Despite numerous studies, there has been little progress in the use of biomarkers for predicting treatment response in patients with tuberculosis (TB). Patients with culture-confirmed pulmonary TB between 2010 and 2014 were prospectively recruited. Blood samples were taken upon diagnosis and 2 months after the start of standard anti-TB treatment. A pilot study utilizing measurement of TB-antigen-stimulated cytokines was conducted to select potential biomarkers for further testing. Outcome was defined as persistent culture positivity at 2 months into treatment. Of 167 enrolled patients, 26 had persistent culture positivity. RANTES, IL-22, MMP-8, IL-18, MIG, and Granzyme A were selected as potential biomarkers. For predicting persistent culture positivity, receiver-operating characteristics (ROC) analysis showed that initial RANTES (AUC: 0.725 [0.624-0.827]) and 2-month MMP-8 (AUC: 0.632 [0.512-0.713]) had good discriminative ability. Using a logistic regression model, low initial RANTES level (< 440 pg/mL), initial smear positivity, and high 2-month MMP-8 level (> 3000 pg/mL) were associated with persistent culture positivity. Low initial RANTES level and initial smear positivity had a positive predictive value of 60% (12/20) for persistent culture positivity, compared with 4% (3/75) among patients with high RANTES level and smear negativity upon diagnosis. In the 72 patients with either low RANTES/smear negativity or high RANTES/smear positivity upon diagnosis, the 2-month MMP-8 level had a positive and negative predictive value of 24 and 94%, respectively, for 2-month culture status. Aside from an initial sputum smear status, serum RANTES level at diagnosis and MMP-8 level at 2 months of treatment may be used to stratify risk for culture persistence.