Diagnostic accuracy of the Yokohama system for reporting breast fine-needle aspiration cytology and breast imaging-reporting and data system in breast lesions

Study design and setting

This cross-sectional study was conducted over a period of 18 months, in collaboration with the Departments of Pathology, Surgery, and Radiology. The study was done after approval of the Institutional Ethics Committee (Vardhman Mahavir Medical College and Safdarjung Hospital, New Delhi) vide no. IEC/VMMC/SJH/Thesis/2023-03/CC-245 and in accordance with the Declaration of Helsinki. Patients presenting with breast lesions who underwent FNAC followed by biopsy/resection were included. Patients who had received neo-adjuvant therapy were excluded. All eligible breast cancer patients were included, and cases were selected consecutively. In this study, FNAC was performed in 188 cases, of which 155 cases had subsequent histopathology reports available for correlation. In 33 cases, FNAC was done without a follow-up histopathology report due to a lack of biopsy/resection. In addition, 12 cases had histopathological reports available, but did not undergo FNAC. FNAC was reported without knowledge of histopathological findings to ensure blinding. Similarly, histopathology was reported without prior knowledge of FNAC results to maintain objectivity and prevent bias. The baseline clinical and lesion characteristics (such as age, lesion size, and clinical symptoms) of the 33 cases without histopathology follow-up were not systematically different from those included (n = 155). These cases were excluded solely due to a lack of histopathology follow-up and not because of any intrinsic clinical or pathological differences.

Sample size calculation

The sample size was calculated to determine the diagnostic accuracy of FNAC in detecting malignancy, using histopathology as the gold standard. The Wilson Score Interval method was used for a proportion-based accuracy study. Assuming a 95% confidence interval, an expected accuracy of 90% based on Ahuja and Malviya, and an absolute error margin of 5%, the required sample size was estimated using the formula:^[6] −

Where Z = 1.96 (for 95% confidence), accuracy = 0.90, and d = 0.05. Substituting values, the calculated sample size was 138.2, which was adjusted for a 10% margin of error, resulting in a final sample size of 153, rounded to 155 participants. The calculation assumes a proportion-based estimation of diagnostic accuracy using FNAC and histopathology results, application of the Wilson Score Interval method for confidence interval estimation, random sampling of patients presenting with breast lesions, independence of observations, and a sufficiently large sample to ensure a precise and stable estimate of diagnostic accuracy.

BI-RADS and Yokohama system classification

BI-RADS Classification includes 7 categories (0 = incomplete, 1 = negative, 2 = benign, 3 = probably benign, 4 = suspicious [4a: low suspicion; 4b: moderate suspicion; 4c: high suspicion], 5 = highly suggestive of malignancy, and 6 = proven malignancy).^[7,8] Yokohama System includes 5 categories (1 = insufficient, 2 = benign, 3 = atypical, 4 = suspicious, and 5 = malignant).^[9]

Data collection and preparation

Clinical details were obtained from patient records. Mammography was performed to assign the BI-RADS score. FNAC was done following standard procedures after obtaining informed consent. FNAC was performed using a 22-23-gauge disposable needle attached to a 10 mL syringe. The puncture site was selected based on clinical and radiological guidance, with preference for the most representative and accessible area of the lesion. Multiple passes (typically 2-3) were made to ensure adequate sampling. The aspirated material was expelled onto clean glass slides, smeared, and immediately fixed in 95% ethanol for Papanicolaou (Hi-Media; RM5375; India) staining, while air-dried smears were prepared for May Grunwald Giemsa (Hi-Media; GRM945; India) staining. In cases where sample adequacy was uncertain, rapid on-site evaluation (ROSE) was employed when feasible. Biopsy specimens were fixed in formalin, embedded in paraffin, and stained with Hematoxylin and Eosin (H&E, Biolab; CY1577; India). When necessary, additional immunohistochemical staining was performed. However, in the present study, the criteria for defining “when necessary” were not pre-specified. The cases that underwent immunohistochemistry typically included those with atypical cytological features or where the nature of the lesion remained uncertain after routine staining. The specific markers applied (such as Estrogen receptor, Progesterone receptor, HER2, and Cytokeratin 5/6) were chosen on a case-by-case basis, but the exact selection strategy and interpretation criteria were not standardized. Furthermore, the study did not analyze the incremental impact of immunohistochemistry on the diagnostic accuracy of either the Yokohama System or BI-RADS classifications.

Data analysis

For analysis, cytological classifications were grouped into three categories based on their degree of malignancy suspicion to facilitate comparison with histopathological diagnoses (Scenario A: malignant category considered positive, Scenario B: malignant and suspicious category considered positive, and Scenario C: malignant, suspicious, and atypical category considered positive). This categorization assessed the impact of expanding diagnostic thresholds on diagnostic accuracy calculations. All statistical analyses were performed using the Statistical Package for the Social Sciences version 26.0. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and diagnostic accuracy were calculated using histopathological results as the gold standard. The ROM was calculated for each diagnostic category. Similarly, the diagnostic accuracy of BI-RADS scores was assessed using the same criteria.

Demographics and lesion distribution

A total of 188 patients were enrolled in the study, with ages ranging from 14-75 years (mean age: 35.3 years). The cohort predominantly consisted of younger patients, with the majority falling into the age groups of 14-30 years (40.4%) and 31-45 years (38.3%). The sex distribution was predominantly female (99.5%), with one male patient diagnosed with invasive papillary carcinoma.

Malignant lesions were more commonly observed in patients aged ≥45 years, with the mean age for malignant lesions being 44.5 years, compared to a mean age of 28.9 years for benign lesions. This age difference between benign and malignant lesions was statistically significant, highlighting the role of age as a factor in breast cancer diagnosis.

Lesion characteristics

Histopathology results showed that, out of the 188 lesions, 77 (41%) were malignant, while 111 (59%) were benign. Among the benign lesions, the most frequent types were fibroadenomas, which accounted for 33.5% of the total cases. Other benign lesions included benign ductal hyperplasia (11.7%) and fibroadenomatoid hyperplasia (6.9%). The most common malignant lesions were invasive ductal carcinoma (IDC) (27.1%) and intraductal carcinoma (21.1%).

The breakdown of the malignant lesions included both in situ and invasive cancers, with 45.4% of malignant lesions being invasive and the remaining 54.6% being in situ, which reflects the varied presentation of breast malignancies. Histopathology findings confirmed the presence of various tumor subtypes, including IDC, invasive lobular carcinoma, and invasive papillary carcinoma, with IDC being the most prevalent.

BI-RADS and Yokohama classification

According to the Yokohama classification, 5.9% (11 cases) of lesions were classified as non-diagnostic (Category 1), 52.1% (98 cases) as benign (Category 2), while 4.8% (9 cases) were categorized as atypical (Category 3), 8.5% (16 cases) as suspicious (Category 4), and 28.7% (54 cases) as malignant (Category 5).

The majority of cases were classified as BI-RADS 3 (23.9%) and 5 (23.9%), followed by 4a (22.3%). Categories 4b (10.6%) and 2 (9.0%) were less common, while 4c (5.3%) and 6 (4.8%) had the lowest frequencies. No cases were assigned to categories 0 or 1 [Figures 1 and 2].

Close

Figure 1:

Cytopathological, breast imaging-reporting and data system (BI-RADS), histopathological images in a case of fibroadenoma. (a) Smears exhibit monolayered sheet of ductal cells with overlying myoepithelial cells in a case of fibroadenoma (Yokohama category 2) (scale bar: 50 µm; 400x, MayGrünwald-Giemsa), (b) Radiological images show a BI-RADS 3 lesion – probably benign, (c) Sections show an intracanalicular pattern with compressed ducts surrounded by proliferating stroma (scale bar: 100 µm; 200x, Hematoxylin and eosin).

Export to PPT

Close

Figure 2:

Cytopathological, breast imaging-reporting and data system (BI-RADS), histopathological images in a case of ductal carcinoma. (a) Smears exhibit a discohesive cluster of atypical ductal cells with moderate pleomorphism in a case of ductal carcinoma (Yokohama category 5) (scale bar: 50 µm; 400x, May-Grünwald-Giemsa), (b) Radiological images show BI-RADS 5 lesion, highly suggestive of malignancy, (c) Sections show sheets and clusters of atypical ductal cells with enlarged hyperchromatic nuclei, moderate cytoplasm, and prominent nucleoli (scale bar: 100 µm; 200x, Hematoxylin and eosin).

Export to PPT

Diagnostic performance of FNAC

FNAC was performed on all lesions, with the results compared against histopathology as the gold standard. On histopathology, 111 (59%) cases turned out to be non-malignant, and the remaining 77 (41%) were malignant. Among patients with malignant histopathology diagnosis, the higher Yokohama category was more prevalent than the lower Yokohama category. A total of 50 (64.9%) patients with malignant diagnoses in histopathology were in the Yokohama category 5. Among patients with malignant histopathology diagnosis, 5 (6.5%) belonged to Category 1 of the Yokohama classification. 9 (11.7%) belonged to Category 2, whereas 3 (3.9%) and 10 (13.0%) belonged to Categories 3 and 4, respectively.

Among patients with non-malignant histopathology diagnosis, 6 (5.4%) belonged to Category 1 of the Yokohama classification. 89 (80.2%) belonged to Category 2, whereas 6 (5.4%) belonged to Categories 3 and 4, respectively. Only 4 (3.6%) belonged to Category 5. The ROM for the Yokohama system was 45.5%, 9.2%, 33.3%, 62.5%, and 92.6% in Categories 1, 2, 3, 4, and 5, respectively [Table 1].

The analysis of false positives and false negatives revealed that FNAC yielded false-positive results in 4 cases (3.6%), wherein lesions classified as malignant by FNAC were found to be benign on histopathology. These cases primarily consisted of fibroadenomas with atypical features and sclerosing adenosis, both of which can mimic malignancy cytologically. Cytologically, these cases exhibited high cellularity with overlapping epithelial fragments, mild nuclear pleomorphism, and occasional prominent nucleoli, features that led to their categorization as malignant on FNAC. These findings highlight specific benign entities that may mimic carcinoma on cytology, particularly in the absence of architectural correlation. Inclusion of these entities in malignant reporting thresholds lowered the specificity to 96.4%, underscoring the importance of cautious interpretation when such cytomorphological overlaps are present.

False-negative results were observed in 9 cases (11.7%), where lesions initially categorized as benign by FNAC were later confirmed to be malignant on histopathology. The predominant factors contributing to false-negative diagnoses included sampling errors, low cellularity of aspirates, and the presence of well-differentiated carcinomas that exhibited minimal cytological atypia.

A higher BI-RADS score was more prevalent among malignant cases. No malignancies were observed in BIRADS Categories 0, 1, or 2. The distribution of cases across BIRADS scores has been detailed in Table 2. The ROM for each BI-RADS category was 0.0% for Category 2, 6.7% for Category 3, 16.7% for Category 4a, 40.0% for Category 4b, 100% for Category 4c, 88.9% for Category 5, and 100% for Category 6 [Table 2].

When using the Yokohama classification, sensitivity was highest for Scenario C, with a value of 81.8%. Specificity was highest in Scenario A with 96.4%. Maximum PPV was achieved by Scenario A with a value of 92.6%, and maximum NPV was achieved by Scenario C with a value of 87.2%. The Scenario with the highest diagnostic accuracy was Scenario B with 85.6% [Table 3].

The highest sensitivity and NPV (87.0% and 90.4%) were seen in Scenario C when BI-RADS scores 4b, 4c, 5, and 6 are considered malignant. The specificity was maximum in both Scenarios A and B, with 95.5% each. Maximum PPV was seen in Scenario B with 92.2%. Maximum diagnostic accuracy was seen in Scenario B, which was 87.7% [Table 4].