Interviews are the most frequently used elicitation methods in requirements engineering (RE); nevertheless, there are but few experimental research papers focused on such methods. Recently we have carried out experiments in order to analyze the effectiveness of structured and unstructured interviews. One of those papers has joined others in the field of information systems forming a basis for study. Although meta-analyzed studies seem similar in their design, a focus on threats to validity uncovers more differences than similarities. The analysis of such threats can be a tool for identifying moderator variables and understanding how to improve the design of future replications in order to create new evidence and improve the results of meta-analyses. This paper shows a procedure to achieve this.