Computer Aided Screening for Tuberculosis in Low Resource Environments (CASTLE)

Liverpool School of Tropical Medicine, Kamuzu University of Health Sciences, Malawi-Liverpool-Wellcome Trust Clinical Research Programme

People living with HIV (PLHIV) who require admission to hospital in WHO Africa region have poor outcomes. TB is very common in this group, but can be difficult to diagnose. The CASTLE trial aims to determine whether systematic screening for tuberculosis using digital chest X-ray with computer-aided diagnosis (DCXR-CAD) plus urine lipoarabinomannan testing with Fujifilm SILVAMP TB LAM (FujiLAM) plus usual care can improve admission outcomes for hospitalised PLHIV, compared to usual care alone. Our study is a single centre, unblinded, cluster-randomised (by day of admission) trial of DCXR-CAD plus FujiLAM plus usual care vs. usual care alone for screening for TB in unselected adult PLHIV admitted to a district general hospital in Malawi. The primary outcome is the proportion of people starting TB treatment by the time of death or hospital discharge. The secondary outcomes are all-cause mortality at 56 days from enrolment, proportion of people starting TB treatment within 24 hours from enrolment, and proportion of people with undiagnosed TB. In the CASTLE study we collect a single sputum sample for M. tb culture from participants and undiagnosed TB specifically refers to a person who did not start TB treatment by the time of death or discharge from hospital and has a M. tb cultured from their sputum sample. Alongside the two trial arms, a third smaller diagnostic cohort arm (1 in 9 of admission days / trial clusters) will explore the range of underlying infectious pathology. The diagnostic cohort does not contribute to trial outcomes.

Participants in the usual care arm will receive no specific trial intervention. Usual care includes tests routinely available at Zomba Central Hospital, including (but not limited to) conventional (plain film) chest X-ray, urine Alere LAM and sputum Xpert Mtb/Rif on treating clinician request.

Participants randomized to the intervention arm will receive TB screening using DCXR-CAD and urine FujiLAM. The CAD score and FujLAM results will be appended into their medical notes for treating clinicians to see. If patients have a CAD score above a pre-determined threshold the study team will attempt to collect sputum for Xpert Mtb/Rif. Chest X-ray images will be available for clinicians to view. This is in addition to usual care (detailed above).

Patients in the observational enhanced diagnostic arm will receive an enhanced package of diagnostics. This is a smaller arm (1 in 9 of all clusters) and is observational only - participants in this arm do not contribute to trial outcomes.

CAD4TB is a Computer Aided Diagnosis (CAD) image processing algorithm that can aid interpretation of Chest X-ray images to accurately detect tuberculosis.

Fujifilm SILVAMP TB LAM is a high sensitivity test for mycobacterial lipoarabinomannan (LAM) in urine samples.

Positive sputum culture for mycobacterium tuberculosis at reference lab and participant is not on TB treatment at time of hospital discharge or death.

Censored at 56 days for those who are still alive and admitted to hospital at 56 days from enrolment.

The proportion of TB diagnoses that are microbiologically confirmed vs. clinically diagnosed without microbiological confirmation.

Proportion of people randomised to DCXR-CAD plus FujiLAM who have a valid CxR and CAD score recorded, and a FujiLAM result.

Sensitivity, specificity, positive and negative predictor value compared to a composite microbiological gold standard.

To describe the proportion of patients meeting a clinical or clinical / microbiological description for the following: Sepsis; Invasive bacterial disease; Cryptococcal disease; Pneumocystis jirovecci pneumonia; Bacterial pneumonia; Immune reconstitution inflammatory syndrome (IRIS); HIV treatment failure

Inclusion Criteria: Requires acute admission to a hospital medical ward at Zomba Central Hospital for any reason Is living with HIV (existing or new diagnosis, irrespective of ART status) Willing and able to give informed consent Exclusion Criteria: Aged <18 years Has been admitted to a medical ward for longer than 18 hours Taking TB treatment before admission or has received treatment for TB within the preceding 6 months. Has already been in the study during an earlier hospital admission.

Burke RM, Nyirenda S, Twabi HH, Nliwasa M, Joekes E, Walker N, Nyirenda R, Gupta-Wright A, Fielding K, MacPherson P, Corbett EL. Design and protocol for a cluster randomised trial of enhanced diagnostics for tuberculosis screening among people living with HIV in hospital in Malawi (CASTLE study). PLoS One. 2022 Jan 10;17(1):e0261877. doi: 10.1371/journal.pone.0261877. eCollection 2022.