Objectives. This study evaluated the mechanical performance, under low-load cyclic loading, of two different knotless suture anchor designs: sutures completely internal to the anchor body (SpeedScrew) and sutures external to the anchor body and adjacent to bone (MultiFIX P). Methods. Using standard suture loops pulled in-line with the rotator cuff (approximately 60°), anchors were tested in cadaveric bone and foam blocks representing normal to osteopenic bone. Mechanical testing included preloading to 10 N and cyclic loading for 500 cycles from 10 N to 60 N at 60 mm/min. The parameters evaluated were initial displacement, cyclic displacement and number of cycles and load at 3 mm displacement relative to preload. Video recording throughout testing documented the predominant source of suture displacement and the distance of ‘suture cutting through bone’. Results. In cadaveric bone and foam blocks, MultiFIX P anchors had significantly greater initial displacement, and lower number of cycles and lower load at 3 mm displacement than SpeedScrew anchors. Video analysis revealed ‘suture cutting through bone’ as the predominant source of suture displacement in cadaveric bone (qualitative) and greater ‘suture cutting through bone’ comparing MultiFIX P with SpeedScrew anchors in foam blocks (quantitative). The greater suture displacement in MultiFIX P anchors was predominantly from suture cutting through bone, which was enhanced in an osteopenic bone model. Conclusions. Anchors with sutures external to the anchor body are at risk for suture cutting through bone since the suture eyelet is at the distal tip of the implant and the suture directly abrades against the bone edge during cyclic loading. Suture cutting through bone may be a significant source of fixation failure, particularly in osteopenic bone. Cite this article: Y. Ono, J. M. Woodmass, A. A. Nelson, R. S. Boorman, G. M. Thornton, I. K. Y. Lo. Knotless anchors with sutures external to the anchor body may be at risk for suture cutting through osteopenic bone. Bone Joint Res 2016;5:269–275. DOI: 10.1302/2046-3758.56.2000535