 years of age, particularly if they were symptomatic or if colposcopy was normal and satisfactory.

Between the study period of June 2006 and August 2008, 38 244 cervical LBCs were screened from the Gateshead population. Seventy-two (0.19%) of all LBCs screened were reported as borderline glandular cells. They were identified from the cervical cytology database. On the basis of NHSCSP guidelines and current recommendations, all 72 women were referred for colposcopic assessment after one reported incident of borderline glandular cytology. Three women were excluded from this study as they did not attend colposcopy appointments. This study reviewed the remaining 69 women.

The case notes and cytology, pathology and colposcopy databases of the 69 women were reviewed to obtain the following data: age, menopausal status, use of exogenous hormones, use of intrauterine devices, previous cervical treatment, indication for the index smear resulting in the report of borderline glandular cells, colposcopic findings, procedure performed at first and subsequent assessments, subsequent histopathology reports and final diagnosis.

Final histological diagnoses were based on cervical punch biopsy, loop excision of the cervix, pipelle endometrial sampling or surgical hysterectomy specimen. Some women had diagnostic loop excisions performed on subsequent visits after showing persistent abnormal cytology despite normal and satisfactory colposcopy.

All data were tabulated and analysed statistically using SPSS® (SPSS Inc, Chicago, IL, USA). Fisher’s exact test of significance was used to correlate significant final histology (defined as all grades of intraepithelial neoplasia or worse) with menopausal status, use of exogenous hormones, previous cervical treatment, appearance of the cervix and indication for the index smear. A value of P < 0.05 was considered to be significant. All high-grade histological diagnoses were analysed separately and further age-based group analyses were performed.

Results

The average age of the 69 women was 39.6 years, ranging between 24 and 63 years. Twenty-three women (33.3%) were on some form of hormonal treatment: 12 (17.4%) were on the combined oral contraceptive pill, two (2.9%) on Depo-Provera, one (1.4%) on progestogen implant, one (1.4%) on Mirena intrauterine system, one (1.4%) on progestogen-only pill, three (4.3%) on hormone replacement therapy and three (4.3%) on progesterone therapy for menorrhagia. One woman (1.4%) was under 3 months postpartum and one (1.4%) was using an intrauterine contraceptive device. Forty-four (63.8%) women were not on any hormones or intrauterine contraceptive device. Sixty-two women (89.9%) were premenopausal and seven (10.1%) were postmenopausal. Table 1 shows various variables and their association with the final histology.

Table 1.   Variables and association with histology
 n (%)HG (CIN ≥ 2)LG (CIN 1)Normal or benignNo histology
  1. CIN, cervical squamous intraepithelial neoplasia; HG, high-grade; LG, low-grade.

  2. *Some women had more than one symptom.

Contraception/hormones
None44 (63.8)9 (20.5)3 (6.8)29 (65.9)3 (6.8)
Combined oral contraceptives12 (17.4)8 (66.7)04 (33.3)0
Depo-Provera2 (2.9)01 (50.0)1 (50.0)0
Implant1 (1.4)1 (100)000
Progestogen-only pill1 (1.4)01 (100)00
Mirena intrauterine system1 (1.4)1 (100)000
Intrauterine device (nonhormonal)1 (1.4)001 (100)0
Other oral progestogens3 (4.3)1 (33.3)1 (33.3)1 (33.3)0
Hormone replacement therapy3 (4.3)1 (33.3)02 (66.7)0
Postpartum (<3 months)1 (1.4)001 (100)0
Menopausal status
Premenopausal62 (89.9)20 (32.3)5 (8.1)35 (56.4)2 (3.2)
Postmenopausal7 (10.1)1 (14.3)1 (14.3)4 (57.1)1 (14.3)
Indication
Routine + recall50 (72.5)18 (36.0)3 (6.0)28 (56.0)1 (2.0)
Previous abnormal smear9 (13.0)3 (33.3)2 (22.2)4 (44.4)0
Previous inadequate smear2 (2.9)001 (50.0)1 (50.0)
Opportunistic2 (2.9)01 (50.0)1 (50.0)0
Follow-up after treatment3 (4.3)002 (66.7)1 (33.3)
Symptoms2 (2.9)002 (100)0
Suspicious-looking cervix1 (1.4)001 (100)0
Symptoms*
Asymptomatic36 (52.2)15 (41.7)5 (13.9)13 (36.1)3 (8.3)
Postcoital bleed20 (29.0)4 (20.0)016 (80.0)0
Intermenstrual bleed9 (13.0)009 (100)0
Menorrhagia6 (8.7)1 (16.7)05 (83.3)0
Vaginal discharge10 (14.5)3 (30.0)1 (10.0)6 (60.0)0
Appearance of cervix
Normal47 (68.1)11 (23.4)4 (8.5)30 (63.8)2 (4.3)
Ectropion19 (27.5)9 (47.4)2 (10.5)7 (36.8)1 (5.3)
Suspicious2 (2.9)1 (50.0)01 (50.0)0
Polyp1 (1.4)001 (100)0

Eight women (11.6%) had a history of previous cervical treatment. This was LLETZ in seven women and laser treatment of the cervix in one woman.

The analysis of indication for the index smear revealed that 50 smears (72.5%) were part of routine screening recalls, nine (13%) were follow-ups after previous abnormal smear, two (2.9%) were previous inadequate smears, two (2.9%) were opportunistic, three (4.3%) were yearly recalls from previous abnormal histology, two (2.9%) were for symptoms and one (1.4%) was for a woman who underwent cervical smear outside of the screening programme because of a suspicious looking cervix when she attended the general practitioner for other gynaecological reasons.

Thirty-six women were asymptomatic. Among the 33 symptomatic women, 20 had postcoital bleeding, nine had intermenstrual bleeding, six had menorrhagia and 10 had vaginal discharge. Of the 20 women who had postcoital bleeding, two were subsequently found to have cervical carcinoma (one squamous cell carcinoma and one adenocarcinoma) and two had high-grade CIN. One had a benign cervical polyp, two had cervical endometriosis, seven had other benign changes and six had normal histology. Two women with cervical and one woman with endometrial cancer were asymptomatic.

The appearance of the cervix was recorded at the time of the index borderline glandular LBC smears. Forty-seven women had a normal-looking cervix, 19 had ectropion, two had a suspicious-looking cervix (one noticed during routine recall smear testing and one from a smear because of a suspicious-looking cervix identified outside of the screening programme) and one had a cervical polyp. Of the 19 women with ectropion, there were two cervical carcinomas (one squamous cell carcinoma and one adenocarcinoma), two CGIN, eight CIN and five benign abnormalities. Of the two women with a suspicious-looking cervix, one had adenocarcinoma of the cervix and the other had chronic cervicitis. The remaining woman who was found to have cervical cancer had a normal ectocervical appearance.

Histology was available for 66 women. The remaining three women underwent cytological surveillance and, to date, have not been diagnosed with any abnormality. There were five (7.2%) cancers in total, including three cervical adenocarcinomas, one cervical squamous carcinoma and one endometrial adenocarcinoma. Twenty-two women (31.9%) had intraepithelial neoplasia (27.5% CIN and 4.3% CGIN) and 30 (43.5%) had benign and inflammatory changes. Nine women had normal histology (13.0%). Table 2 details the final diagnoses of all the women.

Table 2.   Final diagnoses of the study group
DiagnosisFrequencyPercentage
  1. CGIN, cervical glandular intraepithelial neoplasia; CIN, cervical squamous intraepithelial neoplasia.

Malignant57.2
Adenocarcinoma of the cervix34.3
Squamous cell carcinoma11.4
Endometrial adenocarcinoma11.4
Intraepithelial neoplasia2231.9
CIN 168.7
CIN 268.7
CIN 3710.1
High-grade CGIN34.3
Benign/inflammatory3043.5
Inflammation1420.3
Human papillomavirus changes68.7
Tubo-endometrial metaplasia45.8
Endometriosis22.9
Microglandular hyperplasia22.9
Polyp22.9
Normal913.0
No histology34.3
Total69100.0

Twenty-one women were under 35 years of age and 48 were aged 35 years or more. All seven women under 35 years of age with suspected high-grade lesions on colposcopy had histologically confirmed CIN 2 or worse. No women under 35 years of age with normal and satisfactory colposcopy had premalignant or malignant lesions. However, three of 19 women over 35 years of age had high-grade CIN, CGIN or invasive squamous cell carcinoma despite normal and satisfactory colposcopy. Of the seven women over 35 years of age with high-grade colposcopic impressions, three had a normal or benign final histological diagnosis. Figure 1 shows the correlation of colposcopic impression with the final cervical histology.

Figure 1.

 Correlation of colposcopy and cervical histology. SCJ, squamocolumnar junction.

Final cervical histological diagnoses were based on at least LLETZ or a larger specimen in 48 of the 69 women (70%): 42 LLETZ, three total hysterectomy, two radical hysterectomy and pelvic node dissection and one radical abdominal trachelectomy specimen.

Of the 21 women (30%) who underwent either a punch biopsy (18) of the cervix or cytological surveillance alone (three), all but one had subsequent negative cytology and were discharged from the colposcopy clinic after a median follow-up of 4 months (range, 1–17 months) and after a colposcopy correlation meeting review. One remaining woman with low-grade squamous abnormality on repeat cytology continues to be followed up in the colposcopy clinic. Figure 2 shows the source of final cervical histology with follow-up information relating to all conservatively managed cases.

Figure 2.

 Final cervical histology and subsequent follow-up.

Twenty-seven women (39.1%) aged between 30 and 61 years underwent endometrial sampling. Sixteen had menstrual symptoms recorded. Twenty-three of these 27 women were premenopausal, and four were postmenopausal. Four samples were inadequate for conclusive analysis. Two had hysteroscopy and endometrial biopsy, which were normal. The one case of endometrial carcinoma in this series was suspected on pipelle endometrial sampling. She was a 40-year-old premenopausal asymptomatic woman with normal colposcopy. Following discussions in the multidisciplinary team meeting, she underwent laparoscopy-assisted vaginal hysterectomy and bilateral salpingo-oophorectomy, which confirmed International Federation of Gynecology and Obstetrics (FIGO) stage 1A endometrial adenocarcinoma. She has had regular follow-up since then and is currently alive and well.

Of the four cervical cancers, two women with FIGO stage 2A and 1B1 cancers had radical surgery. Of the remaining two women with FIGO stage 1B1 cancers, one had a radical abdominal trachelectomy and the other was managed by repeat LLETZ and laparoscopic bilateral pelvic lymph node dissection. None of the four women needed further treatment and all four women are currently alive and well.

Analyses of menopausal status, history of previous abnormal cytology or histology, indication for the index smear and abnormal appearance of the cervix did not identify any statistically significant association with histological outcome. However, the use of exogenous hormone preparations was associated with significant cervical lesions (P* = 0.004). The ability of colposcopy to differentiate high-grade lesions (CIN 2/3) from all others (normal and low grade) was analysed. The overall positive predictive value of colposcopy for high-grade disease was found to be 78.6%, with a negative predictive value of 81.3%.

Discussion

The incidence of cervical cytology representing ‘borderline glandular cells’ has been reported to be between 0.07% and 1.80%.8,9,12 The incidence of borderline glandular cells in our series based on the BSCC classification and following the introduction of LBC was 0.19%. This is higher than the previous series from our department of borderline glandular cells after conventional cytology, which was 0.07%.9 This increase in incidence using LBC appears to reflect an increase in the identification of benign lesions being reported as borderline glandular cells. Although increased accuracy in the discrimination of benign lesions, and therefore reduced over-reporting, was seen in an earlier study on LBC,13 our results do not reflect this finding. The increase in incidence with apparent greater identification of benign lesions may also explain the relative reduction in premalignant (31.9%) and malignant (7.2%) lesions on LBC relative to the previous series using conventional cytology (34.9% and 16.2%, respectively).

Nevertheless, in this review, 39.1% of borderline glandular cells reported on LBC were associated with clinically significant lesions, representing either a premalignant abnormality (either CIN or CGIN) or a malignancy. This represents a much higher association with clinically significant lesions in comparison with its squamous counterpart, borderline squamous change (reported rates of clinical lesions after one report of borderline squamous cells are 5–13%),14,15 and confirms the need for prompt referral and management.

Of the 27 cases diagnosed with intraepithelial neoplasia or invasive neoplasia, only seven (25.9%) were confirmed to be true glandular lesions (10.1% of the 69 cases), with the remaining 20 cases (74.1%) of squamous origin. This finding highlights the need for a thorough and detailed colposcopic examination and, to a large degree, explains the positive predictive value of 78.6% for colposcopy in this series, a result which is not very dissimilar from the positive predictive value expected to be found following colposcopy in women referred with purely squamous cytological abnormalities. Although the majority of glandular lesions were either found in the transformation zone or contiguous to the transformation zone, the sensitivity of colposcopy for the detection of glandular lesions has been questioned and considered to be limited.14 The incidence of glandular lesions increases with age, and they are relatively less common in younger women. Glandular lesions are also more likely to be multifocal and deeper in the endocervical canal in older than younger women.16,17

Despite this, the colposcopic impression in our series was the most important predictive factor for the presence or absence of significant disease in women under 35 years of age. No women in this group had an unsatisfactory colposcopy, and a negative colposcopy was associated with a 100% negative predictive value. That is, colposcopy was able to confidently exclude significant pathology in all women under 35 years of age by a normal and satisfactory colposcopic examination. As the avoidance of performing unnecessary LLETZ procedures in this group of women is considered to be particularly important, we would recommend that, if colposcopy is normal and satisfactory, adequate sampling of the endocervical canal with an endobrush and the avoidance of a loop procedure would be an acceptable step in an otherwise asymptomatic woman of less than 35 years of age. Follow-up with repeat endocervical cytology in this group of women can be reasonably relied upon. The endocervical cell component of cytology with endobrush, in conjunction with broom, has been shown to be significantly superior relative to broom alone in many reports,18 including a Cochrane review.19 If repeat cytology continues to be abnormal, a diagnostic LLETZ should be carried out following cytological review and discussion in a colposcopy correlation meeting held jointly with the cytopathologists.

The use of LLETZ at the first visit, however, appears to be justifiable in women under 35 years of age when a high-grade lesion is seen on colposcopy, or if colposcopy is unsatisfactory in the presence of other risk factors.

In women aged 35 years or more, colposcopy did not identify all clinically significant lesions. In this series, one case each of low-grade CIN, high-grade CIN, CGIN and cancer were not identified after normal and satisfactory colposcopy. In addition, there were seven unsatisfactory colposcopies in this group in which the squamocolumnar junction was not visible; four of these women were premenopausal. On the basis of these findings, we would recommend a diagnostic LLETZ/loop cone biopsy in women aged 35 years or more, regardless of the colposcopic impression. Recommendations on endometrial sampling are difficult to make in this group of women on the basis of only one case of endometrial cancer in a 40-year-old asymptomatic women in our series. However, we would suggest that this group of women should undergo endometrial sampling as part of their overall assessment, at least until further evidence becomes available on LBC and, in particular, if they are symptomatic and when colposcopy is normal and satisfactory.

These recommendations are based on a small case series of 69 women who attended colposcopy after a single smear showing borderline glandular cells on LBC. Despite being a small series, it is the first report regarding the outcome of this relatively rare category of smear abnormality following the introduction of LBC in the UK.

The introduction of LBC across all NHS laboratories in the UK has presented significant challenges for cytoscreeners, advanced practitioners and cytopathologists, and has necessitated the development of new skills through retraining for the widespread transition from conventional cytology. Uncertainties on the impact of this transition, specifically on the reporting of glandular lesions and the histological and clinical outcomes of women reported to have borderline glandular cells, have been identified previously.10 Although this current review provides reassuring data that the clinical outcomes of women with borderline glandular smears is broadly similar to that seen previously following conventional cytology,9 larger series from other UK laboratories and colposcopy clinics are desirable. In particular, they would provide more comprehensive information on the use of age and colposcopic findings to determine the management and use of additional diagnostic procedures, including LLETZ, loop cone biopsy, endometrial sampling and repeat cytology.

Although the results of this case series can provide useful information to formulate local colposcopy protocols, we recommend that, as is our current practice, all cases of borderline glandular cells should be discussed at multidisciplinary colposcopy correlation meetings to ensure that optimal management is performed in each case and that the prompt diagnosis of clinically significant lesions occurs without resorting to the excessive use of excisional procedures, especially in younger women.

In conclusion, the incidence of LBC in the current UK cervical screening programme showing ‘borderline glandular cells’ is 0.19% in our series; 82.6% of reported borderline glandular cells on LBC are associated with histologically confirmed clinical lesions; 39.1% of women with borderline glandular cells on LBC have premalignant (31.9%) or malignant (7.2%) lesions. On the basis of our results, it would be considered to be acceptable to manage women under 35 years of age with normal and satisfactory colposcopy conservatively. In women above 35 years of age, we would recommend a diagnostic LLETZ procedure, irrespective of the colposcopic findings. All cases should be discussed at multidisciplinary colposcopy correlation meetings.

Disclosure of interests

None of the authors have any conflict of interest.

Contribution to authorship

RN conceived and designed the project. AP, NT and JDH collected cytological and histological data. AP collected clinical data and performed the analysis. JDH and RN contributed to the analysis and appraisal of the data. AP drafted the paper which was critically appraised by all authors. RN and JDH contributed to the final draft of the paper. All authors approved the final manuscript.

Details of ethics approval

Not applicable.

Funding

Not applicable.

Acknowledgements

We would like to acknowledge Mr Martin Jones (Bio-medical Scientist, Department of Pathology, Queen Elizabeth Hospital, Gateshead, UK) for his help with data collection from cytopathology databases.

Ancillary