Determining accurate left ventricular (LV) function is clinically important. Three-dimensional echocardiography (3DE) achieves better estimation than 2-dimensional echocardiography. However, underestimation of LV volumes has often been reported without a systematic attempt to synthesize these data. This meta-analysis aimed to assess the bias of 3DE in evaluating LV volumes and ejection fraction (EF) and to investigate factors affecting that bias. Studies that compared LV volumes and/or EF between 3DE and magnetic resonance imaging were eligible. Meta-analysis of 95 studies including 3,055 subjects revealed significant underestimation of LV end-systolic volume (−4.7 ml, p <0.0001) and end-diastolic volume (−9.9 ml, p <0.0001), whereas measurement for EF revealed excellent accuracy (−0.13%, p = 0.41). Meta-regression analysis for factors of systematic bias in volumetry revealed that female gender and existence of cardiac disease were associated with more underestimation, whereas use of semiautomatic tracking and matrix-array transducers counteracted the underestimation. In conclusion, by meta-analysis synthesizing many small studies, we found underestimation of LV volumes and factors affecting the systematic bias by 3DE. These data provide a more detailed basis for analyzing and improving the accuracy of 3DE, an indispensable step toward further clinical application in LV assessment.