SYNOVITIS AND THE RISK OF KNEE OSTEOARTHRITIS: THE MOST STUDY

INTRODUCTION

Osteoarthritis (OA) of the knee is the most common cause of knee pain in middle-aged and older persons¹ and its prevalence is increasing².

Persons with symptomatic knee OA have much intra-articular structural pathology including cartilage loss, meniscal damage, bone marrow lesions and synovitis. Unfortunately, little treatment is available that could reverse most of these pathological joint features. And, to make it worse, in those with knee OA, these features often coexist³. Furthermore, the abnormal joint mechanics that develop as part of OA, resulting in excessive focal loads across the knee play a major role in disease progression^{4, 5}. These pathomechanics may be extremely difficult to treat or reverse, putting affected knees beyond the reach of structure modifying therapy.

Given the constellation of findings that often coexist in knees with extant disease, it makes sense to move prevention and treatment efforts in OA to an earlier stage of disease. At an earlier stage, it is more likely that focal loads across the knee joint would not be excessive, and that tissue pathology within the joint would be limited. Even though we may not yet be able to treat abnormalities within a single tissue, it would inform efforts at prevention to determine which pathologic lesions trigger the onset of disease. While cartilage loss and meniscal damage are not yet clearly treatable, especially in the face of mechanopathology, treatments targeting inflammation within the joint are available.

The effect of synovitis, inflammation of the synovial lining of the joint on incident OA is not known. Evidence on its contribution to progressive structural disease in knees that already have disease has been inconsistent and has suggested that only severe forms of synovitis increase risk^{6, 7} or that it plays no role⁸. In general, studies examining synovitis and its relation to OA have not adjusted for the structural pathology known to cause synovitis. Synovitis can be caused by release of cartilage fragments that activate synovial lining cells⁹ and can be caused by meniscal damage¹⁰. Thus, in theory, any effect of synovitis on disease occurrence could be confounded by other known causes of disease: cartilage and meniscal damage. In one recent report¹¹, synovitis was found to increase risk of incident OA, but these findings were not adjusted for the co-occurrence of cartilage and meniscal damage, and there was no association of synovitis at baseline before disease incidence, suggesting that it arose secondary to damage. In another recent report, Wang et al¹² reported that, among persons without OA, synovitis increased the risk of worsening cartilage defects but did not increase risk of quantitative cartilage loss. To our knowledge, there have been no studies examining synovitis and its relation to incident radiographic OA after adjustment for structural pathology known to cause synovitis. This issue is critical because, if there is no independent effect of synovitis on OA, then treating synovitis will have no structural benefit. Also, reports have suggested that only synovitis which appears advanced or severe on MRI¹⁰ or arthroscopy⁶ is associated with structural worsening and that there is not a ‘dose response’ relation of synovitis to OA.

The acquisition of baseline MRIs in large numbers of subjects in the Multicenter Osteoarthritis Study (MOST) who were subsequently followed for the development of OA created the opportunity to examine which structural pathology triggers disease development. Many subjects did not have OA at baseline but were at high risk of disease, and developed disease during follow-up, enabling investigators to examine MRI findings prior to disease occurrence. We took advantage of these opportunities to ask three questions: 1. Which structural features increased the risk of disease? 2. Was synovitis related to disease onset after adjusting for other risk factors thought to cause synovitis? and 3. Was more synovitis associated linearly with risk of disease or rather, was there a threshold of synovitis above which risk of disease increased?

METHODS

RESULTS

Of subjects in MOST, some were ineligible for MRI (see Figure 2), others had radiographic OA at baseline and were ineligible for this analysis of incident OA and yet others did not come for follow-up. There were 239 incident cases and 731 controls with baseline MRI data (see Table 1 and Figure 2). Baseline body mass index and female gender were increased in cases. Risk of incident OA was increased in knees with baseline KL grade 1 as opposed to grade 0 findings.

In MOST, hyaline cartilage lesions increased risk of disease. However, small lesions extending to bone (grade 2.5) conferred a substantially higher risk of disease than small lesions that did not reach bone (grade 2) (Table 2). Bone marrow lesions also conferred an increased risk of OA. Meniscal tears increased risk of disease as did meniscal extrusion. When tested as a single measure unadjusted for the co-occurrence of other MRI findings, synovitis conferred a modest increased risk of later disease (OR = 1.2 per unit increase of synovitis, 95% CI 1.1, 1.3, p<.001). The odds ratios associated with incidence did not change appreciably when we adjusted for mechanical axis.

When we then adjusted for cartilage defects, meniscal tears and extrusion using the WORMS scores for these findings, we found that synovitis was still associated with incident OA. For the synovitis score as a continuous measure, a one unit increase in the score was associated with an odds or 1.1 of OA (95% CI 1.0, 1.2, p =.02) (see Table 3). However, increasing synovitis score on a continuum was not necessarily associated with an increase in OA risk. Rather scores of 3 or more were associated with a substantial increased risk (adjusted OR 1.6, p =.003) whereas scores of 1 or 2 did not confer a significantly increased risk. Increased risk of OA was seen both for knees with synovitis in the suprapatellar pouch and for knees with synovitis in the Hoffa’s fat pad.

Of knees with baseline synovitis scores of 3 or greater, 149/195 (76.4%) had at least one subregion synovitis score of at least 2, whereas that was uncommon for knees with total synovitis scores of 2 (35/214; 16.4%). Of knees with baseline synovitis scores at Hoffa’s fat pad of 3 or greater, all had at least one subregion synovitis score of at least 2, whereas that was uncommon for knees with summary scores of 2 (37/113; 32.7%).

When we restricted our analyses to knees with baseline Kellgren and Lawrence scores of 0, our results were similar. For the synovitis score as a continuous measure, a one unit increase was associated with an odds or 1.1 of OA (95% CI 1.0, 1.2, p =.04). A synovitis score of 1 was not associated with an increase in OA risk compared with a synovitis score of 0 (adjusted OR = 0.9, 95% CI 0.7, 1.3), but synovitis scores of 2 or greater were associated with an increased risk of OA after adjustment.

DISCUSSION

Our findings strongly suggest that synovitis is an independent risk factor for new onset knee OA even after adjusting for other structural pathology that has been linked causally to synovitis. Our findings suggest further that small increases in synovitis confer little if any risk, but that risk is limited to knees with substantial increases in synovial volume.

The etiology of OA is complex. Synovitis is triggered by other pathology in the joint including not only cartilage and meniscal damage but by crystal deposition and also by ligamentous damage²¹. Bivariate analyses examining the relation of synovitis with OA could show an association even if the relation of synovitis with OA were completely confounded by factors that cause both synovitis and disease, and strong evidence already exists that cartilage and meniscal damage cause synovitis. Even so, our analysis suggests that synovitis is an independent cause of disease. While bivariate analyses suggested that even small increases in synovitis score increased risk (as shown in Table 2), the multivariable analyses (Table 3) did not confirm this; only when synovitis volume, as assessed by MRI, was increased substantially was OA risk increased.

While the relation of synovitis with OA is confounded by cartilage and meniscal damage, the same approach cannot be used to examine independent effects of cartilage and meniscal damage with incident OA. The causal path tying each these two structures with OA has both an independent line to OA and a line mediated through synovitis and therefore, the relation of cartilage and meniscal damage to OA is not ‘confounded’ by synovitis. It could be argued that synovitis could cause OA through its own effects on cartilage and that the causal model is more complex than the one we tested. However, we regard the cartilage loss effect as a proxy for incident OA, the outcome we tested.

Other studies have also suggested that more than mild synovitis is required for there to be a deleterious effect on disease progression. For example, in a longitudinal arthroscopy-based study, Ayral⁶ reported that reactive synovium (a proliferation of synovial villi of normal morphology) did not confer an increased risk of later cartilage loss whereas ‘inflammatory’ synovium (hypervascularization seen on arthroscopy) markedly increased risk. In an MRI based study from MOST without adjustment for confounding factors and focusing on change in cartilage score, Roemer et al⁷ reported that only synovitis scores of 2 or greater were associated with cartilage loss. It could be argued that we should not have tested synovitis as a ordinal measure if a threshold of synovitis was required to cause disease, but our results for mild synovitis are equivocal, with small increases in risk that may have reached significance with a bigger sample, so that our approach of testing synovitis as an ordinal measure is justifiable.

The degree of synovial thickening on MRI has been correlated especially with both macroscopic and microscopic evidence of inflammation in the synovium^{9, 22}. While non-contrast enhanced MRIs have been validated for the assessment of synovitis⁹, we are not sure how an increased score for synovitis on these images translates into enhanced pathology in the synovium. Without contrast enhancement, nodularity within the synovium which might reflect subsynovial infiltration of inflammatory cells cannot be assessed. Increased synovitis scores within the Hoffa’s fat pad represent increased signal alterations in an otherwise fatty tissue.

While our work and that of others suggests that synovitis has deleterious effects on knee structure, it is well known that synovitis increases the frequency^{9, 23} and severity⁸ of pain in persons with knee OA. This evidence, along with the current evidence on structure, suggests that synovitis constitutes an excellent target for treatment. Observational studies, even with adjustment for confounders, do not prove causal associations and trials of treatments abrogating the synovial response in OA may be necessary to prove that synovitis causes disease. Further, we note that the effect of synovitis on incidence disease, as assessed by the associated odds ratios were less for synovitis than for meniscal tears or bone marrow lesions, pointing to the complexity of OA causation and the critical element of mechanopathogenesis in disease causation.

One of the main limitations of our study was the use of non-contrast enhanced MRIs. While the optimal imaging approach would be contrast enhanced images, it was impractical and possibly unethical to perform contrast enhanced images on such a large number of subjects unaffected at the time by osteoarthritis. Further, studies²⁴ have shown a high correlation between findings on contrast enhanced and non-contrast enhanced images.

In summary, synovitis is an independent cause of incident knee OA, especially synovitis of sufficient severity to increase synovial thickness or increase signal intensity in affected areas of the knee.