"Humanized" mice offer a window into aspects of human physiology that are otherwise inaccessible. The best available methods for liver humanization rely on cell transplantation into immunodeficient mice with liver injury but these methods have not gained widespread use due to the duration and variability of hepatocyte repopulation. In light of the significant progress that has been achieved in clinical cell transplantation through tissue engineering, we sought to develop a humanized mouse model based on the facile and ectopic implantation of a tissue-engineered human liver. These human ectopic artificial livers (HEALs) stabilize the function of cryopreserved primary human hepatocytes through juxtacrine and paracrine signals in polymeric scaffolds. In contrast to current methods, HEALs can be efficiently established in immunocompetent mice with normal liver function. Mice transplanted with HEALs exhibit humanized liver functions persistent for weeks, including synthesis of human proteins, human drug metabolism, drug-drug interaction, and drug-induced liver injury. Here, mice with HEALs are used to predict the disproportionate metabolism and toxicity of "major" human metabolites using multiple routes of administration and monitoring. These advances may enable manufacturing of reproducible in vivo models for diverse drug development and research applications.P reclinical models provide a critical window into human physiology and are important for determining drug safety and efficacy. Bioengineering advances have produced in vivo platforms with improved disease modeling (1, 2), absorption, distribution, metabolism, excretion, and toxicity (ADME/Tox) (3, 4), and circulation-mimicking capabilities (4). "Human-on-a-chip" systems, however, show limited recapitulation of pharmacokinetics and cannot support clinically relevant administration routes. Elsewhere, humanized mice, based on genetic manipulation or transplantation of human cells, have also been highly enabling (5). Mice with chimeric livers show encouraging drug responsiveness (6, 7) and pathogen susceptibility (8-12). However, these models are confined to liver-injury, immunodeficient recipients and depend on the coordinated injection of high-quality human hepatocytes, which must home to the liver from the injection site, engraft, and expand over weeks to months within the injured host (13,14). Given the limitations of current preclinical models, human metabolites and their downstream effects often go undetected until clinical trials, the most costly and dangerous phase of drug development (15).Here we develop an improved humanized mouse model by implantation of tissue-engineered human ectopic artificial livers (HEALs). In the field of tissue engineering, polymeric scaffolds are widely used for cell and drug delivery in vivo (16,17). Primary hepatocytes, however, are challenging to maintain and implant via biomaterials due to their unstable phenotype, variable engraftment efficiencies, and high metabolic needs (18). Use of partial hepatectomy with portal-caval ...