Objectives To determine whether greater effusion-synovitis volume and infrapatellar fat pad (IFP) signal intensity alteration differentiate incident accelerated knee OA (KOA) from a gradual onset of KOA or no KOA. Methods We classified three sex-matched groups of participants in the Osteoarthritis Initiative who had a knee with no radiographic KOA at baseline (recruited 2004–06; Kellgren–Lawrence <2; n = 125/group): accelerated KOA: ⩾1 knee progressed to Kellgren–Lawrence grade ⩾3 within 48 months; common KOA: ⩾1 knee increased in radiographic scoring within 48 months; and no KOA: both knees had the same Kellgren–Lawrence grade at baseline and 48 months. The observation period included up to 2 years before and after when the group criteria were met. Two musculoskeletal radiologists reported presence of IFP signal intensity alteration and independent readers used a semi-automated method to segment effusion-synovitis volume. We used generalized linear mixed models with group and time as independent variables, as well as testing a group-by-time interaction. Results Starting at 2 years before disease onset, adults who developed accelerated KOA had greater effusion-synovitis volume than their peers (accelerated KOA: 11.94 ± 0.90 cm3, KOA: 8.29 ± 1.19 cm3, no KOA: 8.14 ± 0.90 cm3) and have greater odds of having IFP signal intensity alteration than those with no KOA (odds ratio = 2.07, 95% CI = 1.14–3.78). Starting at 1 year prior to disease onset, those with accelerated KOA have greater than twice the odds of having IFP signal intensity alteration than those with common KOA. Conclusion People with IFP signal intensity alteration and/or greater effusion-synovitis volume in the absence of radiographic KOA may be at high risk for accelerated KOA, which may be characterized by local inflammation.