Tuberculosis (TB) is the leading cause of death among HIV-infected persons in sub-Saharan Africa (1). Drug-resistant TB is an emerging public health threat in HIV-prevalent settings, but diagnosis is challenging because of the severely limited laboratory capacity for culture and drug-susceptibility testing (DST). TB diagnosis for HIV-infected patients is particularly challenging because these patients may be more likely to have smear-negative pulmonary disease or extrapulmonary TB (2,3). Extrapulmonary TB often is diagnosed by clinical findings, indirect measures (e.g., chemistry and cell count of cerebrospinal or pleural fluid, ultrasound of lymph nodes, or pericardial effusions), or smear microscopy for acid-fast bacilli from aspirated extrapulmonary fluid. However, drug-resistant TB is impossible to diagnose by these methods, instead requiring mycobacterial culture and DST (4,5).

The prevalence of multidrug-resistant and extensively drug-resistant TB (XDR TB) in South Africa has risen exponentially during the past decade. At our rural study site, ≈10% of all TB cases now are drug resistant, and >90% of TB patients are HIV infected (6). Death from XDR TB exceeds 80%; most infected persons die before sputum culture and DST results are known (6). To improve case detection and decrease diagnostic delay of drug-resistant TB among patients with suspected extrapulmonary TB, we initiated a program to aspirate lymph nodes and pleural fluid for culture and DST. We quantified the yield of these lymph node and pleural fluid aspirates for diagnosing XDR TB.

We performed a retrospective cross-sectional study to determine the proportion of patients in whom TB and XDR TB could be diagnosed by culture of fine-needle aspiration of a lymph node or pleural fluid. Additionally, we sought to determine the yield of lymph node and pleural fluid aspiration beyond culture of sputum alone. All patients were eligible who had an aspirate of a noninguinal lymph node or pleural fluid sent for mycobacterial culture and DST during September 1, 2006–December 31, 2008, from the district hospital in rural Tugela Ferry, South Africa. This 355-bed hospital serves 200,000 Zulu persons.

Hospital protocol was for lymph node aspirates to be obtained at the bedside by using sterile technique and large-bore needle at the point of maximal swelling. Pleural fluid was obtained by standard thoracentesis. Care was taken not to introduce air while injecting the fluid specimen into mycobacterial blood culture bottles (BACTEC MycoF-lytic, Becton Dickinson, Sparks, MD, USA). Smear microscopy was not performed on the fluid aspirate. Bottles were transported to the provincial TB referral laboratory in Durban and cultured by using the automated BACTEC 9240 analyzer (Becton Dickinson) in which growth is continuously monitored for 42 days. Mycobacterium tuberculosis was confirmed with niacin and nitrate reductase tests. DST of positive cultures was performed by using the 1% proportional method on Middlebrook 7H11 agar for isoniazid (critical concentration, 0.2 µg/mL), rifampin (1 µg/mL), ethambutol (7.5 µg/mL), ofloxacin (2 µg/mL), kanamycin (6 µg/mL), and streptomycin (2 µg/mL) (7). XDR TB was defined as M. tuberculosis resistant to at least isoniazid, rifampin, ofloxacin, and kanamycin (8). Aspirate culture results were compared with sputum culture results if the patient also had a sputum culture performed within 2 weeks of the lymph node or pleural fluid culture. Standard practice was for 1 sputum specimen to be collected for smear microscopy and another specimen to be collected for culture, depending on the patient’s ability to expectorate. Sputum was cultured by using the automated BACTEC MGIT 960 system (Becton Dickinson); DST of positive specimens was performed as described above (9).

Medical records were reviewed for basic demographic and clinical data, including age, sex, HIV status, antiretroviral therapy, and TB history. The yield of lymph node and pleural fluid aspirates for detecting M. tuberculosis and drug resistance was described by using simple frequencies. Incremental yield of aspirate was calculated for patients who had collection for sputum and either lymph node or pleural fluid aspirate. Ethical approval was obtained from the University of KwaZulu-Natal, Yale University, and Albert Einstein College of Medicine.

For 77 patients, either a lymph node (n = 34) or pleural fluid (n = 33) was aspirated for culture and DST during the study period (Table 1). No patient had both pleural fluid and lymph node aspirates performed.

Of the 34 lymph node cultures performed, 12 (35%) grew M. tuberculosis, 1 (3%) grew nontuberculous mycobacteria, and 2 (6%) were other bacteria that were not further speciated (Table 1). Of the 12 positive M. tuberculosis cultures, 9 (75%) were drug-susceptible M. tuberculosis, and 1 (8%) was XDR M. tuberculosis; for 2, DST results were missing. Concurrent sputum samples were available for 6 (50%) of the 12 culture-positive M. tuberculosis lymph node aspirates: 3 (50%) were concordant with the aspirate culture (2 drug-susceptible and 1 XDR), and 3 (50%) were sputum culture negative.

Of the 33 pleural fluid cultures performed, 9 (27%) grew M. tuberculosis, 1 grew Cryptococcus sp., and 1 grew another bacterium (Table 1). Of the 9 M. tuberculosis culture-positive pleural fluid specimens, 3 (33%) were drug-susceptible and 6 (67%) were XDR. Among these 9 patients, 5 (55%) had concurrent sputum samples available: 3 (60%) were concordant with the aspirate culture (1 drug-susceptible and 2 XDR), and 2 (40%) were sputum culture negative.

From 17 patients, a sputum sample and either a lymph node or a pleural fluid aspirate was collected for culture and DST (Table 2). For 14, at least 1 specimen was culture positive for M. tuberculosis, of which 9 (64%) were positive for sputum, 11 (79%) were positive in the lymph node or pleural fluid, and 5 (36%) were positive by fluid aspirate alone, including 2 patients with XDR M. tuberculosis (Table 2).

In this study of predominately HIV-infected patients suspected of having extrapulmonary TB, one third of positive M. tuberculosis cultures from lymph node or pleural fluid aspirates were XDR. The additive yield for the diagnosis of any TB of these aspirates above sputum culture alone was 36%. Our findings suggest that strategies of solitary sputum culture or reliance on microscopy of nonsputum fluid analysis would miss opportunities to diagnose drug-resistant TB. Parallel culture of sputum and aspirate fluid appears to be of substantial added benefit for diagnosing XDR TB in this setting.

Our study has several limitations. Aspirates were collected on the basis of the attending physician’s clinical judgment. Therefore, the yield for M. tuberculosis and XDR M. tuberculosis may be overestimated, and other factors that may have influenced the physician’s suspicion of drug-resistant TB or increased the likelihood of a positive aspirate are not known without additional prospective study. It is also not possible to comment on the true incremental yield after comparing with sputum culture; sputum was not collected from all patients, nor were the reasons for lack of collection documented.

Nonetheless, as co-infection with HIV and TB increases in sub-Saharan Africa, the number of persons with extrapulmonary TB, both drug-susceptible and drug-resistant, is anticipated to rise (10,11). Therefore, diagnostic algorithms for extrapulmonary TB must consider the critical importance of extrapulmonary fluid culture and DST for diagnosis of drug-resistant TB, particularly in HIV-infected persons. Furthermore, this study highlights the need to validate novel diagnostic tests for M. tuberculosis drug resistance on nonsputum fluids.