Introduction
Thyroid carcinoma accounts for approximately 1% of all malignancies.1 It is the commonest endocrine malignancy of which papillary thyroid carcinoma (PTC) accounting for 80% of cases in adults1, 2, 3 and 90% in children.3 The peak incidence in adults is between 35-40 years and is commoner in females in the ratio of 3-4:1.4 In Sri Lanka, thyroid gland malignancy is the third commonest malignancy in females with the highest incidence among 15-34 years.4
In the evaluation of a thyroid nodule, fine needle aspiration cytology (FNAC) is a well-established first line diagnostic test.5, 6, 7, 8 The diagnostic accuracy of PTC by FNA accounts for over 90% provided the sample is adequate.8 The accepted criteria for cytodiagnosis of PTC includes true papillary tissue fragments together with nuclear features such as dusty pale enlarged nuclei, fine dusty powdery chromatin, chromatic ridge / bar (nuclear gsingle or multiple micro and macronucleoli, intranuclear cytoplasmic pseudo inclusions with dense stringy ‘bubble-gum’ colloid in the background.9
None of the aforementioned architectural, cytomorphological and background features however are unique to or diagnostic of PTC as they can be present in various other non-neoplastic and neoplastic thyroid lesions such as hyperplastic nodule, papillary hyperplasia, thyroiditis, Hurthle cell neoplasm and hyalinising trabecular adenoma.9, 10 It is challenging to distinguish reactive nuclear changes associated with lymphocytic thyroiditis from PTC since they share certain features like nuclear grooves and intranuclear inclusions.11
None of these features is diagnostic on their own, unless they occur in combination and are relatively widespread.8, 9 The commonest mistake in diagnosing PTC is when the cytopathologist places too much emphasis on a single cytological feature.9
Thus, unequivocal diagnosis of PTCFNA may be difficult to a inexperienced cytopathologists, when the only a few diagnostic features are present and/or if the features are the features are found in low frequency. Furthermore, variants of PTC like follicular variant may exhibit a very few nuclear features which may be present in only a few foci.12 Therefore, it will be helpful to determine which individual cytoarchitecural features are most reliable for the diagnosis of PTC and also to determine which combination of features is more useful in the FNA diagnosis of PTC.
General objective
To determine the usefulness of individual cytoarchitectural features in FNAC smears for the diagnosis of PTC and its variants.
Materials and Methods
Study population and study setting
Fine needle aspiration smears from 50 consecutive histologically confirmed PTC cases, reported at the Departments of Pathology, Faculty of Medicine Colombo and at National Hospital, Sri Lanka during the period of January 2012 to September 2015 were retrieved from the archives. FNAs from 50 consecutive histologically confirmed cases of nodular goiter, chronic autoimmune thyroiditis or hyperplastic nodules reported at the same setting during the period from January 2014 to September 2015 were also retrieved to be used as controls.
Inclusion and exclusion criteria
Haematoxylin and eosin (H and E) stained slides of each case were selected consecutively.
The slides showing extensive drying artifact and slides which were broken or poorly staining were excluded.
Sample size calculation
Sample size was calculated according to the standard formula used to calculate the proportion, taking into consideration confidence interval (CI) of 70% (55-85%) of sensitivity and confidence level at 95%.
n=36 each
n= Sample size
Z = Z value (1.96 for 95% confidence level).
p = Expected proportion in population based on previous studies
0.7 used for sample size needed
c = Confidence interval expressed as decimal (e.g., 0.15= ±15) the slides wa
According to the above formula minimum number of sample for the study is 36 in each group. Therefore, in our study we included 50 consecutive cases (papillary group) and 50 consecutive controls (non-papillary group).
Sampling technique and data collection
The cytology slides of the cases and the controls were mixed and selected randomly for the review.
The presence or absence of the following 31 architectural, cytological and background features were assessed on each H & E stained FNA smear.
Architectural features
Cellularity: defined by the number of cell clusters which is further divided into two groups: i. ≤20 ii. >20
Flat syncytial sheets: defined as flat sheets of follicular cells with lack of distinct cell borders.
Papillary structures with anatomical borders without fibro vascular cores: Anatomical border is defined as a well-defined sharp edge formed by a row of cuboidal or columnar cells.
True papillae with fibrovascular cores (Figure 1 A, B).
Micro acinar structures away from the sheets: defined by the micro acinar arrangement of follicular cells with well–defined lumina.
Cellular swirls: defined as concentrically organized aggregates of about 50-200 tumour cells. Majority of the peripherally situated cells have ovoid nuclei and their long axes are arranged perpendicular to the radius of the swirl (1) (Figure 1 C, D).
Individually dispersed bare nuclei.
Individually dispersed cells with eosinophilic cytoplasm (Figure 1 E).
Cellular crowding within nests and sheets: defined by overlapping of nuclei.
Cytological features
Anisonucleosis/Nuclear enlargement: Enlarged nucleus is defined as a nucleus > two times that of surrounding follicular epithelial cell (Figure 2 A).
Elongated, oval shaped nuclei or oblong nuclei (Figure 2 A)
Fine, powdery chromatin (Figure 2 A)
Longitudinal nuclear grooves/creases: defined as continuous groove or crease which is clearly defined (Figure 2 A).
Nuclear outline irregularity including notched nuclei (Figure 2 A)
Thickened nuclear membranes
Crescent shaped, collapsed nuclei with nuclear moulding (Figure 2 A)
Intranuclear cytoplasmic inclusions (INCI): defined as sharp, well- defined membrane like margin which is not optically clear but is similar to the colour and texture of the cytoplasm and occupies two-thirds of the nucleus.
Nucleoli: marginated and non-marginated
Histiocytoid cells: Cells with abundant and vacuolated cytoplasm and large irregular nuclei with occasional nucleoli, grainy chromatin which lack grooves and pseudo inclusions (Figure 2 B).
Cells with septate vacuoles in the cytoplasm: defined as small intracytoplasmic vacuoles which resemble soap bubbles.
Hurtheloid cells: Cells with abundant eosinophilic and finely granular cytoplasm.C
Metaplastic squamous cells: Cells with moderately abundant eosinophilic cytoplasm (Figure 2 C)s
Columnar cells: Cells with height two times more than their width showing oxyphilic featuress
Mitotic figures.
Background features
Bubble gum’ or ropy colloid: strands and chunks of dense colloid intimately associated with the neoplastic cells.
Multinucleated giant cells
Psammoma bodies: Glassy, refractile, concentric calcified lamellated bodies
Cyst macrophages
Lymphocytes admixed with cell clusters
Plasma cells admixed with cell clusters
Neutrophils within the aspirates
For further assessment, the statistically significant cytological features were then combined into 10 groups as follows;
Combination 1: Flat syncytial sheets, true papillae, cellular swirls, anisonucleosis, elongated oval nuclei and INCI.2, 4, 6, 9, 10, 13
Combination 2: Anatomical borders, true papillae, cellular swirls, elongated oval nuclei, fine powdery chromatin and crescent shaped collapsed nuclei.3, 4, 6, 10, 11, 14
Combination 3: Flat syncytial sheets, true papillae, individually dispersed cells with eosinophilic cytoplasm, nuclear outline irregularity, thickened nuclear membranes and INCI.2, 4, 8, 15, 16, 13
Combination 4: Anisonucleosis, fine powdery chromatin, nuclear outline irregularity, crescent shaped collapsed nuclei and INCI.9, 11, 15, 14, 13
Combination 5: Flat syncytial sheets, True papillae, cellular swirls, Anisonucleosis, elongated oval nuclei, nuclear grooves, thickened nuclear membranes, INCI and bubble gum colloid.2, 3, 4, 6, 9, 10, 12, 16, 13
Combination 6: True papillae and Cellular swirls.3, 4
Combination 7: Nuclear overlapping and Fine powdery chromatin.9, 11
Combination 8: Nuclear grooves an INCI.10, 13
Combination 9: Flat syncytial sheets and Anatomical borders.3, 4
Statistical analysis
The presence or absence of each cytoarchitectural feature was recorded and entered in a Microsoft excel sheet. The final data was then analyzed by SSPS software (version 23). The sensitivity, specificity, positive and negative predictive values (PPV and NPV) of the selected parameters were calculated after constructing two by two tables, considering the histological diagnosis as the gold standard (Table 1). The chi-square test and odds ratio were used to determine the statistical significance (p<0.05) of each cytological parameter. If the former two methods were not applicable like in instances where one or two of the values in 2x2 table is less than 5, the Fisher’s exact test was used (Table 2).
Table 1
Comparison of the presence of cytoarchitectural features of FNAC amongst the patients with papillary carcinoma and non-papillary thyroid lesions (n=100)
Table 2
Detection of significance of each characteristic by chi square and Fisher’s exact test where applicable
Then the sensitivity, specificity, PPV and NPV of the above combinations were calculated (Table 3). The 2x2 table is formed in consideration of the presence or absence of all the features.
Table 3
Calculating sensitivity, specificity, PPV, NPV and significance of combinations of statistically important selected cytoarchitectural features
Then the combinations of the cytoarchitectural features were further analyzed by receiver operating characteristic (ROC) curves to determine the correlation of the combined features with PTC and also to conclude whether the combination of more than two features were more significant than the combination of only two features (Figure 3 and Table 4).
Table 4
ROC of the combinations: area under the curve
According to the ROC curves the results were interpreted as follows:
Table 0
|
Area under the curve |
Quality of the test |
|
0.9-1 |
Excellent |
|
0.8-0.9 |
Good |
|
0.7-0.8 |
Fair |
|
0.6-0.7 |
Poor |
|
0.5-0.6 |
Fail |
The cytological features were categorized according to the specificity and the PPV into the following three groups and the sensitivity of each feature was compared (Table 6).
Table 5
Categorization of cytologic features according to the specificity and the PPV. 100% Specificity and 100% PPV
Results
A total of 100 FNAs were evaluated. In the papillary group 31 cases had been diagnosed as PTC (Thy 5, Bethesda 6) whilst 19 cases had been diagnosed suspicious for PTC (Thy4, Bethesda 5). All FNA cases diagnosed as diagnostic of and suspicious for PTC were confirmed as PTC by histology. In the control group, 2 cases had been diagnosed as follicular proliferation (Thy 3). Rest of the cases were diagnosed as benign (Thy2, Bethesda 2) of which 28 cases had been diagnosed as colloid storing goiter and 20 cases as chronic autoimmune thyroiditis on cytological evaluation. The two cases diagnosed as follicular proliferation were confirmed as hyperplastic nodules by histology. The remaining 48 cases were confirmed histologically as colloid storing goiter (28 cases) or chronic autoimmune thyroiditis (20 cases).
According to the statistical analysis, the highly statistically significant (p=0.000) cytoarchitectural features are flat syncytial sheets, anatomical borders, true papillae, cellular swirls, individually dispersed cells with eosinophilic cytoplasm, anisonucleosis, elongated oval nuclei, powdery chromatin, nuclear outline irregularity, thickened nuclear membranes, crescent shaped collapsed nuclei, intranuclear inclusions, histocytoid cells and cells with soap bubble cytoplasmic vacuolation. Most of these features also had high sensitivity, specificity, PPV and NPV (>70%). However, features like true papillae, cellular swirls, histocytoid cells and cells with soap bubble cytoplasmic vacuolation had low sensitivity but very high specificity whereas features like intranuclear inclusions, anisonucleosis and elongated nuclei had high sensitivity but slightly low specificity.
The other architectural, cytological and background features which were found to be statistically significant (p<0.05) were microacinar structures away from the sheets, cellular crowding with nuclear overlapping, nuclear grooves, squamoid cells and bubble gum colloid. All these features except the latter two had high sensitivity and low specificity.
Squamoid cells and bubble gum colloid were more specific (100% and 78% respectively) than sensitive (12% and 48% respectively). Presence of plasma cells admixed with cell clusters also showed statistical significance (p<0.05) in this study with a sensitivity of 72% and a specificity of 50%. Rest of the features (cellularity, marginated and non-marginated nucleoli, mitosis, multinucleated giant cells, hurtheloid cells, cyst macrophages, lymphocytes and neutrophils within the aspirate) were statistically insignificant (p>0.05). Although psammoma bodies which are generally considered to be significant, failed to prove so in our study. The statistical studies could not be applied on two features; individually dispersed bare nuclei and columnar cells, as two of the four values were less than five. It was not possible to analyze individual cytoarchitectural parameters by ROC since the variables were categorical and not continuous.
All the combinations of more than two features were found to be statistically highly significant in differentiating between PTC and other thyroid pathologies with a statistical significance of p=0.000 with specificity and PPV of 100% in all five combinations. Combination 3 which included flat syncytial sheets, true papillae, individually dispersed cells with eosinophilic cytoplasm, nuclear outline irregularity, thickened nuclear membranes and showed 100% sensitivity and specificity.
Our study also showed that all the combinations of more than two features to be in the first category of 0.9-1 so the test quality indicated as being excellent. All the five combinations are more or less similar with the highest area under the curve to be for combination 2 which is 0.936. The ROC for combinations of only two features completely failed the test thus highlighting the importance of having more than two cytological features for the combinations to be significant.
Although the single cytological parameters could not be compared to the combination parameters in the ROC, it was shown that certain statistically significant single cytological parameters were present in some proportion in the non-papillary group as well. However, when more than two statistically significant cytoarchitectural features were combined, all had 100% specificity and none were present in the non-papillary group as a total. This signifies the use of combination of cytoarchitectural features in the diagnosis of PTC on FNA.
As indicated in Table 6, many individual cytological features with 100% specificity and 100% PPV {except for individually dispersed cells with eosinophilic cytoplasm (72%) and true papillae (56%)} had sensitivity of 50% or less. This highlights the problem in cytological diagnosis of PTC, where individual cytological features with a high specificity and positive predictive value may not always be reliable due to low sensitivity. Nuclear outline irregularity in the category of 90% specificity and 90% PPV also had a high sensitivity (88%). In the category of 80-90% specificity and 80% PPV, most of the features had a high sensitivity except for the mitoses (12%).
Discussion
The main aim of our study was to determine the usefulness of individual cytoarchitectural features in FNA smears for the diagnosis of PTC and its variants and to further determine not only the most useful cytoarchitectural features but also the most reliable combination of cytoarchitectural features for the diagnosis of PTC in cytology smears.
The rationale behind this was that in spite of the well-defined cytological features described in numerous studies the diagnosis of PTC on cytology is often quite difficult and as of date, there has been no international standard which exists for the cytological diagnosis of PTC.
Twenty out of the 31 cytological parameters analyzed in our study were found to be statistically significant in the diagnosis of PTC when compared to the control group. Of these, 14 features had a significance of p=0.000 whilst six features had a significance of p<0.05. The 14 features included flat syncytial sheets, anatomical borders, true papillae, cellular swirls, individual cells with eosinophilic cytoplasm, anisonucleosis, elongated oval nuclei, fine powdery chromatin, nuclear outline irregularity, thickened nuclear membranes, crescent shaped collapsed nuclei with or without nuclear moulding, INCI, histiocytoid cells and cells with septate vacuolation. There was a significant correlation between the histologic diagnosis of PTC and the cytological findings mentioned above.
However, when the statistically significant cytoarchitectural features were combined, all the five combinations of more than four features had 100% specificity and none of the features when combined were all present in the non-papillary group. According to the ROC, combination of more than four cytological features was more significant than combining two features. Conclusion
This study signifies the importance of the use of combination of a larger number of cytoarchitectural features in the diagnosis of PTC on FNA. Further recommendation is to include a larger study group to devise a scoring system which can then be applied to classify the FNA as diagnostic for PTC.
Article DOI
