Liver is a complex organ with numerous key roles in healthy function of the body. Liver function manifests itself across multiple spatial scales, and its understanding in health and disease is key for early diagnosis and effective treatment. From the point of view of in silico medicine, multiscale modeling is a relevant approach, combining relatively detailed description at several levels of organization with computational efficiency. It has been shown that several diseases (e.g. fibrosis, cirrhosis) are linked to changes at the level of liver microarchitecture. However, this scale is usually not fully resolved in mathematical models, and represented either as a porous medium or a compartment model. While some models considering a fully resolved liver microvasculature exist, they are computationally expensive and difficult to parametrize. In this context, we propose a 1D-0D-1D reduced model of flow and transport in lobules, the elementary organizational units of liver. It includes advection-diffusion-reaction of solutes in sinusoidal networks, uptake by hepatocytes and secretion and further advection-diffusion-reaction in the bile canalicular network. The model is numerically implemented in the CompuTiX library for multiscale agent-based simulations to ensure maintainability and modular coupling with cell-level and whole-body models. To embed the proposed model in a multiscale context, this 1D-0D-1D model is confronted with other common representations of liver function, namely compartment models of detoxification at steady state. A theoretical framework is developed to translate information from the resolved microarchitecture to the upscaled model. It enables a systematic study of impact of microarchitecture representation in compartment models on liver function observed in simulations. Limitations of the approach are discussed. This framework is a first attempt in understanding how model parameters and experimental data are related between the whole lobule and its multicellular units.