Background
Osteoarthritis is the most common form of arthritis; it constitutes a leading cause of disability in the adult population [
1] with the knee the most affected joint. Knee Osteoarthritis (KOA) is a heterogeneous pathology characterized by a complex and multifactorial nature [
2]. This multifactorial aetiology contributes to the broad variation in symptoms presentation and treatment response that characterize the KOA subjects and constitutes a challenge for the identification of personalized and effective interventions. Therefore, in order to optimize treatment effect in KOA, the intervention should address this variability and should be tailored to specific subgroups or phenotypes as highlighted in the NICE guidelines on KOA [
3‐
6]. A phenotype in KOA can be defined as a collection of observable traits (i.e. aetiologic factors, risk factors) that can identify and characterize a subgroup in a defined population. The presence of distinct phenotypes within the KOA patient population would suggest distinct underlying causes and mechanisms, which could be highly relevant for understanding and treating the disease [
7,
8].
Previous attempts to identify distinctive KOA phenotypes used different perspectives. Some researchers used disease progression to determine KOA phenotypes, while others looked at pain perception or the degeneration pattern of the cartilage [
9‐
15]. Potentially, hundreds of phenotypes may be identified depending on the definition of phenotypes and on the variables selected. Each approach can be considered equally valid depending on the scope. Only studies focusing on the identification of clinical subgroups characterized by different disease mechanisms can be considered useful to improve treatment allocation and clinical management of the disease. If, as hypothesized, treatments Are highly effective only in one sub-type; the therapeutic effect of the intervention will be lost if tested in KOA population as a whole [
4]. Therefore, the identification of risk and aetiologic factors that can identify specific clinical subtypes of KOA is an important starting point for the implementation of phenotyping research in clinical practice and may be critical for the improvement of treatment allocation and for the development of new treatment strategies.
The aim of this review is therefore to synthesize the current evidence for the existence of distinct sets of variables that may suggest the existence of clinical KOA phenotypes characterized by the presence of different risk and aetiologic factors.
Discussion
The aim of the present study was to synthesize the current evidence for the existence of clinical phenotypes in the KOA population. Six main groups of variables which suggest the existence of different underlying disease mechanisms in the KOA population were identified after a qualitative data analysis. These sets of variables should be further explored in order to confirm and better define the KOA phenotypes emerging from the literature.
In the chronic pain phenotype, high prevalence of widespread pain and psychological disturbs suggests that central sensitization plays a fundamental role in the disease process. Severe pain is often reported in association with low or moderate degeneration of the local joint structures. In these subjects, the joint disease alone is not sufficient to explain the complex symptomatology, thus it is likely that these subjects belong to a specific KOA phenotype rather than to a stage of the disease [
7,
36]. Due to the reversibility of central sensitization combined with the lack of longitudinal studies, it is not yet clear if membership of this subgroup is stable over time. Despite this uncertainty, when patients present symptoms consistent with a chronic pain phenotype, they may need and respond to treatments that differ from those targeted towards joint pain [
4]. Cognitive-behavioural therapy and pain education can be worthwhile in this phenotype and may optimize the results of other traditional intervention such as exercise therapy and joint replacement [
23].
In recent years, a growing body of evidence supports the involvement of local inflammatory mediators in the disease pathogenesis [
41]. Signs of inflammation have been found in a large part of the KOA population. In many cases these signs seem only to characterize specific phases of the disease [
42]. From this literature review emerged evidence that a subgroup of the KOA subjects presents specific inflammatory mechanisms as determinant of the disease. Attur et al. identified a group of KOA subjects with a gene overexpression of inflammatory cytokines in a study with longitudinal design [
18]. This finding suggests that KOA subgroups characterized by specific inflammation mechanisms may exist regardless of disease stage, as found in other studies [
43,
44]. Treatments targeting the inflammation process may be particularly effective in these subjects [
45].
Metabolic alterations seem key factors in two subgroups in which the alterations are present at a systemic level or with regards only to bone and cartilage metabolism in the affected knee joint [
46,
47]. The included studies reporting a metabolic syndrome as key characteristic of a specific KOA subgroup used BMI; blood; and serum biomarkers in their identification process. The use of these features is supported by previous non-phenotyping studies that identified an association between high BMI and OA lesions in non-weight-bearing joints suggesting an underlying systemic pathway [
48]. Moreover, recent studies showed that the combination of cardio-metabolic disturbance and obesity increases the risk of OA and identified an association between OA and hypertension, dyslipidaemia, and hyperglycaemia [
46,
49‐
51]. These findings indicate that systemic metabolic alterations could be one of the main causes for the disease in a specific subgroup of subjects. A multi-stages disease model cannot fully explain the existence of a metabolic syndrome subgroup that instead could be explained as a separate KOA phenotype.
Metabolic alterations in the KOA population have been reported not only at a systemic level, but as specific alterations in cartilage and bone metabolism. Biomarker analysis represents the gold standard for the identification of metabolic alterations in bone and cartilage. The identification of specific biomarkers profiles in the KOA population, as emerged from the studies included in this review, which represents strong evidence in support of the existence of a phenotype in which bone and cartilage metabolism are of primary importance as a determinant of the disease. Drugs aiming to influence bone and cartilage metabolism may see their effect improved if tested in this specific phenotype [
4].
The possibility of a mechanical overload phenotype emerged from this systematic review; however, a large gap in the evidence regarding the existence of this phenotype emerged, due also to the lack of studies with longitudinal design. Among the studies included, malalignment and muscle strength were the biomechanical variables used to define biomechanical phenotypes [
7,
8,
36] in combination with cartilage degeneration, BMI, and previous injuries. Malalignment has been shown to be strongly associated with disease progression and cartilage degeneration in specific compartments of the knee (e.g. varus malalignment is closely associated with medial tibiofemoral compartment disease) [
52]; while high muscle strength has been reported as a protective factor against symptomatic but not radiographic KOA [
53]. The studies included in this systematic review reported subgroups of KOA subjects with high levels of muscle strength. The authors suggested that the presence of high level of muscle strength in combination with other factors (e.g. malalignment, previous injury, BMI) could signify a group of people with high level of physical activity and biomechanical overload [
8]. Therefore, malalignment in combination with other known factors (e.g. muscle strength, previous injury) may confer high local stress in the correspondent joint compartment supporting the hypothesis of biomechanical mechanisms responsible for the disease. For this reason, it is likely that these subjects would respond to, biomechanical interventions (e.g. wedged insoles, knee braces) rather than to drug treatments aiming to protect the cartilage [
4].
Although our study aimed to identify phenotypes based on different disease mechanisms, from the literature a group of subjects with low degeneration and mild clinical symptoms emerged. These subjects seem to suggest the existence of a KOA subgroup characterized by minimal joint disease. Although these features could be considered representing an early stage of the disease; three of the included studies showed stability over time (2–10 years) [
9,
20,
27], supporting the consideration of this subgroup as a phenotype rather than a stage of the disease. Subjects were classified in this group based on the severity and the outcome of the disease regardless of possible mechanisms or aetiology. Despite this, the clinical characteristics of the subjects classified in this subgroup seem to suggest different underling mechanisms of the disease. The inclusion of outcomes in the classification process makes the identification of subjects belonging to this phenotype difficult in clinical practice. Strong evidence of a clinical variable able to predict the non- progression of the disease is still missing.
In this systematic review, six groups of variables that can indicate the presence of six main phenotypes have been identified. These sets of variables seem to suggest the existence of different disease mechanisms and aetiology in specific subgroups of the KOA population. None of the studies analysed here explored the possibility of an overlap between the suggested subgroups. Considering the variables used to identify phenotypes and the pathophysiology of the disease, there is no reason to exclude the possibility of an overlap. For example, patients with chronic pain could present characteristics considered key factors of other phenotypes like metabolic alterations or malalignment. Therefore, while these phenotypes may be distinct, they are not necessarily mutually exclusive. It can be hypothesized that patients with features consistent with more than one phenotype may be more severely affected by the disease and could be regarded as more complex clinical cases.
Another implication of the overlap between phenotypes is the possibility that the phenotypes identified here do not exist as separate entities in the KOA population, but only as result of the choice of specific variables, samples and analysis in the phenotyping process. This represents a limitation of the review that is not able to conclude if these phenotypes can be regarded as separate entities. Therefore, studies that try to identify KOA phenotypes with different disease mechanisms within the same sample are needed to study the possibility and the entity of overlap between phenotypes and verify the existence of phenotypes as distinct groups. Moreover, studies identifying an overlap between phenotypes may be important in the identification of complex KOA cases that may benefit from a combined treatment approach.
Among the 25 studies included, four had a strong risk of bias [
25,
28,
30]. The main source of bias was the presence of confounding factors; of all the studies included in the review, only four studies presented a low risk of bias in that specific area [
7,
8,
27,
40]. Disease duration was the main confounding factor taken into account in this systematic review, whereby differences between patients due to them being in different stages of the same disease process could potentially identify subgroups. These disease-stage subgroups did not fit the definition of phenotypes for the purpose of this review. Therefore, studies in which there were significant differences in disease duration between identified subgroups were regarded as at high risk of bias in this area.
Two of the included studies using blood and serum biomarkers in order to identify phenotypes had a mixed sample of KOA and hip OA [
24,
37]. In both the samples more than 70 % of the subject had a diagnosis of KOA, but nevertheless findings from these studies should be interpreted with caution when applied to the KOA population.
Another important source of bias was the selection of the study sample. Studies that tried to identify specific phenotype may have oversampled high-risk patients, thus leading to elevated prevalence rates. A similar bias was the inclusion of only patients listed for joint replacement [
24,
26,
28]. Furthermore, the evidence presented in this review is limited by the research focus of published studies and their quality. The criterion used to identify a phenotype required the support of two studies with low or moderate risk of bias. This approach implies the possibility that some important phenotypes have not been reported due to a limited number of appropriate studies (as was hypothesized to be the case for the mechanical overload phenotype to some extent).
Because OA is a heterogeneous disease, identifying subgroups for treatments is probably one of the promising ways forward in clinical research [
2]. This can only be achieved when the correct methodology to identify such subgroups is used. For this reason, we focused only on studies that had as a main focus the identification of KOA phenotypes. Some studies looking at the influence of specific risk factors of disease progression and outcome were excluded. We are aware that results emerging from these studies may identify useful evidence, especially in generating new hypotheses regarding phenotypes. Nevertheless, the aim of this review was the identification of phenotypes which have already been broadly studied in the literature and that are supported by evidence emerging from these studies. The absence of a post traumatic KOA as an identified phenotype may work as an example. Only Waarsing et al. analysed the rate of knee injuries to characterize their phenotype. Despite the strong evidence that identifies injuries as an important risk factor in the development of KOA; studies investigating whether patients can be meaningfully grouped based on a history of traumatic injury are absent. It may be that subjects with a history of traumatic knee injury constitute a separate phenotype. Alternatively, injuries may predispose patients to KOA through more than one underlying pathway, and may therefore not be a meaningful phenotypic identifier in itself.
The lack of a clear definition of phenotypes makes synthesis of the current literature difficult; therefore, a clear and shared definition of KOA phenotypes would help to better direct future research in the field. To combine studies, we relied on what was reported by the author and on previous research on KOA risk and aetiologic factors. This approach has intrinsic risks and may be affected by a decisional bias. However, all the data used to draw the conclusions have been reported (see Additional file
1) in the attempt to make the decision process as transparent as possible. We found this methodology the best compromise to deal with the large variability in the field and to provide useful evidence. Finally, the six sets of variables identified in this review may not be able to fully explain heterogeneity of the patient population. Future research may yet lead to the identification of different disease mechanisms suggesting the existence of new phenotypes.