Does Disseminated Nontuberculous Mycobacterial Disease Cause False-Positive Determine TB-LAM Lateral Flow Assay Results? A Retrospective Review Free

Abstract

We retrospectively reviewed the Determine TB-LAM lateral flow assay (LF-LAM) results among human immunodeficiency virus–infected patients with disseminated nontuberculous mycobacterial (NTM) disease. LF-LAM was positive in 19 of 21 patients without evidence of tuberculosis (TB) coinfection. Although TB-NTM coinfection may have been underdiagnosed, our results suggest that disseminated NTM disease may cause false-positive LF-LAM results.

Urine lipoarabinomannan (LAM) is a glycolipid present in the cell wall of all species within the Mycobacterium genus [1, 2]. The Determine TB-LAM Ag lateral flow assay (LF-LAM; Alere Inc, Waltham, Massachusetts) detects LAM in the urine of patients with human immunodeficiency virus (HIV)–associated tuberculosis (TB), and is the only such test commercially available. The World Health Organization (WHO) has recommended LF-LAM for the diagnosis of TB in HIV-infected inpatients with CD4 counts ≤100 cells/μL and in patients who are severely ill [3]. The pooled sensitivity and specificity in patients with CD4 count ≤100 cells/μL are 56% and 90% respectively, though the specificity rises substantially when optimal reference standards are used [4, 5]. Due to its low cost, point-of-care design, and proven mortality benefit [1, 6, 7], LF-LAM is poised for rapid scale-up in high-burden TB/HIV settings.

Whether disseminated nontuberculous mycobacterial (NTM) infections generate false-positive LF-LAM results has not been determined, however [1, 7]. In the case of TB, it has only recently been appreciated that positive urine LF-LAM results likely reflect renal mycobacterial infection, seeded as part of disseminated disease, rather than passive renal filtration of LAM released from remote sites such as the lungs [8]. Thus, while previous studies have found a reassuringly low potential false-positive rate with pulmonary NTM disease [1, 7, 9], it is imperative that the issue be specifically investigated in cases of disseminated NTM infection.

Our aim was to retrospectively review the LF-LAM results and Mycobacterium tuberculosis (MTB) coinfection rate among HIV-infected subjects with microbiologically confirmed disseminated NTM infection seen by the infectious diseases (ID) consultation service at Helen Joseph Hospital (HJH) in Johannesburg, South Africa.

METHODS

All inpatients ≥18 years evaluated by the ID consultation service at HJH between 1 October 2015 and 25 November 2016 were considered. HJH is a 500-bed tertiary referral hospital servicing a high-burden TB/HIV-coinfected population [10]. Approximately 650 of >1600 inpatients seen by the ID consultation service each year are diagnosed with TB, and 40% of LF-LAM tests performed at HJH are positive.

Disseminated NTM infection was considered proven if a mycobacterial blood or bone marrow culture grew an NTM. Subjects with NTM infection confined to a single organ (eg, lungs) were excluded. Cultures were processed at the regional mycobacterial referral laboratory (protocol in Supplementary Materials). Other specimens sent for mycobacterial culture and/or Xpert MTB/RIF (Cepheid, Sunnyvale, California) on the same admission were included in the study, as were similar specimens collected by the referring institution up to 30 days before admission, provided the patient’s admission symptoms had already commenced by that point. Tests were ordered at the discretion of the attending clinicians. Subjects were excluded if they had received specific treatment for either TB or NTM infection within 6 months before admission. Subjects on antiretroviral therapy (ART) were considered failing if an HIV RNA load was >1000 copies/mL after ≥6 months of ART. NTM immune reconstitution inflammatory syndrome (IRIS) was diagnosed if the subject had commenced ART within 90 days and fulfilled the AIDS Clinical Trial Group 2009 IRIS case definition [11]. For subjects admitted more than once over the study period, only the initial presentation with disseminated NTM infection was considered.

LF-LAM testing was performed using Determine TB-LAM lateral flow strips. As per HJH policy, LF-LAM testing was restricted to the ID consultation service, and was considered according to WHO guidelines, but only if initial Xpert MTB/RIF-based assays were negative or unobtainable. Testing was performed by an individual trained according to the manufacturer’s instructions, and who was blinded to all clinical information. A grade 1 band intensity or higher was interpreted as a positive result. This corresponds to the band intensity for grade 2 on strips prior to January 2014. Additional details on the typical workup of TB and NTM patients at HJH are included in the Supplementary Materials.

R software (version 3.3.2) was used for statistical analysis. Confidence intervals for proportions were calculated using the Wilson score interval. Ethical approval for the study was obtained from the University of the Witwatersrand’s Medical Human Research Ethics Committee.

RESULTS

Twenty-six subjects were diagnosed with disseminated NTM infection among 1687 inpatients evaluated by the ID consultation service during the study period. The majority grew Mycobacterium avium (n = 21), followed by Mycobacterium kansasii (n = 3) and Mycobacterium intracellulare (n = 2). The subjects’ median age was 37 (interquartile range [IQR], 33.4–44.5) years. All subjects were HIV-infected, with a median CD4 count of 5 (IQR, 3–14.75) cells/μL, and 14 subjects (54%) were already receiving ART upon admission. Those receiving ART were either failing therapy (n = 5) or presented with NTM IRIS (n = 9), after a median treatment duration of 4 weeks. Twenty-one subjects had chest radiography performed, showing pulmonary abnormalities in 72% (abnormal parenchymal infiltrates in 12 subjects and hilar lymphadenopathy in 5).

Xpert MTB/RIF and mycobacterial cultures were performed on 83 specimens from blood, sputum, cerebrospinal fluid, bone marrow, and urine (Table 1). All subjects had specimens collected from a minimum of 2 different sites (eg, blood and sputum), and the median number of specimens taken per patient was 3. MTB coinfection was microbiologically confirmed in 3 subjects (12%): 1 each from a blood culture (that cultured both M. avium and MTB simultaneously), a bone marrow culture, and a sputum culture. None of the 33 sputum Xpert MTB/RIF tests done in 23 individuals were positive.

LF-LAM was performed on 23 of 26 subjects (88%). Two of the 3 subjects without LF-LAM died after LF-LAM was ordered but before a specimen was collected. Among the 23 subjects with LF-LAM results, 21 were positive (91.3%; 95% CI, 72.1%–97.6%). Excluding subjects with confirmed MTB coinfection, LF-LAM was positive in 19 of 21 subjects with disseminated NTM infection (90.5%; 95% CI, 71.1%–97.4%).

Six (23.1%) subjects died in hospital. The cause of death was disseminated NTM disease in 5 cases, and concomitant Streptococcus pneumoniae bacteremia in 1 case.

DISCUSSION

Our findings demonstrate an unexpectedly high rate of LF-LAM positivity in subjects with disseminated NTM infection. While it cannot be definitively determined whether these findings represent undiagnosed concomitant disseminated TB infection, cross-reactivity with NTM antigens, or a combination of the two, we believe it is plausible that NTM cross-reactivity may account for at least some of the positive LF-LAM results seen in this study.

Previous studies have suggested that NTM infection might cause false-positive LF-LAM results. Cross-reaction with NTM antigens in vitro had been observed in 2 studies using the older LAM enzyme-linked immunosorbent assay, albeit far more weakly than with MTB antigens [1]. A small number of false-positive LF-LAM results have been reported in diagnostic accuracy studies in subjects with NTM cultured from sputum, though these studies were unable to distinguish NTM colonization from active clinical disease [1]. A Danish study of cystic fibrosis patients with confirmed NTM pulmonary infection found a low level of LF-LAM positivity (8.7%; 95% CI, 1.3%–28.1%) [9]. Importantly though, all subjects were HIV uninfected and would not be expected to have had disseminated NTM disease.

LF-LAM diagnostic accuracy studies have demonstrated up to 99% specificity when appropriate reference standards are used [5], suggesting that any potential false-positive results due to NTM were likely to be rare. However, these studies included subjects whose median CD4 counts were 10–40 times higher than in our study [4], and therefore were unlikely to be comprised of sufficient numbers of subjects at substantial risk for disseminated NTM infection. Thus, particularly in patients with CD4 count <20 cells/μL, in whom most disseminated NTM cases are likely to occur, the LF-LAM’s TB specificity may potentially be lower than previously reported, although the specificity for the Mycobacterium genus would nevertheless be preserved.

There are several important limitations to these data. The study is retrospective and thus one cannot exclude biases, including referral bias. Undiagnosed TB coinfection is an important consideration given the high burden of TB infection in South Africa, particularly among HIV-infected patients [10]. Additional mycobacterial cultures and Xpert MTB/RIF on urine would have been important to further exclude this possibility, and to correlate with the LF-LAM results. Nonetheless, the rate of TB/NTM coinfection identified in our study (12%) falls within the 0–30% range seen in 2 other small South African studies conducted in similar populations [12, 13]. Furthermore, although NTM can outgrow MTB in culture media and obscure underlying TB coinfection [14], this reservation would not apply to the 39 negative pulmonary/extrapulmonary Xpert MTB/RIF specimens. Last, the potential cross-reactivity seen with the 3 NTM species represented in our study, all of which share the same LAM subtype as MTB (ManLAM) [2], cannot be assumed with NTM species utilizing other LAM subtypes.

Potential LF-LAM NTM cross-reactivity is important for 2 reasons: laboratory facilities required to diagnose disseminated NTM infection may be unavailable in resource-limited settings where LF-LAM will be heavily utilized, and empiric TB treatment alone is insufficient in NTM infection, where macrolides are the therapeutic backbone. Conversely, whether any LF-LAM NTM cross-reactivity could be harnessed as a diagnostic tool in disseminated NTM disease is a prospect that may warrant further investigation. To the degree that these findings can be confirmed in similar high-burden TB/HIV coinfection settings, they suggest that positive LF-LAM results should be interpreted with caution in patients with very low CD4 cell counts.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. J. S. N. acknowledges the work of the late Professor Stephen Lawn, whose seminal insights into LF-LAM have inspired the current manuscript. The authors acknowledge Right to Care’s support for Helen Joseph Hospital, outside of this study (grant number AID-674-A-12-00020).

Potential conflicts of interest. R. B. has previously received consulting fees from Cepheid Inc. All other authors report no conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References