Although cervical cytology screening has produced a clear reduction in the incidence of and mortality from cervical cancer, this reduction has been restricted to squamous cell carcinoma. A similar improvement has not been seen for cervical adenocarcinoma. At best, screening has had only a minor impact on this disease in limited settings where there is highly developed screening activity, which includes the UK. In fact, numerous studies from recent decades have reported increasing incidence rates of adenocarcinoma in many countries, particularly in young women. The reasons for these increases include better histological assessment, which encompasses the increased recognition of both in situ and invasive lesions. However, increases in the prevalence of high-risk human papilloma virus (HPV) in successive birth cohorts and an inability to accurately screen for precursor lesions are the most important factors underlying the observed increases in the incidence of invasive adenocarcinoma.1
There is strong circumstantial evidence that high-grade cervical glandular intraepithelial neoplasia is a precursor to invasive disease. The accurate identification of these lesions is more difficult than for squamous precursors, and it is generally thought that many glandular lesions remain undetected by cytological screening, accounting for their relative infrequency compared with their squamous intraepithelial counterparts.2
Lesions of endocervical cell origin are more difficult to detect because of their location, but it is also recognised that precise histological and cytological diagnosis of glandular pathology are areas of particular challenge for pathologists.
The underperformance of screening may be further explained by the relative rarity of the lesion: <1% of abnormal samples show glandular abnormality, and these abnormalities can be overshadowed by their common association with more obvious squamous cell lesions.
The cytological prediction of glandular abnormalities was first described in the 1970s,3 and evidence has accrued gradually over the ensuing 40 years, first for identification in conventional cytology and latterly for liquid based cytology (LBC).4
Most of the data concerning these lesions comes from small, single-institution, hospital-based observational studies, often with poor histological correlation. Cytological diagnosis of glandular abnormality is generally agreed to be less precise than for squamous cell precursors. The first problem is of poor specificity.5 Cytological prediction of high-grade glandular neoplasia is associated with a glandular pathology only in approximately one half of cases.6 Confusingly, squamous lesions, non-cervical glandular neoplasia and certain benign lesions, can generate samples resembling cervical glandular atypia. The awareness of this has led to the introduction of the borderline glandular category, which although less predictive of glandular neoplasia nevertheless appears to be associated with a significant risk of high-grade pathology of both glandular and squamous differentiation.4
The second problem concerns the reproducibility of cytological diagnosis, which has been shown to be poor even between expert cytopathologists. Thirdly, international comparisons of data are undermined by the varied reporting systems for these lesions, and variations in quality control between laboratories. Most data relates to the various incarnations of the Bethesda system, which differs from the NHS Cervical Screening Programme.4
These difficulties with cytology are clearly of concern to colposcopists. Although colposcopy is recommended for investigating atypical glandular cytology, it also performs poorly in identifying glandular pathology.7 Hence, colposcopists may feel obliged (albeit reluctantly) to perform fairly radical cervical excision procedures, with their accompanying surgical and reproductive morbidities, often in young women wishing to retain their fertility, in order to ascertain whether disease is present and adequately removed. Inevitably, some women will ultimately be shown to have had unnecessary treatment. In contrast, it has been observed that there can be reluctance to perform excisions in the presence of negative colposcopic evaluations, even where guidelines clearly advocate removal,8 which risks missing significant lesions. This highlights the dilemma of managing suspected glandular abnormality.
More recently, concern has been generated by the introduction of liquid-based cytology. This relates to the observation that the architectural features that enable the diagnosis of this lesion might be altered by this method. Fortunately, early studies indicate that the principal diagnostic features persist with cervical broom devices and liquid-based cytology, and although subtle differences between conventional and liquid-based presentation are observed, these differences are minimal.9 Indeed, improved discrimination between true glandular neoplasia and benign look-alikes has been reported.10 It is, however, not clear whether the different preparation systems (i.e. ThinPrep® or SurePath®) might influence the accuracy of cytodiagnosis.
In the NHS Cancer Screening Programmes (NHSCSP), and within the Bethesda cytological classification, glandular abnormalities are categorised as suspected high-grade lesions when there is clear evidence of glandular dyskaryosis. Within NHSCSP, a single category of ?glandular neoplasia is used for suspected cervical glandular intraepithelial neoplasia (cGIN) or invasive endocervical, endometrial or extra uterine adenocarcinoma. In the Bethesda system, there is a separate category for suspected adenocarcinoma in situ.
For those in which there are morphologic changes in glandular cells falling short of frank dyskaryosis, a ‘borderline’ report, with additional free text stating that the abnormality is within glandular cells, is issued in the UK. In the Bethesda system the category of ‘atypical glandular cells’ (AGC) is broadly equivalent, although this category is further subdivided according to the likely cell of origin and whether or not the cytopathologist ‘favours neoplasia’.
The main problem relates to the borderline glandular category, for which there are limited robust clinical outcome studies. In contrast to (squamous) borderline or atypical squamous cells of uncertain significance (ASCUS) samples, the limited evidence suggests a substantial likelihood of significant pathology in the borderline glandular group.11 Although the most frequent lesion encountered is a squamous intraepithelial abnormality, there are also instances of invasive cervical and endometrial neoplasms presenting with borderline/atypical glandular cells.
It is impossible to audit the clinical outcomes following the diagnosis of borderline glandular abnormality in the UK because of a lack of a distinct category identifiable through the Exeter system (a national database of women’s screening history).5 There is divergence of opinion between cytologists and colposcopists, and between Scotland and the rest of the UK, on the precise management of women with borderline glandular reports. Furthermore, reporting may be influenced by local referral practice. There is considerable variation between cytologists in their reporting of the borderline category, and there is support from cytologists for the introduction of a distinct code for borderline nuclear change/glandular smears.
Therefore, the paper by Patel et al.12 in this issue is a useful contribution to the limited literature on this grade of glandular abnormality in the era of liquid-based cytology. The study group appear to represent a routinely screened group of women (and 93% of samples were appropriately screened); however, over 50% of the women in this study reported abnormal or excessive bleeding, 33% were on exogenous hormones and the mean age was 39.6 years. This suggests a considerable degree of selection bias. There were good histological ascertainment rates (96%), although only 70% had, as a minimum, a transformation zone excision. Such an excision should be the minimum prerequisite for the confident exclusion of a glandular lesion.
The laboratory in question appears to process around 18 000 samples a year, reporting 32 borderline nuclear change/glandular samples annually. This, in the main, appears to be associated with the inclusion of a high proportion of cases found to be benign look-alikes on histological examination, and contradicts the results from an audit of glandular reporting from before and after the liquid-based cytology rollout.10
The study findings underline the need to promptly refer women with borderline nuclear change/glandular results to the colposcopy clinic, and to involve the multidisciplinary team, but the specific management plans advocated are debateable.
The limitations of colposcopy were highlighted in the group of women aged over 35 years, where one invasive cancer and two intraepithelial lesions were not recognised. There was also overcalling of some high-grade lesions in this group (three of seven predictions).
The recommendation for continued monitoring rather than excision in women aged under 35 years, where colposcopy appeared more reliable, should be interpreted with caution because of the small numbers involved (21 subjects), and four high-grade lesions were undercalled in this group. Also, for the 30% of the group who were not subjected to transformation zone excision, given the unreliability of colposcopy and punch biopsy to diagnose glandular lesions, it is not possible to absolutely exclude the presence of a glandular lesion, especially as follow-up information is limited to 4–6 months. Further careful follow-up of this group is clearly desirable. The study also fails to detail the degree of experience of the colposcopists. It would be appropriate for all such cases to be assessed by a senior colposcopist.
The recommendation for excisional treatment in women aged over 35 years on the basis of a single borderline nuclear change/glandular result is surprising, given the data. The authors report that the liquid-based cytology technique has led to an increase in diagnosis of benign lesions, compared with previous experience. Approximately 19% of this group had a high-grade intraepithelial (14.6%) or invasive (4.4%) lesion. Of the four cervical cancers in the study, two reported post-coital bleeding and one had a suspicious looking cervix at referral. Hence, the possibility of more selective use of excision biopsy after one colposcopy visit, with further review by a senior colposcopist, and multidisciplinary team involvement, might be considered. We agree with the authors that additional larger series are essential to help define management policies for the future.
Can adjunctive tests help to refine diagnosis? HPV testing has been suggested to have a role in the triaging of women with atypical cervical samples.13 Given the probable continuing shortcomings of colposcopy and cytology, the role of HPV screening as an adjunct to cytological diagnosis of glandular lesions should be assessed.
Some strains of HPV have an established role in the aetiology of both intraepithelial and invasive glandular lesions. In recent studies high-risk subtypes were found in 100% of cervical adenocarcinoma in situ (AIS) and in 93.5–97.5% of invasive adenocarcinomas.14,15 HPV types 16 and 18 are present in more than 93% of AIS lesions. HPV-18 appears to be preferentially associated with the development of glandular neoplasia,14,15 whereas HPV-16 can be an aetiological factor for both squamous and glandular disease. The evidence suggests that HPV-18-associated AIS may be more rapidly progressive, and/or that HPV-16-associated disease is easier to detect and to therefore eradicate.15
The corollary of these observations is that high-risk HPV testing might act as an important safety net for the failure of cytology-based screening techniques to detect glandular pre-malignancy. The incorporation of such testing offers the potential to reduce the incidence of adenocarcinoma. HPV negativity may also be a useful aid to diagnosis in the presence of possible glandular neoplasia, in distinguishing endocervical from endometrial neoplasia, and other benign lesions.
Although the data on primary prevention of adenocarcinoma by vaccination is encouraging,16 uncertainty remains about the duration of immunity, and the developing world may not be able to afford the vaccine. Hence, secondary preventive efforts should not be abandoned, but they need to be pursued with discretion. Reducing the incidence and mortality of adenocarcinoma is one of the remaining major challenges for cervical screening programmes; however, an excessively aggressive approach will lead to reproductive morbidities, such as an increase in the incidence of subsequent preterm labour. Hence, collaborative studies involving cytopathologists and gynaecologists are vital to improve diagnosis, and to limit therapy to those of highest risk.
An NHS evaluation of atypical glandular cytology has been sanctioned and supported by the NHSCSP, with the aim of improving the precision of diagnosis of glandular lesions. This will include HPV analysis of the residues of cytological samples, a central slide review of presenting cytology, and an evaluation of the contribution of colposcopy to diagnosis. It is hoped that a panel review of these rare cytological diagnoses will lead to more robust morphological definitions of grades of atypical glandular cytology, to help cytopathologists reach an accurate diagnosis, and to determine whether a particular type of liquid-based preparation is superior to others. At the very least it should provide accurate data on the predictive value of atypical glandular cytology of high and borderline grades, which might influence subsequent guidelines. This is an ambitious project that requires early identification and registration of these uncommon cases by cytopathologists liaising with their colposcopist colleagues.