We identified relevant articles, reviews, and abstracts in a search of PubMed/MEDLINE and EMBASE for English-language articles published between 1990 and October 2013. The initial search strategy included the terms osteoarthritis, risk factor, predictor, progression, guidelines, biomarkers, MRI, and phenotype, and yielded 463 items. Separate subsearches were also performed using a cross-search of the above terms combined, and additional references were selected from the reference lists of selected articles and the presentations made during the working meeting. The narrative is therefore largely based upon expert opinion. Overall, 73 relevant items were selected by the authors according to their quality and pertinence for discussion by the ESCEO working group.
Epidemiology and Risk Factors for Progression
The general definition of osteoarthritis is a group of overlapping disorders with similar structural and clinical outcomes . Indeed, osteoarthritis can affect articular cartilage, subchondral bone, synovium, meniscus, muscle, capsule, and ligaments. This definition of osteoarthritis may be important for the characterization of patient profiles, since it determines what sort of information is to be captured, i.e., structural, clinical, or surgical. Osteoarthritis has been classified into three subsets according to distinct etiological, clinical, and therapeutic characteristics: estrogen deficiency-related, genetically-induced, and age-related (Fig. 1) [1, 2]. This proposal is based on the etiology and pathogenesis of osteoarthritis , and combines the three main biological processes crucial for the development of osteoarthritis, together with additional risk factors such as obesity, the metabolic syndrome, and trauma. Understanding the pathogenesis of the disease is important for determining patient profiles in order to establish rational treatment in osteoarthritis. Another important consideration is that osteoarthritis is usually an insidiously progressive disease, and a patient’s profile may evolve over the course of the disease . It has been proposed that all of the clinical forms are interchangeable in the early stages of disease and may appear as differing clinical profiles in various tissues . In more advanced disease, clinical and imaging presentations become more generalized and profiles may overlap.
In the research setting, osteoarthritis is generally classified radiographically, most typically using the Kellgren–Lawrence (KL) grading system [10, 11]. Although the correlation is not always linear [12, 13], the osteoarthritis patients with the most pain tend to have the highest KL grades.
Osteoarthritis has a high incidence. A recent study in a Spanish population including more than 3 million individuals reported incidence rates of clinically diagnosed osteoarthritis of 6.5, 2.1, and 2.4 per 1,000 person-years for knee, hip, and hand, respectively . The incidence of osteoarthritis increases with age, rising sharply beyond the age of 50 years and leveling off after the age of 80 years [7, 14–16]. It is considerably more common in women than in men [7, 15, 16]. For example, in the Spanish study, the incidence rate of knee osteoarthritis in women was 8.3 per 1,000 person-years versus 4.6 per 1,000 person-years for men . In view of the high prevalence and high cost of the surgical consequences of joint failure, osteoarthritis constitutes a major healthcare burden. The situation may even be worsening, since the rates of total hip replacement have increased and there are signs that joint replacement may be occurring at an earlier age . The notion of healthcare burden is also important since patients with osteoarthritis are at higher risk for mortality than the general population, and there have been calls for a more unified approach to healthcare in these patients, including effective management of cardiovascular risk factors and co-morbidities [18, 19].
There are many risk factors for osteoarthritis, including systemic factors, such as age, sex, body mass index (and obesity), genetic factors, bone density, and estrogen status, and local biomechanical factors, such as obesity, physical activity or occupation, intense sporting activity, joint injury or deformity, and muscle weakness . Some of these are also risk factors for progression of the disease. The three risk factors that appear to most consistently predict progression are obesity, generalized osteoarthritis, and synovitis/effusion . Joint space narrowing over 5 years has recently been demonstrated to predict future knee replacements up to 15 years later [21, 22]. There are many epidemiological studies on progression of osteoarthritis. In one analysis performed in the Chingford Women’s Study , the progression of radiographic knee osteoarthritis, i.e., KL grade, was recorded in a sample of 561 patients. Although more than half of patients had no progression with stable KL grades over 15 years, patients with KL grade 1 were twice as likely to progress as those with KL grade 0 . While the evidence points to a rapid progression of patients with early disease, it should be treated with caution due to the possibility of collider bias, which is a potential confounder in any analysis involving selection of patients at baseline on the basis of a characteristic that is also a risk factor.
Current treatment guidelines in osteoarthritis generally agree that it requires a combination of non-pharmacological and pharmacological modalities [25–31]. Management should start with non-pharmacological therapy, and if symptoms persist a stepwise increase in treatment intensity should follow, starting with over-the-counter treatments, then prescription treatments for pain control, and then, if needed, referral for surgery; opioid analgesics are generally reserved for patients who cannot receive surgery. Regulatory guidelines are also available for drug development in the symptomatic and structural management of osteoarthritis . None of these guidance documents provide any information on which patients should be treated.
Identification of Patient Profiles in Osteoarthritis
The characterization of patient profiles appears as an important priority for osteoarthritis for a number of reasons. First, it would help to better orientate research and understand the disease, which would in turn improve its management. Second, it would facilitate the design of randomized clinical trials in the field and the development of new pharmacological strategies. Third, it would help to better determine which patients are the most likely to benefit from which treatment, leading to more personalized medicine and more effective use of healthcare .
Osteoarthritis is a heterogeneous disorder and the various patient profiles will be influenced by many different factors (Table 1), all of which may impact response to treatment and some of which are interdependent. These include factors related to morphology and anatomy, the type of tissue that is affected, the presence of co-morbidities, and the clinical presentation.
The identification of patient profiles will necessarily involve the use of biomarkers and imaging markers, as has been discussed previously by the ESCEO [4, 5]. Although there are a number of promising candidates for biomarkers [5, 33, 34], such as urinary C-terminal telopeptide of collagen type II (CTX-II) and serum cartilage oligomeric protein (COMP), none is sufficiently discriminating for diagnosis or prediction of prognosis in patients or for use as a surrogate outcome in clinical trials. As regards imaging, radiographic joint space width or narrowing remain the recommended parameters according to the regulatory bodies . On the other hand, MRI markers also provide a good measure of cartilage morphometry [4, 36], meniscal damage , bone marrow lesion , and synovial effusion [38, 39].
Profiles According to Articular Involvement: Local or Generalized Inflammation-Driven Osteoarthritis
Generalized osteoarthritis may be a different profile from local osteoarthritis. The term ‘generalized osteoarthritis’ is widely used in the literature with a variety of conflicting definitions . In the context of our discussion, we consider generalized osteoarthritis as representing a systemic disease affecting a number of different joints at the same time, i.e., it is polyarticular. This is a feature of osteoarthritis co-morbid with inflammation or the metabolic syndrome. On the other hand, local osteoarthritis occurs in a single joint, i.e., it is monoarticular, and is typical of post-traumatic osteoarthritis. It could be surmised that biological markers would be more suitable for identification of the profile of generalized osteoarthritis, while imaging markers would be more effective for exploration of local effects post-trauma. Similarly, within the range of biomarkers, serum measurements may provide a better measure of generalized osteoarthritis (cartilage and bone sources), while intra-articular biomarkers may better characterize local osteoarthritis [41, 42]. Ongoing genetic studies in more than 2,500 patients with generalized osteoarthritis are set to detect linkage of circulating biomarkers to osteoarthritis-associated genes, and may shed further light on the possibilities of genetic profiling for the disease . There is contradictory evidence surrounding the role of inflammation, with reports that serum high-sensitivity C-reactive protein may or may not be associated with structural progression and symptoms [44–48]. An alternative explanation is that this is related to a high body mass index or the presence of the metabolic syndrome.
Early Versus Late: Profiles According to Structural Damage and Response to Treatment
Patients in the early stages of osteoarthritis, i.e., those with less severe disease, may have a separate profile from patients with advanced osteoarthritis, i.e., those with very severe disease. This is an important point since there is evidence that response to treatment may depend on a number of factors related to severity, e.g., extent of structural damage or the site of osteoarthritis (hand, knee, or hip). A dedicated PubMed search using “predictor” as a MeSH term yielded 13 studies on predictors of response in knee osteoarthritis (Table 2) (including five studies in a single systematic review on steroid injection) [49–57]. All of the studies involved assessment of structure via radiography, usually with some evaluation of symptoms. They cover a wide variety of potential treatments including doxycycline, weight loss, cindunistat, corticosteroid injection, chondroitin sulphate, glucosamine, and celecoxib. The overall conclusion from the studies is that patients with the least severe knee osteoarthritis (i.e., those with the largest joint space width early disease or with a low KL grade) are more likely to benefit from effective treatment than those with more severe disease (a smaller joint space width or higher KL grade). This implies that the most powerful predictor of treatment response is severity of osteoarthritis, with the least symptomatic patients, i.e., those with early stage disease, likely to have the best response to treatment. Patients with more advanced osteoarthritis, with extensive structural damage to the joint or malalignment, appear less likely to respond.
There are a number of methodological issues with the studies included in Table 2 (e.g., heterogeneity of the data, small studies, low statistical power, placebo effects, regression to the mean, confounding factors, ceiling effects with the use of radiography, and collider bias). These observations therefore require careful interpretation and further research to ascertain the validity of any conclusion.
Profiles According to the Biomechanical Properties of the Joint
There is an increasing weight of evidence that biomechanical aspects may affect the incidence and progression of osteoarthritis [58, 59]. One example of this comes from an MRI study, which indicated that the presence of cam deformity femoroacetabular impingement appeared to predict the onset of hip osteoarthritis . The most important contributors to progression appear to be joint malalignment, loss of meniscal function, and ligament injury, and so it is likely that the same biomechanical aspects could also determine response to treatment. Other mechanical factors such as joint injury, obesity, and sport may also play a role [61, 62]. This raises the possibility of different profiles according to joint morphology and biomechanical function.
Biomechanical factors should be taken into account in the management of osteoarthritis. For example, correction of malalignment using a wedge insole has been reported to be highly effective in relieving pain and function in valgus knee osteoarthritis . Similarly, use of a medial collagen meniscus implant in patients with meniscus injuries was shown to improve pain, activity, and radiological outcomes over 10 years compared with patients with partial medial meniscectomy . There is also evidence that deformity or malalignment could have an effect on the efficacy of treatments for osteoarthritis. In one of the studies in Table 2, patients with neutral joint structure at baseline responded better to the structural effects of treatment with doxycycline than those with varus malalignment .
Profiles According to the Role of the Subchondral Bone
There is evidence that there are two separate patient profiles related to lesions in the subchondral bone. Although alterations in the subchondral bone appear to play a crucial role in the development and progression of osteoarthritis , the literature on the relationship between osteoarthritis and osteoporosis remains mixed on the subject [66–68]. We should recall that despite increase bone volume fraction, subchondral bone is hypomineralized and of inferior quality in osteoarthritis. One hypothesis is that the subchondral bone remodeling observed in osteoarthritis is different from that in osteoporosis [69, 70]: increased bone mineral density (i.e., increased bone formation) is associated with subchondral bone sclerosis, while decreased bone mineral density (i.e., increased resorption) is associated with subchondral bone osteopenia. Animal experiments indicate that osteoporosis aggravates the progression and severity of osteoarthritis [71, 72], possibly due to subchondral bone fragility [73, 74].
The presence of two patient profiles may have direct implications for the effect of bone-active treatments in osteoarthritis. Indeed, response to treatment may vary according to whether the subchondral bone lesion has a sclerosis phenotype or an osteoporotic phenotype [75, 76], and patients with the osteoporotic profile could gain more benefit from bone-active treatments [69, 77–79]. More research into this is necessary, especially since varying results have been found with antiresorptive agents in osteoarthritis [80, 81]. On the other hand, subchondral bone mineral density should be assessed in osteoporosis patients with other risk factors for osteoarthritis in order to establish an early treatment .