1This paper presents a multi-laboratory comparison study of in vitro models assessing 2 bioaccessibility of soil-bound lead in the human gastrointestinal tract under simulated fasted 3 and fed conditions. Oral bioavailability data from a previous human in vivo study on the same 4 soil served as a reference point. In general, the bioaccessible lead fraction was significantly 5 (P<0.05) different between the in vitro methods and ranged for the fasted models from 2% to 6 33% and for the fed models from 7% to 29%. The in vivo bioavailability data from literature 7 were 26.2 ± 8.1% for fasted conditions, compared to 2.5 ± 1.7% for fed conditions. Under fed 8conditions, all models returned higher bioaccessibility values than the in vivo bioavailability, 9 whereas three models gave a lower bioaccessibility than bioavailability under fasted 10 conditions. These differences are often due to the method's digestion parameters that need 11 further optimization. An important outcome of this study was the determination that the 12 method for separating the bioaccessible lead from the non-bioaccessible fraction 13 (centrifugation, filtration, dialysis) is crucial for the interpretation of the results. 14 Bioaccessibility values from models that use more stringent separation methods better 15 approximate in vivo bioavailability results, yet at the expense of the level of conservancy. We 16 conclude from this study that more optimization of in vitro digestion models is needed for use 17 in risk assessment. Moreover, attention should be paid to the separation method since it 18 largely influences what fraction of the contaminant is considered bioaccessible. 19 20