Accuracy of endobronchial ultrasound-guided transbronchial needle aspiration in HIV-infected patients with thoracic lymphadenopathy in a tuberculosis low-burden area

Endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) is an innovative technique to explore hilar and mediastinal lymphadenopathy. We aimed to assess its diagnostic accuracy in HIV-infected patients in a tuberculosis low-burden area. A retrospective review was performed of all HIV-infected patients with thoracic lymphadenopathy referred for EBUS-TBNA between January 2012 and January 2019 in 3 Northern French Hospitals. A total of 15 patients was included during the study period. Our patients were predominantly male (80%), with a mean age of 50 ± 11 years. Six patients (43%) had a CD4 cells count of less than 200/mm3. Eleven patients (73%) were receiving antiretroviral therapy, and 7 (47%) reached undetectable viral load. Adequate lymphnode sampling was accomplished in all patients. No serious complications were reported. EBUS-TBNA led to a definitive diagnosis in 12 out of 15 patients (80%). It identified 4 neoplasia, 3 atypical mycobacterial diseases, 2 tuberculosis, 1 Castleman disease, 1 sarcoidosis, and 1 professional dustiness. In 3 cases, sampling revealed normal lymphoid tissue. Active surveillance confirmed the suspected diagnosis of HIV adenitis with regression of lymphadenopathy on antiretroviral therapy in 2 cases. In one case of negative sampling, thoracoscopy led to the diagnosis of tuberculosis. In our cohort, accuracy of EBUS-TBNA was 92%. EBUS-TBNA appeared to be a safe and accurate tool in the investigation of mediastinal lymphadenopathy in HIV-infected patients in settings of tuberculosis low-prevalence. It can avoid more invasive procedures such as mediastinoscopy.

www.nature.com/scientificreports/ of mediastinal structures and enhances site selection for transbronchial sampling 5 . EBUS-TBNA has become an essential tool and is now one of the recommended first steps in the mediastinal staging of lung cancer 6 . It is also a reliable way to diagnose conditions such as sarcoidosis and tuberculosis 7,8 . Data concerning use of EBUS-TBNA in HIV-infected patients are limited. Few studies, all conducted in tuberculosis high or intermediate burden area reported case series of HIV-infected patients who underwent EBUS-TBNA and demonstrated high accuracy [9][10][11] . Majority of patients were diagnosed with tuberculosis. The authors assumed that the diagnostic yield of EBUS-TBNA in HIV patients, even in a TB low burden area would be high. We performed a retrospective review of all HIV-infected patients with thoracic lymphadenopathy referred for EBUS-TBNA in 3 Northern French Hospitals, a tuberculosis low-incidence area.
patients and methods patients and hospital setting. We conducted a multicentre retrospective cohort study in 3 Northern French Hospitals (Lille, Roubaix, Tourcoing), over a period of 7 years (1 January 2012 through 1 January 2019). Lille hospital is an academic hospital. Both other hospitals are general hospitals. All consecutive HIV-infected patients with thoracic lymphadenopathy referred for EBUS-TBNA during the study period were retrospectively included. In all cases, EBUS-TBNA was indicated to establish the etiologic diagnosis of an enlarged lymph node of unknown cause. All the patients were evaluated by CT chest with contrast before the procedure, with or without positron emission tomography-computed tomography (PET/CT). Mediastinal or hilar lymph nodes that measure > 1 cm in short axis are considered as enlarged.
ethics. According to French Law, written informed consent was obtained for all patients who signed consent forms for participation to clinical database (Nadis) and later use of the data 12  All patients signed an informed consent before EBUS-TBNA performance. The procedure was performed under conscious sedation with midazolam and nebulized local anaesthesia with lidocaine 1%-2% using a convex probe echo-endoscope (BF-UC160F/180FOL8; Olympus, Tokyo, Japan). Lymph nodes were sampled with a 21-gauge Olympus Vizi Shot single-use aspiration needle, as previously described 8 . Rapid-on-site cytologic evaluation was not performed. Histopathological and microbiological examination was performed. Direct examination for acid-fast bacilli using Ziehl-Neelsen staining and mycobacterial culture were performed in all cases. The Xpert MTB/RIF assay was performed only in case of positive microscopy for acid-fast bacilli. EBUS-TBNA was deemed as diagnostic, if it led to specific diagnosis (tuberculosis, malignancy, lymphoma, sarcoidosis, or others), or if it showed normal lymphatic tissue and a follow-up of at least one year confirmed stability or improvement in patient's condition, ruling out false-negative result.
Statistical analysis. Continuous variables were expressed as mean and standard deviation. They were compared using the Mann-Whitney test. Categorical variables were expressed as number and percentage. They were compared using the Fisher's exact test. Differences between groups were considered to be significant for variables yielding a p value < 0.05.
Bronchoscopic alveolar lavage (BAL) was performed in 6 patients (40%), showing alveolitis with macrophages (n = 4), polynuclear neutrophils (n = 1), or lymphocytes (n = 1). During EBUS-TBNA, a mean of 4.9 ± 0.9 punctures was performed. Subcarina and paratracheal stations were the most common sampled sites in respectively 47% and 40% of our patients. There were no complications during the procedure, except one bronchospasm leading to anticipated interruption of the exam. Adequate lymphoid tissue was obtained in all patients. A positive histological result of malignancy was found in 4 patients, showing epidermoid cancer (n = 2), adenocarcinoma (n = 1), non-small cell carcinoma (n = 1). Histological analysis confirmed Castleman's disease in one case and professional dustiness in one case. Gigantocellular epithelioid granulomas were found in 4 patients with caseous necrosis in 3 patients. Microbiological analysis revealed positive mycobacterial culture for M. tuberculosis in 2 patients and for M. avium-intracellulare in 3 patients. Retained diagnosis was sarcoidosis in one patient, showing negative mycobacterial culture. This patient underwent mediastinoscopy, which confirmed the diagnosis of sarcoidosis and the absence of mycobacterial infection. Finally, EBUS-TBNA established a specific diagnosis in 12 patients (80%). In 3 patients, histological analysis revealed normal lymphatic tissue. One patient benefited from a lung segmentectomy by open thoracotomy leading to the diagnosis of tuberculosis. Two others patients became asymptomatic with antiretroviral therapy alone. Follow-up for more than 1 year revealed no additional disease and they were assumed to have www.nature.com/scientificreports/ non-specific lymphadenitis. In our case-series, EBUS-TBNA diagnostic sensitivity, specificity and accuracy were respectively 92%, 100% and 93%.

Discussion
In our settings, EBUS-TBNA was rarely performed in HIV-infected patients with thoracic lymphadenopathy, even if its use increased since 2015. In our case-series, the procedure was safe and accurate in more than 90% of cases.
Only three studies have previously reported EBUS-TBNA in HIV-infected patients. They were all conducted in countries of tuberculosis high or intermediate burden: Mexico, Singapour, and India 9-11 . The WHO's global tuberculosis database last update estimates incidence of tuberculosis to be respectively 23, 47 and 199 per 100,000 population per year in Mexico, Singapour and India 13 . To our knowledge, our case-series is the first to analyse EBUS-TBNA performance in HIV-infected patients in a tuberculosis low-incidence area. Last estimation of tuberculosis incidence in France is 9 per 100,000 population per year, and even lower in Northern area of France.
Despite the tuberculosis low-incidence in our area, tuberculosis was diagnosed in 20% of our patients, confirming the still high-burden of this infection in HIV-infected patients. Previously published studies reported tuberculosis in 29 to 77% of HIV-infected patients who underwent EBUS-TBNA [9][10][11] . If we considered all mycobacterial infections, including tuberculosis and atypical mycobacterial infections, these infections represented the In our study, lung cancer was the second most frequent diagnosis, established in 4 patients (27%). In the era of combination antiretroviral therapy, a high prevalence of non-HIV related malignancy has been reported in HIV-infected patients with mediastinal lympadenopathy, even if lymphoma remained the most common neoplasia 14 . None of our patients was diagnosed with lymphoma while it accounted for half of the neoplasms in previously published studies evaluating EBUS-TBNA in HIV-infected patients [9][10][11] . Of note our patients had higher median CD4 cells count than patients recruted in those studies. While 73% of our patients were receiving antiretroviral therapy, only 18% of patients have initiated antiretroviral therapy in the only study which made this data available 11 . Another reason for the lack of diagnosis of lymphoma in our study could be the choice of performing a medianoscopy in case of suspicion of lymphoma rather than EBUS-TBNA. Many clinicians still have concern that the small size of the samples obtained with EBUS-TBNA would not be sufficient to properly determine the pathologic subtype, essential for the management of lymphoma 15 . The frequency of malignancy justified lymphadenopathy sampling for histological analysis whenever possible in HIV-infected patients with mediastinal lymphadenopathy.
Accuracy of EBUS-TBNA has been established for most of the affections which can account for mediastinal lymphadenopathy in HIV-infected patients. Others diagnosis established after EBUS-TBNA performance in our cohort were sarcoidosis, Castleman's disease and professional dustiness. Each of these diseases were diagnosed in one of our patients. The variety of diagnosis is another justification for etiological investigation of mediastinal lymphadenopathy in HIV-infected patients in tuberculosis low-incidence area.
To date, mediastinoscopy is still considered the gold standard for diagnosis of mediastinal lymphadenopathy. But compared with mediastinoscopy, EBUS-TBNA can be performed using moderate sedation. None of our patients needed general anesthesia. It is a safe procedure, even in HIV-infected patients with low CD4 cell count. In our case-series, one patient out of fifteen (7%), experienced moderate bronchospasm. Sanchez-Cabral et al. reported one pneumomediastinum (2%) 10 . No mediastinis had been described after EBUS-TBNA performance in HIV-infected patients. In a large published EBUS-TBNA series, including 1891 patients, only a very low rate (0.74%) of post-procedure minor complications were described, with no major complication 16 . EBUS-TBNA has the advantage over mediastinoscopy to routinely access posterior mediastinal (station 7) and hilar lymph nodes (stations 10 and 11). In our cohort, station 7 was the most common sampled site.
EBUS-TBNA was first validated for mediastinal staging in patients with lung cancer 6 . It was then used for diagnosis confirmation of various pathologies with mediastinal lymphadenopathies including tuberculosis, sarcoidosis and lymphoma 8,9,17 . EBUS-TBNA accuracy was high in selected patients with a high pre-test probability of a specific diagnosis. In a study including 158 patients with high clinical suspicion of mediastinal tuberculous lymphadenitis, EBUS-TBNA demonstrated a diagnostic accuracy of 98% using culture and histological results. Similarly, the diagnostic accuracy of EBUS-TBNA was 96% in patients with suspicion of lymphoma raised because of history of lymphoma, known lymphoma elsewhere, or clinical and radiological compatible presentation 17 .
But the overall diagnostic yield of EBUS-TBNA performed for intrathoracic lymphadenopathy in non-HIV patients dropped to 50-60% 18,19 . All 3 previously published studies performed in HIV infected patients reported higher global accurancy of EBUS-TBNA, ranging from 60-89% 9-11 . The author attributed these results to the high frequency of infections, especially tuberculosis. Tuberculosis can be diagnosed with multiple modalities including smear, culture and nucleic acid amplification 20 .
As the authors assumed before the initiation of this study, the accuracy of EBUS-TBNA in HIV patients, even in a TB low burden area was higher than the overall accuracy observed in non-HIV patients 18,19 . Our study confirmed the clinical significance of mediastinal lymphadenopathy in HIV-infected patients in the era of combination antiretroviral therapy and the justification of invasive sampling performance if necessary to reach a specific diagnosis. In our case-series, EBUS-TBNA diagnostic sensitivity, specificity and accuracy were respectively 92%, 100% and 93%. Among our patients, 80% underwent a PET/CT prior EBUS-TBNA performance revealing FDG-avid mediastinal lymphnodes in all patients. Choice of sampling sites guided by PET/CT probably helped to improve EBUS-TBNA diagnostic yield 21 . Use of more invasive procedures was avoid in the majority of our patients. Mediastinoscopy was performed in one patient, confirming sarcoidosis diagnosis already retained by EBUS-TBNA results. Thoracotomy was performed in one patient with associated pulmonary consolidation for whom EBUS-TBNA results were negative. It led to the diagnosis of tuberculosis.
Our study as several limits. Because of its retrospective nature, the patients who benefited from EBUS-TBNA were selected by the prescribers as being the most likely to take advantage of the procedure. But others patients could have been good candidates for EBUS-TBNA performance. The number of patients included was low. Analysis on a larger sample of patients is required to confirm the high yield of EBUS-TBNA in HIV-infected patients with thoracic lymphadenopathy. But, the limited number of patients in our study reflected the need to promote EBUS-TBNA in HIV setting and improve knowledge about this technique among infectious disease specialists.
In conclusion, our study conducted in the setting of tuberculosis low-incidence confirmed the safety and high diagnostic yield of EBUS-TBNA in the investigation of mediastinal lymphadenopathy in HIV-infected patients. EBUS-TBNA may replace more invasive procedures such as mediastinoscopy as first-line technique.

Data availability
Data are available from the authors upon request (contact: Dr Agnès Meybeck, agnesmeybeck@yahoo.fr).