Identification of knee osteoarthritis disability phenotypes regarding activity limitation: a cluster analysis

Knee osteoarthritis (OA) is a condition with great heterogeneity by its nature. Despite similar structural involvements among people with knee OA, different outcomes of management were observed [1‐3]. Classifying this population in homogeneous subgroups might lead to more direct and specific treatments [4]. Related studies have attempted to identify the common phenotypes of knee OA using various methods such as cluster analysis [5], latent class analysis [6], and predefined [7] methods. Considering the definition of phenotype as “the observable properties of an organism that are produced by the interaction of the genotype and the environment” [8], studies had attempted to identify the phenotypes of individuals with knee OA. Impairment physiognomies commonly used as phenotypic variables in research studies involve clinical findings such as pain sensitivity [9, 10], knee alignment [7], and gait parameters [11].

Considering the International Classification of Functioning, Disability and Health (ICF) framework, not only impairment but activity limitation and participation restriction were also affected by the pathology of knee OA [12, 13]. However, participation restriction is usually individualized and depends largely on personal as well as environmental factors. Therefore, functional disability or activity limitation is more commonly used to guide goal setting in knee OA management in rehabilitation. Related studies regarding physical therapy interventions usually evaluated the physical functions of these patients [14‐17].

Using activity limitation to identify the phenotypes among patients with knee OA would lead to more specific levels of disability related to the physical therapy treatment goals. Few studies have identified knee OA phenotypes based on variables of activity limitation [18, 19]. Two studies used the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) function subscale as the phenotypic variables [18, 19]. They reported that adults experiencing risk of knee OA and those with symptomatic knee OA demonstrated a variety of functional decline measured by the WOMAC function subscale. However, the phenotypes based on the sum score of 17 functional activities in WOMAC might not be accurately linked to the physical therapy management [20, 21]. The further analysis of the specific items of activities with greater difficulty would be needed.

Identifying specific activities impacted by knee OA should be more useful to guide management and prognosis. However, many activities are associated with knee OA symptoms. Stair climbing, rising from a chair and walking were reportedly the most common functional limitations among individuals with knee OA [22]. Moreover, the practice guidelines of knee OA suggest evaluating the functional capacity of walking, stair climbing, sit-to-stand, and balance ability [23]. Therefore, this study aimed to verify these activity limitations as reported in literature which could specifically classify the phenotypes of individuals with knee OA. The impairments and participation restrictions among phenotypes were also compared to validate the subgroup classification. We hypothesized that specific activity limitations would be identified and these activities could also classify individuals with symptomatic knee OA in homogeneous subgroups regarding level of disability. Moreover, the severity of impairments and participation restrictions would differ among phenotypes and these variables would be able to explain more variances when subgrouping of participants was considered.

In all, 250 participants with symptomatic knee osteoarthritis (OA) were enrolled in this study. Age, BMI, duration of knee pain and proportion of males and females of overall participants and the subgroups from cluster analysis are presented in Table 1. The BMI were significantly different among subgroups.

Fig. 2 presents the distributions of the responses of 23 activities limitation assessed. Three activities chosen based on literature to possibly be the cluster variables including maintaining standing position, walking long distance and stair climbing met the criteria of 100% responses and had appropriate distributions of difficulties among 250 participants. Therefore, they were taken into the cluster analysis to further determine the phenotypes of individuals with knee OA. The results of performance tests of timed self-paced walking and stair climbing test were used for the analysis.

This study aimed to identify the activity limitations appropriate to identify homogeneous subgroups of knee OA. The impairment and participation restriction were then compared among the established subgroups. The results were congregant with our hypotheses. Three variables including difficulty levels of maintaining a standing position, walking time, and stair climbing time represented distinct phenotypes. Four phenotypes were identified with differing levels of disability in knee OA. The impairments and participation restriction levels significantly differed among individuals in each phenotype. Greater variance was accounted for when using the subgroups according to the phenotype compared with the overall group. This implied that the subgroup could more precisely identify people not only activity but also their impairments and participation levels. The related literature also confirmed that among patients with knee OA, these three activities were the common ones being assessed and used as functional goals in physical therapy clinics [13, 22, 29]. Therefore, the three activity limitations identified in this study had the power to classify subgroups of people with knee OA.

The activity limitation variables had been used to classify individuals with knee OA in homogeneous subgroups [18, 19]. However, the number of phenotypes varied among studies. In this study, activity limitation variables could cluster the patients with knee OA in four phenotypes with the disability levels of no, mild, moderate and severe. A related study conducting 5-year follow-up among people with early symptomatic knee OA identified three phenotypes of knee OA consisting of good, moderate, and poor outcome subgroups [18]. The different characteristics regarding activity limitation decline over time among subgroups was reported. The authors suggested that their homogenous identification of individuals with knee OA could be used to develop specific interventions [18]. Another 7-year follow-up study proposed five phenotypes of people with knee OA differentiated by functional decline. The subgroups comprised high functioning, minimal limitation, late worsening, remitting and progressive worsening trajectory [19]. Secondary analysis demonstrated an association between decline of activity and contributing factors of activity limitation including radiographic disease severity, knee pain, obesity and depressive symptom [19].

Other studies have proposed using commonly used functional scores such as WOMAC to identify subgroups of knee OA [6, 30]. However, the WOMAC function subscale could not appropriately predict phenotypes [6, 30]. The study by Egsgaard et al used mixed phenotypic variables including WOMAC subscales, Lequesne index, quality of life, pain catastrophizing, quantitative sensory testing and inflammatory profiles as the clustering variables. Only the pain sensitization profile could identify four distinct phenotypes comprising low sensitivity to pain, early phase sensitization, presence of pain sensitization and presence of pain sensitization and catastrophizing [30]. On the other hand, the secondary comparison of activity limitations among the phenotypes of impairment variables also significantly differed [31, 32]. This implied a close relationship among the activity limitations and impairments among individuals with knee OA.

In this study, the results of secondary analysis supported that activity limitation variables identified in the cluster analysis could classify participants with knee OA in homogeneous subgroups. The comparisons of impairment and participation restriction variables demonstrated significant differences among phenotypes of all variables. Pain intensity, limited ROM, muscle power and participation restriction levels were worse in subgroups with greater disability accordingly. The impairments related to the activity limitations of each phenotype might be used to develop specific treatment guidelines. Similarly, the related study demonstrated that knee pain and knee flexion ROM also differed among phenotypes based on activity limitations [18].

The major strength of this study was that it identified the common activities which would be disable in persons with knee OA. These activities would be useful in the goal setting process of rehabilitation in patients with different levels of disabilities. However, the study had some limitations. First, the participants included were the patients registered to the primary healthcare settings and diagnosed using the knee OA clinical criteria. No radiographic or other investigations were performed to confirm the structural lesions. Second, due to the cross-sectional nature, the cause and effect of the impairments and participation variables on activity limitations could not be identified.

The cluster analysis confirmed the heterogeneity nature of knee OA. The three activity limitation variables of maintaining a standing position, stair climbing time and walking time could be used to identify homogeneous subgroups of knee OA. Goal setting and treatment planning could be guided by the characteristics of phenotype. However, specific physical therapy management guidelines related to phenotypes are required for further study to identify those completely related factors of these activity limitations.