Background: Blood derivative injections have been recently proposed to address osteoarthritis (OA) with overall positive results, although long-term data on their efficacy are lacking. A novel blood derivative has been developed to concentrate growth factors and antagonists of inflammatory cytokines and shown promising early findings. Purpose: To investigate if the positive effects of a single intra-articular injection of autologous protein solution (APS) in patients affected by knee OA—previously documented at 1 year in a multicenter double-blind randomized saline-controlled trial—last up to 3 years. Study Design: Case series; Level of evidence, 4. Methods: A total of 46 patients with Kellgren-Lawrence 2 or 3 knee OA were randomized into 2 groups: 1 ultrasound-guided APS injection (n = 31) or 1 saline injection (n = 15). At 1 year, the saline group was allowed to cross over. Patients were re-evaluated at 24 and 36 months through the visual analog scale for pain (VAS), Western Ontario and McMaster Universities Osteoarthritis Index Likert 3.1 (WOMAC LK 3.1), Knee injury and Osteoarthritis Outcome Score (KOOS), 36-Item Short Form Health Survey (SF-36), and Outcome Measures in Rheumatology–Osteoarthritis Research Society International (OMERACT-OARSI) responder rate. Magnetic resonance imaging evaluation was performed with the MRI Osteoarthritis Knee Score (MOAKS) before and at 24 months after treatment, and radiographs were assessed per Kellgren-Lawrence before and annually after treatment. Results: In the APS cohort, WOMAC pain improved from 11.5 ± 2.4 (mean ± SD) to 4.3 ± 4.0 at 1 year and to 5.7 ± 5.0 at 3 years ( P < .0001 vs baseline). The APS cohort also showed a statistically significant improvement in its KOOS pain score from 39.4 ± 13.1 to 70.6 ± 21.5 at 1 year and to 64.1 ± 24.6 at 3 years ( P < .0001 vs baseline) and VAS pain scores from 5.5 ± 2.2 to 2.6 ± 2.5 at 1 year and to 3.4 ± 2.9 at 3 years ( P = .0184 vs baseline). VAS pain score significantly worsened from 12 to 36 months ( P = .0411). All patients in the saline group decided to cross over to APS, and their final scores were better than baseline, although not significantly better than at the crossover point. Overall, 7 of 26 (26.9%) APS cases and 4 of 14 (28.6%) crossover cases were considered failures as patients underwent further injective treatments or surgical procedures between the 12- and 36-month follow-up. MOAKS findings showed no statistically significant differences. Patients with better cartilage had greater WOMAC pain improvement when their baseline scores were worse, whereas the trend was reversed for patients with cartilage loss at baseline. Conclusion: Intra-articular use of APS for mild to moderate knee OA was safe, and significant pain improvement was documented 3 years after a single injection. Patients with better cartilage status seem to respond better than patients with more cartilage loss, with more clinical improvement even when starting from more painful conditions. Registration: NCT02138890 (ClinicalTrials.gov identifier).