Skip to main content
Fig. 4 | BMC Infectious Diseases

Fig. 4

From: A comparison of clinical development pathways to advance tuberculosis regimen development

Fig. 4

Seamless BAR optimization. A Heatmap quantifying the aggressiveness of Bayesian adaptive randomization across a range of ɣ and η values. B Variability of patient distribution across simulations represented as % CV. C Representations of aggressive and less-aggressive BAR designs shows the suitability of less-aggressive adaptation for the seamless design. Less-aggressive adaptation allows relapse data to be taken into account and adjust randomization. At 1100–1200 patients randomized Arm 1 recruitment slows down while Arm 6 speeds up, reflecting incoming relapse data that reduces confidence in Arm 1 and increases confidence in Arm 6. D Graduation and stop plot comparing Seamless MAMS and BAR. The continuous nature of the BAR recruitment was translated into a semi-discrete outcome for comparison to the MAMS design, graduation of an arm was defined as greater than 350 patients recruited into the arm, stopping of an arm defined as less than 200 patients recruited (analogous to stopping at MAMS interim) and in-between defined as 200–350 patients recruited into the arm. Both designs are comparable and are suitable for our purposes, but BAR excels at penalizing and thus stopping poorly performing arms

Back to article page