A notable issue with many of the reviewed studies pertains to their use of functional ROI analyses that sometimes lack the more stringent statistical corrections of whole-brain analyses. For example, to overcome issues of low power, reported results are sometimes restricted to post-hoc analyses in regions that showed significant results across all subjects to all task conditions; wholebrain analyses of the main (for example, group or type of stimulus) or interaction effects, or of correlations with task performance or clinical end-points, are not consistently performed. Therefore, such ROI results could represent a Type I error but they could also miss the key neural substrates that are involved in the phenomenon under investigation, for example, craving or control of craving. A way to circumvent the limitations of post-hoc analyses is to perform both whole-brain analyses and use a priori defined anatomical ROIs148,149, which could also help to standardize the nomenclature of ROIs across studies. Other common issues pertain to incomplete presentation of the actual data (such as not providing both mean and variance, or not providing scatterplots when reporting correlations), which
