The purpose of this study is to introduce a novel approach to incorporate patient-specific breathing variability information into 4D dose simulation of volumetric arc therapy (VMAT)-based stereotactic body radiotherapy (SBRT) of extracranial metastases. Feasibility of the approach is illustrated by application to treatment planning and motion data of lung and liver metastasis patients. The novel 4D dose simulation approach makes use of a regression-based correspondence model that allows representing patient motion variability by breathing signal-steered interpolation and extrapolation of deformable image registration motion fields. To predict the internal patient motion during treatment with only external breathing signal measurements being available, the patients' internal motion information and external breathing signals acquired during 4D CT imaging were correlated. Combining the correspondence model, patient-specific breathing signal measurements during treatment and time-resolved information about dose delivery, reconstruction of a motion variability-affected dose becomes possible. As a proof of concept, the proposed approach is illustrated by a retrospective 4D simulation of VMAT-based SBRT treatment of ten patients with 15 treated lung and liver metastases and known clinical endpoints for the individual metastases (local metastasis recurrence yes/no). Resulting 4D-simulated dose distributions were compared to motion-affected dose distributions estimated by standard 4D CT-only dose accumulation and the originally (i.e. statically) planned dose distributions by means of GTV indices (dose to 98% of the GTV volume). A potential linkage of metastasis-specific endpoints to differences between GTV indices of planned and 4D-simulated dose distributions was analyzed.