The results of our study indicate that changes in severity of BMLs and synovitis were associated with the fluctuation of frequent knee pain and pain severity, and of the two BMLs was the better predictor for fluctuation of knee pain than synovitis. Furthermore, improvement of BMLs was associated with a decreased risk of frequent knee pain and of less severe pain, whereas worsening of synovitis and effusion was associated with an increased risk of frequent knee pain and of more severe pain.
Previous studies have focused on risk factors for radiographic knee OA24
, and few have attempted to study risk factors for pain from knee OA; thus, factors associated with symptomatic knee OA are largely unknown. Clinically, physicians often notice that patients with knee OA experience pain exacerbation over the course of the disease25
. Such pain patterns were also observed in the Framingham OA study; roughly a third of participants with symptomatic knee OA had improvement of their symptoms over time26
. Recently, two studies3, 27
reported that joint pain among persons with knee or hip OA is often intermittent and variable, and adaptation and avoidance strategies modify the experience of pain. All these findings indicate that pain from knee OA often fluctuates and suggest that factors that vary over time may play roles in pain fluctuation.
To our knowledge, only a few studies have been conducted to evaluate the association between structural lesions detected by MRIs and the presence of pain or on pain severity from knee OA4-13
; results, however, are inconsistent. In a case-control study, the proportion of BML scores worsening was higher in the knees with onset of frequent pain (49.1%) than in knees without frequent pain (26.8%), and two or more units increase in BMLs score were associated with a 3-fold increased risk of frequent pain12
. In a longitudinal study Hill et al. found that, compared with the baseline visit, about 15-20% of knees' synovitis scores improved and 20% deteriorated at a 30-month follow-up visit; changes in the synovitis score were modestly associated with the change in pain severity, with a one unit increase in summary score of synovitis resulting in a 3.2 mm increase in VAS pain score (0-100 scale)11
. However, this study did not examine whether reduction or complete resolution of synovitis was associated with a decreased risk of knee pain or pain severity. Our study demonstrated that improvement of BMLs, but not improvement of synovitis or effusions, was associated with a decreased risk of frequent knee pain and pain severity. Given the vast number of Americans with knee OA and the dearth of effective treatment for pain among those suffering with the disease, a better understanding of risk factors for pain is sorely needed. If our findings are confirmed by future studies and these structural lesions do indeed contribute to the occurrence of frequent knee pain and to pain severity, treatments and therapies that target these specific features can be developed.
The use of traditional epidemiologic study designs to assess the dynamic relation of structural changes to the risk of knee pain and pain severity poses significant methodological challenges. Pain is a subjective experience. Various risk factors, such as genetic predisposition28
, prior experience29
, personal perception30
, and social-cultural environment32
, play important roles in a subject's response to painful stimuli. Unless all of these risk factors are measured and adjusted for, studies aiming to evaluate the effect of pathological lesions in or around the knee on the risk of knee pain or on pain severity are susceptible to confounding bias. Second, the high cost of reading knee MRIs makes it expensive to examine the relation of changes in the structural lesions detected by MRIs to the fluctuation of knee pain in a large scale of longitudinal study.
We used the self-matched case-control study design15, 33
, akin to the case-crossover study34
, to evaluate the association between three structural lesions and fluctuation of knee pain. Since each knee serves as its own control, confounders, either known or unknown, that do not vary over time within a knee but that may differ among persons are eliminated. The self-matched case-control study design allows us to examine knee-specific rather than population-average effect of change in reversible structural lesions over time on the fluctuation of frequent knee pain or pain severity; thus it would allow us to have more confidence in making causal inference. Our findings, if confirmed by other studies, will be an incentive to invest in treatments that target these structural abnormalities.
While cartilage is aneural, we postulate that cartilage loss could contribute to pain indirectly. For example, cartilage debris could be ingested by synovial cells, inducing synovitis35, 36
. Further, since an effusion must arise from the synovium, cartilage loss may indirectly be related even to effusion. A few studies have found that cartilage defects are associated with knee pain8, 10
. However, when we controlled for cartilage morphology score, the association between BMLs, synovitis and effusion with knee pain did not change materially. In other work, we will explore the relation of cartilage loss with pain but note here that synovitis would be a mediating factor in tying cartilage loss to pain. Thus, one could not validly examine the role of cartilage loss with pain while adjusting for synovitis. We did not find that psychological factors, indicated by mental health component score from the SF-12, were associated with presence of structural lesions; however, psychological factors may affect pain reporting. Nevertheless adjustment for mental health score did not alter the effect of BMLs, synovitis and effusion on knee pain. Interestingly, when adding knee injury into the multivariable regression model, the effect of synovitis and effusion was attenuated, suggesting that knee injury may cause the occurrence of synovitis or effusion of the knee.
Our study has some limitations. First, the weighted kappa statistic of the measurement of each of the three structural lesions was slightly above 0.60; thus our assessment of these lesions may be susceptible to misclassification. However, the readers were unaware of the pain status of the knee when scoring structural lesions; thus the true effect sizes associated with these structural lesions are likely to be attenuated owing to non-differential misclassification bias. Second, we used non-contrast MRIs to assess severity of synovitis. However, as of today large epidemiological studies have mostly not used contrast for assessing synovitis in view of potential toxicity involved with the intravenous administration of contrast agents and associated costs. In addition, some studies have shown that synovitis assessed on non-contrast MRIs is associated with knee pain11
. It is beyond the scope of our study to determine what caused the occurrence and resolution of BMLs, synovitis and effusion. Nevertheless, future studies aimed at understanding the causes of these reversible lesions will guide investigators to search for more rational targeted therapies for knee pain. Third, other potential risk factors for knee pain, such as physical activity, muscle weakness, and weather-related factors, were not collected at each clinic visits. Finally, we used R2
value, albeit small, to rank the importance of each structural lesion on frequent knee pain severity. However, it is well known that R2
generated from the logistic regression model is of marginal use, and such measure should not be interpreted as actual percent of variance of variable explained.
In conclusion, our study showed that changes in scores of BMLs and synovitis were associated with the fluctuation of frequent knee pain and pain severity and the effect of BMLs was greater than synovitis. Improvement of BMLs over time was associated with concomitant reduction in pain presence and pain severity, whereas worsening of synovitis and effusion over time was associated with an increase in knee pain presence and severity. These findings have implications for treatment and prevention in symptoms of knee OA.