Improving the diagnosis of tuberculosis - clinical evaluation of four new diagnostics

The United Nations Sustainable Development Goals target to end tuberculosis-related deaths, transmission and catastrophic costs by 2030. In order to reach this very ambitious aim, a diverse set of various actions by country governments, civil society organisations and research communities is needed. Central components of the collective efforts are discovery, development and evaluation of new tools, interventions and strategies. Novel diagnostic tools are of particular interest since an early and accurate diagnosis is considered key to improve tuberculosis control and to accelerate the fight against tuberculosis globally.
Therefore, the overall goal of this doctoral thesis was to evaluate four new diagnostic tests based on different approaches for tuberculosis detection. The clinical accuracy studies were conducted in sub-Saharan countries with high burden of tuberculosis disease.
This thesis is based on four manuscripts:
Manuscript 1: Evaluation of giant African pouched rats for detection of pulmonary tuberculosis in patients from a high-endemic setting
The first manuscript depicts the evaluation of trained giant African pouched rats for detecting Mycobacterium tuberculosis in sputum of 469 well-characterised patients with presumptive tuberculosis in a rural setting of Tanzania. The rats could detect tuberculosis with a sensitivity of 57% and a specificity of 81%. The diagnostic performance was negatively influenced by low burden of bacilli, but independent of the HIV status. Based on evidence collected in this and other similar studies, trained tuberculosis detection rats are not suitable as a standalone sputum-based diagnostic test. Research and development should focus on improving the diagnostic accuracy, and but also on issues such as scalability, deliverability and cost-effectiveness.
Manuscript 2: Diagnostic accuracy of computer-aided detection of pulmonary tuberculosis in chest radiographs: a validation study from sub-Saharan Africa
The CAD4TB software, a computer-aided system for detection of tuberculosis on chest radiographs, has been evaluated in 861 adults with suspected pulmonary tuberculosis in rural Tanzania. The area under the receiver operating characteristic curve for the detection of culture-positive pulmonary tuberculosis was 0.84, which indicates that CAD4TB can accurately differentiate between tuberculosis cases and controls. The performance of CAD4TB was better in HIV-negative compared to HIV-infected individuals. Overall CAD4TB outperformed a clinical officer reading x-ray images, but did not reach the accuracy of an expert. After adjusting the threshold for test positivity, higher sensitivity and acceptable specificity were achieved. This qualifies CAD4TB as a potential triage test for pulmonary tuberculosis in adults before further molecular confirmation test.
Manuscript 3: Xpert MTB/RIF assay for diagnosis of pulmonary tuberculosis in children: a prospective, multi-centre evaluation
The microbiological confirmation of tuberculosis in children is a challenge because the collection of specimen is problematic. In particular, the paucibacillary nature of the disease rarely allows detection of microorganisms by smear microscopy, and culture -having also an imperfect sensitivity- is usually not available in high burden settings. Thus, diagnosis in children relies typically on clinical symptoms, tuberculin skin test, radiological findings, and history of tuberculosis contact.
The presented multi-centre evaluation of Xpert MTB/RIF in 451 children with presumptive pulmonary tuberculosis established an overall sensitivity of 68% and specificity of 100% compared to culture. Cross-reactivity with non-tuberculous mycobacteria has not been observed. Xpert MTB/RIF detected ten children with smear negative but culture-positive for Mycobacterium tuberculosis and three additional cases in absence of mycobacterial growth on culture. Overall, the results indicate that Xpert MTB/RIF can diagnose pulmonary tuberculosis in children better than smear microscopy, but the sensitivity is suboptimal in comparison to culture. Therefore, Xpert MTB/RIF cannot be used as a standalone test to rule out paediatric tuberculosis because many children will still need empirical antituberculosis therapy in case of Xpert-negative results. Yet, the technology is being further developed to a more sensitive format that will hopefully resolve the dilemma of diagnosing tuberculosis in children.
Manuscript 4: Performance of the novel TAM-TB assay to diagnose active tuberculosis in children: a prospective, proof of concept, multi-centre study
Non-sputum-based diagnostics are considered as promising solutions in the direction towards improved diagnosis of paediatric tuberculosis. The accuracy of a novel immunodiagnostic test -the TAM-TB assay- to identify children with active tuberculosis has been evaluated in a proof-of-concept study with 130 children in Tanzania. The assay uses the loss of CD27 expression on Mycobacterium tuberculosis-specific CD4 T cells as a marker of active tuberculosis. Standardised clinical case classifications based on microbiological and clinical findings served as a comparator. The TAM-TB assay detected culture-confirmed cases with a sensitivity of 83% and a specificity of 97%. Compared with Mycobacterium tuberculosis culture, the combination of Xpert MTB/RIF and TAM-TB, two assays with a turnaround time of less than 24 h, had a sensitivity of 94%, missing only one culture-confirmed case. Furthermore, five of eight children who were culture-negative and TAM-TB assay positive had highly probable or probable tuberculosis. An analysis of the five clinical cases suggested that they probably did have active tuberculosis, suggesting that the TAM-TB assay is at least as sensitive as Mycobacterium tuberculosis culture. However, the flow cytometry based assay is currently too complex for routine use. Simplification, optimisation, miniaturisation and broad clinical validation are necessary before TAM-TB can become part of diagnostic algorithms in paediatric tuberculosis.
This doctoral thesis also analyses the four clinically evaluated diagnostic approaches in the context of target product profiles for diagnostic needs of highest priority, as established by the World Health Organisation and key stakeholders. Currently, none of the diagnostics fully concurs with the requirements of the target product profiles. After further advancement, CAD4TB holds the potential for triage testing. A more sensitive version of the Xpert assay could become a smear-replacement test for the diagnosis of tuberculosis in children. A simpler and faster version of the TAM-TB assay would have prospects of becoming a non-sputum-based biomarker test.
In the last part of this doctoral thesis, accuracy studies are described in relation to current diagnostic evaluation pathways as an overarching concept. Special emphasis has been put on the important, but still underutilised studies on health impact and cost-effectiveness, which are needed beyond accuracy assessment. Further improvement of the development and evaluation of new diagnostic tests or strategies will bring the end of the global tuberculosis epidemic closer.
The results of this doctoral thesis improve our understanding of four diagnostic approaches and the key components for effective diagnostic evaluation. The lessons learnt will help to inform future clinical studies of new diagnostics for tuberculosis.