Purpose: The purpose of this study was to evaluate the quality of meta-analysis regarding exercise using Assessment of Multiple Systematic Reviews (AMSTAR) as well as to compare effect size according to outcomes. Methods: Electronic databases including the Korean Studies Information Service System (KISS), the National Assembly Library and the DBpia, HAKJISAand RISS4U for the dates 1990 to January 2014 were searched for ‘meta-analysis’ and ‘exercise’ in the fields of medical, nursing, physical therapy and physical exercise in Korea. AMSTAR was scored for quality assessment of the 33 articles included in the study. Data were analyzed using descriptive statistics, t-test, ANOVA and χ2-test. Results: The mean score for AMSTAR evaluations was 4.18 (SD=1.78) and about 67% were classified at the low-quality level and 30% at the moderate-quality level. The scores of quality were statistically different by field of research, number of participants, number of databases, financial support and approval by IRB. The effect size that presented in individual studies were different by type of exercise in the applied intervention. Conclusion: This critical appraisal of meta-analysis published in various field that focused on exercise indicates that a guideline such as the PRISMA checklist should be strongly recommended for optimum reporting of meta-analysis across research fields.