As alternatives to in vivo assays, physiologically based in vitro methods have been developed to measure bioaccessibility of hydrophobic organic contaminants (HOCs) in soils. However, bioaccessibility can usually be underestimated since in vitro tests fail to provide sufficient affinity for HOCs. Sorption sink was therefore included to simulate intestinal cell absorption and to promote the mobilization of HOCs from soils. In this study, polydimethylsiloxane (PDMS) vials, widely used as passive dosing, were introduced as a sorption sink to improve the performance of physiologically based extraction test (PBET). The bioaccessibility of PCBs (representatives of HOCs) in 13 lab-spiked soils measured by PBET coupled with PDMS vials ranged from 56.5 ± 2.7% to 109.3 ± 1.5%. Correlation was conducted between the bioaccessibility and relative bioavailability (RBA) of PCBs assessed using an in vivo mouse model. A significant correlation (p < 0.001, R2 = 0.72, slope = 0.85 ± 0.16) was observed between in vitro and in vivo data, indicating that the proposed method here can be a robust in vitro method to predict PCB RBA in soils. The accuracy of this novel method was further shown by extracting one field contaminated soil with environmental relevant levels of PCBs. The relative standard deviation of bioaccessibility measured by PBET with PDMS vials was 1.2–9.8%, and much lower than those by PBET alone with values of 17.1–63.6%. In addition, the PDMS vials can be reusable as sorption sink, and no significant variation (p = 0.44) in PCB bioaccessibility was observed among 5 cycles of extracting soils with PBET coupled with PDMS vials. Due to the high sorption capacity of PDMS and flexibility of PDMS mass used for vials, the novel method here is expected to be applicable in soils with a wide range of contamination levels.