Reply to Boyles

Abstract

We thank Boyles et al. [1] for the interest shown in our work. Determine TB-LAM comprises a simple low-cost lateral-flow strip test that can provide a rapid point-of-care diagnosis of HIV-associated tuberculosis (TB) by detecting mycobacterial lipoarabinomannan (LAM) in urine samples [2]. Our first report on this assay evaluated diagnostic accuracy and predictive values when screening patients prior to antiretroviral therapy (ART) in South Africa [3]. The second, to which Boyles et al. [1] address comments, reported on the prognostic significance of positive urine LAM test results among those with confirmed TB [4]. It is important to note that, in these studies, urine samples were tested retrospectively for LAM and the results were not used to inform clinical management. The questions raised by Boyles regarding the adverse consequences of false-positive Determine TB-LAM results can only be addressed by prospective studies in which the assay informs treatment decisions. Speculation regarding the potential adverse consequences of false-positive Determine TB-LAM assay results raises two key issues: the reported specificity of the assay and the patient populations in which the assay should be used. In our clinical studies, we have found that three different generations of the LAM assays, the Chemogen MTB ELISA, the Alere Clearview TB-ELISA and the Determine TB-LAM point-of-care assay, have each had a very high specificity of 98–100% [3,5]. However, some other clinical studies of urinary LAM assays have reported lower specificity [2,6] and elsewhere we have discussed in detail potential reasons for this [2]. These include possible deficiencies in the sputum sampling and culture methods for the gold standard, lack of other sample testing to strengthen the microbiological gold standard (especially for cases in which there is extrapulmonary disease but no pulmonary involvement), cross-reactivity with nontuberculous mycobacterial species, bacterial contamination of stored urine samples and variations in manufacturing lots of the products [2]. Further studies employing rigorous methodologies with a robust TB diagnosis gold standard are needed to clarify the false-positivity rate and positive predictive value (PPV) in different populations and settings. The false-positivity rate will also depend on whether LAM assays are used in the appropriate patient populations. It is abundantly clear that these assays only have useful diagnostic accuracy in HIV-infected patients with advanced immunodeficiency [3,5–7]. In our cohort with high TB prevalence, the PPV exceeded 91% in patients with CD4 cell counts less than 150 cells/μl and those with WHO stage 3 or 4 disease [3]. In marked contrast, diagnostic accuracy is very poor at CD4 cell counts more than 200 cells/μl and in HIV-negative patients. When used appropriately to test patients with very advanced immunodeficiency in settings with high TB burden, the potential benefits of the assay are obvious. Such patients have very high mortality risk; rapid diagnosis and immediate initiation of TB treatment are therefore essential and yet conventional TB diagnostics perform poorly [8]. Even when novel rapid molecular tests such as the Xpert MTB/RIF assay are used, location away from the direct clinical interface may crucially undermine impact, as it may be operationally difficult to connect positive results with the patient in a timely way [9]. Point-of-care testing is clearly needed. In our cohort, most of those with HIV-associated TB who died had considerable delays in initiation of TB treatment and yet all had urine samples that retrospectively tested positive using Determine TB-LAM. Such patients potentially had much to gain from point-of-care diagnosis at the initial screening visit and same-day initiation of TB treatment. In this township environment, the vast majority of adults are infected with Mycobacterium tuberculosis and patients in ART services have a cumulative risk of TB of over 80% [10]. Thus, although inadvertent overtreatment of patients with false-positive results is undesirable, a proportion of such patients may nevertheless benefit from a TB-preventive effect. Indeed, this is precisely the rationale underpinning two large multicentre studies (NCT01380080 and NCT01417988) examining a strategy of empiric TB treatment for patients with advanced immunodeficiency in African settings with high TB prevalence [11]. Such patients potentially have much to gain from empiric TB treatment and relatively little to lose. Compared with this strategy, testing with Determine TB-LAM may actually provide a much more refined approach, allowing TB treatment to be targeted to the subset of patients who are most likely to benefit. Clinical impact studies that compare different strategies are needed. Acknowledgements Conflicts of interest S.D.L. was funded by the Wellcome Trust, London, UK. R.W. was funded in part by the International Epidemiologic Database to Evaluate AIDS with a grant from the National Institute of Allergy and Infectious Diseases (NIAID: 5U01AI069924–02); Cost-Effectiveness of Preventing AIDS Complications (CEPAC) funded by the National Institutes of Health (NIH, 5 R01AI058736-02); USAID Right to Care (CA 674 A 00 08 0000 700) and the South African Centre for Epidemiological Modeling and Analysis (SACEMA). There are no conflicts of interest.