Many researchers have explored the impact of digital games on learning effects in different STEM subjects. The purpose of this meta-analysis is to examine the effect of digital game-based STEM education on the learning achievement of K-12 or higher education students. The analysis results of effect sizes from 33 studies (N = 3894) published from 2010 to 2020 showed that digital games contributed to a moderate overall effect size (ES = 0.667, 95% CI [0.520-0.814], p < 0.001) when compared with other instructional methods. Furthermore, the study explored multiple moderator variables and their potential impacts on learning outcomes such as control treatment, subject discipline, educational level, game type, gaming platform, and intervention duration. The findings suggest that digital games are a promising pedagogical method in STEM education that effectively improves learning gains. Additionally, the study concludes with three recommendations for future research and practices on digital games in STEM education.