Clinical features of symptomatic patellofemoral joint osteoarthritis

Osteoarthritis (OA) is not a single disease [1] and distinct phenotypes are believed to exist even within a single joint complex like the knee. Among the various approaches to subclassifying knee OA, the recent European League Against Rheumatism (EULAR) Task Force on diagnosis of knee OA recognised that subsets with different risk factors and outcomes can be defined by compartmental distribution, but pointed out that the ability to discriminate between these subsets in routine practice and the utility of doing so had not been formally tested [2]. Perhaps unsurprisingly, the diagnosis of knee OA subsets is rarely seen in current primary care. For example, in a total population of 57,555 adults registered with UK general practices, only 13 cases had a recorded diagnosis by the general practitioner of patellofemoral joint OA; less than 1% of knee consulters in a year [3].

There are several reasons why distinguishing patellofemoral from tibiofemoral joint OA phenotypes may be important. There is growing evidence indicating that patellofemoral joint OA impacts independently on symptoms and function [4‐9], that it also frequently occurs in the absence of tibiofemoral disease [4, 6, 10‐13], and that its aetiology and, therefore, risk profile and management, may differ [12, 14‐17]. For example, a history of knee injury or meniscectomy may tend to indicate tibiofemoral joint OA [14, 15] while a history of anterior knee pain in young adulthood may suggest patellofemoral joint OA [18]. The direction of frontal plane knee malalignment may serve to indicate patellofemoral joint OA and tibiofemoral joint OA (valgus malalignment being associated with the predominant pattern of lateral patellofemoral joint OA, varus malalignment with medial tibiofemoral joint OA) [16, 19‐21]. Although a recent systematic review revealed only two randomised controlled trials of interventions specifically for isolated patellofemoral joint OA [22], more are now emerging [23, 24]. In the context of recommendations that OA can often be confidently diagnosed without the need for imaging [2, 25], these developments pose a fundamental question: can patellofemoral joint OA be identified in routine clinical practice and, if so, which features are most informative?

Expert clinical accounts of the clinical manifestations and typical features of patellofemoral OA are available in medical textbooks and review articles [17, 26‐31] but there has been very little empirical research. One exception, a hospital-based case-control study, documented the comparative clinical features of patellofemoral and tibiofemoral joint OA in only 42 knees [32]. Previous research on the clinical features, classification criteria and diagnosis of knee OA, including that for the American College of Rheumatology (ACR) classification criteria and EULAR Task Force, has tended to be based on knee OA as a whole [33‐38] and there is some evidence that the features derived from these studies may selectively reflect tibiofemoral disease [39]. A recent editorial concluded that "little is known about how best to separate patellofemoral symptoms from those arising from the tibiofemoral joint" [40].

In this paper we set out to extend our previous work [37] on pursuing rational clinical diagnosis of knee OA in primary care. We investigate the comparative clinical features of symptomatic patellofemoral and tibiofemoral joint OA and we explore their ability, when used in combination, to allow confident diagnosis of subsets of symptomatic knee OA in the community setting.

While there is little to distinguish mild isolated patellofemoral joint OA from simple knee pain, moderate to severe isolated patellofemoral joint OA is indicated by a history of dramatic swelling in the past, valgus malalignment, markedly reduced quadriceps strength and pain on patellofemoral joint compression. Tibiofemoral joint involvement is indicated by previous injury, varus malalignment, bony enlargement, reduced knee flexion range of motion and fixed flexion deformity. However, in the community setting, confident clinical diagnosis of any subset of radiographic knee OA will often not be possible.

Using comprehensive plain radiographic views - the reference standard recommended by the EULAR Task Force [2] - the current study applied two different thresholds for defining knee OA subsets. We considered a wide range of potential indicators derived from a review of previous literature and consensus development with clinicians and gathered by trained assessors using simple, practicable techniques according to standardised protocols.

Our findings on the pattern of associations between individual risk factors and different subsets of knee OA are largely consistent with those identified in previous longitudinal studies on patellomoral and tibiofemoral joint OA. Age and BMI are confirmed as strong indicators of knee OA but, as McAlindon et al. [15] observed, are equally important to patellofemoral and tibiofemoral joint disease subsets. Malalignment is a strong indicator of moderate to severe knee OA subsets with varus malalignment indicating isolated tibiofemoral joint OA and valgus malalignment indicating isolated patellofemoral joint OA. Due to limited numbers of participants, we did not separately define medial and lateral compartment disease for either tibiofemoral or patellofemoral joint OA. Among participants with moderate to severe isolated patellofemoral joint OA in the current study, the ratio of lateral to medial compartment involvement was greater than 2:1. This tendency towards lateral patellofemoral joint disease and its association with valgus malalignment is consistent with previous work [16, 19‐21, 32]. In those with isolated tibiofemoral joint OA, the ratio of medial to lateral compartment disease was greater than 4:1 based on joint space narrowing from the PA view. Given this predominance of medial tibiofemoral joint disease, the association with varus malalignment is consistent with the role of malalignment in the progression of tibiofemoral joint OA [53]. When we narrowed the definition of isolated patellofemoral joint OA to lateral compartment disease only and isolated tibiofemoral joint OA to medial compartment disease only (based on grade 2 to 3 joint space narrowing) the relationship with malalignment became even stronger (see Additional file 5). What our study adds is that the association between malalignment and subsets of knee OA is still detectable even by crude clinical assessment (a gap between the knees or ankles when instructed to stand with the feet together).

Our findings regarding the clinical manifestations of patellofemoral and tibiofemoral joint OA confirm many of those found in previous studies of undifferentiated knee OA. Functional limitation, bony enlargement, coarse crepitus, fixed flexion deformity and reduced flexion range of motion appear to be relatively robust indicators of knee OA [2]. Palpable effusion showed a strong and consistent association with OA in our study, particularly for tibiofemoral joint OA; a finding that contrasts with the EULAR Task Force summary based on two studies [32, 34] but which is in agreement with a recent Canadian study [38]. In addition, we found a history of previous dramatic swelling ("came up like a balloon"), markedly reduced quadriceps strength and pain on patellofemoral joint compression were informative indicators of moderate to severe isolated patellofemoral joint OA. Our finding that quadriceps weakness is selectively a feature of more advanced patellofemoral joint OA and not tibiofemoral joint OA appears to support similar recent findings in cross-sectional [54, 55] and longitudinal analyses [56] although the markedly reduced performance on isometric testing found as the distinctive feature in the current study may be indicative of painful/fearful inhibition more than weakness per se. The finding of a substantially increased risk of dramatic swelling was unanticipated. Without further information we can only speculate on whether this is related to the sorts of previous episodes of subluxation/dislocation reported in hospital cases by Iwano [32].

Our study has several limitations. Plain radiography captures a relatively limited and late view of OA pathology [1]. For that reason, there is the potential for misclassification (specifically due to pre-radiographic tibiofemoral disease) and for the misattribution of clinical features to isolated patellofemoral joint OA [38]. It remains possible, for example, that a recent recalled episode of dramatic swelling, quadriceps weakness or inhibition and reduced knee flexion may be signs not of isolated patellofemoral joint OA but of early tibiofemoral joint OA. While a more sensitive imaging modality would be able to detect this, there is still a fundamental issue that cross-sectional diagnostic studies provide only a snapshot on current status, and in knee OA this is in the context of an evolving sequence (or multiple sequences) of disease [57]. In spite of quality assurance and control procedures, the reliability of the assessment of some clinical signs and symptoms was still poor and would be expected to result in an underestimation of their informativeness. Nevertheless, we feel that this provides a reasonable reflection of what might be expected from non-specialist assessment. Whilst extensive, our list of potential diagnostic indicators was not comprehensive - tenderness on palpation of the patella facets [17] and abnormal gait [38] are two clinical features in particular that might be worth including in future investigations. We initially attempted to image the PF joint in a weight-bearing position according to the Buckland-Wright protocol but this resulted in poor quality films which failed to demonstrate the joint space well in a significant minority of participants who had difficulty adopting the weight-bearing position. Our imaging of the PF joint in the supine position excludes the impact of muscle forces on the joint space width, which may miss minor joint space narrowing. However, this would not affect the classification of 'any' PF OA which relies on the presence of osteophytes. Furthermore, our analysis of "moderate/severe OA" required moderate or worse narrowing and we feel that this is unlikely to have been missed even when the PF joint was imaged in the supine position. Finally, with regard to our multivariable analyses, it should be pointed out that these were based on a high number of variables to cases and with considerable univariable analysis and variable reduction. They require external validation in separate samples. Nonetheless, they do suggest marginally informative clinical features that can be used as the basic building blocks for clinical diagnosis. Our analyses also highlight the oft-neglected issue that even given several 'statistically significant' associations and 'substantial' areas under the ROC curve, one cannot assume that this will translate into the correct classification of a very high proportion of patients or into a confident diagnosis in the majority of cases [58, 59]. The best that may be currently achieved by the generalist in routine practice in the absence of definitive imaging is a knowledge of the likely pattern of knee OA based on a 'balance of probabilities'.

In the case of moderate to severe disease, the clinical profile of symptomatic, radiographically-confirmed patellofemoral joint OA is distinct from tibiofemoral joint OA. However, in the community setting, a confident diagnosis will seldom be possible without imaging. Most signs and symptoms of knee OA reported in the medical literature are predominantly indicators of tibiofemoral joint OA. Selectively effective non-surgical treatments for patello-femoral joint OA are unlikely to be able to be appropriately targeted on clinical grounds alone to the majority of patients presenting with isolated patellofemoral joint OA to primary care.