Abstract Background Diagnosis of paucibacillary tuberculosis (TB) including extrapulmonary TB is a significant challenge, particularly in high-income, low-incidence settings. Measurement of Mycobacterium tuberculosis (Mtb)-specific cellular immune signatures by flow cytometry discriminates active TB from latent TB infection (LTBI) in case-control studies; however, their diagnostic accuracy and clinical utility in routine clinical practice is unknown. Methods Using a nested case-control study design within a prospective multicenter cohort of patients presenting with suspected TB in England, we assessed diagnostic accuracy of signatures in 134 patients who tested interferon-gamma release assay (IGRA)-positive and had final diagnoses of TB or non-TB diseases with coincident LTBI. Cellular signatures were measured using flow cytometry. Results All signatures performed less well than previously reported. Only signatures incorporating measurement of phenotypic markers on functional Mtb-specific CD4 T cells discriminated active TB from non-TB diseases with LTBI. The signatures measuring HLA-DR+IFNγ + CD4 T cells and CD45RA−CCR7−CD127− IFNγ −IL-2−TNFα + CD4 T cells performed best with 95% positive predictive value (95% confidence interval, 90–97) in the clinically challenging subpopulation of IGRA-positive but acid-fast bacillus (AFB) smear-negative TB suspects. Conclusions Two cellular immune signatures could improve and accelerate diagnosis in the challenging group of patients who are IGRA-positive, AFB smear-negative, and have paucibacillary TB.