An open-label, randomized study was conducted at Sappasithiprasong Hospital, Ubon Ratchathani, northeast Thailand. The study protocol was approved by the Ethical Committee of the Ministry of Public Health, Royal Government of Thailand. Recruitment was undertaken between July 1998 and October 2002 during twice-daily rounds of the medical, surgical, and intensive care wards. Patients with suspected melioidosis were examined, and microbiological samples (blood, sputum, throat swab, urine, and pus) were obtained for bacterial culture. Patients were enrolled into the study if they were 15 years of age or older, had completed an intravenous course of appropriate antibiotics for culture-confirmed melioidosis, had provided written informed consent to participate, and agreed to outpatient follow-up for at least 20 weeks. Patients with mild disease who did not require intravenous antibiotics were also eligible. Exclusion criteria were being pregnant or lactating; known intolerance or allergy to TMP-SMX, doxycycline, or chloramphenicol; and infection with a strain of B. pseudomallei
with in vitro resistance to doxycycline or chloramphenicol as determined by disk diffusion. Diabetes was defined by contemporary definitions (1
). Renal impairment was defined as a creatinine level greater than 2.0 mg/dl prior to admission.
Randomization was performed by a computer-generated sequence and was balanced after each block of 10 patients. Sealed envelopes were prepared by a member of the research unit who was not involved in enrollment. These were opened by a member of the study team following a decision to start oral treatment. Patients were randomly allocated to receive TMP-SMX (8 mg TMP and 40 mg SMX/kg of body weight daily; maximum dose, 160 mg TMP and 800 mg SMX twice daily), doxycycline (4 mg/kg daily; maximum dose, 100 mg twice daily), and chloramphenicol (40 mg/kg daily; maximum dose, 500 mg four times daily for the first 4 weeks of oral therapy) or TMP-SMX and doxycycline without chloramphenicol at the same doses. All study drugs were manufactured either by Thai pharmaceutical companies or the Government Pharmaceutical Organization and were provided free to study patients. Minimum duration of oral treatment was 12 weeks, but total duration was based on clinical history and the course of illness. Patients that had evidence of deep-seated infection (such as undrained visceral abscesses or infections involving bones or joints) or those with more severe disease on presentation tended to receive longer courses of antibiotics. Conversely, patients with mild disease, such as infections involving skin and soft tissue, or patients with parotid abscesses tended to receive shorter courses of antibiotics. Duration of antibiotic treatment was determined by a single clinician with extensive experience in the clinical management of melioidosis (W. Chaowagul), who undertook outpatient review within 4 weeks after discharge and periodically thereafter based on clinical progress.
Follow-up was continued until July 2004. Patients were primarily monitored through the outpatient department. Nonattendees were sent letters inquiring about their subsequent clinical course and visited at home by the study team if necessary.
Patients who failed to attend clinic appointments were deemed to have ceased their medications when last reviewed in the outpatient department. Recurrence of disease was treated as for the primary episode, with readmission for investigation and intensive intravenous therapy if required, followed by the four-drug conventional combination. Amoxicillin-clavulanate with supplemental amoxicillin was used for patients who developed suspected drug allergies.
The primary outcome measure was culture-confirmed relapse or time to death attributable to melioidosis. “Clinical relapse” was also determined; this was defined as possible relapse characterized by clinical features compatible with melioidosis that was treated with antibiotics active against B. pseudomallei but was culture negative. Mortality was categorized as attributable to primary infection or relapse or attributable to other causes.
Time to outcome was measured as the start of oral treatment to relapse or attributable death. Censoring events were death from causes other than culture-confirmed or clinical relapse and loss to follow-up. A comparison of adverse drug events was made between each group.
The original study design called for 226 patients to achieve an 80% power to detect a difference in relapse rates of 25% and 10% at the 0.05 significance level. All analyses were done on an intention-to-treat basis and then repeated per protocol. In the per-protocol analyses, patients who switched treatment to the other study arm were reallocated according to the final treatment they received. In this analysis, patients who changed treatment regimens to antibiotics that did not include a study regimen were not reallocated.
Statistical tests were performed using the statistical program STATA/SE, version 8.0 (StataCorp LP, College Station, Tex.). Comparisons of continuous data were performed using a Student's t test or the Mann Whitney U test where appropriate. Proportions were compared using Fisher's exact test. Time-to-event endpoints were compared using the log-rank test or Wilcoxon test as appropriate and were depicted graphically using a Kaplan-Meier graph. A Cox proportional hazards model was used to adjust the treatment effect for potential confounding factors; the following variables were considered in the forward-variable selection procedure: age, distribution of infection, initial use of ceftazidime, history of chronic renal failure, diabetes mellitus, and failure to complete at least 12 weeks of therapy for analysis of relapse after cessation of oral treatment. The final Cox proportional hazards models were reanalyzed with TMP-SMX resistance as a variable to explore its effect.
Susceptibility to ceftazidime, imipenem, chloramphenicol, doxycycline, and amoxicillin-clavulanate was performed by the Kirby-Bauer disk diffusion test. As TMP-SMX susceptibility testing by this method is not reliable and testing for MICs was not in routine use at this time (10
), test results were not reported to clinicians during this trial. Susceptibility testing using the Etest (AB Biodisk) was retrospectively performed on isolates stored at −70°C in trypticase soy broth with 15% glycerol using NCCLS methodology and breakpoints for B. pseudomallei
(susceptible, ≤2/38 μg trimethoprim-sulfamethoxazole; resistant, >4/76 μg) (9