Purpose.Recent advances aimed at preventing and treating complications after keratoprosthesis surgery have improved prognosis, but it has been suspected that various preoperative diagnoses may carry substantially different postoperative outcomes. This article attempts to clarify the ranking of prognostic categories for patients undergoing keratoprosthesis surgery.Methods.A retrospective review of the outcome in a recent series of 63 patient eyes operated at the Massachusetts Eye and Ear Infirmary between 1990 and 1997 and followed up for a minimum of 21 months. Anatomic retention of the device and the loss of vision caused by complications were recorded. The patients were divided into four categories according to preoperative cause.Results.Anatomically, one keratoprosthesis extruded spontaneously. Another 10 were permanently removed because of complications. Of the 63 eyes, 10 never achieved a visual acuity of at least 20/200 vision because of preexisting retinal or optic nerve damage. The remaining 53 had a visual acuity of 20/200 to 20/20 as follows: Stevens–Johnson syndrome (n= 7), after 2 years: 33%, after 5 years: 0%; chemical burn (n= 17), after 2 years: 64%, after 5 years: 25%; ocular cicatricial pemphigoid (n= 20), after 2 years: 72%, after 5 years: 43%; graft failure in noncicatrizing conditions (dystrophies, degenerations, or bacterial or viral infections) when a repeat graft was expected to have a poor prognosis (n= 19), after 2 years: 83%, after 5 years: 68%. The difference in outcome between the Stevens–Johnson syndrome outcome group and the graft failure group or the ocular cicatricial pemphigoid group was statistically significant. In the group of 53 eyes, visual acuity was restored to 20/200 to 20/20 for a cumulative total of 138 years.Conclusion.Outcome of the keratoprosthesis surgery varied markedly with preoperative diagnosis. Most favorable was graft failures in noncicatrizing conditions, whereas Stevens–Johnson syndrome was the worst. Ocular cicatricial pemphigoid and chemical burns occupied a middle ground. The difference between the groups seemed to correlate with the degree of past preoperative inflammation.