PURPOSE: Although most in vitro studies indicate that collagen is a suitable biomaterial for tendon and ligament tissue engineering, in vivo studies of implanted collagen for regeneration of these tissues are still lacking. The objectives of this study were the following: (1) to investigate the regeneration of the central third of the ovine patellar tendon using implants made of an open array of collagen fibres (reconstituted, extruded bovine collagen); and (2) to compare two collagen crosslinking chemistries: carbodiimide and carbodiimide associated with ethyleneglycoldiglycidylether. METHODS: Forty-eight Welsh Mountain sheep were operated on their right hind leg. The central third of patellar tendon was removed and substituted with carbodiimide (n = 16) and carbodiimide-ethyleneglycoldiglycidylether-crosslinked implants (n = 16). In the control group the defect was left empty (n = 16). The central third of contralateral unoperated tendons was used as positive controls. Half of the sheep in each group were killed at 3- and 6-month time points. After proper dissection, tendon sub-units (medial, central and lateral) were tested to failure (n = 6 for each group), whilst 2 non-dissected samples were used for histology. RESULTS: Both the implants had significantly lower stress to failure and modulus with respect to native tendon at both 3- and at 6-month time points. The implants did not statistically differ in stress to failure, whilst carbodiimide-crosslinked implants had significantly higher modulus than carbodiimide-ethyleneglycoldiglycidylether-crosslinked implants both at 3 and at 6 months. Histology showed carbodiimide-crosslinked implants to have a better integration with the native tendon than carbodiimide-ethyleneglycoldiglycidylether-crosslinked implants. Carbodiimide-crosslinked implants appeared partially resorbed and showed increased tissue ingrowth with respect to carbodiimide-ethyleneglycoldiglycidylether-crosslinked implants. CONCLUSIONS: To deliver collagen implants as an open array of fibres allows optimal tendon-implant integration and good ingrowth of regenerated tissue. In the present study the resorption rate of both the examined implants was too low due to the high level of crosslinking. This led to only minor substitution of the implant with regenerated tissue, which in turn produced a low-strength implanted region. Further studies are needed to find the right balance between strength and resorption rate of collagen fibres.