ObjectiveTo investigate the prognostic value of illness perception (IP) on knee pain, quality of life (QoL) and functional level in elderly individuals reporting knee pain. DesignA prospective cohort study of 1552 elderly with knee pain comparing two previously established clusters based on the Brief Illness Perception questionnaire. Cluster 1 (“Concerned optimists” [hypothesized unfavorable profile]; n = 642) perceived their knee pain as a greater threat to them than Cluster 2 (“Unconcerned confident” [hypothesized favorable profile]; n = 910). Primary outcome was the change from baseline to year 2 in the KOOS Pain subscale. Secondary outcomes were changes from baseline in quality of life (EuroQol-5 Domain and EQ VAS) and in the KOOS subscales Symptom, Activities of Daily Living, Knee-related QoL and Sports and recreation. Analyses were done on the original Intention-To-Survey (ITS) population, using repeated measures mixed linear models. ResultsAmong the ITS population, 841 (54%) responded to the 2-year survey. There was a statistically significant but clinically irrelevant cluster difference in the 2-year change from baseline in KOOS pain (mean difference: 6.0 KOOS points [95% CI: 7.3 to −4.7]) explained by a minor improvement in Cluster 1: (6.2 points) and no changes in Cluster 2: (0.2 points). Comparable results were found across the secondary outcomes. Clinically irrelevant cluster changes in IP were seen. ConclusionIn a cohort of people with knee pain, IP phenotype (i.e., Clusters) were of no prognostic value for the 2-year changes in pain, function, and QoL. Targeting IP may not be relevant in this patient population. Trial registration number and date of registrationThe Frederiksberg Cohort study was pre-registered at clinicaltrials.gov (NCT03472300) on March 21, 2018.