Point-of-care ultrasound for tuberculosis diagnosis in children: a Médecins Sans Frontières cross-sectional study in Guinea-Bissau

**Background:** Tuberculosis (TB) remains a major cause of morbidity and mortality in children, especially in low-resource settings where diagnostic confirmation is low (<30%). Point-of-care ultrasound (POCUS) is a non-invasive, inexpensive imaging tool that can detect features of both pulmonary and extrapulmonary TB. This study aimed to describe TB-focused POCUS findings in children with presumptive TB in Guinea-Bissau, a setting with high TB, HIV, and malnutrition burdens.

**Methods:** This observational cross-sectional study was conducted at Simão Mendes National Hospital in Bissau, Guinea-Bissau, from July 2019 to April 2020. Children aged 6 months to 15 years with presumptive TB (defined by persistent cough >2 weeks, unexplained fever >1 week, extrapulmonary signs, or failure to improve after 1 week of admission) were enrolled. Participants underwent clinical evaluation, laboratory testing (HIV, GeneXpert Ultra), and clinician-performed POCUS. POCUS assessed eight signs: lung consolidation, subpleural nodules (SUNs), pleural effusion, pericardial effusion, hepatic focal lesions, splenic focal lesions, ascites, and abdominal lymphadenopathy. A positive POCUS was defined as the presence of any sign. Ultrasound images were reviewed by expert reviewers blinded to clinical data; discordant cases were reviewed by a second expert. Children were categorized as confirmed TB (GeneXpert Ultra positive), unconfirmed TB (clinical diagnosis with treatment initiation), or unlikely TB (no TB treatment, clinical improvement). Statistical analyses included univariable and multivariable logistic regression, with p<0.05 considered significant.

**Key Results:** A total of 139 children were enrolled (62 [45%] female, 55 [40%] aged <5 years). Of these, 27 (19%) had confirmed TB, 62 (45%) unconfirmed TB, and 50 (36%) unlikely TB. HIV infection was present in 59 (42%) and severe acute malnutrition (SAM) in 83 (60%). POCUS was positive in 83 (93%) of TB patients vs. 17 (34%) of unlikely TB patients. For confirmed TB, POCUS sensitivity was 85% (23/27; 95% CI 67.5% to 94.1%) and specificity 66% (33/50; 95% CI 52.2% to 77.6%). The most common signs in TB patients were lung consolidation (57%), SUNs (55%), pleural effusion (30%), and focal splenic lesions (28%). In unlikely TB, consolidation was seen in 13 (26%), but pleural/pericardial effusion, focal liver lesions, and abdominal lymph nodes were absent. SAM was significantly associated with POCUS positivity (p=0.006), while HIV status and age were not. Children with SAM had more SUNs (65.5% vs. 35.7%, p=0.009). HIV-negative children had more pleural effusion (46.9% vs. 10%, p=0.0003). Inter-reader concordance (Cohen's kappa) ranged from 0.6 (abdominal lymph nodes) to 0.9 (pericardial effusion), with overall POCUS positivity concordance of 0.8. Mean POCUS examination time was 23.5 minutes (95% CI 16 to 35).

**Clinical Implications:** This study demonstrates that POCUS can detect TB-related signs in a high proportion of children with TB, particularly lung consolidation and subpleural nodules. The high sensitivity (85%) in confirmed TB suggests a supportive role for POCUS in diagnosis, though moderate specificity (66%) indicates potential for overtreatment. The association of POCUS positivity with severe malnutrition highlights the need for heightened suspicion in malnourished children. POCUS was feasible in a resource-limited setting, with moderate-to-high inter-reader agreement and acceptable examination duration. However, the unblinded design may overestimate accuracy, and larger, blinded studies are needed to confirm these findings. POCUS could help reduce the diagnostic gap for pediatric TB in low- and middle-income countries.