Cytopathological expression of different types of urothelial carcinoma in situ in urinary bladder washings

Authors


W. Ryd, Department of Pathology and Clinical Cytology, Sahlgrenska University Hospital, S-413 45 Göteborg, Sweden.
e-mail: walter.ryd@path.gu.se

Abstract

OBJECTIVE

To evaluate and compare the cytopathological expression of the five major histological types of carcinoma in situ (CIS) in urinary bladder washings from patients with flat urothelial lesions.

MATERIALS AND METHODS

Seventy-five cases of primary and secondary urothelial CIS with no concomitant tumours, and having tissue and cytological samples, were identified. Biopsies were evaluated based on the consensus classification as: large-cell pleomorphic; large-cell non-pleomorphic; small-cell; clinging; and cancerization of the urothelium. In the cytological classification the ‘clinging’ category was excluded, as its definition depends on the histological appearance. κ statistics were used to evaluate the correlation between histopathology and cytology.

RESULTS

More than one subtype of CIS could often be identified in both the histological and cytological specimens. Cytology often showed more subtypes than did histopathology. Statistically, there was only a moderate correlation between histopathology and cytology for recognising different patterns.

CONCLUSION

Different patterns of CIS can be identified by cytology; it is important for cytologists to be aware of the cytological spectrum of CIS and not to under-diagnose monomorphic, pagetoid (cancerization) and small-cell forms. Studies on treatments for CIS and of the clinical significance of different subtypes of CIS should include both cytopathology and histopathology.

Abbreviations
CIS

carcinoma in situ

H&E

haematoxylin and eosin

LC-P

large-cell pleomorphic

LC-NP

large-cell not pleomorphic

SC

small cell.

INTRODUCTION

In clinical practice the primary application of urinary cytology is to detect high-grade urothelial tumours or carcinoma in situ (CIS), and to monitor patients with a former history of urothelial precancerous or malignant lesions [1–6]. Cytology has the advantage of sampling the entire urothelium and may be more sensitive than many biopsies in predicting tumour recurrence or progression [1,7]. The diagnosis of papillary urothelial neoplasms is not the main task of the cytopathologist. These tumours are usually easily detected by cystoscopy and graded histopathologically. Urothelial CIS are easily missed at cystoscopy, but often better detected on cytology. Thus, cytology, cystoscopy and histopathology are complementary methods for diagnosing urothelial CIS [3]. There are ongoing studies on the molecular mechanisms, genetic and molecular markers for the early detection of urothelial premalignancy, and on markers for the differential diagnosis, prevention and treatment of flat lesions [8–12].

The WHO/International Society of Urological Pathology (WHO/ISUP) consensus classification of urothelial neoplasms describes different types of CIS, and includes severe dysplasia and some forms of dysplasia previously considered as moderate [13,14]. McKenney et al.[15] suggested five major subtypes of urothelial CIS in their recent detailed study, i.e. large-cell CIS with pleomorphism, large-cell CIS with no pleomorphism, small-cell CIS, clinging CIS, and ‘cancerization’ of the urothelium. The last pattern includes pagetoid and undermining or overriding subtypes. It was anticipated that different morphological patterns would present different cytological characteristics and therefore in the present study we determined the contribution of urinary cytology in bladder washing material to the classification of urothelial CIS. However, the clinical significance and any difference in therapeutic response of these subtypes of CIS remain to be determined.

MATERIALS AND METHODS

Details of patients histologically diagnosed as having CIS or severe dysplasia between 1985 and 2002, and with previous urinary bladder washing cytology, were retrieved from the archives of the Department of Pathology and Clinical Cytology, Sahlgrenska University Hospital, Goteborg; 75 were selected for the study (14 women and 61 men, mean ages 72.7 and 73 years, respectively, overall 72.9). The criteria for the selection were: urinary bladder washing material prepared by a filter technique; a histological diagnosis of CIS or severe dysplasia histologically, and malignant or atypical cells cytologically; urinary bladder washing material taken within 3 months before or together with the biopsy; no invasive and/or papillary urothelial carcinoma within 3 months before or after the diagnosis of CIS; no papillary or polypoid lesions seen at cystoscopy or in the biopsies. The histological sections were stained with haematoxylin and eosin (H&E), and cytology material prepared with the filter technique and stained according to Papanicolau.

All cases were selected, evaluated and classified histologically and cytologically by one of the authors (M.A.D.), who also re-examined the histology biopsies together with the uropathologist (F.A.) and the cytological specimens with the cytopathologist (W.R.). Biopsies and cytology of low quality were excluded. Biopsies were classified according to the categories of McKenney et al.[15]. Cytologically, the bladder washings were divided into four groups, large-cell pleomorphic (LC-P), large-cell not pleomorphic (LC-NP), small cell (SC) and cancerization pattern. The κ statistic was used to evaluate any agreement between the histological and cytological subtypes.

RESULTS

The cytological characteristics of the different patterns are given in (Table 1); LC-P is characterized by pronounced polymorphism and a high dissociation tendency, with frequent ‘cell-in-cell’ formations. Nuclei often show folds and/or the shape is irregular (Fig. 1a,b). The LC-NP type is less polymorphic and tumour cells often form three-dimensional groups. Dissociated cells are more uniform than in the LC-P type and the nuclei smoother (Fig. 1c,d). The SC type is composed of small cells in small groups. The cells show some nuclear irregularities. The groups are three-dimensional and there is some nuclear crowding (Fig. 1e,f). Cancerization of urothelium has almost the same characteristics as in the tissue sections. Flat sheets of benign mucosa contain one or more large, atypical cells (Fig. 1g,h). Benign flat sheets with attached atypical large cells at the periphery were also apparent. The clinging pattern was usually a part of a lesion. Only three ‘pure’ clinging examples were found histopathologically. Cytologically, one of these three cases was LC-NP, and the other two a mixture of LC-NP and LC-P. In one of the cases with a histologically extensive clinging pattern (Fig. 2a) there was a pemphigus-like detachment of atypical urothelial cells (Fig. 2b). The cell-in-cell phenomenon (‘cannibalism’) was seen in 65% of the cases.

Table 1.  Cytological features of different types of urothelial carcinoma in situ
FeatureLC-PLC-NPSCCancerization
CellularityHighHighLowUsually with other LC types
General appearance
and grade
Polymorphic high-grade atypia
in large cells with severe
atypia
Large cells with moderate
severe atypia
Low-grade atypiaBenign sheets of urothelium
with scattered large atypical
cells and a dissociated
mainly large cell component
Dissociation
tendency
High and characteristic but
clusters also seen
Moderate: mixture of cell
clusters + dissociated cells
with similar appearance
Low: cells mostly in small
groups, no obvious
dissociation
Cohesive sheets of benign
urothelial cells
NucleusLarge cells with vesicular
chromatin pattern and
convoluted nuc. membrane
‘cell-in-cell’ common.
Smaller cells with dark
eccentric nuclei (‘ink drop’)
Large nuclei with moderate
polymorphism. Nucleus
rounder and membrane
less convoluted than LC-P
Chromatin pattern
usually granular
Small, rounded or oval
nuclei with small folds or
irregularities. Chromatin
pattern fine, dusty or like
ground glass
Large nuclei with dark or
vesicular chromatin pattern,
irregular nuclear contours
NucleolusOne or several prominentVariable but usually not as
prominent as in LC-P
Inconspicuous or invisibleProminent
CytoplasmRelatively richRelatively richRelatively poorRelatively rich
Figure 1.

(a) LC-P sections; histology × 200, H&E; (b) cytology × 400, Papanicolau, showing the cell-in-cell phenomenon in the upper right corner. (c) LC-NP sections; histology × 200, H&E; cytology, × 400, Papanicolau; (d) SC sections; histology × 200, H&E; (e) cytology × 400, Papanicolau; (f) Cancer of the urothelium, histology × 200, H&E; cytology × 200, Papanicolau.

Figure 2.

The typical clinging appearance (a) and pemphigus-like appearance (b). × 200, H&E.

In the histopathology biopsies, there were up to three different CIS patterns in the same biopsy. Thus in the 75 cases, 120 histological patterns were detected. The results are shown in Table 2; κ values showed fair agreement between the histological and cytological results of LC-P and LC-NP types (0.21–0.40).

Table 2.  Subtypes detected by either histopathology or cytopathology or both, and the agreement according to type and method
MethodLC-PLC-NPSCCancerizationTotal
  • *

    Clinging type excluded.

  • H(–)C(–), pattern not seen by either method; H(–)C(+) only cytologically; H(+)C(+) on histology and cytology; H(+)C(–) only histopathologically.

N (%):
Histopathological42 (56)31 (41) 7 (9) 8 (11) 88*
Cytopathological52 (69)40 (53) 11 (15) 6 (8)109
Both59 (79)46 (61)15 (20)12 (16)132
Type, %
H(–)C(–)21.338.780.084.0 
H(–)C(+)22.720.010.7 5.3 
H(+)C(+)46.733.3 4.0 2.7 
H(+)C(–) 9.3 8.0 5.3 8.0 
Agreement68.072.084.086.7 
κ 0.32 0.44 0.24 0.21 

DISCUSSION

Intra-urothelial dysplastic lesions were classified by Koss et al.[2] as intra-urothelial neoplasia I, II and III; the last grade included severe dysplasia and CIS [2]. Urothelial dysplasia is a precursor of CIS and has a high potential for progression to invasive cancer [16]. In the WHO/ISUP consensus classification of urothelial neoplasms [13] and in the proposed nomenclature for subtypes of CIS [15], the CIS concept includes severe dysplasia of the urothelium. Thus, although by definition all CIS are considered as high-grade lesions, there is a morphological spectrum from severe dysplasia of the SC type to the LC-P. The cytological features of these patterns should differ considerably and thus we compared the histological morphology of different subtypes of urothelial CIS with the corresponding cytology in bladder washing material.

The material was assessed retrospectively and it was not possible to consider the effect of different therapeutic measures. According to Koss [2] and Murphy [17], cytotoxic drugs or BCG do not cause epithelial cell changes that could be confused with cancer, but thiotepa may cause minor atypia in the form of cellular and nuclear enlargement.

Some LC CIS could not be subclassified in some of the biopsies and were described only as the LC type; we excluded two such cases from the study.

Fixation may interfere with cell morphology, i.e. alcohol fixation may cause a SC appearance in LC CIS. Milord et al.[18] recently reported a histological morphometric study where they compared CIS cells with lymphocytes and normal urothelium. As a useful and simple morphological criterion they stated that the nuclear area of the 25% largest CIS nuclei was five times the size of lymphocytes, whereas normal urothelial nuclei are twice the size of lymphocytes. Different patterns of CIS were not considered in their report. In the present study, in cytology the nuclear area of the SC type were estimated to be 3–6 times that of a lymphocyte, while large cells in the pagetoid pattern and in the non-pleomorphic type were ª 10 times larger. In the LC-P type the nuclei were 2–20 times larger, or more the size of a lymphocyte.

The LC types were significantly more common in the urinary bladder washings than in biopsies. This increase may be a result of desquamation from poorly adhesive cells in the clinging form. However, as urinary washing material represents more or less the entire surface of the mucosa, patterns not seen in the biopsy may be detected cytologically. Some LC CIS patterns were difficult to classify by cytology; these were usually LC-NP in the biopsy. The typical cytological features of LC-P are an obvious dissociation tendency and pronounced polymorphism. Darkly stained (ink-drop like or coal black) usually eccentrically located nuclei that frequently occur in this type must not be misinterpreted as degeneration in benign cells. If the nuclear membrane is intact and irregular these cells are most likely not degenerated [19].

Cancerization of the urothelium can be seen as pagetoid growth and overriding and undermining forms histopathologically [15], and is not a difficult pattern to recognize in bladder washings. The pagetoid form of cancerization is as typical as that in tissue samples. Furthermore, the pagetoid pattern is perhaps the only one that can be given a definite diagnosis by cytology in pure CIS cases, and in CIS-associated papillary or invasive tumours. Artificially attached atypical cells on the borders of benign mucosal sheets can be seen, but cause diagnostic problems for the overriding or undermining forms of the ‘cancerization’ pattern. Pagetoid growth must not be confused with cell polymorphism within mucosal fragments of CIS. We only accepted cases with scattered atypical cells within benign epithelium as ‘cancerization’. Recently Lopez-Beltran et al.[20] reported the outcome of 11 patients with pagetoid CIS and found it to be the same as for patients with no pagetoid changes.

The main differential diagnoses of SC CIS in bladders with no cystoscopic tumour are other low-grade atypias: dysplasia, reactive atypia or atypia of unknown significance, which remain controversial terms [5,21]. Although the atypia observed in SC CIS was lower than in the LC types, the cells are usually 3–6 times larger than lymphocytes and the nuclear membrane shows slight but discernible folds. The cytoplasm is relatively scant and cell membranes are not distinct; cell clusters are common. For differential diagnosis, slightly enlarged cells and a slightly elevated nuclear/cytoplasmic ratio are not sufficient for a diagnosis of ‘atypia’. A nuclear membrane irregularity is essential and almost always present in neoplastic cells in urinary cytology [19].

The clinging pattern was very frequent in the biopsies (43%) and in three cases was the only pattern. A totally denuded mucosa makes the diagnosis of CIS impossible. Although this pattern is presented as one of those of CIS, it is to some extent an artefact caused by biopsy procedures. Cytopathology is especially useful for this pattern and may permit a more accurate recognition and diagnosis of CIS. In the present series the cytological expressions of three ‘pure’ clinging cases was LC-NP in one and mixed LC-NP and LC-P in two. Further studies are needed to establish whether the clinging subtype has a characteristic pattern or not. De Bellis and Schumann [22] suggested cytology of the cystoscopic biopsy supernatant to be used as an adjunctive method for diagnosing urothelial CIS, which might be valuable for the clinging form.

Kojima et al.[23] considered cell ‘cannibalism’ to be an indicator of dedifferentiation and invasion of transitional cell carcinomas, and that this phenomenon might be used as an indication of progression of CIS to invasive cancer. Suprun [24] stated that cell cannibalism is a multifactorial process developing in invasive cancers. In the present study, cell-in-cell phenomenon or ‘cannibalism’ was detected in 65% of all CIS cytology cases, regardless of the pattern. In the LC-P pattern this phenomenon can be easily detected, especially in cellular samples.

In conclusion, cytology alone is insufficient for a diagnosis of urothelial CIS. An invasive lesion or papillary growth must be excluded by cystoscopy. However, we show that by using criteria from the histopathological classification of CIS [15], urothelial CIS can be divided into four subgroups cytopathologically. For cytopathologists it is important to be aware of the morphological spectrum of CIS and not to underdiagnose monomorphic, pagetoid and small cell forms. The statistical correlation between histopathology and cytology is relatively poor. Thus a combination of cytology and histopathology seems to better reflect the spectrum of lesions than histopathology alone. We suggest that studies on the behaviour of the subtypes of CIS and on the effects of therapy should be based on a combined histological and cytological evaluation of the lesion.

ACKNOWLEDGEMENTS

The authors thank the cytotechnologists and laboratory technicians of Sahlgrenska University Hospital Pathology and Clinical Cytology Department, and Dr Gonul Dinc for the evaluation of the statistics.

Ancillary