According to a post I located on the Box2D forums here, it uses your renderer's coordinate system, with some coefficient relating world space units to renderer units.

The testbed for Box2D supposedly uses OpenGL, in which Y+ is up, X+ is right, and Z+ is projecting out from the monitor's surface.

Also, an article I found here mentions the SCREEN_TO_WORLD and WORLD_TO_SCREEN global parameters, and provides an implementation of a Camera class to overcome the challenges of the somewhat-poor implementation of this in Box2D. Even if you're already past this point, it may provide further insight into what you're trying to accomplish.

Sorry if this isn't really a true answer, but I hope I've at least pointed you in the right direction.

According to a post I located on the Box2D forums here, it uses your renderer's coordinate system, with some coefficient relating world space units to renderer units.

The testbed for Box2D supposedly uses OpenGL, in which Y+ is up, X+ is right, and Z+ is projecting out from the monitor's surface.

Also, an article I found here mentions the SCREEN_TO_WORLD and WORLD_TO_SCREEN global parameters, and provides an implementation of a Camera class to overcome the challenges of the somewhat-poor intentionally-separated implementation of this in Box2D. Even if you're already past this point, it may provide further insight into what you're trying to accomplish.

Sorry if this isn't really a true answer, but I hope I've at least pointed you in the right direction.