Randomized clinical trials to identify optimal antibiotic treatment duration

BACKGROUND: Antibiotic resistance is a major barrier to the continued success of antibiotic treatment. Such resistance is often generated by overly long durations of antibiotic treatment. A barrier to identifying the shortest effective treatment duration is the cost of the sequence of clinical trials needed to determine shortest optimal duration. We propose a new method to identify the optimal treatment duration of an antibiotic treatment regimen. METHODS: Subjects are randomized to varying treatment durations and the cure proportions of these durations are linked using a logistic regression model, making effective use of information across all treatment duration groups. In this paper, Monte Carlo simulation is used to evaluate performance of such a model. RESULTS: Using a hypothetical dataset, the logistic regression model is seen to provide increased precision in defining the point estimate and confidence interval (CI) of the cure proportion at each treatment duration. When applied to the determination of non-inferiority, the regression model allows identification of the shortest duration meeting the predefined non-inferiority margin. CONCLUSIONS: This analytic strategy represents a practical way to develop shortened regimens for tuberculosis and other infectious diseases. Application of this strategy to clinical trials of antibiotic therapy could facilitate decreased antibiotic usage, reduce cost, minimize toxicity, and decrease the emergence of antibiotic resistance.