BACKGROUND: Evidence on the association between sitting for extended periods (i.e. prolonged sedentary time (PST)) and cardio-metabolic health is inconsistent in children. We aimed to estimate the differences in cardio-metabolic health associated with substituting PST with non-prolonged sedentary time (non-PST), light (LIPA) or moderate-to-vigorous physical activity (MVPA) in children. METHODS: Cross-sectional data from 14 studies (7 countries) in the International Children's Accelerometry Database (ICAD, 1998-2009) was included. Accelerometry in 19,502 participants aged 3-18 years, together with covariate and outcome data, was pooled and harmonized. Iso-temporal substitution in linear regression models provided beta coefficients (95%CI) for substitution of 1 h/day PST (sedentary time accumulated in bouts > 15 min) with non-PST, LIPA or MVPA, for each study, which were meta-analysed. RESULTS: Modelling substitution of 1 h/day of PST with non-PST suggested reductions in standardized BMI, but estimates were > 7-fold greater for substitution with MVPA (- 0.44 (- 0.62; - 0.26) SD units). Only reallocation by MVPA was beneficial for waist circumference (- 3.07 (- 4.47; - 1.68) cm), systolic blood pressure (- 1.53 (- 2.42; - 0.65) mmHg) and clustered cardio-metabolic risk (- 0.18 (- 0.3; - 0.1) SD units). For HDL-cholesterol and diastolic blood pressure, substitution with LIPA was beneficial; however, substitution with MVPA showed 5-fold stronger effect estimates (HDL-cholesterol: 0.05 (0.01; 0.10) mmol/l); diastolic blood pressure: - 0.81 (- 1.38; - 0.24) mmHg). CONCLUSIONS: Replacement of PST with MVPA may be the preferred scenario for behaviour change, given beneficial associations with a wide range of cardio-metabolic risk factors (including adiposity, HDL-cholesterol, blood pressure and clustered cardio-metabolic risk). Effect estimates are clinically relevant (e.g. an estimated reduction in waist circumference of ≈1.5 cm for 30 min/day replacement). Replacement with LIPA could be beneficial for some of these risk factors, however with substantially lower effect estimates.