Background: Despite the use of directly observed therapy (DOT) by tuberculosis control programs, patient treatment failure, relapse, and acquired drug resistance remain problematic in a small number. We investigated serum drug levels in non-HIV-infected tuberculosis patients who were receiving DOT by the health department and did not respond to treatment as expected. Methods: The indications for checking levels were as follows: (1) slow clinical response or failure to convert the sputum culture within 12 weeks; (2) treatment failure, early disease relapse < 13 months since being declared cured; (3) relapse, late disease reactivation ≤13 months since being declared cured; and (4) acquired drug resistance while receiving DOT. Baseline characteristics of control subjects who responded to therapy as expected were compared. Venous blood for analysis was obtained at 2 h after directly observed ingestion and measured by high-performance liquid chromatography. Results: Twenty-four patients receiving daily or twice- weekly standard therapy with isoniazid (INH, 300 or 900 mg) and rifampin (RMP, 600 mg) were identified; 22 had drug levels evaluated at 2 h. For INH, 15 of 22 patients (68%) had levels less than the reported target range. For RMP, 14 of 22 patients (64%) had low levels. Among the 14 patients receiving INH, 900 mg, and RMP, 600 mg, 4 (29%) had very low levels of both. Use of a combination INH/RMP tablet was associated with lower INH levels (p=0.04); however, RMP levels were higher (p<0.02). Alcohol use was associated with significantly higher RMP (p<0.01) serum concentrations. Conclusions: Important questions remain concerning the utility and timing of serum drug measurements. However, if a patient is not responding to therapy as expected and one is assured that the Myobacterium tuberculosis organism is susceptible to the drugs given and that the patient is taking the medication as prescribed, drug level monitoring should be considered.