Increased Tuberculosis Case Detection - DiOpTB

2 Objectives 2.1 Primary objective

1\. Diagnostic yield of active TB within ten days, comparing Enhanced Usual Diagnostic Procedure (EUDP) to Optimized Diagnostic Procedure (ODP).

2.2 Secondary objectives

1. Number of patients treated for TB within two weeks comparing EUDP to ODP. 2. The additional diagnostic yield of CAD CXR compared to Xpert and culture. 3. Improved follow-up (FU) rates in the cascade of care (i.e., one week and six months FU for all included and treatment start and outcome for all TB diagnosed). 4. Differences in diagnostic yield of active TB between routine sputum samples, instructed sputum samples and non-sputum samples (faecal and saliva combined with buccal/tongue swabs). 5. Feasibility of including Oxford Nanopore sequencing for detection and molecular resistance patterns measured as rate of analysed samples within two weeks.

3 Background In 2022, the WHO estimated that of a total 10.6 million new TB cases, more than three million went undiagnosed and of the remaining seven million only 57% were bacteriologically confirmed (1). In sub-Saharan African settings such as Ethiopia and Guinea Bissau, smear microscopy remains the major diagnostic tool in most areas despite the rollout of rapid diagnostic tests such as Xpert. In a multi-centre trial by Theron et al, the implementation of Xpert in African settings did indeed reduce diagnostic delay but unfortunately without any effect on the numbers who were initiated on TB treatment nor on mortality (2). The trial showed that Xpert rollout was not superior to enhanced, well-equipped, microscopy-based diagnostic facilities. In an editorial to the Lancet written by our group, it was concluded that TB elimination could be better advanced by improving currently available tools than by expanding Xpert testing to peripheral health facilities (2, 3). Nevertheless, when TB is diagnosed with a point of care test, there is a need to ensure that correct treatment is provided. However, culture-based drug susceptibility testing is very scarce in high endemic areas and often takes several weeks or months to perform (4). Recently, the Cryptic study (5) has shown that genotypic drug susceptibility testing (gDST) may guide treatment with a sensitivity and specificity well above 90% for key drugs. New techniques such as MinION (Oxford Nanopore Technologies) have also made it possible to perform sequencing and gDST directly from sputum samples. Such point-of-care-based sequencing technology is comparable in size to a USB flash drive (6) and may be attached to a laptop computer at a health centre in a high endemic area.

A cluster-randomised trial implementing the TBscore recently showed a fourfold increase in case detection rate in Ethiopia but not in Guinea-Bissau (7). It identified that factors such as laboratory capacity and routines in collecting and examining sputum smear samples may have a high impact on case detection rate and could be optimized based on the available resources. Surprisingly, the sensitivity of sputum smear microscopy ranges from 20-80% with an average of about 50% (8), which may partly be due to patient selection but also depend on considerable variability in sputum collection strategies and/or laboratory procedures. In a comparison of sputum collection methods by Datta et al (9), pooling of sputum and structured instructions before sampling on average led to a twofold higher diagnostic yield whereas a spot versus morning sample showed no difference. A multi-centre study including Ethiopia, comparing fluorescence microscopy to conventional light microscopy showed a small but significant increase in sensitivity (72.8 vs 65.8%)(10). Further, recent research has shown that buccal and tongue swabs, that are easily obtained, can hold valuable diagnostic potential. (11) In smear-negative patients with presumed TB, the available diagnostic tools in high endemic countries include CXR but standardized procedures for evaluation of CXR have been scarce. Recently, CAD software based on artificial intelligence algorithms such as qXR (Qure.ai, India) have improved detection of microbiologically confirmed TB from 50-60% by experienced radiologists to 70-84% with a specificity of 80% by CAD (12). However, a recent systematic review concluded as did the WHO that there are too few high-quality studies to fully assess its diagnostic accuracy (13). We now propose to conduct a multi-centre cluster-randomised clinical trial to evaluate whether improvements on available diagnostic resources can increase the diagnostic yield of active TB and decrease mortality for patients diagnosed with TB.

4 Methods 4.1 Location and nature of sites The present study will be conducted in two African countries: Guinea-Bissau and Ethiopia. In Bissau, the capital of Guinea-Bissau, The Bandim Health Project has been a Health and Demographic Surveillance Site (HDSS) for 45 years and has a well-defined study population of approximately 100,000 under continued surveillance. Within the study area there are two health centres (HCs) from where patients with presumed TB will be enrolled (Bandim HC, Belem HC).

In Ethiopia the study will be conducted in collaboration with the University of Gondar in the region of North-Gondar, which has a population of more than two million. Two health centres located in North Gondar Zone, namely Azezo HC and Gondar HC will participate. The Gondar University Hospital is a teaching and referral hospital and will be used for further management of severe TB cases during this study.

4.2 Epidemiology and study population 4.2.1 Guinea-Bissau The epidemiology of TB in the Bissau study population has been extensively described (14-18). The overall incidence of TB has declined only slightly since 2004 and was estimated at 273/100,000 population in 2020, while TB/HIV co-infection declined from 108 per 100.000 to 14 per 100,000 over the period (19). Smear negative cases and case fatality rate likewise declined over the period. The incidence of smear positive TB remained stable at 188 per 100,000 between 2004 and 2011 (20). All HCs have basic laboratory facilities to carry out sputum smear microscopy and all provide TB treatment. The national referral hospital for TB, Hospital Raoul Follereau, is located adjacent to the study area and is a close collaborating partner. The incidence rate of TB in Guinea-Bissau as a whole is 361 per 100,000 with a case detection rate estimated at 35% (21, 22).

4.2.2 Ethiopia TB continues to be a major public health concern in Ethiopia fuelled by the expansion of the HIV epidemic since the 1990s. According to the 2022 WHO TB report (1), Ethiopia is among high-burden countries for both TB and TB/HIV and has an estimated incidence rate of 119 per 100,000, and a TB mortality of 17.7 per 100,000 (21). HIV-positive TB incidence is 6.2 per 100,000 and case detection rate is currently estimated at 73% (21, 22). These figures are high considering that Ethiopia is the second most populous country in Africa with an estimated total population size of more than 100 million. HIV screening is carried out as a routine.

4.3 Design The present study is designed as an open-label, stepped-wedge cluster-randomised controlled trial (23) to investigate an optimized diagnostic procedure for active TB in healthcare centres in Guinea-Bissau and Ethiopia. Applying the stepped-wedge design ensures that all participating HCs will implement the intervention during the study period. This design is particularly useful for evaluating the population-level impact of an intervention, which is of interest in this study. All clusters (i.e., HCs) start with Enhanced Usual Diagnostic Procedure (EUDP) and are then randomized to switch to the intervention phase at predefined time points (see table 1).

See below for detailed description of sample size calculations.

4.4 Bandim TBscore The Bandim TBscore (TBscore) (Table 2) consists of five symptoms (cough, haemoptysis, dyspnoea, chest pain, and night sweats) and six signs (pale inferior conjunctivae, pulse \>100 per minute, positive finding at lung auscultation, temperature \>37°C (axillary), body mass index (BMI) \<18/\<16, and mid-upper-arm circumference (MUAC) \<220 mm/\<200 mm) (24). Each variable contributes one point while BMI and MUAC contribute an additional point if BMI\<16/MUAC\<200 mm; hence, the maximum score is 13. The score divides patients into three severity classes (SC): SC-I, TBscore 0-5; SC-II, TBscore 6-7, and SC-III, TBscore≥8. A simplified version of the score - TBscoreII - with a maximum score of 8 points has also been developed (25). The advantage of the latter score is that it can be performed without a physician present. The TBscore has been assessed in both Gondar and Bissau and found to be a useful add on in the diagnostic cascade of care.(7)

4.5 Buccal, tongue swap and saliva sample Buccal and tongue samples will be collected using the Omniswab (Whatman, catalogue #WB100035) and added to a container where patients leave a saliva sample. Samples will be collected by trained laboratory staff, who gently brush the inside of each cheek and then the tongue of the participant for 10 seconds with the OmniSwab. The OmniSwab has a breakpoint and the head will be ejected into 500 µl buffer containing 50 mM Tris pH 8.0, 50 mM EDTA, 50 mM sucrose, 100 mM NaCl, and 1% SDS, and transported to the laboratory at 4˚C. (11) There, the OmniSwab-collected samples will be vortexed in the saliva, and the swabs heads removed. One part of the sample will be analyzed using Xpert Ultrawhile the other part will be stored at - 80 °C until further processing.

4.6 Computer-aided detection chest X-ray (CAD CXR) applying artificial intelligence (AI) A preliminary study using an AI based CAD CXR software (qXR, Qure) compared to two Ethiopian radiologists included 498 CXRs from a previously performed randomized controlled trial on the TBscore. Of those, the less experienced radiologist found 50, the more experienced radiologist found 100 and CAD CXR found 83 to be indicative of TB. Using Xpert PCR as the gold standard for TB diagnosis, the overall AUC for the CAD CXR was 0.84 while the less experienced radiologist performed at a sensitivity of 41.4% and a specificity of 94.1% and the experienced radiologist's assessments were 55.2% sensitive and 85.0% specific. The agreement between the radiologists was moderate (kappa=0.45), as was the agreement between each radiologist and the software (kappa=0.36, kappa=0.59).

In the present study we will include a mobile phone app to guide photographing analog X-ray films. These photographs will then be uploaded to a locally placed box (qbox) and analyzed on site.

4.7 Enrolment At all sites adult patients will be screened during consultations carried out at primary healthcare centres. All patients presenting with cough of any duration, sputum production, or weight loss will have their TBscore assessed. All participating health centres have previous experience collecting the symptoms and signs necessary for the TBscore and completing a score chart from which the TBscore can be calculated. All patients with a TBscore≥4 will be referred for TB diagnostics. Patients with 4≤TBscore\<6 will be referred to fluorescence microscopy while patients with TBscore≥6 will be referred to Xpert PCR. The staff at the sites will receive general training in TB diagnosis and then the healthcare facilities will, following a random sequence, switch from Enhanced Usual Diagnostic Procedure (EUDP), consisting of standard TB program diagnostics but ensuring availability of all reagents, to intervention (i.e., OPD).

4.7.1 Enhanced Usual Diagnostic Procedure (EUDP) The standard TB diagnostics in both settings consist of performing the sputum smear analysis by the clinical routine. Smear-negative cases will be followed as per standard routine (Figure 1A).

4.7.2 The Optimized Diagnostic Procedure (ODP) intervention

A three-step package which involves (Figure 1B):

1. O

• 1. Number of patients on active TB treatment comparing EUDP clinics to ODP clinics. [Срок оценки: 1.5 years]
• 2. Diagnostic yield of CAD CXR compared to smear microscopy, Xpert PCR, and culture. [Срок оценки: 1.5 years]
• 3. Follow-up rates in the cascade of care (i.e. one-week and six-months follow-up for all included and treatment start and outcome for all diagnosed with TB) [Срок оценки: 1.5 years]
• 4. Differences in diagnostic yield between instructed sampling, buccal samples, fecal samples and routine sputum sample. [Срок оценки: 1.5 years]