This examination of the literature has revealed a wide variation in the degree to which knee pain relates to radiographic knee osteoarthritis and vice versa. We postulated that there might be three particular reasons as to why discordance between x rays and symptoms might arise, from which we can now draw three main conclusions.
Firstly there may be insufficient x ray numbers or views used to estimate the association. The studies show that the prevalence of radiographic knee OA will be underestimated in persons with knee pain in studies that do not obtain all potential x ray views of the knee. This is supported by the finding that knee studies including x rays of the patello-femoral joint (PFJ), improve the sensitivity with which symptoms such as pain can identify radiographic knee OA [8
]. By adding a lateral or skyline to the A/P view, overall prevalence of radiographic knee OA in pain positive persons increases to 80% [8
]. A recent paper from our group, subsequent to this review, has confirmed this conclusion by showing directly that the prevalence of overall radiographic OA of the knee increases with the number of radiographic views in a population with knee pain [16
]. However, much discordance remains between pain and x ray findings, and no combination of views reaches a point where patients with knee pain invariably have radiographic knee OA. This is also true for studies examining the prevalence of pain in populations with radiographic knee OA. There is a great deal of discordance evident amongst these studies as highlighted in Table . Overall these studies support the conclusion that the lack of association between radiographic knee OA and pain is to some extent a real one.
Secondly, the way pain is defined (e.g. whether disability is included or not) and the grading of radiographic severity, have important influences upon estimates of association between knee pain and radiographic OA and vice versa. Table and table examined this relationship with respect to pain definition and demonstrate the wide variation in pain definitions used, and the correspondingly wide variations in the associations between knee pain and x ray findings. It seems likely that the often observed discrepancy between pain and radiographic knee OA has something to do with this variation in definition of pain, and that, if similar methods of pain definition were used, some consistency in the level of discrepancy might emerge. However, the variation between studies is quite marked, so one cannot be wholly convinced of the idea that using one standard uniform definition will lead to x rays and pain becoming more concordant. Other reasons might play their part here. Figure shows the sources of chronic knee pain in the older person that as a whole make up the knee 'pain picture' we encounter in general practice. Pain in the knee is more than just the result of the pathological changes reflected in the x ray. Other factors may account for knee pain which will not be evident on the knee x ray. Figure clearly shows this, indicating that the pain may be the result of other bone problems, not visible on an x ray such as oedema, or non-OA conditions such as ligament injury or tendonitis. Indeed, some chronic knee pain might be more strongly linked to issues of cognitive or emotional state such as depression rather than local pathology at the knee joint. Of course, all these things can coexist at the same time, making up multiple layers of causality of knee pain
The complementary problem concerns the variation in definitions of radiographic OA in any particular view. Some would argue that an isolated osteophyte is not osteoarthritis, although whether the mildest form of osteophyte is included in the definition of OA or not seems to make little difference to the association with pain. However what is clearer from the papers we reviewed is that, with respect to the x ray grade, at the severe end of the spectrum there is a closer association of pain and x rays as shown in table and table , but milder disease is more common and the discordance evident at lower levels of K/L grade is important to consider in studies of knee pain and OA. The way the x ray is taken is also important. Between studies the radiographic technique employed may have differed. This will have encompassed whole protocols which might involve the position of the knee (semi-flexed or straight knee). In addition reading the radiographs requires consistency. The studies described go to great lengths to attain intra-study consistency, but we are unable to comment on inter-study consistency and this must be taken into account when evaluating the findings between studies.
Thirdly, the nature of the study population is important since variations in the association of knee pain and radiographic knee OA may be influenced by characteristics of the population sampled. Younger age groups with knee pain are less likely to have radiographic knee OA (table ), and there is also some variation with age in the proportion of persons with radiographic knee OA and one study suggests that younger patients with radiographic knee OA are less likely to be symptomatic [10
]. Ethnicity also has some influence over the relationship [13
]. Study populations are of course more diverse than in age, gender and ethnicity alone, and it may be that other characteristics than these may both influence the link between x rays and pain, and vary between the populations studied. We did not investigate the effect of other characteristics in this study.
The major issue for future research is that commitment to more uniformity and standardisation in definitions is needed to allow comparability between studies, and to remove variability between studies as a factor obscuring accurate estimates of the 'true' association between x rays and symptoms at the knee. This would almost certainly involve x raying multiple views of the knee, in a standardised way using consistent protocols across research groups. Pain analysis needs to be similarly standardised, and as recently used in one paper [29
], the WOMAC scale allows detailed analysis of pain and dysfunction. Pain grading is essential and might be achieved through using the von Korff Chronic Pain Grade to allow combined measurement of pain and disability severity [30
]. Finally using a sampling frame that identified people with a wide range of severity and duration of knee pain, and unselected for their use of healthcare, would deliver a population that truly would be free of selection bias and comparable across study groups.
We conclude, inevitably, that knee pain is an imprecise marker of radiographic knee osteoarthritis, even in older age groups, but the extent of this imprecision depends heavily on the extent of radiographic views of the joint obtained. Radiographic knee osteoarthritis is likewise an imprecise guide to the likelihood that knee pain or disability will be present, although the more severe the radiographic osteoarthritis, the more likely there are to be accompanying symptoms. Both associations are affected by the definition of pain used and the nature of the study group. The experience of pain is multi-factorial in its origin, and factors such as patient depression play an important part in its manifestation, and this is as true of osteoarthritis and joint pain in older people as it is for pain of uncertain pathology in younger people [31
]. Using x rays as a means for investigating knee pain, particularly in older people, requires these other factors to be taken into consideration, and the results of knee radiographs should not be used in isolation when assessing individual patients with knee pain.