Software Product Lines (SPLs) enable and maximize reuse of software artefacts, using software variability as central technique. In Model-Based Safety Analysis, system and software models are annotated with failure models that are used to produce safety analysis artefacts like fault trees and FMEAs. However, little work has been done to show MBSA in product lines, exploiting failure models to create safety analyses for variants in the product line. State machines have been widely used to support both fault propagation and probabilistic system safety analysis. In this paper, we introduce an approach to support variability modeling and reuse of state-machine diagrams used for system safety analysis. The approach enhances traditional software product line cycle with new activities aimed to support the reuse of safety information using state-machine diagrams and facilitates the management of the diversity of functional safety across system configurations using variability models. We evaluate our approach using an automotive braking system where we show reduction of the burden of safety analysis and improvements in traceability between safety artifacts and variability abstractions.