Significance of non‐granulomatous cytomorphology on fine needle aspirate in lymphadenitis cases classified as tuberculous by using a composite reference standard

Fine needle aspiration cytology (FNAC) is established as a first line investigation for tuberculous lymphadenitis (TBLA). We aimed to describe the various cytomorphologic features of tuberculosis (TB) on FNAC and their contribution in the diagnostic decision‐making in suspected TBLA cases.


| INTRODUCTION
Tuberculosis (TB) remains a major cause of mortality and morbidity worldwide, with extrapulmonary TB (EPTB) accounting for approximately 16% of all TB cases globally. 1 Tuberculous lymphadenitis (TBLA) is the most common presentation of EPTB, accounting for approximately 35%, and is one of the most common causes of lymphadenopathy in a high TB endemic setting. 2 In determining whether a lymph node swelling is due to TB infection or other etiologies, e.g. malignancy, histopathology has been central, with an emphasis on the presence of granulomas. In recent decades, however, fine needle aspiration cytology (FNAC) has emerged as a first line of investigation for routine practice 3 and is recommended by the TB control programme in India for the diagnostic work up in suspected TBLA. 4 The test has clear advantages in being relatively cheap, without the need for advanced laboratory facilities, and could provide an alternative diagnosis within hours. Furthermore, it is minimal invasive and with reduced risk of serious complications compared to excision biopsies. 5 Albeit not as accurate as a biopsy to differentiate between m. tuberculosis and other granulomatous conditions, the high cost of molecular based tests, and poor sensitivity of routine TB diagnostic tools to analyze paucibacillary lesions supports FNAC as a first choice of investigations for clinicians in routine practice in low-resource settings. [6][7][8] Similarly to histopathology, the typical cytomorphology of TB on FNAC have been centered around the findings of granulomas or epithelioid cells with or without necrosis. This reliance on the presence of epithelioid cells as a surrogate of granulomatous patterns, however, might be a simplification, as several studies have found that presence of some atypical features on FNAC can be the only findings suggestive of TB. 2,[9][10][11] In a TB endemic setting, features resembling suppurative lymphadenitis are known to be a challenge for pathologists to rule out TB, especially in immunocompromised patients. 9 Furthermore, reactive patterns may represent TB due to the relative small amount of tissue samples on FNA. 12 In this study, from the high TB burden setting of India, the aim was to describe the various cytomorphologic features of TBLA on FNAC and the contribution of non-granulomatous patterns in the diagnostic decision-making in suspected TBLA cases.

| Study design and participants
This cross-sectional study was nested in a prospective cohort study Patients with a clinical presentation considered suggestive of TBLA by clinicians at CRGH were included in the study, and were prospectively enrolled from in-and outpatient departments between April 2018 and February 2020. Exclusion criteria were insufficient material for investigation on FNAC or receiving anti-tuberculous treatment at the time of enrollment. Informed written consent was obtained from all participants. None of the invasive procedures were performed for research purposes only. Verbal and written consent was obtained prior to HIV testing. The study was approved by the Regional Committee for Medical Research Ethics in Norway, REK Helse-Vest (2014/46/REK vest) and by the ethical committee in India (IEC 10/2018).

| Data collection
Detailed demographic information, clinical history and examination of all recruited patients was recorded by using a pre-designed questionnaire. Patients were interviewed in Hindi or in their local language.
The presence, nature and duration of constitutional and local symptoms were specifically inquired.
Site, size and appearance of lymph nodes were recorded, blood samples were collected and radiological investigations including chest X-ray and/or ultrasound were performed for all patients. In some cases, computed tomography was available.
All patients were followed up regularly until the completion of treatment. At follow-up, clinical parameters for assessing response to treatment, including signs and symptoms, weight, and lymph node size were recorded. Patients not on anti-tuberculous treatment were followed up for 6 months after enrollment.

| Diagnostic samples and tests
Fine needle aspiration from lymph nodes was performed under sterile conditions, using a 21 G needle attached to a 10 mL disposable syringe. Specimen from lymph nodes were subjected to the routine diagnostic work-up; one of the smears was stained with May Grunwald-Giemsa for cytology, and one was stained with Ziehl-Neelsen (ZN) for demonstration of acid-fast bacilli (AFB). Criteria for a positive result ranged from 1-2 bacilli per 100 fields (100X) (weakly positive) to >9 bacilli per field (100X) (strongly positive). 13 The material from the syringe was used for Lowenstein-Jensen (LJ) culture, and GeneXpert MTB/RIF assay (Cepheid GeneXpert ® System -Xpert).
Sputum was also collected from most patients for Xpert, LJ culture and AFB microscopy. Mycobacterial culture from lymph nodes was performed at the microbiology laboratory at RDGMC. The smears were incubated at 37 C on Lowenstein-Jensen medium and cultured for 8 weeks. The Xpert assay was performed according to protocols by the WHO at RDGMC. 15 Gram staining was performed by the pathologist in some cases.
When a Mantoux test was performed, a skin reaction size of >15 mm at 48 h was considered positive. 16 A random blood glucose or two-hour plasma concentration after oral glucose tolerance test of above 11.1 mmoL/L was considered diagnostic for diabetes mellitus (DM). 17

| Patient categorization
A composite reference standard (CRS) was used to categorize the study participants into the bacteriologically confirmed TB, clinically confirmed TB and non-TB groups based on various diagnostic criteria ( Table 1). The categorization was done at the end of follow-up and without the FNAC result included in the CRS.

| Statistical analysis
Statistical analysis was performed using Statistical Package for the Social Sciences (SPSS) for Windows, version 28. Chi square test was done to assess differences in categorical variables. Sensitivity, specificity, positive predictive value, negative predictive value and accuracy was calculated using cross-tabulation. A p-value <0.05 was considered statistically significant.

| RESULTS
A total of 266 patients were included in the study. Based on the CRS, 158 (59%) were classified as TB patients, with 56 (21%) categorized as bacteriologically confirmed TB and 102 (38%) as clinically confirmed TB. The non-TB group consisted of 108 (41%) patients, serving as control group.

| Baseline characteristics
There was a significantly higher proportion of females in the TB group compared to the non-TB group ( Table 2). The age distribution varied significantly between the groups with most of the TB patients (54%) being between 15-29 years. Few patients presented with HIV or DM comorbidity in both groups. Most of the patients in our cohort were underweight.
A significantly higher proportion of TB patients (85%) presented with systemic symptoms compared to 75% in the non-TB group.
Regarding local signs, a significantly higher proportion of TB patients presented with lymph node enlargement in the posterior cervical triangle. No significant difference between the groups was observed in lymph node tenderness, discharge, or matted lymph node. The most common finding was unilateral lymph node enlargement, found in more than 90% in both groups.
The mortality in the cohort was low, 0.63% and 0.83% among TB and non-TB patients respectively.

| Cytologic findings and diagnostic validity of FNAC
On microscopic examination (Figure 1), the cytological findings compatible with TB were classified into four patterns according to necrosis and arrangement and composition of cells. The distribution of these four patterns according to TB category based on the CRS is shown in Table 3. The most common pattern was granulomatous inflammation with necrosis found in 59% of the TBLA cases, followed by necrosis only (21%). Granulomatous inflammation without necrosis (6%) was the least common finding. Non-TB cytomorphology was reported in 15% of CRS positive cases, with reactive patterns constituting 13% of these cases. In the bacteriologically confirmed TB group 54/56 (96%) of cases presented with cytomorphology considered tuberculous by the pathologist.
The sensitivity, specificity and 95% C.I. for the various patterns on FNAC compared to a CRS is shown in Table 4 Clinical presumptive EPTB patient started on anti-tuberculous treatment (ATT) with a good response a at 2/3 months and/or at end of treatment.
Non-TB case Patient started on ATT based on clinical presumptive EPTB but did not respond to treatment OR Improvement without ATT and/or response to specific non-tuberculous therapy OR Alternative diagnosis concluded by the clinician a A good response to treatment was considered when a minimum of three of the following was recorded at 2/3 months and/or at end of treatment: (1) Improvement in systemic and local symptoms, (2) any weight gain, (3) reduction in lymph node size, (4) improvement in subjective score using VAS and EQ-5D. Abbreviations: EPTB, Extrapulmonary tuberculosis; LJ culture, Lowenstein-Jensen culture.

| Correlation between cytomorphologic patterns and various TB diagnostic tests
The correlation between cytomorphologic patterns and a positive ZN stain, Xpert and culture result is shown in  (20) 23 (21) ≥45 16 (10) 22 (20) HIV positive a 6 (4) 2 (2) .480 DM comorbidity 7 (4) 9 (8) .189  Figure 1A: Epithelioid cells without necrosis (X40). Figure 1B: Epithelioid cells with necrotic background (X40). Figure 1C: Necrosis with scanty neutrophils (X40). Figure 1D: Caseous necrosis with presence of lymphocytes (X40). Figure 1E: Non-TB abscess, showing abundant polymorphs in extensive necrotic background with nuclear debris (X40). Figure    Other e-f 1 (1. In a high TB endemic setting, the major differential diagnosis to TBLA are acute infections, nontuberculous mycobacterial (NTM) infections, and malignancies, with very different treatment options. 18 By using FNAC, an invasive biopsy procedure for diagnostic purposes could be avoided, thus sparing the patient for complications and scarring. However, the features of granulomas on FNAC may be missing or atypical, resulting in reduced sensitivity. Moreover, non-specific findings resulting in a low specificity may potentially lead to a delayed diagnosis of an underlying condition.
The presence of necrosis in the absence of granulomatous inflammation in a high TB burden setting represents a challenge for the pathologist to rule out TB. Classically, caseous necrosis has been considered TB, however, a more atypical necrotic pattern, resembling abscess has been described, especially in immunocompromised. 19 In our study we proposed that the two necrotic patterns; necrosis without predominance of neutrophils and caseous necrosis with or without presence of lymphocytes could suggest TB. A high proportion of TB cases present with only necrosis in our study (21%), which is similar to other studies from high burden settings, as shown in  Even in the presence of granulomatous inflammation, it can be difficult to differentiate TB from other diseases causing granulomas on FNAC, e.g., sarcoidosis, leprosy and fungal infections. 28 In our study, granulomatous inflammation was present in the majority, 102/158 (65%), of TB cases. When analyzing the diagnostic accuracy of the various morphologic patterns, granulomatous inflammation with necrosis showed a higher sensitivity (59%) than the other patterns, and a specificity of 85%. Still, even this pattern, generally con- In our cohort, we found a significantly higher proportion of females in the TB group, with a male to female ratio of 1:2. A predominance of females among TBLA patients has also been shown in other studies. 30,31 Also in line with existing literature, 31 we found a significantly higher proportion of TBLA in the younger age group of 15-29 years. Regarding cytomorphology, gender did not affect the findings on FNAC. When stratifying the cohort into age groups, we found a higher proportion of patients above 45 years of age presenting with a necrotic pattern, which could be explained by a weakened immune system in the elderly.
EPTB has been shown to be more prevalent among HIV coinfected and underweight, however, in our study there was no significant difference between the TB and non-TB group regarding these known risk factors. Perhaps surprisingly, the prevalence of DM was higher in the non-TB group, although not statistically significant. This contrasts with studies on pulmonary TB (PTB) where DM is known to increase the risk of developing the disease, 32 and screening for DM is currently recommended in the TB control programme in India. 33  we found a significantly higher proportion of TBLA cases presenting with systemic symptoms when compared to the non-TB group, while no difference was observed between TB and non-TB cases regarding local findings. As TBLA is widely diagnosed clinically in rural lowresource settings, the diagnostic criteria should improve.
The strengths of this study are the relatively high number of participants being prospectively enrolled in a TB endemic setting, a long follow-up, and the comparison of FNAC performance to a CRS consisting of a wide range of diagnostic tests and clinical response to treatment. By using a CRS, we wanted to compensate for the lack of a perfect gold standard in diagnosing TBLA. However, the limitation of a CRS is the risk of misclassification bias, especially in the clinically confirmed group where response to treatment is a major criterion.
Keeping in mind that rifampicin, used in first-line TB treatment, is a rather broad-spectrum antibiotic, this could result in improvement among patients with other bacterial infections. Another limitation of the study is that the protocol did not guide the clinicians on what symptoms or clinical findings were compatible with TBLA, which could lead to a variation in pre-test probability of TB for the patients enrolled. On the other hand, we wanted the data to mimic real-life practice, providing valuable data on routine TB diagnostics and management in high TB burden and limited-resource settings.
In conclusion, we found that about one-third of TBLA patients presented without granulomas on FNA, highlighting the importance of considering TB in a wide spectrum of cytomorphology in a high TB burden setting. Our study supports the use of FNAC as a first-line investigation tool for diagnosing TBLA in a low-resource setting due to its relative simplicity and good sensitivity. However, the low specificity of FNAC, emphasizes the need for a second-tier confirmatory test with improved specificity.

AUTHOR CONTRIBUTIONS
OMBH performed data curation, formal analysis, drafted the manuscript, performed literature review and edited the manuscript. MK performed data collection, investigations, sampling and reviewed the manuscript. NA performed data collection, investigations and sampling. MRP developed the methodology, supervised and edited the manuscript. TM conceptualized the study, acquired funds for the study, developed the methodology, supervised and edited the manuscript. All authors read and approved the final manuscript.