Comparative Efficacy of Chest Ultrasound, Chest X-Ray and Computer-aided Diagnostic (CAD) for Tuberculosis Diagnosis in Low-Resource Setting: A Cross-Sectional Study From Ethiopia (CUSTB)
Comparative Efficacy of Chest Ultrasound, Chest X-Ray and Computer-aided Diagnostic (CAD) for Tuberculosis Diagnosis in Low-Resource Setting: A Cross-Sectional Study From Ethiopia (CUSTB)
Pulmonary Tuberculosis (TB) remains a significant global health concern, particularly in low- and middle-income countries (LMIC), where resources for healthcare are often limited. While CXR is the standard imaging modality for TB diagnosis, its sensitivity and specificity can vary depending on factors such as the stage of the disease and the quality of the image obtained. This study endeavors to assess the diagnostic precision of Chest Ultrasound (CUS) relative to Chest X-ray (CXR) and CAD score in the detection of Pulmonary Tuberculosis (TB) among both index cases and household contacts.
BACKGROUND Pulmonary Tuberculosis (TB) remains a significant global health concern, particularly in low- and middle-income countries (LMIC), where resources for healthcare are often limited. Early and accurate diagnosis of TB is crucial for timely initiation of treatment and prevention of transmission, yet it presents challenges due to various factors including the complexity of symptoms and limitations in diagnostic tools. While CXR is the standard imaging modality for TB diagnosis, its sensitivity and specificity can vary depending on factors such as the stage of the disease and the quality of the image obtained. Specifically, in Ethiopia, a recent study on the facilitators of pulmonary tuberculosis diagnosis emphasizes the importance of integrating radiographic screening with symptom-based screening in health facilities, while acknowledging the high cost of this implementation. In recent years, CUS has emerged as a promising adjunctive tool in TB diagnosis, especially in at-risk populations, such as People Living with HIV. Its advantages include portability, lack of radiation exposure, and potential for bedside use, making it particularly valuable in resource-limited settings where access to advanced imaging techniques may be limited. The initial evidence indicates that the use of CUS exhibits high sensitivity in detecting microbiologically confirmed TB among adults. Additionally, Computer-Aided Diagnosis (CAD) systems utilizing artificial intelligence algorithms have shown promise in improving diagnostic accuracy by assisting clinicians in interpreting medical images. However, despite these advancements, limited research has directly compared the diagnostic performance of CUS, CXR, and CAD score in the diagnosis of TB, with no evidence at all in household contacts of index cases. Understanding the comparative effectiveness of these diagnostic modalities is essential for optimizing TB diagnosis strategies and improving patient outcomes, especially in high-risk populations such as household contacts who are at increased risk of TB transmission11. Therefore, this study aims to fill this gap by evaluating the diagnostic accuracy of CUS, CXR, and CAD score in identifying TB among both index cases and household contacts. By employing a cross-sectional design, the study seeks to determine the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of each diagnostic modality, as well as explore potential correlations and discrepancies among them. The findings from this study have the potential to inform clinical practice guidelines and contribute to the development of more effective TB diagnosis and management strategies tailored to the needs of diverse populations and healthcare settings.
STUDY DESIGN AND METHODS This study will employ a cross-sectional design to compare the diagnostic accuracy of CUS with CXR and CAD score in identifying TB among index cases and household contacts. The study will last 12 months at the end of which data analysis will be performed by the research team. No interference with the routinary activities related with the admission and the care of the patient is expected in the research. The participants will be enrolled at the admission in Outpatient Department (OPD) or Medical Ward, after the inclusion criteria are spontaneously encountered thanks to independent clinician evaluation. Vital parameters and clinical signs will be recorded CXR will be performed as standard of care with capture images in the posteroanterior view, ensuring adequate visualization of the chest area. Position the participant in an upright or standing position, facing the X-ray machine. The acquired chest X-ray images will be transferred to the CAD software platform, which may highlight regions of interest and assess a score according to the findings. During CUS, the participant will stay in a supine or seated position, exposing the chest area for ultrasound examination. A thin layer of ultrasound gel will be applied to the skin to facilitate acoustic coupling and improve image quality. STATISTICAL ANALYSIS Two interim analyses are planned at 3 and 6 months of enrollment to verify the assumptions about the sensitivity and specificity of CUS, CXR and CAD in diagnosing TB. The interim analyses will not include any formal statistical testing. Categorical data will be summarized as absolute and relative frequencies. Numerical data will be summarized using mean and standard deviation (SD), or median and interquartile range (IQR). In accuracy investigation, the standard measures will be calculated (sensitivity, specificity, positive predictive value, negative predictive value). The non-inferiority hypothesis will be testes using a RMLE-based score test according to Liu et al.12 Adjustment for multiple testing will be performed according to Benjamini- Hochberg procedure. Concordance between CUS and CXR, and between CUS and CAD will be assessed using Cohen's kappa and Gwet's AC1. Comparisons between variables will be performed with exploratory purpose using Pearson's or Spearman's correlation coefficients, Student's t-test, paired Student's ttest, Mann-Whitney test, Wilcoxon test, Chi Square test, or Fisher's test, as appropriate. Estimates will be reported with 95% confidence intervals were appropriate. Statistical significance will be set at 5%. The statistical analysis will be carried out using R 4.3 (R Foundation for Statistical Computing, Vienna, Austria).
Inclusion Criteria:
Exclusion Criteria:
giacguido@gmail.com3202842602 ext. +39