CHAPTER 24 Glandular neoplasms of the cervix
Chapter contents
Introduction
Glandular tumours of the cervix present diagnostic and management difficulties affecting both pathologist and clinician. This chapter examines the current classification, terminology, diagnostic cytological features and pitfalls in the diagnosis of preinvasive and invasive neoplasia of endocervical cell origin, including the less common cell variants that reflect the müllerian derivation of the lining of the endocervical canal. Modern diagnostic techniques are reviewed and, in conclusion, the clinical management of cytological glandular abnormalities of the cervix in relation to the role of the cytopathologist is discussed.
Epidemiology
Incidence
Evidence is accumulating which shows that the incidence of adenocarcinoma of the cervix is rising1–5 but following analysis of data from 60 population-based cancer registries, it appears that the trends are complex.6 While an increase in incidence of adenocarcinoma and adenosquamous carcinoma has been shown in young women in many countries, including America, Australia and the UK, in other developed countries, including the Netherlands, Germany and New Zealand, there has been no significant change. A study from Southampton showed no rise or fall in the incidence of adenocarcinoma over a 12–year period in that region.7 In Finland, France and Italy there have been falls in incidence. The increase, where present, is seen particularly in women born from 1935 onwards, with women born around 1955 having a three times greater risk than those born in 1935.
Recent data from Sweden8 showed no clear benefit in the detection of glandular compared with squamous carcinoma through screening. In Denmark,9 a reduction in the incidence of adenocarcinoma has been reported in women over 40 years of age but an increased incidence in 20–29 year old women in spite of improved coverage over a 15-year period. Some10 have suggested that it takes 10–15 years to develop expertise in accurately identifying glandular lesions and others11 predict a decrease in the cumulative incidence of cervical adenocarcinoma in the current decade.
Risk factors
Epidemiological factors for cervical adenocarcinoma are less well-defined than for squamous carcinoma. As with squamous carcinoma, the increase in cervical adenocarcinoma is related to the number of sexual partners and intercourse at an early age.12 In a meta-analysis by Berrington de González13 high parity and long duration of oral contraceptive usage were associated with both histological types; however, adenocarcinoma showed a significantly lower association with current smoking than squamous carcinoma. Other factors affecting recorded incidence include changes in reporting practice, with increasing awareness of glandular lesions,14 changes in choice of sampling devices leading to better sampling of the endocervical canal15 and the effects of organised screening programmes leading to a relative decrease in squamous lesions.6
There is strong evidence that up to 50% of endocervical adenocarcinomas are associated with cervical intraepithelial neoplasia (CIN). It is tempting to speculate that both lesions share a common pathogenesis, namely an aberrant proliferation of reserve cells, which may result in the formation of either a glandular or a squamous neoplasm. Human papillomavirus (HPV) types 16 and 18 have been demonstrated in invasive and intraepithelial endocervical neoplasia, supporting the concept of a common pathogenesis for at least some lesions.16–20 HPV testing is likely to prove useful in the investigation of borderline glandular lesions,21 although on histology many of these will reveal high grade CIN.22
Finnish16 and Dutch studies23 demonstrated that adenocarcinoma is more likely to be associated with HPV 18 than with HPV 16. Adenocarcinoma and squamous carcinoma also differ in intratypic variants of HPV strains.24 It has also been shown that HPV-negative cervical carcinoma is more likely to be adenocarcinoma.25
Endocervical adenocarcinoma precursor lesions
The endocervical canal is lined by columnar epithelium, which forms a single layer over the stromal ridges, villi and crypts. With the onset of neoplasia, the single layering is disturbed, initially by the development of pseudostratification. Cytologically this appears as overlapping and crowding of nuclei in sheets of epithelial cells. Most endocervical adenocarcinomas are of an endocervical type, but müllerian epithelium has the capacity to differentiate along several pathways, with the result that some tumours have histological features that closely resemble those usually arising in the endometrium or ovary. müllerian epithelium also readily undergoes metaplastic change and so, for example, an adenocarcinoma of enteric type may be seen.26 This diversity of histological pattern is reflected in the classification of endocervical tumours.27
Cervical glandular intraepithelial neoplasia (CGIN): adenocarcinoma in situ (AIS)
This lesion was first described histologically in 1953, by Friedell and McKay.28 They noted that the average age of patients with CGIN was several years less than that of patients with invasive disease, suggesting that the in situ form precedes the development of invasive cancer. This concept has been supported by other studies.29–31
On histology CGIN retains the architectural pattern of normal endocervical crypts. Usually the surface epithelium and both superficial and deep crypts are involved at or near the squamocolumnar junction. Partial crypt involvement is a frequent finding, with an abrupt transition from normal to neoplastic epithelium (Fig. 24.1).32 High-grade (HG-) CGIN is usually a single lesion, less often multifocal and frequently extends into the endocervical canal.33
Cases with an endometrioid or intestinal pattern (Fig. 24.2)29 may be seen together with other histological variants, usually as focal areas within a lesion which is predominantly of endocervical type.33 When high-grade CGIN is associated with an in situ or invasive squamous carcinoma, the two cell types may be seen as adjacent areas of abnormality (Fig. 24.3). This almost certainly leads to underdiagnosis of the glandular component in lesions of mixed squamous and glandular types.34
While features of high-grade CGIN are relatively well described, low-grade CGIN remains a poorly reproducible entity. Histological criteria for low-grade CGIN have been proposed32 and cytological criteria have also been published, but these are controversial.35–37 There is currently no proof that low-grade glandular atypia is a precursor to adenocarcinoma.38 Atypia in glandular cells often relates to non-neoplastic conditions39 that mimic high-grade CGIN rather than represent low-grade CGIN. Nowadays, this dilemma can be investigated by HPV testing and the use of molecular markers.31
Cytological findings: CGIN/AIS
Cytological manifestations of CGIN were first published by Barter and Waters in 1970.40 Since then, evidence has accrued on which to base a prediction of glandular neoplasia.41–45 In conventional cytology, prediction is usually made on the presentation of cells in sheets and clusters and rarely on abnormality in individual cells alone. Raab et al. identified in conventional cytology the three most useful individual cellular criteria on which to discriminate between benign and neoplastic changes in endocervical cells. These are irregularity of nuclear membrane thickness, raised nuclear/cytoplasmic (N/C) ratio and presence of atypical single cells (Fig. 24.4).44,46,47 Using liquid-based cytology (LBC) there are slight differences compared to conventional smears as the nuclear details are more pronounced and architectural features more subtle.39,48–51 However, comparing conventional cytology and the two most commonly used LBC methods, SurePath (SP) and ThinPrep (TP), Belsley et al. concluded that differences between the three methods are minimal.52
Exfoliation pattern (Fig. 24.5)41,53
Whereas feathering is frequently the most useful criterion to distinguish between squamous lesions and glandular neoplasia in conventional cytology45 with the rounding up of cell clusters in LBC, it is less common in TP and infrequent in SP.52
Nuclear features
Cytoplasmic features
This feature helps in discrimination between the rosettes of CGIN and those of benign entities such as tuboendometrioid metaplasia. The absence or near absence of cytoplasm around the periphery of clusters of glandular cells acts as a trigger to alert the microscopist to the likelihood of glandular neoplasia. Particularly in SurePath, the absence of discernible cytoplasmic clothing of the nuclei frequently leads to a sharp, guttate (rain drop) appearance in clusters from CGIN (Fig. 24.12). In individual cells, and those involved in feathering, the stretching of the cytoplasm over the nucleus is sometimes termed the ‘snake and egg’ effect (Fig. 24.11). Occasionally, goblet cells characteristic of intestinal differentiation are seen (see pitfalls, below).
The classic architectural features of rosette formation, feathering and pseudostratification are manifestations mainly of endocervical-type differentiation of CGIN but co-existing subtypes may be present in up to two-thirds of glandular lesions, although they rarely occur alone.43
Diagnostic accuracy of CGIN/AIS
Interobserver variability in interpretation of glandular entities is well documented.60,61 In both conventional cytology and LBC, sensitivity in detection of endocervical glandular lesions is lower than for CIN.62 The positive predictive value (PPV), which may be used as a surrogate marker for specificity, is also lower.14,63–65 While the former may be the result of sampling failure changes in sensitivity and PPV may be manifestations of interpretation error66 and is influenced over time by changes in awareness leading to alteration in cytological and histological recognition6,67 Initially in the UK, pilot sites examining direct-to-vial screening population samples found lower sensitivity to glandular abnormalities68 but post-pilot audits have demonstrated no change in sensitivity and increased accuracy in discrimination between glandular neoplasia and benign mimics.69
For LBC, use of residual material has been implicated as contributing to poorer sensitivity in split-sample studies70,71 and, although lack of familiarity with the cellular presentation, in particular the absence of feathering, has been deemed likely to be a significant additional factor,65,72 some workers have found little change in architectural features traditionally used for identification of glandular neoplasia and recognition of benign look-alikes.45,39,59 In one centre, changeover from conventional smear preparation to LBC reduced the false negative rate for glandular neoplasia from 43.6% in conventional to 15.4% in thin-layer samples and reduced false positive reports associated with squamous look-alikes from 30.4% to 11.1%, thereby demonstrating both enhanced sensitivity and accuracy.73
Cell block preparations have been used from residual material from LBC vials as a further aid to interpretation.74,75 Identification of glandular neoplasia based on examination of well-presented three-dimensional microbiopsies has been found to be more useful than reliance on feathering and other artefacts used in conventional preparations.
Diagnostic pitfalls: CGIN/AIS
A comprehensive account of the interpretive difficulties in conventional cytology was published by Crum et al. in 1997.76 These difficulties are reflected in the published positive predictive value (PPV) estimations for glandular prediction. PPV estimations vary from 77.4% for firm predictions to 25.9% for samples showing less definite features of glandular neoplasia.63,77 With the advent of LBC techniques several authors have reported reduction in false positive reporting associated with mimics.39,59,60,69,78,79 The following however, may cause some diagnostic confusion (Table 24.1).
Table 24.1 Cytological prediction of CGIN: Diagnostic difficulties and look-alikes (see Ch. 25)
| Non-neoplastic | Neoplastic |
| Cervicitis | Gland crypt involvement in CIN |
| Endocervical polyps | Type II CGIN |
| Tubal metaplasia | Early invasive adenocarcinoma |
| Endometrial hyperplasia and neoplasia | |
| Endometriosis | Extrauterine carcinoma |
| Microglandular hyperplasia | |
| Arias-Stella reaction | |
| Isthmic (lower uterine segment) sampling |
Severely dyskaryotic squamous cells
Cells from gland crypts involved in CIN pose major diagnostic difficulties.14,46,53,62,80,81 In conventional cytology, these are often associated with the use of the more pointed samplers such as the Aylesbury spatula and endocervical brushes. Selvaggi has described the cytological manifestations of CIN with crypt involvement in both conventional and LBC preparations.81,82 In hyperchromatic crowded groups the central areas show architectural anarchy with piling up of dyskaryotic nuclei. In LBC, mitotic figures and nucleoli are more readily seen. Nucleoli, if present in squamous lesions tend to be smaller than those seen in glandular neoplasia. These features are seen in both ThinPrep and SurePath preparations (Figs 24.16, 24.17).
Although cellular disorder is to be anticipated at the periphery of cell groups from CIN, occasionally architecture akin to CGIN is present with elongated nuclei protruding from cell clusters,82 with peripheral palisading and discernible pseudostratification and even an impression of feathering (Fig. 24.18).76 Features that help to identify squamous neoplasia include smooth chromatin texture in nuclei which have evenly distributed chromatin, or chromatin clumping and clearing in those with maldistribution of chromatin. The peripheral nuclei show variation in shape, size and chromasia. A relative denseness of cytoplasm with absence of peripheral cytoplasmic tags also favours squamous over endocervical neoplasia. Irregularity in thickness of nuclear membranes and the presence of nucleoli are more in keeping with glandular neoplasia.44 The nuclear outline in squamous neoplasia is more likely to show sharp notches, not associated with nuclear folds (Table 24.2) (see Ch. 23).
Table 24.2 Characteristics of cell clusters in crypt involvement in CIN versus HG-CGIN
| Crypt involvement | CGIN | |
|---|---|---|
| Group contour | Thick steep-sided microbiopsies | Shallow clusters 2–3 cells deep |
| Group centre | Crowded disordered | Residual honeycomb pattern |
| Group periphery | Haphazard cell arrangement | Palisading or feathering |
| Nuclear morphology | Variable shape, size, chromasia | Relatively even shape, size, chromasia |
| Nuclear membrane | Irregular thickness | Irregular thickness |
| Nuclear outline | Irregular may be notched | Smooth round/oval |
| Chromatin | Usually fine granules | Coarse irregular sized granules |
| May be maldistributed | Commonly evenly distributed | |
| Nucleoli | Small | Prominent, may be large |
| Cytoplasm | Dense smooth edged | Finely vacuolated, wispy edged |
Inflammatory change in endocervical cells
Normal endocervical cells are usually present singly or in small clusters and sheets with minimal nuclear overlapping. In inflammatory conditions, including polyps and cervicitis (Fig. 24.19),44,78,83 they may present in moderately crowded groups but with discernible inter-nuclear spacing. Cytoplasm may be dense, cyanophilic or eosinophilic; terminal bars and occasionally cilia may be seen.84 Where there is peripheral palisading, with crowding there may be a spurious impression of nuclear stratification. In LBC this is sometimes seen also in short strips of glandular cells with well-demarcated cytoplasm and angular borders (Fig. 24.20).51,85 Mild anisonucleosis is common in otherwise unremarkable round or oval nuclei. Chromatin is usually vesicular but may be hyperchromatic and smudged.86 Nucleoli, if present, may be conspicuous and round; very occasional mitoses are seen in regenerating epithelial cells.
Endometrial cells and cervical endometriosis
The presence of endometrial cells in cervical samples can cause confusion. This is particularly so after loop/cone biopsy, trachelectomy or in endocervical brush samples.87–89 The sampler may reach high into the endocervical canal and harvest cells from the lower uterine segment (LUS). These may present as blowzy poorly cohesive cuboidal cells lacking the exfoliative pattern required for a prediction of CGIN, and with delicate vesicular nuclei consistent with cells of endometrial origin. More striking are the large and sometimes branching fragments of crowded glandular tissue. Dense straight-sided tubular microbiopsies with peripheral palisading and associated tangles of delicate stromal cells are characteristic features particularly identifiable on low power inspection (Figs 24.21, 24.22). Capillaries may be observed running through the stromal component and mitotic figures may be evident in samples taken during the first half of the menstrual cycle.63,87,89,90 Adenomyomatous polyps in the lower uterine segment may present cytologically as bland stromal cells in loose frayed clusters. The epithelial content is variable from a few cells to tight packed clusters of crowded nuclei.91
Fig. 24.21 Cervical cytology. LUS. Three-dimensional tube of small uniform cuboidal cells with well-demarcated outline and peripheral palisading in association with stromal fragments (Fig. 26.22) (SurePath).
Fig. 24.22 Cervical cytology. LUS. An untidy tangle of stromal cells with small bland nuclei and ill-defined cytoplasm in association with tube-like microbiopsies of cuboidal cells (Fig. 26.21) (SurePath).
Endometriosis is common in post-cone biopsy crevices but can also be seen in women with no history of previous surgery (Fig. 24.23).92 Conflicting reports have been published of the cytological appearance of superficial cervical endometriosis which have been attributed to hormonal influences. Szyfelbein described similarities with endocervical neoplasia including feathering and Hanau et al. reported macronucleoli which may even lead to an erroneous prediction of adenocarcinoma.93,94 Mulvaney and Surtees described changes similar to those attributed to direct LUS sampling or tubal metaplasia, with absence of feathering and pseudostratification.95 Nevertheless, the endometrioid variant of CGIN should be considered if, in the absence of endometrial stroma, extreme crowding is a feature in samples with small well-preserved endometrial-type cells.57
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
