Cytomorphological Assessment in Aspirates of Ductal Carcinoma in situ: Correlations with Histopathologic Grade, Architectural Pattern, and Invasion

Fine-needle aspiration biopsy (FNAB) of the breast is an effective diagnostic technique with the advantages of high cost-effectiveness, minimal invasiveness, capability of rapid on-site evaluation, requiring only out-patient facilities, and being amenable to an out-patient-based procedure [1]. FNAB demonstrates the highest sensitivity and specificity in diagnosis and differentiation of benign and malignant epithelial lesions [2‒4], but in the breast, accurate grading and assessment of invasion is necessary for accurate treatment planning.

The Van Nuys classification for grading of ductal carcinoma in situ (DCIS) [5, 6], which classifies DCIS into low-, intermediate-, and high-grade based on a detailed description of nuclear grade and presence of necrosis, has prognostic significance. Studies addressing the application of the Van Nuys, or other DCIS classification systems on cytopathology specimens are limited. The few studies cytopathologic grading of DCIS [7‒9] mainly evaluated a limited number of common cytopathologic nuclear features used in histopathologic grading [5, 10, 11]. In this study, a comprehensive panel of cytomorphologic parameters was assessed in a cohort of histopathology-proven FNAB of DCIS to identify parameters correlating with DCIS grading, histopathologic architecture of DCIS, and presence of invasion.

Computerized searches of the departmental pathology archives from the involved institutions were performed for breast aspirates from patients with a previous histopathologic diagnosis of DCIS, from the year 2001 to 2019. Pathology reports, cases notes, and histopathologic slides (if available) were reviewed, and cases were excluded if there was extensive invasion (≥10% tumor area) or mismatch in laterality. The FNAB specimens were reviewed for cellularity, background composition, inflammatory cells, myoepithelial cells, cellular cohesion (assessed as the percentage of cohesive epithelial cells), epithelial fragment architecture (assessed as a percentage of different patterns), and cellular/nuclear features (Table 1). The slides were reviewed by at least two of the authors, and any discrepancies were resolved by reviewing the slides on a multiheaded microscope until a consensus was reached.

Cytomorphologic parameters were compared against histopathologic grading, epithelial fragment architecture, background composition, and presence of focal invasion in subsequent biopsy or excision. Statistical analysis was performed with SPSS version 23.0. The χ² test and Fisher’s exact tests were used to compare categorical variables, and Spearman’s correlation was used for ranked parameters. Cellular cohesion was analyzed as continuous variable using the Mann-Whitney U test. A p value of <0.05 was considered significant. This study was approved by the Institute’s Clinical Research Ethics Committee with waiver of the requirement of written consent.

A total of 104 FNAB specimens were retrieved from 86 patients with an average age of 56.8 (27–91) years. There were 58 and 46 aspirates from the left and right breast, respectively. Eighty-two (78.8%) cases showed no focal invasion in the corresponding histopathology and 22 (21.2%) cases showed focal invasion (<10% tumor area). Excluding one case where histopathologic grading was not available, there were 81 (78.6%) cases of non-high-grade DCIS and 22 (21.4%) cases of high-grade. On review, nine cases had insufficient cellularity for assessment. Presence of focal invasion (p = 1,000) and DCIS grading (p = 0.680) did not demonstrate correlation with insufficient sampling (Table 2). Background features on FNAB, including calcification, comedonecrosis, and apocrine change did not correlate with histopathologic findings (p > 0.05) (Table 3).

Among the nuclear features reviewed, including large nuclear size (p = 0.005), prominent nucleoli (p = 0.011) (Fig. 1a), increased nuclear membrane irregularity (p = 0.043), and high variation in nuclear size (p = 0.025) (Fig. 1b) correlated with a high-grade diagnosis. Furthermore, a large amount of cytoplasm (p = 0.034) and presence of apoptotic figures in epithelial structures (p < 0.001) (Fig. 1c) were also indicative of a high-grade diagnosis (Table 4). A smaller amount of cytoplasm was seen in non-high-grade specimens (p = 0.034). Cytoplasmic vacuolation (p = 0.034) (Fig. 1d) was seen exclusively in non-high-grade specimens. Except for debris (p = 0.033), epithelial fragment architecture, cellular cohesion, epithelial fragment architecture, inflammatory cells, myoepithelial cells, background composition, and cellularity were not associated with grading (online suppl. Table 2; for all online suppl. material, see https://doi.org/10.1159/000535836).

A predominance (≥50%) of solid aggregates (Fig. 2a) on FNAB correlated with histopathologic solid (p = 0.039) and papillary (p = 0.029) patterns, and a predominant composition of papillary fragments (Fig. 2b) correlated with histopathological solid (p = 0.005) and papillary (p < 0.001) patterns (online suppl. Table 1). Monolayer sheets and cribriform structures on FNAB did not correspond to histopathological epithelial fragment architectures (online suppl. Table 1). Calcification, comedonecrosis, and apocrine changes were not reproducible between FNAB and histopathology (Table 3). Cellular cohesion, epithelial fragment architecture, inflammatory cells, cellularity, and background composition did not predict focal invasion on subsequent histopathology (online suppl. Table 2).

In the Van Nuys classification of breast DCIS, high-grade nuclear features were defined as large nuclear size, pleomorphism, prominent nucleoli, coarse clumped chromatin, lack of polarity, lack of architectural differentiation, and generally showing mitosis [5]. Based on this morphologic pattern, low-grade DCIS was defined as cases without high-grade nuclear features or necrosis, intermediate grade as non-high-grade nuclear features with necrosis; and high-grade as DCIS showing high-grade nuclear features accompanied by necrosis [5]. Other grading classifications systems described for DCIS include Holland, modified Black, and modified Lagios systems, all of which are similarly three-tiered and based on nuclear grading [12‒14]. Necrosis is replaced by mitotic count in the Holland and modified Black systems [12, 14]. In general concordance between these classifications was high, and a combination of nuclear features described in the different classification systems was adopted by the WHO classification [10].

Cytopathological grading of aspirates of invasive breast carcinoma has been shown to correlate with histopathologic (Nottingham modification of Bloom-Richardson) grading, with nuclear features including nuclear size and chromatin character identified as statistically significant predictors on multiple regression [15]. Nuclear features in histopathologic specimens are reproducible in breast aspirates [16], and thus it is reasonable to expect that cytopathologic nuclear features to be predictive of histopathologic DCIS grading. A previous smaller study demonstrated correlation between necrosis, cellularity, and chromatin pattern with DCIS grading, but it was based on a “historic” histopathologic classification of DCIS (high, intermediate/low, and cribriform subtypes) [7]. Another more recent study demonstrated 97% concordance between cytopathologic and histopathologic grading using a nuclear size cutoff of two-times the diameter of an erythrocyte for high-grade DCIS [8], without incorporating other nuclear features.

The current study evaluated nuclear features and other cytomorphologic parameters in great detail on the grading of DCIS in breast FNAB. A large nuclear size, prominent nucleoli, coarse chromatin and pleomorphism (corresponding to high nuclear size variation and nuclear membrane irregularity) demonstrable in cytopathology correlated with high histopathologic grading. As both nuclear size and amount of cytoplasm were both positively correlated with grading, nuclear-cytoplasmic ratio was not found to be predictive, this may be explained by the marked pleomorphism, and large cells seen in high-grade DCIS [17]. In addition to the nuclear features described in the Van Nuys classification [7], apoptosis and a large amount of cytoplasm were also observed to correlate with a high-grade diagnosis. Presence of acellular debris was shown to correlate with a high-grade diagnosis, which may be reflective of necrosis on histopathology. Also interestingly, the current study found that cytoplasmic vacuolation was present solely in non-high-grade diagnosis, contrary to previous studies which did not identify any specific features [7, 9]. Other features, including epithelial fragment architectures, background, and composition did show correlation with grading.

In our cohort, no cytomorphological parameter studied showed correlation with presence of focal invasion on histopathology. Many studies have attempted to establish criteria for invasion on FNAB comparing frankly invasive carcinomas with DCIS [18‒20]. Some parameters were suggested to be useful, but results are conflicting. As no reliable feature can be determined, most pathologists refrain from making diagnoses of invasion in malignant aspirates [2].

Breast aspirates usually contain a mixture of different epithelial fragment architectures [8]. We compared the predominant (≥50%) cytopathologic epithelial fragment architecture with the patterns in subsequent histopathology. Aspirates with predominant solid aggregates and papillary fragments showed correlation to presence of both solid and papillary patterns on histopathology. Aspirates from DCIS with a “pure” single pattern are required to establish cytopathological epithelial fragment architectures corresponding to a specific histopathological epithelial fragment architecture. The difficulty lies in recruiting such specimens as DCIS is known to demonstrate heterogeneous epithelial fragment architectures within a single lesion [21]. Regardless, it may not be feasible to predict the composition of epithelial fragment architectures in a variegated lesion based on biopsy sampling.

FNAB is effective in predicting DCIS grading. High-grade histopathologic nuclear features including large nuclear size, pleomorphism, and prominent nucleoli can be adequately reproduced on cytopathology. Apoptosis, background debris, and cytoplasmic vacuolation were additional cytomorphological parameters found to correlate with grading. Epithelial fragment architecture assessment is limited to papillary or solid types. Diagnosis of focal invasion requires histopathologic examination.

The abstract has been presented at the Hong Kong College of Pathologists Trainee Presentation Session 2020.

This study was approved by the Chinese University of Hong Kong – New Territories East Cluster Clinical Research Ethics Committee with waiver of the requirement of written informed consent (reference number 2020.320).

The authors declare that there is no conflict of interest regarding the publication of this paper.

No funding was received for the publication of this article.

J.J.X.L.: conceptualization, investigation, methodology, formal analysis, visualization, and writing – original draft. M.B.C.Y.C.: data curation, investigation, and resources. J.K.M.N.: methodology and investigation. J.Y.T.: formal analysis, methodology, validation, and visualization. G.M.T.: conceptualization, investigation, methodology, supervision, and writing – review and editing.

Data generated and analyzed in the current study are available in the manuscript. Further inquiries can be directed to the corresponding author.