Background Lipoarabinomannan (LAM) is a major antigen of Mycobacterium tuberculosis (MTB). In this report, we evaluated the ability of a novel immunoassay to measure concentrations of LAM in sputum as a biomarker of bacterial load prior to and during treatment in pulmonary tuberculosis (TB) patients. Methods and findings Phage display technology was used to isolate monoclonal antibodies binding to epitopes unique in LAM from MTB and slow-growing nontuberculous mycobacteria (NTM). Using these antibodies, a sandwich enzyme-linked immunosorbent assay (LAM-ELISA) was developed to quantitate LAM concentration. The LAM-ELISA had a lower limit of quantification of 15 pg/mL LAM, corresponding to 121 colony-forming units (CFUs)/mL of MTB strain H37Rv. It detected slow-growing NTMs but without cross-reacting to common oral bacteria. Two clinical studies were performed between the years 2013 and 2016 in Manila, Philippines, in patients without known human immunodeficiency virus (HIV) coinfection. In a case-control cohort diagnostic study, sputum specimens were collected from 308 patients (aged 17-69 years; 62% male) diagnosed as having pulmonary TB diseases or non-TB diseases, but who could expectorate sputum, and were then evaluated by smear microscopy, BACTEC MGIT 960 Mycobacterial Detection System (MGIT) and Lowenstein-Jensen (LJ) culture, and LAM-ELISA. Some sputum specimens were also examined by Xpert MTB/RIF. The LAM-ELISA detected all smear- and MTB-culture–positive samples ( n = 70) and 50% ( n = 29) of smear-negative but culture-positive samples ( n = 58) (versus 79.3%; 46 positive cases by the Xpert MTB/RIF), but none from non-TB patients ( n = 56). Among both LAM and MGIT MTB-culture-positive samples, log 10 -transformed LAM concentration and MGIT time to detection (TTD) showed a good inverse relationship (r = −0.803, p < 0.0001). In a prospective longitudinal cohort study, 40 drug-susceptible pulmonary TB patients (aged 18-69 years; 60% male) were enrolled during the first 56 days of the standard 4-drug therapy. Declines in sputum LAM concentrations correlated with increases of MGIT TTD in individual patients. There was a 1.29 log 10 decrease of sputum LAM concentration, corresponding to an increase of 221 hours for MGIT TTD during the first 14 days of treatment, a treatment duration often used in early bactericidal activity (EBA) trials. Major limitations of this study include a relatively small number of patients, treatment duration up to only 56 days, lack of quantitative sputum culture CFU count data, and no examination of the correlation of sputum LAM to clinical cure. Conclusions These results indicate that the LAM-ELISA can determine LAM concentration in sputum, and sputum LAM measured by the assay may be used as a bi...