On the effect of flexible adjustment of the p value significance threshold on the reproducibility of randomized clinical trials