Challenge-based learning in blended education: a meta-analysis of deep learning outcomes.

Challenge-based learning (CBL) has gained increasing attention in higher education as a student-centered approach within blended and integrated learning, yet its effectiveness in fostering deep learning and the conditions under which it is most effective remain insufficiently established. This meta-analysis synthesized evidence from 22 experimental and quasi-experimental studies conducted in higher education to estimate the overall effect of CBL on deep learning and examine potential moderators, including discipline, grade level, class size, teaching duration, technological support, learning environment, and knowledge type. CBL demonstrated a statistically significant overall positive effect on deep learning [g = 0.792, 95% CI (0.549, 1.035), p < 0.001]. In the dependency-adjusted joint Robust Variance Estimation (RVE) meta-regression, disciplinary domain (p = 0.024) and intervention duration (p = 0.038) were significantly associated with variation in effect sizes as independent moderators, with larger effects observed in STEM contexts and longer implementation durations (>12 weeks). Separate subgroup analyses indicated differences in effect sizes across academic levels and class-size categories; however, these findings were not interpreted as independent moderator effects unless supported by the joint model. Teaching duration, environment (online, offline, or blended), and knowledge type showed no statistically significant moderation. While separate univariate subgroup analyses indicated significant moderation across academic levels (p = 0.002) and class-size categories (p = 0.029), these variables failed to retain independent moderation status in the joint model. Although some subgroup patterns suggested differences across academic levels and learning environments, the available evidence did not establish a linear or monotonic relationship between academic seniority and CBL effectiveness. Overall, CBL is an effective strategy for promoting deep learning, especially when supported by digital infrastructure and implemented in disciplines and contexts aligned with its active, inquiry-driven model. Future implementations should optimize technological integration and pedagogical alignment to maximize impact.