Study design and participants
We conducted a prospective diagnostic accuracy study of Ultra on both FNA and lymph node core-needle biopsy tissue in patients with suspected tuberculosis adenitis. The study was performed at Groote Schuur Hospital, a tertiary referral academic centre in Cape Town, South Africa. Eligible study participants were adults (≥18 years), both in- and outpatients, referred with enlarged lymph nodes of > 20 mm in the widest diameter located in either the cervical, axillary or inguinal region. Patients on tuberculosis therapy were enrolled provided this had been given for <1 month (sub-analyses were done in patients on tuberculosis therapy for <24hr). Patients with contraindications to core-needle biopsy (low platelets, other coagulopathy and bleeding risk, clinically unstable, site of biopsy unsafe) were excluded. Written informed consent was obtained from all participants. The Human Research Ethics Committee of the Faculty of Health Sciences, University of Cape Town, approved the study.
Patients came from within Groote Schuur and from secondary level hospitals and day clinics in the referral area. Results of prior tuberculosis investigations (sputum Xpert or tuberculosis culture from any site within 3 months of referral, or urine lipoarabinomannan (LAM)) were recorded. Details of HIV status, tuberculosis treatment and ART were obtained.
Demographic information, symptoms, symptom duration, HIV test result, and other TB investigations performed were recorded at enrolment. Performance status was graded according to the Eastern European Cooperative Group (ECOG). The site of biopsy was recorded, along with other sites of lymphadenopathy. The presence and duration of constitutional symptoms (cough, loss of weight night sweats) were specifically enquired about as was the duration that the patient had noted the lymphadenopathy. Blood was taken for a full blood count with differential, lactate dehydrogenase and HIV status and, if positive, a CD4 count and a viral load for those on ART.
Study procedures and specimen collection
FNA was performed using a 22G needle and 5 mL syringe; further study procedures were determined by the volume of the aspirate sample obtained as shown in figure 1. A core-needle biopsy was only performed when <0.5mL of caseous material was obtained by the aspirate, due to the risk of causing a draining sinus. Initially, when both an FNA and a biopsy were performed on a participant the Ultra was performed only on the tissue, but there was a protocol change after 25 patients and Ultra was performed on both the FNA and tissue on the same patient. When a patient had both an FNA and a core-needle biopsy the TB culture was only performed on the tissue specimen. For the Ultra on the FNA, the needle and syringe were flushed into a sterile container containing 2 mL of saline. An air-dried smear for AFBs was made at the bedside from a second FNA. For culture from the FNA, the aspirate was flushed directed into mycobacterium culture medium (Middlebrook 7H9 broth medium). The core-needle biopsy was performed by an automated biopsy gun (BARD Magnum™, CR Bard Inc, Covington, GA, USA) with a 14G needle. If the lymph node was not obviously palpable the biopsy was performed under ultrasound guidance. Two or three cores were sent in formalin for histology (10-15 mm long), an additional core was cut in two with a sterile blade and sent for culture and Ultra, both in 2 ml of 0.9% saline. If all of the tests performed were inconclusive, the patient underwent a repeat core-needle biopsy or an excision biopsy at the discretion of the treating clinician.
FNA and tissue specimens were transported within 2 hours of collection to a centralized laboratory and processed individually using standardised protocols by trained laboratory staff. The smear slide made at the bedside was examined by Ziehl-Neelsen (ZN) for AFBs. The tissue obtained by core-needle biopsy was crushed using pestle and mortar. A small portion was smeared on a slide and examined with a ZN stain for AFBs. Mycobacterial culture was performed using an automated liquid mycobacterial culture system (BACTECTM MGITTM 960; Becton, Dickinson and Company, New Jersey, USA). Lymph nodes are considered a sterile site and decontamination was not performed prior to liquid biopsy. If the MGIT flagged positive, one droplet was inoculated on 2% blood agar and incubated for 24hr to check for bacterial growth. If there was no bacterial growth, the MGIT was reported as positive for mycobacterium, the specimen was decontaminated with sodium hydroxide (1%) and N-acety-L-cysteine and then repeat MGIT was performed. Positive isolates from culture media were identified by acid-fast staining followed by MRTBDRplus testing (Hain LifeScience, Hehren, Germany) to confirm the presence of M. tuberculosis and rifampicin and isoniazid sensitivity.
For the Ultra, 1.4 mL of specimen reagent was added to 0.7 mL of aspirate or crushed tissue sample. Results were reported as: invalid (no internal assay control detected); not detected; or detected (with semi-quantitation: trace, very low, low, medium, or high) and rifampicin resistance (detected, not detected, or indeterminate). Staff performing Ultra were blinded to clinical and other microbiological results.
Histological review was performed by a qualified anatomical pathologist who was blinded to the results of the ULTRA and did not have access to the culture, but would have been able to see the results of AFBs on microscopy. If granulomas were identified, a separate ZN stain was performed by the pathologist and a Periodic acid–Schiff (PAS) stains were performed for fungi.
Case definitions and statistical analysis
We assigned participants to one of three diagnostic categories on the basis of clinical, histological and microbiological investigations. Definite tuberculosis (culture-positive on lymph node aspirate or biopsy for M. tuberculosis or AFBs identified on FNA or lymph node tissue). Probable tuberculosis (no other diagnosis that would account for lymphadenopathy with one or more of: macroscopic caseation on lymph node aspirate or biopsy OR tuberculosis confirmed microbiologically [Ultra or culture] at a site other than the lymph node (eg pulmonary) OR granulomas on histology of the lymph node. The third category was not tuberculosis (did not fulfil the criteria for the other two categories). The Ultra diagnostic accuracy was also reported separately using positive culture only as the reference standard.
Sample size estimation in diagnostic accuracy studies depends on the prevalence of disease. We were expecting a high prevalence of tuberculosis based on a pilot study we conducted of core-needle biopsy in HIV-positive patients, 92% of whom had tuberculosis lymphadenitis, but the proportion of patients with tuberculosis in our study was lower than expected (40%) in a pilot phase of our study. We estimated the sensitivity and specificity of Ultra would both be around 90% based on the Cochrane meta-analysis. With a 40% prevalence of tuberculosis a sample size of 87 is needed for 95% CI widths of 10% and with sensitivity and specificity of 90%. We inflated the sample size to 100 because of uncertainties about the diagnostic accuracy of Ultra.
We calculated sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and likelihood rations by defining true or false positives and true or false negatives against the composite reference standard of probable or definite tuberculosis for our primary analysis. As secondary analyses we also determined the accuracy of Ultra using mycobacterial culture alone as the reference standard. Data was entered into a REDCap® database and analysed using the STATAv14 software package (StataCorp, College Station, Texas, USA). Baseline clinical characteristics were compared using the chi-squared or Fisher’s exact test for categorical variables and the Kruskal-Wallis test for continuous variables. We counted invalid tests (i.e. Ultra-error) as negative results. This study is reported in accordance with the Standards for Reporting of Diagnostic Accuracy Studies Guidelines .
Role of the funding source
The funders had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.