BACKGROUND: We designed a prospective, single-center study to assess whether blood metal ion levels could predict implant failure in patients managed with unilateral metal-on-metal hip resurfacing or total hip arthroplasty. METHODS: Five hundred and ninety-seven patients who had received unilateral Articular Surface Replacement prostheses at least twelve months earlier were recruited. Blood metal ion levels were compared between the group of patients with failed implants and the group with non-failed implants. Implant failure was defined as prostheses associated with revision, an intention to revise, or poor patient-reported hip function (Oxford Hip Score, <31 of 48). Specificity, sensitivity, area under the curve, positive and negative predictive values, and odds ratios were calculated. Logistic regression analysis was used to identify other risk factors for implant failure. RESULTS: Patients with failed arthroplasty had significantly higher blood cobalt and chromium ion levels than did patients with non-failed arthroplasty (p < 0.01). Blood cobalt ion levels were disproportionately raised in patients with failed total hip arthroplasty (8.2 μg/L) compared with patients with failed hip resurfacing (2.5 μg/L) (p = 0.018). Blood chromium ion levels were not significantly different in patients with failed total hip arthroplasty and failed hip resurfacing (p = 0.058). The maximum value of either metal ion had good discriminant ability to predict implant failure (area under the curve, 0.76). A 7-μg/L cutoff had a positive predictive value of 0.75 (95% confidence interval, 0.66 to 0.82) and a negative predictive value of 0.82 (95% confidence interval, 0.78 to 0.86). In patients managed with total hip arthroplasty, for each increase of 1 μg/L there was a 23% (p < 0.001) increase in the odds of them being in the failed group. For patients managed with hip resurfacing, the increase in odds was 5% (p < 0.001). CONCLUSIONS: Raised levels of blood metal ions were associated with failed metal-on-metal hip resurfacings and total hip arthroplasties. A threshold level of 7 μg/L had inadequate sensitivity to be used in isolation as a screening test for implant failure, but it provided nearly optimal misclassification rates. No level had a perfect positive predictive value, and so we discourage surgeons from performing revision surgery based on blood metal ion levels alone. Levels of cobalt ions were raised out of proportion to levels of chromium ions in failed total hip arthroplasty and may reflect a different mechanism for metal ion generation. LEVEL OF EVIDENCE: Therapeutic Level III. See Instructions for Authors for a complete description of levels of evidence.