A meta-analysis of findings from 50 independent studies of peer tutoring programs in Mathematics at multiple educational stages showed that 88% of these programs have positive effects on the academic performance of the participants (Hedge's g = 0.333). Some of the variables to be taken into account when developing a peer tutoring experience were analyzed. Results showed that variables such as the ages of the participants, roles, skills of the tutees (disabled or at academic risk vs non-disabled and not at academic risk), length of the sessions and frequency were not significant moderators of the academic achievement. Variables such as educational stage, design of the study, duration of the program, level of knowledge of the tutors, time of the day (school time vs out of school time) and sample size turned out to be significant moderators. Results are discussed and proposals for future research are suggested.