False-Positive Atypical Endocervical Cells in Conventional Pap Smears: Cyto-Histological Correlation and Analysis Free

The incidence of endocervical glandular atypia has previously been reported to vary from 0.1% to 2.1% [1‒11] with an average of 0.29%, according to a significant meta-analysis of 24 studies [12]. Cytological diagnoses of atypical glandular cells, not otherwise specified (AGC, NOS); atypical glandular cells, favour neoplastic (AGC, FN); and adenocarcinoma in situ (AIS) have been shown to have a progressively better association with neoplasia and, notably, specifically to cervical glandular malignancies [1, 5, 6, 11, 13]. However, in Pap smears showing only mild glandular atypia without any squamous atypia, the follow-up histology is benign in 38–79% of the cases [2, 3, 5, 10, 13]. Where benign histology has been specified, cervicitis with or without squamous metaplasia, tubal metaplasia, microglandular hyperplasia, and endocervical or endometrial polyps have been among the most frequently encountered findings, whether alone or in combination [2, 3, 5, 10]. Reserve cell hyperplasia, endometriosis, leiomyomas, IUD effects, radiation changes, decidual reaction of the cervix, chronic endometritis, and cases with no diagnostic histological changes have also been reported [2, 3, 5‒10].

The cytomorphological features predicting AIS and endocervical adenocarcinoma (EAC) have previously been studied and described [7, 14‒17]. The features associated with benign endocervical changes have been less well documented, although reactive and reparative cellular changes have been recognized as a significant source of false-positive Pap smears [1, 18]. Moreover, the cytological features of the benign changes described are somewhat inconsistent and overlapping even with the cytological changes encountered in AIS [19‒23]. The aim of this study was to identify single or combined cytomorphological features that could predict specific benign endocervical changes and help distinguish them from AIS and EAC, thereby preventing overdiagnoses and consequent overtreatment.

A laboratory information system search for cytological diagnosis of AEC, NOS was performed to find 30 cases with cytological endocervical cell atypia and without any squamous atypia. Only cases with a negative follow-up biopsy, including an adequately sampled transformation zone and a negative follow-up history up to the search date, were included in the study. In addition, 15 cases with AEC, NOS only in cytology with AIS or EAC in the follow-up biopsy were selected as the control group. Only cases with an available high-risk human papillomavirus (hrHPV) test result were included.

All 45 cytological samples were taken between 2013 and 2019. The samples were originally analysed in Fimlab Laboratories Oy, which serves the Tampere University Hospital and regional hospitals in Finland’s Pirkanmaa region. The samples consisted of conventional Pap smears and were reported according to the Bethesda Classification for Reporting Cervical Cytology 2014 [23‒25].

The Abbot RealTime hrHPV PCR assay (RealTime; Abbot, Wiesbaden, Germany) was used to detect the hrHPV DNA. The test recognizes 14 hrHPV genotypes, including types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66, and 68. The hrHPV genotypes 16 and 18 were reported separately, and the rest of the genotypes were reported as “other hrHPV than 16 or 18” [26].

The 45 Pap smears were initially diagnosed as glandular atypia by general pathologists in routine practice and, in the present study, were blindly reviewed by two cytopathologists with 25 and 10 years of respective experience in cytopathology. The Pap smears were analysed to search for 38 different cytological features described earlier [17] and consisting of background, cellular, nuclear, and architectural features (Table 1). The findings reported in the study are based on the consensus of the two cytopathologists.

Of the above features, the cellularity was considered scant when the smear was satisfactory according to the Bethesda 2014 criteria, but the amount of endocervical cells was just above the threshold [23‒25]. By contrast, when the number of endocervical cells was plentiful, the cellularity was regarded as high. Degeneration was defined as the loss of well-preserved cell borders with or without partial or complete loss of cytoplasm accompanied by some nuclear swelling. Nuclei were considered enlarged when the size of an endocervical cell nucleus was >2 times the size of a normal endocervical cell nucleus. Nuclear pleomorphism was defined as the >3 times variation of nuclear size and shape irregularity. All cytomorphological features were agreed upon by two observers based on the same slides. The originally benign diagnosed biopsies were revised and analysed in a search for the following histological features: inflammation (subcategorized as chronic, mixed, acute, and follicular), ulceration, granulation tissue, squamous metaplasia, immature squamous metaplasia, atypical squamous metaplasia, microglandular hyperplasia, reserve cell hyperplasia, tubal metaplasia, eosinophilic change, and polyps. The number of inflammatory cells was quantitively evaluated in haematoxylin and eosin slides and graded 0 in case of no inflammatory cells; 1+ in the case of only scattered inflammatory cells or small inflammatory cell groups of <10 cells; 2+ in the case of more diffuse inflammatory cell infiltrates with inflammatory cell groups of >10 cells; and 3+ in the case of dense, diffuse and sheet-like inflammatory cell infiltrates. Categories 2+ and 3+ were considered significant. For the other cytomorphological features, the evaluation was based on the Bethesda 2014 criteria and cytology textbooks [23, 25, 27]. All statistical analyses were performed with SPSS version 28 (IBM SPSS Statistics for Windows, Version 28.0. Armonk, NY: IBM Corp.). Univariate associations were examined using χ² or Fisher’s exact tests. Further analyses of the cytomorphological features were performed to define the combination of features associated with the occurrence of benign histology with a forward stepwise multivariable logistic regression analysis using probability values of <0.05 for the entry of features.

The study was approved by the Ethical Committee of Pirkanmaa Health Care District (R16022). It was not deemed necessary to seek the individuals’ informed consent. The study was conducted in accordance with the Declaration of Helsinki.

The patients in the AIS/EAC group had an average age of 39.0 (SD ± 6.3) years, while those in the benign group had an average age of 43.9 (SD ± 10.2) years. All Pap smears with a negative follow-up history had been taken prior to 2018, resulting in a follow-up time of five or more years, with the exception of one smear taken in 2018, which had a follow-up period of 4 years. Twenty of the 30 patients (66.7%) had a negative hrHPV test result at the time of the cytological diagnosis of AEC, NOS. One patient (3.3%) tested positive for HPV16, 2/30 (6.7%) tested positive for HPV18 and 7/30 patients (23.3%) presented with hrHPV genotypes other than 16 or 18. All hrHPV-positive patients at the baseline, who showed no evidence of malignancies during the follow-up, showed negative hrHPV test results later during the follow-up period.

The most frequent histological finding among this benign group was squamous metaplasia (22/30, 73.3%), followed by significant inflammation (16/30, 53.3%), tubal metaplasia (10/30, 33.3%), and microglandular hyperplasia (6/30, 20.0%) (Table 2). The significant inflammation was mixed in every biopsy and was frequently accompanied by squamous metaplastic changes and ulceration (Table 2). Polyps were not encountered in any of the histological samples, and only scattered cases showed reserve cell hyperplasia (1/30, 3.3%), eosinophilic change (1/30, 3.3%), granulation tissue (2/30, 6.7%), and follicular inflammation (1/30, 3.3%).

The AEC, NOS Pap smears with AIS or EAC in the follow-up biopsy were taken between 2013 and 2019. Each patient in this group was hrHPV-positive. Six of these 15 patients (40.0%) presented with HPV18 infection, 5/15 (33.3%) with HPV16 infection, and 4/15 (26.7%) had infection with hrHPV genotypes other than HPV16 or HPV18. The final diagnoses after the loop electrosurgical excision procedure or hysterectomy in this group were 1/15 (6.7%) EAC, 1/15 (6.7%) EAC + high-grade squamous intraepithelial lesion (HSIL), 10/15 (66.7%) AIS, and 3/15 (20.0%) AIS + HSIL.

In the analysis of cytological features of the Pap smears with benign follow-up histology, the inflammatory background revealed by the smears was correlated with significant inflammation in histology (15/31 [48.4%] versus 2/14 [14.3%], p = 0.029) (shown in Fig. 1). No inflammatory background was encountered in the smears presenting with chronic inflammation only in histology, but inflammatory background was instead more common in smears with mixed inflammatory infiltrates in corresponding histological samples (0/8 [0%] versus 17/37 [45.9%], p = 0.017). Nuclear crowding was more frequent in smears without significant inflammation in histology (22/29 [75.8%] versus 6/16 [37.5%], p = 0.01). Moreover, the nuclei of the glandular cells were more frequently considered to be elongated in smears without significant histological inflammation in the follow-up biopsies (6/14 [42.9%] versus 3/31 [9.7%], p = 0.017). Elongated nuclei were also seen more frequently in smears with histological microglandular hyperplasia (4/6 [66.7%] versus 5/39 [12.8%], p = 0.010).

Degeneration was registered more often in smears without squamous metaplasia or immature squamous metaplasia than in those with squamous metaplastic changes (13/24 [54.2%] versus 4/21 [19.0%], p = 0.015, and 15/30 [50.0%] versus 2/15 [13.3%], p = 0.017). Chromatin pattern was finely granular in 3/15 (20.0%) of the smears with immature squamous metaplasia in their histology compared to 0/30 (0%) in other histotypes (p = 0.032) (shown in Fig. 2). Cytological scant cellularity was associated with granulation tissue formation in histology (1/2 [50.0%] versus 0/43 [0%], p = 0.044), and palisading cell borders were also encountered more often in smears with granulation tissue in histology (2/2 [100%] versus 7/43 [16.3%], p = 0.036) (shown in Fig. 3).

Degeneration and nuclear crowding were the cytological features that best distinguished the glandular atypia with benign follow-up histology from those with histologically proven AIS or EAC (shown in Fig. 4). Degeneration was observed in only 8/30 (26.7%) of the smears in the benign group and in 9/15 (60.0%) of the smears in the carcinoma group (p = 0.030) and nuclear crowding in 15/30 (50.0%) and 13/15 (86.7%) of the smears (p = 0.017), respectively. The other way around, degeneration and nuclear crowding were more common in malignant-proven Pap smears.

Most of the 38 features investigated in the cytomorphological analysis showed no association with specific benign pathologies but presented with similar frequencies in the benign and neoplastic-proven Pap smears (Table 1). Many cytomorphological features, including necrosis, apoptotic debris, cuboidal cell shape, irregular cell borders, nuclear membrane irregularities, oval nuclei, nuclear pleomorphism, nucleoli, macronucleoli, coarsely granular chromatin, single atypical cells, mitotic figures, apoptotic bodies and rosettes, were not encountered in any of the smears. In the statistical analysis, no combinations of cytomorphological features associated with the benign pathologies were encountered.

The most common benign histological changes observed in AEC, NOS Pap smears in the present study were squamous metaplasia, inflammation, or a combination of these, which is in line with several earlier studies [3, 5, 10, 11, 28‒30]. Given that inflammation can induce reactive atypia and, as a protective response, metaplastic squamous changes, it is unsurprising to see them occur in the same specimen [3, 10, 11, 29]. The cause of endocervical cell atypia interpreted as reactive or reparative remains undefined in many previous studies [1, 3, 31, 32]. Reactive or reparative changes in cytology are known to be caused by various benign and even malignant conditions, including infection and inflammation [33‒35]. Therefore, we discuss them in association with inflammatory changes.

In our study, inflammatory background in AEC, NOS Pap smears was associated with significant mixed inflammation in the benign follow-up biopsies. The most significant published cytomorphological features associated with reactive or reparative endocervical cell atypia are listed in Table 3. In summary, nuclear size in reactive/reparative endocervical cell atypia can vary from normal or uniformly enlarged to prominent changes in size and shape. Reactive/reparative endocervical cell atypia may be associated with nuclear membrane irregularities and sometimes even with nuclear hyperchromasia (Table 3). Nuclear/cytoplasmic ratio has been described as ranging from normal to high, and the nucleoli in reactive/reparative endocervical cells may also vary with respect to size (Table 3). Crowding or nuclear overlapping often occur, and mitotic figures may also be encountered occasionally. In most cases, the nucleolar chromatin pattern has been described as even or fine, and the architectural features associated with AIS and EAC, including feathering, rosettes, and palisades, have been absent. In other words, the fine and even chromatin pattern in atypical endocervical cells and the lack of architectural abnormalities seem to be the most constant features differentiating reactive/reparative atypia from the neoplastic endocervical lesions (Table 3). In the case of inflammation-induced reactive cellular changes, the presence of the pleomorphic inflammatory cells and identification of tingible-body macrophages have been suggested to support the benign interpretation of the endocervical cell atypia seen in this group of samples [36].

In the present study, nuclear crowding in Pap smears was negatively correlated with significant histological inflammation. Likewise, nuclear elongation was encountered more frequently in cases without significant histological inflammation. Finely granular chromatin pattern was found to be associated with immature squamous metaplasia but not with significant histological inflammation.

Tubal metaplasia was the third most common benign histological change encountered, but in our study, it showed no association with any of the cytomorphological features investigated. In the literature, tubal metaplasia is most frequently described as strips, sheets, or groups of cells, which may show nuclear crowding with variations in nuclear size, pseudostratification, and even hyperchromatic nuclei or mitotic figures (Table 4). Typically, chromatin abnormalities are absent or mild with no apoptotic bodies or background tumour diathesis, and, when encountered, cilia and terminal bars have been considered characteristic (Table 4).

In our series, when the Pap smears corresponding to the histologically diagnosed tubal metaplasia cases were retrospectively reviewed, cilia were recognized in one of the ten smears (10%), which is the same frequency as that reported earlier by Ducatman et al. [41]. In our case, ciliated cells were seen in a pseudostratified strip interpreted as atypical, though the strip was not the most worrisome group of glandular cells on the slide. Babkowski et al. [40] observed ciliated cells in all smears containing diagnostically challenging endocervical cell groups but only rarely in the groups of cells causing the diagnostic challenges. Tubal metaplasia may be encountered simultaneously with AIS or EAC, and cases of ciliated AIS and EAC have been reported [42, 43]. As such, recognizing cilia in a Pap smear does not rule out the possibility of a malignancy. In tandem with reactive/reparative endocervical cell atypia, fine chromatin pattern together with the lack of some typically AIS- and EAC-associated architectural features appear to be the most helpful cytomorphological features in the differential diagnosis of tubal metaplasia and endocervical malignancies.

In our study, a fifth of the benign biopsies presented with microglandular hyperplasia, which in turn was associated with nuclear elongation in cytology. Investigations of larger series are required to reveal the significance of this finding since nuclear elongation is an important feature of AEC, FN, and AIS, as defined by TBSRCC 2014, and has not been described in association with microglandular hyperplasia in previous studies (Table 4). Instead, cytological features of microglandular hyperplasia that might lead to diagnosis of AEC, NOS reportedly include three-dimensional cell groups, usually with preserved cell polarity, small gland-like spaces and a variation of cell size from small to large. The nuclei in microglandular hyperplasia can be either cubic, cylindrical, or rounded, occasionally hyperchromatic, and their chromatin pattern can be coarsely granular (Table 4). In histology, atypical and mitotically active forms of microglandular hyperplasia have been described, and cells from these lesions may be expected to cause diagnostic problems in cytology [44, 45].

In our study, the overall scant cellularity in Pap smears was correlated with granulation tissue formation in histology, which could be expected in the absence of surface epithelium. In addition, palisading cell borders showed a correlation with granulation tissue. No other reports of the cytological features of granulation tissue were found in the literature.

In line with several previous studies, all our benign follow-up biopsies harboured some histological changes [2, 5, 11, 28]. However, others reported no histological abnormalities, potentially explaining the cytological glandular atypia in 10–71% of the cases [3, 7, 8, 12, 30, 46].

HPV infections are known to have a high spontaneous clearance rate, and most of the histological LSIL, and even some CIN2 lesions, spontaneously regress [47‒50]. In a Brazilian cross-sectional study, 41% of HPV-positive cases with glandular cytological abnormalities had normal Pap smear and colposcopy in a follow-up [51]. Furthermore, the time-honoured concept does not appear to work for glandular abnormalities in the same pattern as that for squamous cell precursor lesions [52]. Although the pathogenesis and natural history of the endocervical glandular malignancies is less well established than that of the squamous lesions, we may nonetheless speculate as to whether HPV infection also induces similar, later regressing changes on the glandular epithelium. This would explain why some of the follow-up histology samples showed no changes that could explain the cytological glandular atypia. Of course, the phenomenon may also be a matter of sampling error or diagnostic error, given that mild cytological atypia is known to be problematic with poor reproducibility [53‒57].

Degeneration and nuclear crowding were identified as the strongest features separating the benign-proven AEC, NOS Pap smears from those harbouring a malignancy since both were more common in cases histologically verified as malignant. In previous studies, degeneration has also been described as one of the features that obscures the neoplastic nature of the atypical cells seen in Pap smears [1, 53]. Likewise, in agreement with our findings, crowded fragments have previously been reported among a combination of cytomorphological features in cases correctly diagnosed as endocervical glandular neoplasia [1, 15, 53].

Owing to our strict inclusion criteria, the present study’s sample size remained relatively small, which is an obvious limiting factor and a possible explanation as to why the investigated cytomorphological features did not show more specific association with any of the histological findings. In addition, we only had conventional smears available for analysis. However, this was a blinded study rather than a retrospective morphological analysis only. Moreover, our benign cases had a relatively long follow-up period, and HPV status history was available, both of which support the interpretation that cellular changes observed in their Pap smears truly were non-neoplastic in origin.

In summary, it is evident that additional tools, besides cytomorphology, are required to reliably distinguish between mild glandular atypia related to benign changes from those harbouring a malignancy. In our study, all the benign-proven AEC, NOS cases were either hrHPV-negative at the study baseline or turned negative during the study period. Most of the AIS and EAC are known to be hrHPV-associated [58‒60]. Thus, hrHPV negativity appears to be the best indicator of benign endocervical glandular cell atypia. Meanwhile, hrHPV-positive cases are problematic and will continue to be a source of overdiagnoses and added costly diagnostic procedures until more specific biomarkers are introduced. Among liquid-based cervical cytological samples with mild squamous atypia, the use of p16/Ki-67 dual stain has been shown to be more specific than the hrHPV testing and cytology in detecting HSILs and carcinomas [61, 62]. Although the data are still limited, the p16/Ki-67 dual stain also appears to be useful in the diagnosis of endocervical glandular atypia [62].

The study protocol was reviewed and approved by the Ethical Committee of Pirkanmaa Health Care District with approval number R16022. The study was based on archive samples and therefore had no impact on the treatment or the follow-up of the patients. The study was conducted without informed consent of each individual, which was approved by the Ethical Committee of Pirkanmaa Health Care District. The study was conducted according to the Declaration of Helsinki.

The authors have no conflicts of interest to declare.

The work was supported by grants from the VTR Funding, Ida Montin Foundation, and Emil Aaltonen Foundation. The funding sources had no role in the preparation of data or the manuscript.

Johanna Pulkkinen: data curation, formal analysis, investigation, project administration, visualization, validation, writing – original draft, and editing. Heini Huhtala: data curation, statistical analyses, supervision, and writing – original draft. Ivana Kholová: conceptualization, data curation, formal analysis, investigation, methodology, project administration, resources, visualization, supervision, validation, writing – original draft, and editing.

The data that support the findings of this study are available from the corresponding author upon reasonable request.