In this paper, we motivate, devise, demonstrate, and evaluate an approach for the research-based development of information systems development methods (ISDMs). This approach, termed "method engineering as design science" (ME-DS), emerged from the identified need for scholars to develop ISDMs using proper research methods that meet the standards of both rigor and relevance. ISDMs occupy a position of central importance to information systems development and scholars have therefore invested extensive resources over the years in developing such methods. The method engineering (ME) discipline has developed different frameworks and methods to guide such development work and, for that purpose, they are well-suited. Still, there remains a need for applications and evaluations of ISDMs based on the demands for knowledge justification. Unfortunately, in many cases, scholars come up short with regard to how ISDMs are generated and empirically validated. While design science (DS) stresses knowledge justification, prominent DS approaches seem to be biased toward the development of IT artifacts, making this approach ill-suited for the development of method artifacts. We therefore propose eight principles that marry ME and DS, resulting in a process model with six activities to support research-based development of ISDMs. We demonstrate and evaluate ME-DS by assessing three existing research papers that propose ISDMs. These retrospectives show how ME-DS directs attention to certain aspects of the research process and provides support for future ISDM development.