Osteoarthritis (OA) is the most common cause of arthritis in the world, and typically presents with joint pain, stiffness and loss of function. OA is associated with a huge burden on society, in terms of both loss of working days and the cost of treatment itself.1 The prevalence of OA is predicted to rise rapidly owing to a combination of an ageing population and the obesity epidemic. The burden of OA in terms of reduced life expectancy is also being recognized, and this further highlights the more malignant nature of the condition.2 Despite the size of the global impact of OA, there is a significant unmet need for effective treatments.
Traditionally, OA has been defined by evidence of structural damage seen radiographically. The heterogeneity of the condition, in terms of aetiology, disease progression, the generation of symptoms and response to treatment, is now being further explored. It is anticipated that a better understanding of the different clinically important subsets of patients will help to identify new targeted therapy as well as optimize the application of currently available treatments.
The aim of this narrative review is to discuss some of the current conundrums in the OA field and how they interplay with the current assessment and treatment of OA and its future development.
Definition of osteoarthritis
Osteoarthritis can be defined based on the presence of characteristic joint symptoms such as pain and stiffness, structural abnormalities, or a combination of both.
Structural changes have been traditionally assessed using radiographs, typical features including joint space narrowing, subchondral sclerosis, bone cysts and osteophyte formation. The most commonly used radiographic measure of OA is the Empire Rheumatism Council system, developed by Kellgren and Lawrence in 1957 [the Kellgren–Lawrence (KL) grading scale].3 This method assumes the sequential development of osteophytes, joint space narrowing and subchondral sclerosis. A radiographic atlas is used to standardize the grading of individual joints from 0 to 4. Studies will typically use grade 2 and above, defined by the presence of at least one significant osteophyte, to define the presence of structural OA (Figure 1).
The most commonly used diagnostic criteria in research for hand, hip and knee OA were developed by the American College of Rheumatology (ACR).4–7 These criteria require the presence of typical joint symptoms for most days of the preceding month, but radiographic changes are not essential.
The selection of the most appropriate criteria to define OA depends entirely on the context of a study. Work carried out in response to questions raised by the United States Food and Drug Administration on defining the disease state suggests that the effects of treatment on structural changes at the joint level, termed disease OA, should be considered separately from the effect on patient-reported symptoms, termed illness OA.8 This terminology is helpful in identifying which patients should be recruited to a particular study. For example, a study of the primary prevention of symptomatic OA should include patients with disease OA but not illness OA.
In order to maximize the translation of research findings to the clinical setting, there should be consistency between how patients, clinicians and researchers define OA. A community-based study showed that the correlation between patient and general practitioner (GP) diagnosis of OA was no better than chance and that neither related to the ACR criteria.9 The lack of a clear definition of a meaningful response to treatment may also have hindered progress in the search for effective therapies, and this is now beginning to be addressed.10 Further alignment and clarity of the definition of OA across the board may help to shape clinically relevant research, allow appropriate assessment of treatment effect size and accelerate the impact on direct patient care.
The size of the problem
It has been estimated that about one-tenth of the world’s population aged > 60 years have symptoms that can be attributed to OA.10 Specifically, symptomatic and radiographically confirmed OA of the knee had a global estimated incidence of 3.8% [95% uncertainty interval (UI) 3.6% to 4.1%] in 2010, with a peak in those around 50 years of age.11 With an ageing population and the obesity epidemic, the number of people suffering from OA of the knee is predicted to rise. Symptomatic knee OA has previously shown an increasing trend, over a 20-year period, with obesity having a contributory effect.12 This is in turn associated with a projected exponential rise in demand for costly surgery. The projected demand for primary total knee arthroplasty is projected to grow by 673% over a 25-year period, equating to 3.48 million procedures (95% prediction interval, 2.95 million to 4.14 million) by 2030 in the USA.13
The actual size of the problem is likely to be significantly underestimated as many people do not seek medical care because they perceive OA to be part of the normal ageing process and that little can be offered to them to help.14 A study showed that over half of people with severe, disabling knee pain had not visited their GP in the previous 12 months.14 As awareness among physicians and the general public increases, it is likely that consultation rates will also increase.
The burden of osteoarthritis
At an individual level, pain in large joints restricts mobility, particularly climbing stairs and walking,15 causes reduced participation16 and also has an impact on mood, functioning and well-being.17,18
The impact on health services is considerable given that the incidence of new GP consultations for knee pain in adults aged ≥ 50 years is approximately 10% per year.19 In the UK, there were 114,500 hospital admissions over a 1-year period, and > 44,000 hip replacements and > 35,000 knee replacements were performed at a cost of £405 million in 2000.1
The impact of OA on society at large is also substantial. In the UK alone, 36 million working days were lost owing to OA, at an estimated cost of £3.2 billion in lost production over 1 year (1999/2000). Over the same period, £43 million was spent on community services and £215 million was spent on social services for OA. The total cost of OA in the UK is estimated as the equivalent of 1% of gross national product per year.1
Osteoarthritis is a global problem. In 2010, OA accounted for 2.2% (95% UI 1.7% to 2.9%) of total years lived with disability worldwide.11 Specifically, OA of the hip and knee was ranked as the eleventh highest contributor to global disability, out of 291 conditions.11
Compared with the general population, patients with OA are at higher risk of death from all causes, but particularly from cardiovascular disease (standardized mortality ratio 1.71, 95% confidence interval 1.49–1.98).2 The burden of OA may, therefore, be higher than currently estimated. Total joint replacement in patients with moderate to severe hip or knee OA has been shown to be cardioprotective,20 which increases the rationalization for surgical intervention and makes it even more crucial to determine the optimum time for surgery.
Risk factors for osteoarthritis
It is important to recognize risk factors for OA so that people at high risk can be identified and, when possible, these factors can be modulated with the aim of improving the patients’ outcome. Defining OA in this context is crucial because risk factors for structural disease may be different from those for symptomatic disease.21
Systemic factors are thought to affect susceptibility to OA in multiple joints, and these include increasing age, female sex, ethnicity/genetics and the presence of the metabolic syndrome.22–24 Local mechanical factors are believed to result in abnormal loading and injury, which affect the risk in an individual joint. Such factors include obesity, repetitive loading and acute injury, joint deformity resulting in abnormal load distribution and muscle strength.24 The relationship with muscle strength is interesting, as increased grip strength is associated with an increased risk of OA in the hand, but poor quadriceps strength increases the risk of structural knee OA in women.24
The evidence so far suggests that the risk factors for the onset, or incidence, of disease differ from those for disease progression. A good example of this is bone density, whereby high bone density protects against further joint space narrowing in knees with features of structural OA but increases the risk of developing new-onset early osteophytes.24
The risk factors associated with the onset of pain per se introduce a distinct category of biopsychosocial mediators in OA, which include genetic predisposition, prior experience of painful conditions, expectation of pain, current mood, coping strategies, catastrophizing and sociocultural environment.25
It is likely that the risk of OA and its progression in any individual depends on the complex interaction between these systemic, local and pain susceptibility risk factors.
Discordance between radiographic osteoarthritis and pain
Traditionally, much emphasis has been placed on the study of radiographic features of OA. For example, longer-term studies of the progression of OA have tended to focus on radiographic rather than clinical features.24 Consistent with this is the fact that the presence of features of radiographic OA remains an important component of the decision-making process in the context of patient care, even in the context of primary care.26
The discrepancy between the presence of radiographic OA and symptoms, particularly in the earlier stages of the disease, is now recognized. A systematic review revealed that the proportion of those with knee pain found to have radiographic OA ranged from 15% to 76%, and, among those with radiographic knee OA, the proportion with pain ranged from 15% to 81%.27 Three main factors were proposed as possible explanations for this variation.27 First, the number and specific views of the radiographs used may be inadequate to detect the structural changes present. This aspect has been further investigated by using more sensitive techniques such as magnetic resonance imaging (MRI), which can detect abnormalities within other tissue structures including subchondral bone marrow lesions, meniscal damage and the presence of synovitis and effusion.28 The evidence obtained so far does not convincingly support the use of MRI in routine clinical practice and, at present, MRI is no better than radiography in terms of discriminating between those with and without pain.29 The second factor described relates to the fact that the way in which pain is defined and assessed has a huge potential impact on the link seen with radiographic disease (see ‘Assessing pain in osteoarthritis’). Third, it has been suggested that characteristics of the population, including age, sex and ethnicity, have an impact on the relationship between structural and symptomatic OA.
Neural processing and the plasticity of pain sensitivity have also been investigated as potential explanations for the observed discord between structural and symptomatic OA.30,31 To date, this work has suggested that central sensitization (see ‘Assessing pain in osteoarthritis’) may account for the high pain reports of those without moderate to severe forms of the structural disease.31
Assessing pain in osteoarthritis
There are a multitude of methods for assessing and recording musculoskeletal pain.32–40 Despite the fact that most patients seek medical input because of pain, quantitative measures of pain are not necessarily routinely used in clinical practice and are usually limited to the visual analogue scale (VAS) or other similarly quick assessment methods.36
Pain is composed of many different qualities, which a simple VAS or other quantitative measure of overall pain severity will not capture. Qualitative work has been conducted to identify potentially important pain characteristics experienced by patients with OA, which can then be further investigated in a more quantitative manner.41–44
Temporal patterns of pain
One of the themes emerging from previous qualitative work is that patients with hip and knee OA describe two temporally distinct patterns of pain. Patients can experience a constant aching or background pain alongside a more intermittent but more severe and disruptive form of pain.43 This discovery was then used to develop a tailor-made questionnaire: the Measure of Intermittent and Constant Osteoarthritis Pain.45 This new measurement tool, which by definition focuses on pain qualities that are important to patients, may be used to provide a more meaningful assessment of response to therapies.46–48
The fluctuation of pain also has an impact on other methods used to measure pain, as severity will vary according to when and how this information is captured. A number of studies have demonstrated that pain severity fluctuates over the course of days, weeks and years.49–51 It is particularly important to be aware of this when designing studies to measure the effectiveness of a treatment intervention. Furthermore, there is evidence to suggest that the consistency or temporal pattern of pain in itself may be associated with different factors and in itself help to predict outcome.50–52 For example, a previous longitudinal study of patients with, or at high risk of suffering from, knee OA showed that consistent knee pain was associated with a higher risk of needing total knee replacement surgery, independent of knee pain severity.51
Neuropathic features of pain
Qualitative focus group work has also identified a subset of patients with knee OA who use neuropathic pain descriptors to describe their pain.44 This led to the use of the PainDETECT Questionnaire (PD-Q)53 in its original54 and modified form (mPD-Q)55 to demonstrate that approximately 20% of patients with OA of the knee describe characteristics suggestive of neuropathic pain (Figure 2). The presence of neuropathic pain on the mPD-Q and PD-Q has also been shown to be significantly associated with neuropathic features suggested by quantitative sensory testing (QST).56,57 Neuroimaging has also further validated these findings by showing that, in patients with hip OA, PD-Q score is significantly associated with periaqueductal grey activation on stimulation with pinprick in the area of referred pain.58 The PD-Q has also been applied in the assessment of postoperative pain, and it was found that 6% of patients with ongoing pain had features of neuropathic pain.59 The next step is to investigate whether or not the presence of this quality can predict response to treatment, or help to stratify treatment strategies.
Pain localization and interindividual variability
The anatomical localization of pain has also been studied, and localized, and regional and diffuse pain patterns have been identified, each one associated with different risk factors.60,61 Further investigation is needed to assess whether or not this method of subgrouping patients will be useful in predicting the natural history of symptoms and response to therapies.
The impact of interindividual variability in pain responses has been previously reviewed, and the effect of factors such as sex, ethnicity and sleep disturbance should also be considered when assessing pain experienced by patients with OA.42,62
Mechanisms of pain in osteoarthritis
The precise mechanisms of pain in OA are not fully understood, and the sources of pain in OA have been the subject of frequent review in the literature.28,63–69 It is now widely accepted that pain is likely to be multifactorial in nature and that it should be considered in the framework of a biopsychosocial model, including a contribution from peripheral structural abnormalities within the joint at one end of the spectrum through to centrally mediated sensitivity and mood at the other.
From a structural joint perspective, the key components in knee OA at present are the articular cartilage, subchondral bone and the presence of synovitis or effusion.28 Although articular cartilage is aneural and, therefore, unable to directly generate pain, it is proposed that the destruction of cartilage releases factors that may result in synovitis or increased intraosseous pressure, thereby generating pain via a range of secondary mechanisms.28 Currently it is thought that the finding that best correlates with the incidence and severity of pain is the presence of bone marrow lesions in the subchondral bone. Bone marrow lesions arise as a result of bone remodelling and histologically show features of fibrosis and microfractures. Although the literature is conflicting, the overall consensus is that these lesions are related to pain.28 Synovitis and effusion are frequently present in OA and correlate with pain.28 This is supported by the fact that intra-articular corticosteroid injections are recommended for short-term pain relief in the recent Osteoarthritis Research Society International guidelines for the non-surgical management of knee OA.70
The response of the nervous system to the stimuli resulting from structural damage within the joint, at both a peripheral and central level, is an important component in the perception of pain in OA.67,71 Nociceptors within the joint are responsible for the transduction of local stimuli. Tissue damage within the joint causes the release of factors that stimulate nociceptors, resulting in pain. These factors are also capable of downstream regulation of the sensitivity to subsequent stimuli. When this leads to an exaggerated response, it is termed peripheral sensitization. This can cause a normally pain-free stimulus, such as walking, to become painful and has been observed in patients with OA of the hand.72
In parallel with peripheral sensitization, similar changes are seen in the central nervous system, known as central sensitization. Evidence for the role of abnormal centrally mediated pain processing in OA arises from a combination of animal studies,73,74 symptom-based assessment,44,54,55 QST75–83 and neuroimaging research.58,84,85 At the level of the neuron, central sensitization results in an increased receptive field, increased responsiveness to a previously recognized stimulus as well as altered responsiveness such that low-threshold stimuli begin to evoke a response.86 Clinically, these changes can manifest as secondary hyperalgesia (increased pain from a stimulus, distant to the primary site of injury, that normally causes pain) and allodynia (pain due to a stimulus that does not normally evoke pain),87 referred pain (pain felt remote to the tissue being stimulated, usually in a segmental distribution)88 and spreading sensitization.89 Many of these features have been documented in pain caused by OA.
The contribution of genetic susceptibility in OA is potentially broad. Genetic predisposition may be used to identify risks for the development of OA as well as the outcome at multiple stages of the disease, varying from pain sensitivity to susceptibility to periprosthetic osteolysis following total hip replacement.90 There is evidence to support the genetic component of both structural changes in OA91 as well as the development of pain.92,93 The additional contribution of genetic profiling beyond basic structural information is not yet clear94 but gene therapy, using local intra-articular delivery in particular, is an exciting and expanding area to be included in future potential OA therapeutics.95–97
Overall, there is likely to be an important interplay between the structural changes shown in OA and the way these changes are relayed by the nervous system, resulting in pain. It is important to recognize the dynamic nature of pain perception, moving away from the traditional model of a hard-wired system for relaying information about peripherally presented noxious stimuli.98
Management of osteoarthritis
A huge number of treatment options are available for the management of OA, with several practice guidelines in place.70,99,100 The current options can be considered to fall into the categories of non-pharmacological, pharmacological and surgical treatments.
A holistic approach to the assessment and subsequent management is the core concept of current treatment strategies in OA.1,99 Comprehensive assessment of patients and the impact of their condition on day-to-day life should be carried out, including physical status, activities of daily living, participation, mood, health education needs, health beliefs and the motivation for self-management.1,99 This information should then be used to derive an individualized treatment package, comprising a combination of treatment strategies rather than a single intervention.99
The three main themes of non-pharmacological therapy in OA are patient education and self-management, weight loss and exercise, and assistive devices, braces and taping. Most guidelines include moderate to strong recommendations for patient education and self-management, with regular contact to promote self-care being a common theme. In general, the recommendations advise low-impact land or water-based aerobic exercise, particularly for hip and knee OA, although the benefit in cases of hand OA is less clear. Many guidelines strongly recommend weight loss in those with hip or knee OA who are also overweight. Some groups have also suggested the addition of manual therapy, which includes massage and joint mobilization/manipulation and exercise, but manual therapy alone has not been recommended. Although walking aids and other assistive devices to improve activities of daily living are generally recommended, there is a lack of consensus on the use of braces and taping.100
First-line pharmacological therapy should comprise paracetamol (acetaminophen). However, recent evidence has suggested associated adverse effects including gastrointestinal adverse events and multiorgan failure, especially associated with prolonged use and, therefore, conservative dosing and treatment duration are recommended.70,100
Second-line therapy includes non-steroidal anti-inflammatory drugs (NSAIDs), topical or oral, and topical capsaicin. The use of NSAIDs has to be carefully balanced against the risk of cardiovascular, gastrointestinal and renal disease, and appropriate NSAID selection and use of proton pump inhibitors, combined with restricted does and treatment duration, are compulsory for patient safety.70,100
Further pharmacological options include opioids such as tramadol, but the evidence suggests a small to moderate effect size only, with many patients withdrawing from treatment because of adverse events.70,100 Glucosamine and chondroitin therapy has received much attention in the past, but current guidelines have deemed these agents as not appropriate for structural disease modification and of uncertain benefit in symptom relief.70
Intra-articular therapy is also an important component of the treatment armamentarium, with corticosteroids being recommended for hip and knee OA.70,100 The guidance on the use of intra-articular corticosteroids in cases of OA of the hand is conflicting.101,102 The use of hyaluronic acid is controversial, with conflicting evidence making it difficult for definitive recommendations to be made.70,100
Potential future pharmacological treatments
Despite the number of pharmacological agents available, the high frequency of medication change among patients suggests a lack of long-term efficacy and tolerability of current therapies.103 Targeted treatment strategies are currently being developed with preliminary evidence of success.
Treatments targeting inflammation are currently under investigation, with preliminary evidence suggesting a positive effect of methotrexate in knee OA104 and anti-tumour necrosis factor therapy in OA of the hand.105
Bone may also be a useful target in OA. There is evidence that bisphosphonates can reduce pain severity in OA106 and that strontium can reduce structural progression with a clinically meaningful effect on symptoms in OA of the knee.107,108
A final future target is nerve growth factor, which is increased in the synovial fluid of patients with OA.109 Tanezumab, a humanized immunoglobulin G2 monoclonal antibody directed against nerve growth factor, has been shown to reduce joint pain and improve function in patients with hip and knee OA.110,111 The investigation of this treatment in OA was temporarily placed on hold owing to concerns regarding accelerated joint destruction and need for arthroplasty, but further study of this target has now been sanctioned.112
Traditionally, the success of joint arthroplasty has been judged based on measures of technical success, such as revision rates and the frequency of loosening around the prosthesis.114 The development of the Oxford Hip and Knee Scores, designed to measure patient perceptions of response to treatment, has led the way towards further emphasis on patient-reported outcome measures.115,116 This change in focus has revealed that approximately 20% of patients experience ongoing long-term pain after knee arthroplasty.117 The risk factors for a poor outcome remain unclear, with numerous possible determinants.114 The most consistent factors in the literature to date are being a young woman and having experienced worse preoperative pain severity.118–120 These features explain only a small proportion of the variation in outcome postoperatively114 and the predictive capability of other features, such as preoperative sensitivity to pain, is being explored with some preliminary success.121–124
It is not easy to ascertain the optimal timing of surgical intervention or to identify who should be considered for surgery.114,125–127 There is currently significant variation in the patterns of referral for, and provision of, surgery, and women are less likely to be recommended for surgery than men.114,128,129 The need for shared decision-making between patients and practitioners in order to narrow this discord has been highlighted, and strategies to facilitate this process are being developed.130
Phenotypes in osteoarthritis
Osteoarthritis is a heterogeneous disease with variation in clinical features, aetiopathology and response to treatments. It is recognized that tailored patient therapy, which is called for in many of the current practice guidelines,1,70,99 is needed in order to optimize therapeutic effects, but the methodology for identifying robust clinically meaningful phenotypes is not well defined.131
A phenotype describes an individual’s set of detectable characteristics as a result of the interplay between the genotype and environment.132 Two main approaches to defining a phenotype in OA have been suggested. The first method uses a hypothesis-free, data-driven method, such as cluster analysis or principal component analysis, to statistically derive significant subgroups within a population (Figure 4).133–135 This method depends entirely on the nature of the data included in the analysis and the threshold set for accepting a separate phenotype. For example, in OA of the knee, this method has been used to derive five phenotypes based on differing levels of joint disease, muscle strength, obesity and depression.134 Equally, a different study used the same approach to define three phenotypes based on varying degrees of pain, fatigue, sleep disturbance and low mood.135 The second method starts with the observation that different clinical patterns of disease or underlying biological mechanisms exist and strives to prove this by demonstrating differences in baseline clinical characteristics or natural history of the disease.133 This approach has been used in OA of the hand to confirm the presence of the following subsets of hand OA: OA at the thumb base (22.4% of patients), nodal interphalangeal joint OA (15.5%), generalized hand OA (10.4%), non-nodal interphalangeal joint OA (4.9%) and erosive OA (1.0%).136
The ultimate challenge once a novel phenotype has been proposed is to validate it and prove its use in the clinical setting. Although clear consensus has not been reached, it has been proposed that validation should involve providing a relation to the response to treatment or prevention strategy.133 As one of the reasons for searching for new phenotypes is to aid the development of novel therapies, this strategy is at risk of being circular and restrictive and supplementary methods are required.
Osteoarthritis is a common and debilitating disease worldwide. Despite this, the underlying mechanisms of the disease are not fully understood and the discordance between structural severity and symptomatic disease cannot be completely explained. By addressing the need for a comprehensive and holistic approach to the assessment and management of patients with OA, using a shared decision-making model, it is anticipated that the utilization of appropriate treatment strategies will improve. Furthermore, investigation of the underlying causes of the disease with the acknowledgement of its heterogeneity is likely to result in the identification of clinically meaningful phenotypes of patients that will also guide therapy.