The present study indicates that being knee-symptomatic, middle-aged, and overweight may predispose for variation in pain and function over two years. These factors are therefore relevant to take into account when deciding patient inclusion and exclusion criteria, outcome measures, and the number needed in clinical trials in subjects with early-stage knee OA.
The Bristol 'OA 500' study is one of few studies on the natural variation of symptoms in knee OA [14
]. In this cohort, several baseline variables including age, gender, and BMI were analyzed as possible predictors of change in a combined index including change in pain, change in index joints, and global change. Neither age, gender, nor BMI were shown to influence the variation over eight years [14
]. In comparison with our study subjects, the Bristol 'OA 500' cohort was on average 10 years older, held more women (69 vs. 16%), and included subjects with OA not only of the knee but also of the hip and hand. The mean BMI or presence of previous knee injuries was not reported.
OA is usually studied in the elderly. However, it is well recognized that OA may develop during middle-age or earlier [15
]. The age range of 27 to 83 for our cohort allowed us to study the variation in symptoms in different age categories. We found an increased variability in knee-related function in patients aged 46–56, compared with the younger age group. Younger subjects scored well at entry, and in general their outcome remained good over time. These age-related findings may reflect the developmental phases of OA and correspond to an increase in the report of knee disability seen in the population during middle-age [16
]. With increased age there may also be large changes in lifestyle due to early retirement or other alterations of the psychosocial situation that may affect self-reported symptoms and knee function. The oldest subgroup (aged >56) seem to form a more stable group than the aged 46–56, but low subject numbers limit interpretation of the results.
In analytic models unadjusted for baseline, overweight or obesity was a predictor of clinically relevant variation in outcome over two years. However, when adjusting for baseline scores, these associations did not remain significant. This is likely an effect of the strong association between high BMI and being symptomatic. In a closely related study we found no evidence that subjects had become sedentary due to symptoms and then obese (as cause and effect) [3
Having a symptomatic knee, in comparison with having a non-symptomatic knee, was strongly associated with large variability over two years. We cannot exclude the contribution of a ceiling effect, i.e., an individual with KOOS Pain ≥ 91 cannot improve by 10 points. There was no floor effect possible, as no study subjects had a score of 10 or worse. The finding of symptoms being a strong predictor of change has less importance for future design in OA trials since a common inclusion criterion is at least moderate pain and functional limitations to be able to detect a clinically relevant improvement from the intervention applied.
We found no significant influence of radiographic status on variation in pain and function in this cohort, consistent with the well-known discordance between radiographic status and pain in population-based studies of OA [18
We arbitrarily applied the same cut-off for clinically relevant change for the generic SF-36 as for the knee-specific measure KOOS. This serves as an important limitation in interpreting the SF-36 results. The pain and function subscales of the KOOS and SF-36 measures hold different number of items, and the different items may have different numbers of response options. A change in response of one step, e.g. from mild to moderate, will have greater impact on the final score if the scale has fewer items, or fewer response options. For the KOOS subscale Pain, a one step change in response option will result in a score change of 2.8 points. For the SF-36 subscale Bodily Pain the corresponding change in response would yield a score change of 6–20 points (depending on the item/response option changed), explaining the larger proportion of patients changing on the SF-36, compared with the KOOS.
Other limitations of our study include low subject numbers and not taking into consideration several factors in that may be important to the outcome studied, most importantly medication, co-morbidities, educational level, coping, and mood status. Also, the use of self-estimates of weight and length from the second assessment and the low proportion of women serve as limitations. Meniscectomy is more frequently performed in men [20
], the reasons for which are not clear. We were unable to evaluate the possible influence of "regression to the mean" phenomenon as we only had data from two time points. Strengths of our study include the use of validated self-reported questionnaires which help avoid investigator bias, and the high follow-up rate of 73%.