Fig. 9: QR values for control/placebo and actively treated individuals in trials reported with positive outcomes inform ability to distinguish actively treated from placebo.

a QR values for control/placebo (n = 448) participants with corresponding percentiles. Blue line and shaded region indicate mean and standard deviation (SD). b QR values for active treatment individuals in trials reported with positive outcome (n = 259). Blue line and shaded region indicate mean and SD. c Probability (bold blue line) of identifying actively treated individuals with 95% confidence intervals (blue shaded region). The distribution shown among participants treated with drugs from positive trials (b) or placebos (a) was used in a logistic regression model with active treatment or placebo as the outcome. To account for imbalance between active (n = 259) and placebo (n = 448) groups, observations were weighted so that active and placebo participants were equally represented in producing the probability curve (c) shown. Active treatment observations were weighted more heavily than placebo observations; weights were determined as the target proportion divided by the actual proportion, where the target proportion is equivalent to the proportion of the dataset composed of placebo participants (i.e., the larger group, 63%). Quantile box plots (a, b) show the distribution of data; whiskers from top to bottom indicate percentiles: 0, 0.5, 2.5, 10, 90, 97.5, 99.5, 100. The box is bounded at the 25th and 75th percentiles (interquartile range) and horizontal line within box indicates median 50^th percentile. Percentiles are also annotated. Violin plots (a, b) visualize data distribution.