Defining the precise clean-up goals for lead (Pb) contaminated sites requires site-specific information on relative bioavailability data (RBA). While in vivo measurement is reliable but resource insensitive, in vitro approaches promise to provide high-throughput RBA predictions. One challenge on using in vitro bioaccessibility (BAc) to predict in vivo RBA is how to minimize the heterogeneities associated with in vivo-in vitro correlations (IVIVCs) stemming from various biomarkers (kidney, blood, liver, urinary and femur), in vitro approaches and studies. In this study, 252 paired RBA-BAc data were retrieved from 9 publications, and then a Bayesian hierarchical model was implemented to address these random effects. A generic linear model (RBA (%) = (0.87 ± 0.16) × BAc + (4.70 ± 2.47)) of the IVIVCs was identified. While the differences of the IVIVCs among the in vitro approaches were significant, the differences among biomarkers were relatively small. The established IVIVCs were then applied to predict Pb RBA of which an overall Pb RBA estimation was 0.49 ± 0.25. In particular the RBA in the residential land was the highest (0.58 ± 0.19), followed by house dust (0.46 ± 0.20) and mining/smelting soils (0.45 ± 0.31). This is a new attempt to: firstly, use a meta-analysis to correlate Pb RBA and BAc; and secondly, estimate Pb RBA in relation to soil types.