Estimates were calculated using bivariate random effects models, preferentially using data excluding clinical cases (analysis A), otherwise using strict microbiological reference standards (analysis B). I²: inconsistency. ^#: if urine was frozen before lipoarabinomannan testing was performed, it was considered “frozen” whether processing of urine was done before or after freezing; ^¶: univariate random effects regression performed because too few studies to converge bivariate model; ⁺: Chemogen Inc., Portland, ME, USA; ^§: Alere Inc. (formerly Inverness Medical Innovations Inc.), Waltham, MA, USA.