SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Objective

To identify subgroups or phenotypes of knee osteoarthritis (OA) patients based on similarities of clinically relevant patient characteristics, and to compare clinical outcomes of these phenotypes.

Methods

Data from 842 knee OA patients of the Osteoarthritis Initiative were used. A cluster analysis method was performed, in which clusters were formed based on similarities in 4 clinically relevant, easily available variables: severity of radiographic OA, lower extremity muscle strength, body mass index, and depression. Univariable and multivariable regression analyses were used to compare phenotypes on clinical outcomes (pain and activity limitations), taking into account possible confounders.

Results

Five phenotypes of knee OA patients were identified: “minimal joint disease phenotype,” “strong muscle phenotype,” “nonobese and weak muscle phenotype,” “obese and weak muscle phenotype,” and “depressive phenotype.” The “depressive phenotype” and “obese and weak muscle phenotype” showed higher pain levels and more severe activity limitations than the other 3 phenotypes.

Conclusion

Five phenotypes based on clinically relevant patient characteristics can be identified in the heterogeneous population of knee OA patients. These phenotypes showed different clinical outcomes. Interventions may need to be tailored to these clinical phenotypes.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

The population of knee osteoarthritis (OA) patients is heterogeneous (1, 2), characterized by variations in pathophysiologic etiologies (2, 3), disease progression (4), underlying mechanisms of OA pain (5), severity of pain and activity limitations (clinical outcome) (1), functional decline (6), and treatment response (1). It has been hypothesized that the heterogeneous population of knee OA patients actually consists of different subgroups or phenotypes (1, 3).

The identification of different phenotypes of knee OA patients may be highly relevant to disease treatment (2). Current treatment recommendations for knee OA are intended to provide symptomatic relief (7), but not all patients respond similarly to treatment (1). Therefore, it has been theorized that treatment may need to be tailored to phenotypes of knee OA patients to optimize its effect (1–3), preferably in a cost-effective manner (8). If a treatment is exclusively effective in one particular phenotype, this effect might be missed when examined in a population consisting of more than one phenotype (3, 9).

To identify clinically relevant phenotypes, phenotypes could be formed based on similarities in characteristics that: 1) are regularly assessed in clinical practice (3), 2) show large variations between subjects (10), and 3) have a major impact on disease progression and/or clinical outcome (3). In addition, characteristics should preferably be amenable to treatment. Based on their relevance to clinical practice, we chose 4 patient characteristics to identify clinical phenotypes in a knee OA population: radiographic severity of OA, muscle strength, body mass index (BMI), and depression.

Radiographic OA is the primary modality for disease diagnosis and classification in clinical practice (11). Large variations in radiographic OA are common among symptomatic knee OA patients ranging from no to severe structural disease (12). There is evidence for a relationship between radiographic OA and clinical outcome, although this evidence is conflicting (4, 11, 12).

Muscle strength of the lower extremity is an important characteristic in clinical practice in knee OA (13). Muscle weakness is highly prevalent among knee OA patients (14), and large variations in muscle strength between knee OA patients have been reported (15). Muscle weakness is a risk factor for knee OA onset (14) and progression (16), poor clinical outcome (17), and functional decline in knee OA (6).

Obesity is a common comorbidity in knee OA (18), and large variations in BMI in knee OA patients have been reported (19). High BMI is strongly related to knee OA onset and progression (18, 20), poor clinical outcome (17, 19), and functional decline in knee OA (6).

Depressive symptoms are more prevalent in knee OA patients than in healthy persons (21) and are varying from no depressive symptoms to major depression among knee OA patients (17). Depression is an important risk factor for poor clinical outcome (17) and functional decline (6) in knee OA.

We hypothesize that clinical phenotypes can be identified in the heterogeneous population of knee OA patients based on similarities in the 4 patient characteristics mentioned above. We hypothesize the existence of a phenotype of nonobese knee OA patients with high muscle strength demonstrating relatively good clinical outcomes, regardless of the presence of severe radiographic OA. In addition, we hypothesize a phenotype of obese knee OA patients with low muscle strength (i.e., dynapenic obesity [22]) demonstrating poor clinical outcomes, even in the absence of severe radiographic OA. Finally, we hypothesize the existence of a phenotype of knee OA patients with depressive symptoms demonstrating poor clinical outcomes, even in the absence of severe radiographic OA.

The first aim of the present study was to identify clinical phenotypes based on similarities on 4 patient characteristics. The second aim was to compare clinical outcomes of the identified phenotypes.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

We used data from the 2-year followup visits of the “progression subcohort” of the Osteoarthritis Initiative (OAI) database. This database is available for public access online at http://www.oai.ucsf.edu (data set version 9, version 3.1, and version 3.2.1). Subjects were eligible for the “progression subcohort” if they had both of the following in at least 1 knee at baseline: frequent knee symptoms in the past 12 months defined as “pain, aching, or stiffness in or around the knee on most days” for at least 1 month during the past 12 months, and radiographic tibiofemoral knee OA, defined as definite tibiofemoral osteophytes (Osteoarthritis Research Society International atlas grades 1–3 [23]) on the baseline fixed-flexion radiograph. From a total of 1,389 participants from the “progression subcohort,” 842 participants with complete data on radiographic OA, muscle strength, BMI, and depression at the 2-year followup visits were included in our study.

Clustering variables.

Radiographic severity of knee OA was scored by an overall severity score (Kellgren/Lawrence [K/L] score) (24) of the most severely damaged knee ranging from 0–4, based on the weight-bearing fixed-flexion knee radiograph. The mean score of left and right lower extremity isometric muscle strength (quadriceps and hamstring strength in newtons) was used as a measure of muscle strength (13, 25). BMI was calculated by dividing weight (in kilograms) by squared height (in meters). A person with BMI greater than or equal to 30.0 kg/m2 is defined as obese (26). Depression was based on the score on the Center for Epidemiologic Studies Depression Scale (CES-D) questionnaire, which is one of the most frequently used questionnaires for depressive experiences with very strong psychometric properties (27). The CES-D is a 20-item measure with a sum score ranging from 0–60, in which a score of >16 is indicative of mild depression and a score of ≥27 is indicative of major depression (28). Data from the 2-year followup visits were used.

Clinical outcome variables.

Two variables for knee pain and 2 variables for activity limitations, all measured at the 2-year followup, were used to assess clinical outcome. Knee pain was measured by an 11-point numerical rating scale (NRS) ranging from 0–10 (where 0 indicates no pain) for self-reported knee pain during the last week (29) and by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain subscale ranging from 0–20 (where 0 indicates no pain) (30). Activity limitations were assessed by the WOMAC physical function subscale ranging from 0–68 (where 0 indicates no activity limitations) (30) and by a 20-meter walking test (31). The NRS questionnaire and both WOMAC questionnaires had been obtained for each knee separately; we used the mean score for the left and right knees for our analyses. The 20-meter walking test was performed twice; again we used the mean score for our analyses.

Possible confounders.

Age, sex, duration of knee symptoms, bilateral knee pain, co-occurrence of hip pain, and number of comorbidities were obtained by questionnaires at the 2-year followup visit. Duration of knee symptoms was categorized as follows: 1 year or less, 2–5 years, and more than 5 years. Co-occurrence of hip pain was assessed as follows: any pain, aching, or stiffness in the hip in the past 12 months, including pain in the groin and in the front and sides of the upper thigh. Bilateral knee pain was present when the NRS for self-reported knee pain during the last week was 1 or higher in both knees. Number of comorbidities was assessed by the Charlson Comorbidity Index (32).

Statistical analysis.

First, descriptives of the total study group were analyzed on clustering variables and possible confounders.

Second, a cluster analysis method was used to identify relatively homogeneous phenotypes in the knee OA population. Cluster analysis is a statistical technique that identifies subgroups of individuals that are similar to each other in certain clustering variables, but different from individuals in other groups. The technique aims to achieve the highest within-group homogeneity, as well as the lowest between-group homogeneity. K-means clustering is a specific cluster analysis method that does not require specifying the number of clusters in advance. A K-means algorithm will calculate the most optimal solution with K number of clusters (33).

The 4 clustering variables (radiographic OA, muscle strength, BMI, and depression) were standardized as

  • equation image

before cluster analysis because of differences in measurement unit (10). Cluster analysis techniques are sensitive to outliers (34); therefore, outliers were first identified (based on a principal component plot of the 4 clustering variables) and removed from the data set.

The FASTCLUS procedure, i.e., a K-means algorithm with a nearest centroid sorting method, from the SAS package (35) was used to perform cluster analysis. Determination of the optimal number of clusters (3 or more) in the data set was based on 2 recommended clustering criteria (36), both using a ratio of the sum of squares within clusters and the sum of squares between clusters: Calinski and Harabasz F (37), which is suggested to be the best clustering criterion (36) and frequently used in other studies (38, 39), and Beale's F (40, 41). A cluster solution of c clusters was the most optimal solution if both F values were the highest and if a significant increase (P < 0.05) in Beale's F after comparison with c − 1 clusters was found. Furthermore, validity of the cluster analysis was analyzed by randomly splitting the data set into 2 halves and repeating the same cluster analysis on each half. Similar (number of) clusters on each half indicates that the results of the cluster analysis were valid (33).

Third, we compared clinical outcomes of the identified phenotypes by univariable and stepwise multivariable regression analyses in SPSS, version 15.0. Univariable linear regression analyses were used to test the null hypothesis of similar clinical outcomes of the phenotypes, with a level of significance of 0.05. A significant regression coefficient indicates a statistical difference between 2 phenotypes. The independent variable in these analyses was the phenotype classification (by using dummy variables). The dependent variable was 1 of the 4 clinical outcome variables (NRS pain, WOMAC pain, WOMAC physical function, and 20-meter walking test). Multivariable linear regression analyses were used to test if possible confounders were (partly) responsible for differences in clinical outcome between phenotypes. Six possible confounders (age, sex, duration of knee symptoms, bilateral knee pain, co-occurrence of hip pain, and number of comorbidities) that may affect clinical outcome were added to the multivariable regression analysis one at a time. If the addition of a possible confounder to the model resulted in a 10% change (or more) of the regression coefficient, it was kept in the model (42). After correction for possible confounders, differences between phenotypes on the clinical outcome variable were statistically tested again with a level of significance of 0.05.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

Identification of phenotypes.

Table 1 displays the characteristics of the study sample. The mean ± SD age of the total group was 63.2 ± 9.1 years and 55% of the 842 included patients were women. Knee symptoms were present for more than 5 years in one-half of the persons (46%) and almost two-thirds of the study group had bilateral knee pain (62%). Radiographic evidence of knee OA (i.e., K/L score ≥2) was found in 84% of the persons, while 60% had evidence of moderate to severe knee OA (i.e., K/L score 3 or 4). Almost half of the persons were obese (46%) and 1 of 10 persons (10%) had mild or major symptoms of depression according to the CES-D.

Table 1. Characteristics of the study group*
 Value (n = 842)
  • *

    IQR = interquartile range; OA = osteoarthritis; K/L = Kellgren/Lawrence; BMI = body mass index; CES-D = Center for Epidemiologic Studies Depression Scale.

Age, mean ± SD years63.2 ± 9.1
Women, %55
Duration of knee symptoms, % 
 Less than 1 year15
 Between 2 and 5 years39
 More than 5 years46
Bilateral knee pain, % yes62
Co-occurrence of hip pain, % yes56
No. of comorbidities, median (IQR)0 (1)
Radiographic knee OA, % 
 K/L score 09
 K/L score 17
 K/L score 224
 K/L score 340
 K/L score 420
Muscle strength, mean ± SD N223.8 ± 89.0
BMI, mean ± SD kg/m229.9 ± 4.8
 Obesity (BMI ≥30.0 kg/m2), %46
CES-D score, median (IQR)5 (7.5)
 Depression (CES-D score >16), %10

We identified 1 outlier and removed this case from the data set. As shown in Table 2, F values of both clustering criteria were the highest in a 5-cluster solution, indicating that the optimal number of clusters in this data set was 5. Beale's F statistics also revealed that a 5-cluster solution showed a significant improvement of the model's fit compared to a 4-cluster solution, while a 6-cluster solution did not show a significant improvement of the model's fit compared to a 5-cluster solution.

Table 2. Clustering criteria for solution of 3 or more clusters
No. of clustersCalinski and Harabasz F (37)Cluster comparisonBeale's F (df numerator, df denominator) (40)Beale's F P (41)
  • *

    Five-cluster solution.

3183.01
4211.553 vs. 41.44 (65, 419)0.02
5*245.77*4 vs. 5*1.99 (45, 374)*< 0.001*
6217.605 vs. 60.61 (33, 341)0.96

Validity of the 5-cluster solution appeared to be good, as this solution was the most suitable solution in both (randomly split) halves of the data set as well. Furthermore, almost all of the subjects (99%) were allocated in similar clusters when cluster analysis was performed in one of both halves, compared to the cluster analysis of the original data set (data not shown).

Descriptives of the clustering variables of the 5 phenotypes and the total study group are shown in Table 3. The 5 phenotypes were characterized by: 1) mostly no or doubtful evidence of radiographic OA, average muscle strength, and low prevalence of obesity (26%) and depression (5%) (“minimal joint disease phenotype”); 2) high muscle strength, moderate to severe radiographic OA, obesity in 48% of participants, and low prevalence of depression (1%) (“strong muscle phenotype”); 3) muscle weakness, low prevalence of obesity (18%), moderate to severe radiographic OA, and low prevalence of depression (0%) (“nonobese and weak muscle phenotype”); 4) obesity in all participants (100%), muscle weakness, mild to moderate radiographic OA, and low prevalence of depression (1%) (“obese and weak muscle phenotype”); and 5) prevalence of depression in almost all participants (88%), mild to moderate radiographic OA, muscle weakness, and obesity in 58% of participants (“depressive phenotype”).

Table 3. Descriptives of phenotypes on clustering variables*
 “Minimal joint disease phenotype” (n = 140)“Strong muscle phenotype” (n = 189)“Nonobese and weak muscle phenotype” (n = 261)“Obese and weak muscle phenotype” (n = 168)“Depressive phenotype” (n = 83)Total study group (n = 841)
  • *

    OA = osteoarthritis; K/L = Kellgren/Lawrence; BMI = body mass index; CES-D = Center for Epidemiologic Studies Depression Scale; IQR = interquartile range.

Radiographic OA, %      
 K/L score 04910149
 K/L score 13520257
 K/L score 2162121392424
 K/L score 305247464740
 K/L score 402433122120
Muscle strength, mean ± SD N241.5 ± 76.9339.1 ± 63.5170.4 ± 43.2180.6 ± 53.6188.1 ± 72.7223.8 ± 89.0
BMI, mean ± SD kg/m227.5 ± 3.929.9 ± 3.427.0 ± 3.035.9 ± 3.531.0 ± 4.629.9 ± 4.8
 Obesity (BMI ≥30.0 kg/m2), %2648181005846
CES-D score, median (IQR)5 (6)3 (4.5)4 (6)5 (6)21 (7)5 (7.5)
 Depression (CES-D score >16), %51018810

The “strong muscle phenotype” was mostly male (6% women), while the “nonobese and weak muscle phenotype,” “obese and weak muscle phenotype,” and “depressive phenotype” were mostly female (ranging from 70–80%) and the “minimal joint disease phenotype” was mixed (49% women). The mean age in the “nonobese and weak muscle phenotype” was higher (67.2 years) than in the other phenotypes (ranging from 60.0–62.1 years). Long-lasting knee symptoms (symptom duration >5 years) were less present in the “minimal joint disease phenotype” (39%) compared to the other phenotypes (ranging from 45–52%). A majority of the persons from the “strong muscle phenotype” ever had knee surgery or arthroscopy (56%), while the prevalence of knee surgery or arthroscopy in other phenotypes ranged from 22–36%.

Differences in clinical outcomes.

Figure 1 shows the 4 clinical outcome variables of each of the 5 phenotypes. Overall, the “minimal joint disease phenotype” was clinically the least affected phenotype, followed by “strong muscle phenotype,” “nonobese and weak muscle phenotype,” “obese and weak muscle phenotype,” and “depressive phenotype.”

thumbnail image

Figure 1. Four clinical outcome variables of each of the 5 phenotypes. A, numerical rating scale for pain (range 0–10), B, Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain subscale (range 0–20), C, WOMAC physical function subscale (range 0–68), D, 20-meter walking test (seconds). 1 = “minimal joint disease phenotype”; 2 = “strong muscle phenotype”; 3 = “nonobese and weak muscle phenotype”; 4 = “obese and weak muscle phenotype”; 5 = “depressive phenotype.”

Download figure to PowerPoint

Univariable regression analyses revealed 33 significant differences between phenotypes on the 4 clinical outcome variables, while 7 comparisons were not significantly different (data not shown). In the multivariable regression analyses, 31 phenotype differences appeared to be significant after adjusting for possible confounders (age, sex, duration of knee symptoms, bilateral knee pain, co-occurrence of hip pain, and number of comorbidities) (Table 4). The clinically most affected phenotypes were the “obese and weak muscle phenotype” and the “depressive phenotype,” with significantly worse outcomes on both pain outcomes and both activity limitation outcomes than the other 3 phenotypes, as shown in Table 4. Furthermore, the “depressive phenotype” demonstrated significantly higher WOMAC pain scores than the “obese and weak muscle phenotype” as well (5.3 versus 4.5). The “minimal joint disease phenotype” demonstrated significantly lower pain levels and less severe activity limitations than the other phenotypes, except for the “strong muscle phenotype.”

Table 4. Clinical outcomes of phenotypes*
 “Minimal joint disease phenotype” (n = 140)“Strong muscle phenotype” (n = 189)“Nonobese and weak muscle phenotype” (n = 261)“Obese and weak muscle phenotype” (n = 168)“Depressive phenotype” (n = 83)Total study group (n = 841)
  • *

    NRS = numerical rating scale; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

  • Significantly worse clinical outcome compared to “minimal joint disease phenotype,” corrected for confounding (P < 0.05 for all).

  • Significantly worse clinical outcome compared to “minimal joint disease phenotype” and “strong muscle phenotype,” corrected for confounding (P < 0.05).

  • §

    Significantly worse clinical outcome compared to “minimal joint disease phenotype,” “strong muscle phenotype,” and “nonobese and weak muscle phenotype,” corrected for confounding (P < 0.05 for all).

  • Significantly worse clinical outcome compared to “minimal joint disease phenotype,” “strong muscle phenotype,” “nonobese and weak muscle phenotype,” and “obese and weak muscle phenotype,” corrected for confounding (P < 0.05).

Pain      
 NRS pain2.5 ± 2.42.8 ± 2.03.3 ± 2.24.3 ± 2.6§4.3 ± 2.6§3.4 ± 2.4
 WOMAC pain2.1 ± 2.22.6 ± 2.23.5 ± 3.04.5 ± 3.4§5.3 ± 3.93.4 ± 3.1
Activity limitations      
 WOMAC physical function6.8 ± 8.58.1 ± 8.111.3 ± 9.816.2 ± 12.0§17.4 ± 12.5§11.4 ± 10.7
 20-meter walking test, seconds15.1 ± 2.014.7 ± 1.916.2 ± 2.617.9 ± 3.4§17.3 ± 5.0§16.1 ± 3.1

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

It has been suggested that the heterogeneous knee OA population can be divided into clinical phenotypes (1, 3). As far as we know, we are the first to identify clinical phenotypes in a knee OA population. To identify phenotypes, we used a statistical technique, i.e., cluster analysis, that is designed to find meaningful homogeneous subgroups (10). We used 4 clinically relevant, easily available characteristics (radiographic OA, muscle strength, BMI, and depression) to identify phenotypes. Hypothetically, the 4 characteristics could have divided the study population into numerous phenotypes. Cluster analysis revealed that in reality, only 5 phenotypes exist.

Prior to the study, we hypothesized the existence of some particular phenotypes of knee OA patients. These particular phenotypes concerned a group of nonobese knee OA patients with high muscle strength, a group of knee OA patients with dynapenic obesity (i.e., combination of low muscle strength and high BMI) (22), and a group of depressive knee OA patients. Our study demonstrated that these phenotypes actually do exist in the knee OA population by identifying a “strong muscle phenotype” (although this phenotype showed average BMI), an “obese and weak muscle phenotype,” and a “depressive phenotype,” respectively. Furthermore, 2 other phenotypes of knee OA patients were identified: a “minimal joint disease phenotype” and a “nonobese and weak muscle phenotype.” This classification into 5 phenotypes reduced heterogeneity in the knee OA population, as clinical patient characteristics that were used for cluster analysis showed lower variations within each phenotype, compared to the variations in the total study group (Table 3). Almost all 5 phenotypes differed significantly from each other in their level of pain and activity limitations (Table 4). Furthermore, most of these phenotype differences are clinically relevant and meaningful as well, based on a minimum clinically important difference of 15% (43, 44). Differences between phenotypes had been analyzed based on person-specific outcomes on pain and activity limitations (i.e., mean score for both knees). Similar results were found when using knee-specific outcomes (data not shown).

The 5 identified phenotypes may represent different pathophysiologic etiologic subtypes of OA. In the “minimal joint disease phenotype,” structural disease is hardly present. Furthermore, knee symptom duration is lower than in the other 4 phenotypes, which could imply an early OA subtype. The “strong muscle phenotype” might be a subtype of severe radiographic OA patients (76% with moderate to severe radiographic OA) with posttraumatic OA who are still physically active, as shown by strong lower extremity muscles. In our study, a history of knee surgery or arthroscopy was indeed more common in this phenotype (56%) compared to other phenotypes (22–36%). The 2 weak muscle phenotypes might consist of physically inactive persons, in which inactivity has resulted in muscle weakness only (“nonobese and weak muscle phenotype”) or in dynapenic obesity (“obese and weak muscle phenotype”) (22). Because the age of persons from the “nonobese and weak muscle phenotype” was on average 6 years higher than all other phenotypes, this phenotype may represent an age-induced knee OA subtype, in which age-related changes in the musculoskeletal system (45) may have initiated knee OA development. The “obese and weak muscle phenotype” could represent a biomechanically-induced knee OA subtype, in which overload of the knee joint may have caused OA (14, 16, 46). In addition, metabolic changes that are linked to obesity could have played a role in the knee OA etiology of this subtype, as demonstrated by recent literature (46). Finally, persons of the “depression phenotype,” who are demonstrating the highest levels of pain and activity limitations, might represent a subtype of knee OA persons in which depression has resulted in an intensified focus on pain-related stimuli (cognitive effect), increased sensitization of the central nervous system on pain stimuli (neurophysiologic effect), and physical inactivity (behavioral effect) (47). These hypotheses of possible knee OA subtypes need further support from evidence.

The phenotype differences in all 4 clinical outcomes in our study may underscore the well-known importance of knee OA risk factors for clinical practice. The presence of only one risk factor (i.e., severe radiographic OA in “strong muscle phenotype” and low muscle strength in “nonobese and weak muscle phenotype”) or no risk factors at all (i.e., “minimal joint disease phenotype”) seems to result in only mild pain levels and few activity limitations. On the other hand, the presence of multiple knee OA risk factors (i.e., obesity and muscle weakness in “obese and weak muscle phenotype”) is likely to have a stronger effect on clinical outcome. Psychosocial factors seem to have even larger effects on clinical outcome in knee OA, as indicated by the outcomes of the “depressive phenotype.”

We would like to make a number of comments on our study. First, we demonstrated that the identification of 5 phenotypes was a valid result in our data set (i.e., the same results were obtained in randomly split halves of the data set). Because our study population might not be generalizable to other knee OA populations, studies in other data sets are needed to replicate our findings. Additionally, it would be interesting to include healthy age- and sex-matched controls in these studies, to be able to compare phenotype characteristics with normative scores. Second, although we explained some of the heterogeneity of the knee OA population by identifying phenotypes, there is still overlap between phenotypes, as shown by SDs in clustering variables and clinical outcome variables. Third, we identified phenotypes based on similarities in 4 patient-specific characteristics, specifically radiographic severity of OA, muscle strength, BMI, and depression. We chose these characteristics based on their clinical relevance. Also, other patient characteristics could be chosen, for example, biomechanical (e.g., alignment, laxity, proprioceptive accuracy), psychosocial (e.g., anxiety, self-efficacy), or genetic factors. These characteristics could have explained some of the variation in clinical outcome within the identified phenotypes in our study. Furthermore, knee-specific or compartment-specific variables (for example, features from radiography or magnetic resonance imaging [MRI]) might be appropriate to identify meaningful phenotypes as well. MRI features may be preferred over radiographic features, since the former have been shown to be more strongly associated with clinical outcome than the latter (48). Fourth, we included persons from the OAI study, a longitudinal cohort study. These persons could have had treatments prior to or during the OAI followup visits. We did not adjust for differences in treatments between phenotypes because of “bias by indication,” i.e., treatment application depends on clinical patient characteristics that have been used for cluster analysis. Consequently, it is not feasible to correct for differences in treatment. Finally, no statements can be made on causal relationships between phenotype characteristics and clinical outcome because of the cross-sectional design of our study.

Our classification into 5 phenotypes of knee OA patients may have major implications for clinical practice. Because the knee OA population is heterogeneous, interventions may need to be tailored to specific homogeneous phenotypes of knee OA patients to optimize overall effectiveness of knee OA interventions (1–3) in a cost-effective manner (8). International guidelines recommend a combination of pharmacologic and nonpharmacologic interventions for the general knee OA population (7). Each of the 5 identified phenotypes may need to receive different tailored interventions. For example, patients from the “depressive phenotype” may benefit most from a combination of interventions, in which an educational program, a cognitive–behavioral therapy, and an exercise therapy with graded activity levels are included (47), while in some patients the additional use of antidepressants may also be indicated. Patients from the “obese and weak muscle phenotype” may need an intervention that focuses on both weight loss and muscle strengthening (49), while muscle strengthening exercises only could be sufficient for persons from the “nonobese and weak muscle phenotype.” Most of the persons from the “strong muscle phenotype” are not likely to benefit from exercise therapy. These persons, diagnosed with severe radiographic OA, might be candidates for orthopedic surgery if pain and activity limitations are severe. Finally, persons from the “minimal joint disease phenotype” may not be in need of specific treatment, although an educational program and instructions for home-based exercises could have beneficial effects for long-term clinical outcomes. Knowledge on tailored interventions is limited and more research will be needed to evaluate differences in response to specific interventions in various phenotypes and to evaluate the effectiveness of tailored interventions in specific phenotypes.

To conclude, 5 phenotypes can be identified in the heterogeneous population of knee OA patients based on clinically relevant patient characteristics. These phenotypes showed different clinical outcomes. Interventions may need to be tailored to these phenotypes.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Mr. Knoop had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Knoop, Steultjens, Dekker.

Acquisition of data. Knoop, Knol.

Analysis and interpretation of data. Knoop, van der Leeden, Thorstensson, Roorda, Lems, Knol, Steultjens, Dekker.

ROLE OF THE STUDY SPONSOR

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES

The Osteoarthritis Initiative (OAI) is funded by the National Institutes of Health, a branch of the Department of Health and Human Services. Private funding partners of the OAI include Merck Research Laboratories, Novartis Pharmaceuticals Corporation, GlaxoSmithKline, and Pfizer, Inc. The study sponsors had no involvement in the present study.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. ROLE OF THE STUDY SPONSOR
  9. REFERENCES
  • 1
    Driban JB, Sitler MR, Barbe MF, Balasubramanian E. Is osteoarthritis a heterogeneous disease that can be stratified into subsets? Clin Rheumatol 2010; 29: 12331.
  • 2
    Hinman RS, Crossley KM. Patellofemoral joint osteoarthritis: an important subgroup of knee osteoarthritis. Rheumatology (Oxford) 2007; 46: 105762.
  • 3
    Felson DT. Identifying different osteoarthritis phenotypes through epidemiology. Osteoarthritis Cartilage 2010; 18: 6014.
  • 4
    Dieppe PA, Cushnaghan J, Shepstone L. The Bristol ‘OA500’ study: progression of osteoarthritis (OA) over 3 years and the relationship between clinical and radiographic changes at the knee joint. Osteoarthritis Cartilage 1997; 5: 8797.
  • 5
    Maly MR, Costigan PA, Olney SJ. Mechanical factors relate to pain in knee osteoarthritis. Clin Biomech (Bristol, Avon) 2008; 23: 796805.
  • 6
    Dekker J, van Dijk GM, Veenhof C. Risk factors for functional decline in osteoarthritis of the hip or knee. Curr Opin Rheumatol 2009; 21: 5204.
  • 7
    Zhang W, Nuki G, Moskowitz RW, Abramson S, Altman RD, Arden NK, et al. OARSI recommendations for the management of hip and knee osteoarthritis: part III. Changes in evidence following systematic cumulative update of research published through January 2009. Osteoarthritis Cartilage 2010; 18: 47699.
  • 8
    Bitton R. The economic burden of osteoarthritis. Am J Manag Care 2009; 15 Suppl: S2305.
  • 9
    Lohmander LS, Felson D. Can we identify a ‘high risk’ patient profile to determine who will experience rapid progression of osteoarthritis? Osteoarthritis Cartilage 2004; 12 Suppl: S4952.
  • 10
    Everitt BS, Landau S, Leese M. Cluster analysis. 4th ed. London: Arnold; 2001.
  • 11
    Bedson J, Croft PR. The discordance between clinical and radiographic knee osteoarthritis: a systematic search and summary of the literature. BMC Musculoskelet Disord 2008; 9: 116.
  • 12
    Neogi T, Felson D, Niu J, Nevitt M, Lewis CE, Aliabadi P, et al. Association between radiographic features of knee osteoarthritis and pain: results from two cohort studies. BMJ 2009; 339: b2844.
  • 13
    Steultjens MP, Dekker J, van Baar ME, Oostendorp RA, Bijlsma JW. Muscle strength, pain and disability in patients with osteoarthritis. Clin Rehabil 2001; 15: 33141.
  • 14
    Slemenda C, Brandt KD, Heilman DK, Mazzuca S, Braunstein EM, Katz BP, et al. Quadriceps weakness and osteoarthritis of the knee. Ann Intern Med 1997; 127: 97104.
  • 15
    Segal NA, Glass NA, Felson DT, Hurley M, Yang M, Nevitt M, et al. Effect of quadriceps strength and proprioception on risk for knee osteoarthritis. Med Sci Sports Exerc 2010; 42: 20818.
  • 16
    Thorstensson CA, Petersson IF, Jacobsson LT, Boegard TL, Roos EM. Reduced functional performance in the lower extremity predicted radiographic knee osteoarthritis five years later. Ann Rheum Dis 2004; 63: 4027.
  • 17
    Maly MR, Costigan PA, Olney SJ. Contribution of psychosocial and mechanical variables to physical performance measures in knee osteoarthritis. Phys Ther 2005; 85: 131828.
  • 18
    Felson DT, Anderson JJ, Naimark A, Walker AM, Meenan RF. Obesity and knee osteoarthritis: the Framingham Study. Ann Intern Med 1988; 109: 1824.
  • 19
    Rogers MW, Wilder FV. The association of BMI and knee pain among persons with radiographic knee osteoarthritis: a cross-sectional study. BMC Musculoskelet Disord 2008; 9: 163.
  • 20
    Reijman M, Pols HA, Bergink AP, Hazes JM, Belo JN, Lievense AM, et al. Body mass index associated with onset and progression of osteoarthritis of the knee but not of the hip: the Rotterdam Study. Ann Rheum Dis 2007; 66: 15862.
  • 21
    Axford J, Butt A, Heron C, Hammond J, Morgan J, Alavi A, et al. Prevalence of anxiety and depression in osteoarthritis: use of the Hospital Anxiety and Depression Scale as a screening tool. Clin Rheumatol 2010; 29: 127783.
  • 22
    Bouchard DR, Janssen I. Dynapenic-obesity and physical function in older adults. J Gerontol A Biol Sci Med Sci 2010; 65: 717.
  • 23
    Altman RD, Hochberg M, Murphy WA Jr, Wolfe F, Lequesne M. Atlas of individual radiographic features in osteoarthritis. Osteoarthritis Cartilage 1995; 3 Suppl: 370.
  • 24
    Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis 1957; 16: 494502.
  • 25
    Rantanen T, Era P, Heikkinen E. Maximal isometric strength and mobility among 75-year-old men and women. Age Ageing 1994; 23: 1327.
  • 26
    Obesity: preventing and managing the global epidemic: report of a WHO consultation. World Health Organ Tech Rep Ser 2000; 894: 1253.
  • 27
    Weissman MM, Sholomskas D, Pottenger M, Prusoff BA, Locke BZ. Assessing depressive symptoms in five psychiatric populations: a validation study. Am J Epidemiol 1977; 106: 20314.
  • 28
    Zich JM, Attkisson CC, Greenfield TK. Screening for depression in primary care clinics: the CES-D and the BDI. Int J Psychiatry Med 1990; 20: 25977.
  • 29
    Turk DC, Melzack R. Handbook of pain assessment. 2nd ed. New York: Guilford Press; 2001.
  • 30
    Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1988; 15: 183340.
  • 31
    Hochberg MC, Altman RD, Brandt KD, Moskowitz RW. Design and conduct of clinical trials in osteoarthritis: preliminary recommendations from a task force of the Osteoarthritis Research Society. J Rheumatol 1997; 24: 7924.
  • 32
    Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40: 37383.
  • 33
    Clatworthy J, Buick D, Hankins M, Weinman J, Horne R. The use and reporting of cluster analysis in health psychology: a review. Br J Health Psychol 2005; 10: 32958.
  • 34
    Jain AK, Dubes RC. Algorithms for clustering data. 1st ed. Englewood Cliffs (NJ): Prentice-Hall; 1988.
  • 35
    SAS Institute. SAS/STAT(R) user's guide. 2nd ed. 2008.
  • 36
    Milligan GW, Cooper MC. An examination of procedures for determining the number of clusters in a data set. Psychometrika 1985; 50: 15979.
  • 37
    Calinski RB, Harabasz J. A dendrite method for cluster analysis. Commun Stat 1974; 3: 127.
  • 38
    Nouwen A, Gingras J, Talbot F, Bouchard S. The development of an empirical psychosocial taxonomy for patients with diabetes. Health Psychol 1997; 16: 26371.
  • 39
    Patterson RE, Haines PS, Popkin BM. Health lifestyle patterns of U.S. adults. Prev Med 1994; 23: 45360.
  • 40
    Beale EM. Euclidean cluster analysis. Bull Int Stat Inst 1969; 43: 924.
  • 41
    Johnson DE. Applied multivariate methods for data analysis. Pacific Grove (CA): Brooks/Cole Publishing Company; 1998.
  • 42
    Kleinbaum DG, Kupper LL, Muller KE. Applied regression analysis and other multivariable methods. Boston: PWS-Kent Publishing Company; 1988.
  • 43
    Salaffi F, Stancati A, Silvestri CA, Ciapetti A, Grassi W. Minimal clinically important changes in chronic musculoskeletal pain intensity measured on a numerical rating scale. Eur J Pain 2004; 8: 28391.
  • 44
    Angst F, Aeschlimann A, Stucki G. Smallest detectable and minimal clinically important differences of rehabilitation intervention with their implications for required sample sizes using WOMAC and SF-36 quality of life measurement instruments in patients with osteoarthritis of the lower extremities. Arthritis Rheum 2001; 45: 38491.
  • 45
    Loeser RF. Age-related changes in the musculoskeletal system and the development of osteoarthritis. Clin Geriatr Med 2010; 26: 37186.
  • 46
    Sowers MR, Karvonen-Gutierrez CA. The evolving role of obesity in knee osteoarthritis. Curr Opin Rheumatol 2010; 22: 5337.
  • 47
    Edwards RR, Calahan C, Mensing G, Smith M, Haythornthwaite JA. Pain, catastrophizing, and depression in the rheumatic diseases. Nat Rev Rheumatol 2011; 7: 21624.
  • 48
    Yusuf E, Kortekaas MC, Watt I, Huizinga TW, Kloppenburg M. Do knee abnormalities visualised on MRI explain knee pain in knee osteoarthritis? A systematic review. Ann Rheum Dis 2011; 70: 607.
  • 49
    Christensen R, Bartels EM, Astrup A, Bliddal H. Effect of weight reduction in obese patients diagnosed with knee osteoarthritis: a systematic review and meta-analysis. Ann Rheum Dis 2007; 66: 4339.