Correlation does not equal causation.
There's a correlation there where hs-CRP predicts fracture risk (that's generally independent as far as the other variables tested), but it's not clear (aka there's no evidence) whether hs-CRP directly causes fracture risk. There's probably some unknown variables at play here.