Background and purpose — Patellofemoral arthroplasty (PFA) has been debated since early studies showed poor implant survival. Recent studies show better results. This review reports failure modes for PFA and investigates differences in data reported from registries and clinical studies. Additionally, we report differences in failure modes among implant designs. Methods — A systematic search was performed in September 2018. All studies and registers describing failure modes of PFA were included and implant design was noted for each revision. Results — This review includes 1,299 revisions of a primary PFA reported in 47 clinical studies and 3 registers. The failure modes were: 42% OA progression, 16% pain, 13% aseptic loosening, 12% surgical error, 4% wear, 2% infection, 2% broken patellar component, 1% stiffness, 1% fracture, and 7% other. The data from registries and cohort studies differed statistically significantly in 7 out of 12 failure modes. Significant differences were found in several failure modes among implant designs. Interpretation — OA progression is the most common failure mode of PFA. There are significant differences in data on failure modes between registers and protocolled studies, notably for surgical error. The implant design significantly influences several of the failure modes. In conclusion, indication, surgical technique, and implant design are important for a successful PFA, and register-based failure modes should be interpreted with caution.