In people with HIV (PWH), the WHO-recommended tuberculosis four-symptom screen (W4SS) targeting those who need molecular rapid test may be suboptimal. We assessed the performance of different tuberculosis screening approaches in severely immunosuppressed PWH enrolled in the guided-treatment group of the STATIS trial (NCT02057796). Ambulatory PWH with no overt evidence of tuberculosis and CD4 cell count <100/µL were screened for tuberculosis prior to antiretroviral therapy (ART) initiation with W4SS, chest X-ray, urine lipoarabinomannan (LAM) test and sputum Xpert MTB/RIF® (Xpert). Correctly and wrongly identified cases by screening approaches were assessed overall and by CD4 count threshold (≤50 and 51-99 cells/µL). Of 525 enrolled participants (median CD4 cell count: 28/µL), 48 (9.9%) were diagnosed with tuberculosis at enrollment. Among participants with a negative W4SS, 16% had either a positive Xpert, a chest X-ray suggestive of tuberculosis or a positive urine LAM test. The combination of sputum Xpert and urine LAM test was associated with the highest proportion of participants correctly identified as tuberculosis (95.8%) and non-tuberculosis cases (95.4%), with proportions equally high among participants with CD4 counts above or below 50 cells/µL. Restricting the use of sputum Xpert, urine LAM test or chest X-ray to participants with a positive W4SS reduced the proportion of wrongly and correctly identified cases. There is a clear benefit to perform both sputum Xpert and urine LAM tests as tuberculosis screening in all severely immunosuppressed PWH prior to ART initiation, and not only in those with a positive W4SS. NCT02057796.< Réduire