The study was performed to assess the factors affecting the effectiveness of autologous adipose-derived mesenchymal stem cell (ADMSC) therapy for the treatment of knee osteoarthritis. Included were patients not responding to nonoperative treatment for ≥3 months with Kellgren and Lawrence grade 3. Patients received an intra-articular injection of either autologous serum or ADMSCs. The effectiveness was determined by both improvements in the scores of Western Ontario and McMaster Universities Osteoarthritis index (WOMAC) and 100-mm visual analogue scale (VAS) at 0.5, 1, 2 and 3 years post dose using 2-sample t-test and chi-square test. Included were 125 patients (50 non-obese, 75 obese) and 127 patients (47 non-obese, 80 obese) in ADMSC and serum groups, respectively. ADMSC showed better improvements in WOMAC and VAS after 6 months compared with serum group (WOMAC, -8.2, p = 0.001; VAS, -9.5, p = 0.002) attributed not to obese (-4.6, p = 0.13; -4.2, p = 0.27) but non-obese patients (-13.7, p = 0.002; -17.5, p = 0.001). Within ADMSC group, the improvement in WOMAC was also greater in non-obese (-27.5, -31.0, -32.4, -33.1) compared with obese patients (-19.4, -19.3, -20.0, -19.9, at 0.5, 1, 2, 3 years, respectively). Findings were similar in the improvement of VAS (non-obese, -32.5, -36.1, -37.2, -37.4; obese, -20.6, -22.7, -20.0, -22.7). Such results indicate that the effectiveness of ADMSCs was more remarkable in non-obese than obese patients both in a short- and long-term.