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. Acknowledgements
  9. REFERENCES

Objective

To predict the 2-year course of activity limitations in patients with early knee and/or hip osteoarthritis (OA).

Methods

The Cohort Hip & Cohort Knee (CHECK) study is a prospective followup study. The CHECK cohort, comprising participants (n = 1,002) with early OA-related knee and/or hip symptoms, was followed for 2 years. Participants completed questionnaires and underwent physical, laboratory, and radiographic examination. Regression models were used to examine whether baseline variables predicted the course of activity limitations as measured with the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Analyses were performed separately for participants with knee symptoms and participants with hip symptoms.

Results

After 2 years of followup, activity limitations slightly decreased. Large between-subject variation was observed in WOMAC change scores. In participants with knee symptoms, young age, non-Western ethnicity, bilateral hip pain, morning stiffness in the knee, high comorbidity count, high body mass index, high bodily pain, poor general health perception, and pain coping strategy were associated with a poor 2-year outcome on activity limitations. In participants with hip symptoms, few activity limitations at baseline, bilateral hip pain, morning stiffness in the knee, high comorbidity count, low active hip flexion, poor general health perception, and pain coping strategy were associated with a poor 2-year outcome on activity limitations.

Conclusion

After 2 years of followup, large between-subject variation was observed in the course of activity limitations. The course of activity limitations is to some extent already predictable at an early stage of knee and hip OA.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Osteoarthritis (OA) of the knee and/or hip is a leading cause of activity limitations in Western countries (1–3). Activity limitations are defined as difficulties an individual may have in executing activities in the International Classification of Functioning, Disability and Health (4). The course of activity limitations differs considerably between patients; some patients seem to be stable or even improve, whereas others deteriorate (5–7). Several sociodemographic factors, comorbidities, physical impairments, psychological factors, social factors, and health behaviors have been linked to an increase in activity limitations in patients with OA (8). However, these factors have been found in populations with established OA. Less research has been directed at early OA. Therefore, in early OA, less is known about factors affecting activity limitations. This makes it difficult for physicians to distinguish between patients with poor and good prognoses at an early stage of the disease. As a result, many patients with early OA are not sufficiently informed about their likely course of activity limitations, and are not referred for appropriate treatment (9, 10). This is unfortunate, because developing activity limitations is one of the main fears of patients with OA (9).

To reduce this uncertainty about the course of activity limitations for both physicians and patients, identification of factors affecting activity limitations is needed. This would enable physicians to identify patients at risk for developing activity limitations and to select patients eligible for treatment. The most efficient prevention efforts in OA may well target patients with early disease (11). Therefore, it is important to investigate whether, already at an early stage of knee and hip OA, predictors for the course of activity limitations can be identified.

The purpose of the present study was to predict the 2-year course of activity limitations in patients with early knee and/or hip OA. To enhance the pragmatic applicability of the study results, the focus was on predictors that are routinely measured or easily implemented by physicians at an early stage of the disease.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Study design and population.

The Cohort Hip & Cohort Knee (CHECK) study is a prospective followup study of 1,002 individuals with early symptoms of the knee and/or hip (12). The CHECK cohort was formed from October 2002 until September 2005. At baseline, the majority of participants (83%) reported knee symptoms (12), of whom 76% fulfilled the American College of Rheumatology (ACR) criteria for the classification of knee OA (13). Hip symptoms were reported by 59% of the participants (12), of whom 24% fulfilled the ACR criteria for the classification of hip OA (14). Two-year followup data show an increase in these percentages, as well as an increase in radiologic signs (Wesseling J, et al: unpublished observations). Therefore, the CHECK cohort can be considered an early OA cohort, but definite description can only be made at a later stage. Nationwide in The Netherlands, 10 general and academic hospitals located in urbanized and semiurbanized regions are participating. Participants were referred by general practitioners, referred from the outpatient clinics of the participating centers, or recruited through advertisements in local newspapers and on the Web site of the Dutch Arthritis Association. During the baseline measurements, the physicians in the participating centers checked whether each patient fulfilled the inclusion criteria. Individuals were eligible for inclusion if they had pain and/or stiffness in the knee and/or hip (Figure 1), were ages 45–65 years, and had never or ≤6 months ago visited a general practitioner for these symptoms for the first time. Exclusion criteria were any other pathologic condition that could explain the symptoms, comorbidity that did not allow physical evaluation and/or followup of ≥10 years, malignancy in the last 5 years, and inability to understand the Dutch language. Based on symptoms, a knee stratum (n = 832) and a hip stratum (n = 588) were defined. Participants with both knee and hip symptoms were included in both strata. The study was approved by the medical ethics committees of all participating centers, and all participants gave their written informed consent before entering the study.

thumbnail image

Figure 1. Participants were excluded if they only experienced pain in ≥1 of the marked areas in the left pain drawing. For inclusion, participants had to experience pain in ≥1 of the marked areas in the right pain drawing. (a) = iliac crest; (b) = ischial tuberosity; (c) = anterior superior iliac spine; (d) = pubic tubercle; (e) = one-third of the way down the spine; (f) = cranial border of the patella; (g) = tibial tuberosity.

Download figure to PowerPoint

Outcome measurement.

Activity limitations were assessed with the physical functioning (PF) scale of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (15). The WOMAC-PF is a disease-specific, self-administered instrument consisting of 17 items. Scores range from 0 to 68, with higher scores indicating more activity limitations. The WOMAC-PF is widely used in clinical research and has been shown to be reliable, valid, and responsive for use in patients with OA (15–17).

The primary outcome, a dichotomous measure of activity limitations, was constructed according to Sharma et al (7). Baseline and 2-year WOMAC-PF scores were categorized into quintile groups (knee stratum: 0–5, 6–11, 12–18, 19–26, and 27–68; hip stratum: 0–6, 7–12, 13–19, 20–27, and 28–68). For each participant, a poor outcome on activity limitations was defined as moving into a higher group or remaining within the 3 highest groups after 2 years. A good outcome on activity limitations was defined as moving into a lower group or remaining in one of the 2 lowest groups after 2 years. In addition, a continuous outcome measure of activity limitations was calculated by subtracting the baseline WOMAC-PF scores from the 2-year scores.

Potential prognostic factors measured at baseline.

Based on clinical applicability, biologic plausibility, expert opinions (physiatrists, rheumatologists, a general practitioner, and expert researchers in the field were consulted), and the literature, the prognostic value of 47 variables for activity limitations in knee symptoms and 41 variables for activity limitations in hip symptoms was studied. Baseline characteristics were categorized into 7 groups: demographics, symptoms, comorbidity and interventions, participation and lifestyle, physical examination, laboratory/radiographic examination, and self-report questionnaires. The order in which these 7 groups of variables were studied was based on the usual clinical routine of the physician. According to this routine, information about demographics, symptoms, comorbidity, interventions, and participation and lifestyle were obtained from patient history. Therefore, the prognostic value of these factors was examined first. After anamnesis a physical examination was performed, followed by laboratory and radiographic examination. Because questionnaires are not routinely assessed in clinical practice, the prognostic value of variables measured with questionnaires was examined last. All variables assessed are listed in Tables 1 and 2.

Table 1. Baseline characteristics of the knee stratum (n = 832) and results of the prognostic factor selection steps 1 and 2 to select potential prognostic factors for a poor 2-year outcome on activity limitations*
 ValuePoor outcome on activity limitations
Prognostic factor selection step 1, OR (95% CI)Prognostic factor selection step 2, OR (95% CI)
  • *

    Values are the median (interquartile range) unless otherwise indicated. Prognostic factor selection steps 1 and 2 are described in the Statistical analysis section. References for the measurement instruments are indicated in the first column. OR = odds ratio; 95% CI = 95% confidence interval; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; NRS = numeric rating scale; ESR = erythrocyte sedimentation rate; K/L = Kellgren/Lawrence; SF-36 = Short Form 36 health survey; EQ-5D = EuroQol 5 domains; VAS = visual analog scale.

  • P < 0.2.

  • Categories are based on the definition of ethnic minorities of Statistics Netherlands, whereby Western immigrants of Indonesian and Japanese origin are analyzed as a separate group.

Block 0: WOMAC function (15) (range 0–68)14 (7–23)1.03 (1.02, 1.05) 
Block 1: demographics   
 Age, mean ± SD years56.0 ± 5.10.97 (0.96, 0.98)0.97 (0.94, 1.00)
 Women, no. (%)663 (79.7)1.14 (0.80, 1.62) 
 Ethnicity, no. (%)   
  Native or immigrant of Western origin775 (93.1)ReferenceReference
  Immigrant of non-Western origin16 (1.9)4.21 (1.18, 14.98)3.77 (1.05, 13.54)
  Western immigrant of Indonesian origin20 (2.4)1.59 (0.65, 3.94)1.80 (0.71, 4.54)
 Household composition >1, no. (%)706 (84.9)0.83 (0.54, 1.28) 
 Educational level, no. (%)   
  Primary school23 (2.8)ReferenceReference
  Secondary school571 (68.6)0.49 (0.20, 1.19)0.46 (0.19, 1.14)
  Higher professional education/university214 (25.7)0.32 (0.13, 0.81)0.29 (0.11, 0.75)
Block 2: symptoms   
 Knee pain, no. (%)   
  Unilateral370 (44.5)Reference 
  Bilateral with index knee344 (41.3)1.17 (0.86, 1.59) 
  Bilateral with equal symptoms118 (14.2)1.32 (0.85, 2.04) 
 Hip pain, no. (%)   
  No pain414 (49.8)ReferenceReference
  Unilateral260 (31.3)1.38 (1.00, 1.90)1.34 (0.92, 1.95)
  Bilateral with index hip113 (13.6)1.80 (1.14, 2.84)1.74 (1.05, 2.89)
  Bilateral with equal symptoms45 (5.4)3.16 (1.59, 6.31)2.89 (1.40, 5.97)
 NRS for pain intensity during the past week (range 0–10)3 (2–5)1.15 (1.07, 1.23)1.11 (1.03, 1.21)
 Pain during sitting/lying (WOMAC item) (15), no. (%)   
  None/slight603 (72.5)ReferenceReference
  Moderate162 (19.5)1.40 (0.98, 2.01)1.02 (0.69, 1.52)
  Severe/extreme44 (5.3)1.68 (0.89, 3.16)1.06 (0.52, 2.13)
 Morning stiffness knee <30 minutes (13)518 (62.3)1.69 (1.26, 2.27)1.57 (1.14, 2.16)
 Morning stiffness hip ≤60 minutes (14), no. (%)247 (29.7)1.51 (1.11, 2.05)0.99 (0.67, 1.47)
Block 3: comorbidity and interventions   
 Comorbidity count, no. (%)   
  0210 (25.2)ReferenceReference
  1251 (30.2)1.18 (0.81, 1.71)1.14 (0.78, 1.67)
  2177 (21.3)1.54 (1.02, 2.31)1.49 (0.98, 2.26)
  ≥3177 (21.3)2.63 (1.71, 4.03)2.28 (1.45, 3.58)
 Body mass index, kg/m225.6 (23.6–28.4)1.08 (1.04, 1.12)1.07 (1.03, 1.11)
 Use of pain medication, no. (%)309 (37.1)1.51 (1.13, 2.02)1.24 (0.91, 1.70)
 Knee or hip surgery during past year, no. (%)6 (0.7)0.65 (0.10, 3.99) 
Block 4: participation and lifestyle   
 Paid employment, no. (%)419 (50.4)0.79 (0.59, 1.04)0.77 (0.58, 1.02)
 Demanding physical work often/always, no. (%)64 (7.7)1.30 (0.76, 2.20) 
 Physical activity during leisure, no. (%)   
  0–2 days per week381 (45.8)ReferenceReference
  3–5 days per week315 (37.9)0.81 (0.60, 1.11)0.80 (0.59, 1.10)
  6–7 days per week106 (12.7)0.75 (0.48, 1.16)0.74 (0.47, 1.15)
 Weekday alcohol consumption, no. (%)   
  0 glasses244 (29.3)Reference 
  1–3 glasses543 (65.3)0.82 (0.60, 1.12) 
  ≥4 glasses22 (2.6)0.68 (0.28, 1.69) 
 Tobacco use, no. (%)119 (14.3)1.33 (0.88, 1.99)1.30 (0.86, 1.96)
Block 5: physical examination   
 Knee flexion, degrees135 (130–140)0.99 (0.97, 1.00)0.99 (0.98, 1.01)
 Pain during knee flexion, no. (%)   
  None/slight744 (89.4)ReferenceReference
  Severe/extreme81 (9.7)1.55 (0.95, 2.53)1.43 (0.86, 2.38)
 Knee extension, degrees2 (0–5)1.00 (0.95, 1.04) 
 Pain during knee extension, no. (%)   
  None/slight781 (93.3)Reference 
  Severe/extreme44 (5.3)1.42 (0.74, 2.73) 
 Palpable warmth, no. (%)46 (5.5)0.78 (0.42, 1.45) 
 Crepitus during knee flexion, no. (%)392 (47.1)0.96 (0.73, 1.27) 
 Pain on palpation of the joint line, no. (%)374 (45.0)1.52 (1.16, 2.01)1.47 (1.10, 1.95)
 Bony enlargement, no. (%)36 (4.3)1.41 (0.70, 2.84) 
 Positive refill test, no. (%)61 (7.3)0.53 (0.31, 0.93)0.45 (0.25, 0.78)
 Pain on patellofemoral joint compression, no. (%)235 (28.2)1.39 (1.02, 1.90)1.32 (0.96, 1.81)
 Presence of Heberden's nodes, no. (%)409 (49.2)1.10 (0.83, 1.45) 
Block 6: laboratory/radiographic examination   
 ESR, mm/hour8 (5–14)1.01 (0.99, 1.03) 
 K/L knee score (19), no. (%)   
  0566 (68.0)ReferenceReference
  1200 (24.0)1.14 (0.82, 1.59)1.14 (0.82, 1.59)
  2 or 359 (7.1)1.53 (0.87, 2.68)1.53 (0.87, 2.68)
Block 7: self-report questionnaires   
 SF-36 scores (27, 28), mean ± SD (range 0–100)   
  Bodily pain67.3 ± 17.70.97 (0.97, 0.98)0.98 (0.97, 0.99)
  Vitality63.9 ± 16.80.98 (0.97, 0.99)1.00 (0.99, 1.01)
  Mental health76.5 ± 14.70.99 (0.98, 1.00)0.99 (0.98, 1.01)
  General health53.3 ± 18.30.98 (0.97, 0.99)0.99 (0.98, 1.00)
 EQ-5D (29, 30)   
  Item on anxiety/depression, no. (%)   
   Not anxious/depressed658 (79.1)ReferenceReference
   Anxious/depressed153 (18.4)1.37 (0.94, 1.99)0.97 (0.58, 1.62)
  VAS for health (range 0–100)78.0 (70.0–85.0)0.97 (0.96, 0.99)1.00 (0.98, 1.01)
 Social support (31) (range 12–60)15 (12–22)1.03 (1.01, 1.04)1.01 (0.99, 1.03)
 Pain Coping Inventory (32)   
  Distraction (range 5–20)11 (9–13)1.07 (1.03, 1.12)1.04 (0.98, 1.10)
  Transformation (range 4–16)8 (7–10)1.09 (1.03, 1.15)1.05 (0.99, 1.12)
  Reducing demands (range 3–12)6 (5–7)1.05 (0.97, 1.14) 
  Resting (range 5–20)9 (7–11)1.10 (1.04, 1.16)1.05 (0.97, 1.13)
  Retreating (range 7–28)10 (8–13)1.03 (0.99, 1.07)0.98 (0.93, 1.03)
  Worrying (range 9–36)13 (11–16)1.03 (1.00, 1.07)0.95 (0.90, 1.00)
Table 2. Baseline characteristics of the hip stratum (n = 588) and results of the prognostic factor selection steps 1 and 2 to select potential prognostic factors for a poor 2-year outcome on activity limitations*
 ValuePoor outcome on activity limitations
Prognostic factor selection step 1, OR (95% CI)Prognostic factor selection step 2, OR (95% CI)
  • *

    Values are the median (interquartile range) unless otherwise indicated. Prognostic factor selection steps 1 and 2 are described in the Statistical analysis section. References for the measurement instruments are indicated in the first column. See Table 1 for definitions.

  • P < 0.2.

  • Categories are based on the definition of ethnic minorities of Statistics Netherlands, whereby Western immigrants of Indonesian and Japanese origin are analyzed as a separate group.

Block 0: WOMAC function (15) (range 0–68)16 (8–25)1.01 (1.00, 1.03) 
Block 1: demographics   
 Age, mean ± SD years55.8 ± 5.30.98 (0.95, 1.01)0.98 (0.95, 1.01)
 Women, no. (%)475 (80.8)1.00 (0.65, 1.53) 
 Ethnicity, no. (%)   
  Native or immigrant of Western origin557 (94.7)ReferenceReference
  Immigrant of non-Western origin5 (0.9)0.95 (0.12, 7.76)0.97 (0.11, 8.22)
  Western immigrant of Indonesian origin12 (2.0)2.81 (0.75, 10.62)2.86 (0.75, 10.95)
 Household composition >1, no. (%)499 (84.9)0.97 (0.59, 1.61) 
 Educational level, no. (%)   
  Primary school11 (1.9)ReferenceReference
  Secondary school407 (69.2)0.52 (0.13, 2.06)0.47 (0.12, 1.88)
  Higher professional education/university155 (26.4)0.38 (0.09, 1.55)0.33 (0.08, 1.38)
Block 2: symptoms   
 Knee pain, no. (%)   
  No pain170 (28.9)ReferenceReference
  Unilateral182 (31.0)1.32 (0.86, 2.03)1.03 (0.63, 1.66)
  Bilateral with index knee178 (30.3)1.57 (1.01, 2.44)1.00 (0.59, 1.69)
  Bilateral with equal symptoms58 (9.9)1.36 (0.74, 2.49)0.76 (0.38, 1.51)
 Hip pain, no. (%)   
  Unilateral380 (64.6)ReferenceReference
  Bilateral with index hip159 (27.0)1.21 (0.82, 1.78)1.23 (0.82, 1.86)
  Bilateral with equal symptoms49 (8.3)2.32 (1.21, 4.43)2.31 (1.16, 4.61)
 NRS for pain intensity during the past week (range 0–10)4 (2–5)1.05 (0.97, 1.14) 
 Pain during sitting/lying (WOMAC item) (15), no. (%)   
  None/slight387 (65.8)Reference 
  Moderate147 (25.0)0.93 (0.63, 1.37) 
  Severe/extreme38 (6.5)1.15 (0.57, 2.31) 
 Morning stiffness knee <30 minutes (13), no. (%)274 (46.6)2.02 (1.43, 2.84)1.89 (1.24, 2.90)
 Morning stiffness hip ≤60 minutes (14), no. (%)326 (55.4)1.42 (1.02, 1.98)1.22 (0.85, 1.76)
Block 3: comorbidity and interventions   
 Comorbidity count, no. (%)   
  0142 (24.1)ReferenceReference
  1170 (28.9)1.29 (0.81, 2.04)1.24 (0.78, 1.98)
  2113 (19.2)1.59 (0.95, 2.66)1.46 (0.86, 2.46)
  ≥3152 (25.9)2.85 (1.75, 4.63)2.51 (1.52, 4.13)
 Body mass index, kg/m225.4 (23.2–28.2)1.07 (1.02, 1.12)1.06 (1.01, 1.10)
 Use of pain medication, no. (%)250 (42.5)1.54 (1.09, 2.16)1.29 (0.90, 1.85)
 Knee or hip surgery during past year, no. (%)7 (1.2)1.54 (0.30, 7.84) 
Block 4: participation and lifestyle   
 Paid employment, no. (%)296 (50.3)0.74 (0.53, 1.04)0.71 (0.50, 1.00)
 Demanding physical work, often/always, no. (%)54 (9.2)0.88 (0.49, 1.58) 
 Physical activity during leisure, no. (%)   
  0–2 days per week254 (43.2)ReferenceReference
  3–5 days per week233 (39.6)0.70 (0.48, 1.01)0.67 (0.45, 0.98)
  6–7 days per week83 (14.1)0.72 (0.43, 1.19)0.68 (0.41, 1.14)
 Weekday alcohol consumption, no. (%)   
  0 glasses157 (26.7)ReferenceReference
  1–3 glasses399 (67.9)0.75 (0.51, 1.10)0.77 (0.52, 1.14)
  ≥4 glasses18 (3.1)1.13 (0.41, 3.12)1.11 (0.40, 3.09)
 Tobacco use, no. (%)80 (13.6)1.25 (0.77, 2.03) 
Block 5: physical examination   
 Hip internal rotation, degrees30 (23–35)0.99 (0.98, 1.01) 
 Pain during hip internal rotation, no. (%)   
  None/slight460 (78.2)Reference 
  Severe/extreme125 (21.3)1.15 (0.76, 1.76) 
 Hip flexion, degrees118 (110–125)0.98 (0.96, 0.99)0.98 (0.96, 0.99)
 Pain during hip flexion, no. (%)   
  None/slight468 (79.6)ReferenceReference
  Severe/extreme113 (19.2)1.36 (0.89, 2.07)1.15 (0.74, 1.78)
 Presence of Heberden's nodes, no. (%)296 (50.3)0.94 (0.68, 1.32) 
Block 6: laboratory/radiographic examination   
 ESR, mm/hour8 (5–13)1.00 (0.98, 1.02) 
 K/L hip score (19)   
  0454 (77.2)Reference 
  187 (14.8)1.23 (0.74, 2.02) 
  2 or 344 (7.5)0.92 (0.48, 1.76) 
Block 7: self-report questionnaires   
 SF-36 scores (27, 28), mean ± SD (range 0–100)   
  Bodily pain67.2 ± 17.60.98 (0.97, 0.99)0.99 (0.98, 1.01)
  Vitality63.0 ± 16.80.99 (0.98, 0.99)1.00 (0.98, 1.01)
  Mental health76.3 ± 14.00.99 (0.98, 1.00)1.01 (0.99, 1.02)
  General health53.2 ± 18.00.97 (0.96, 0.98)0.98 (0.97, 0.99)
 EQ-5D (29, 30)   
  Item on anxiety/depression, no. (%)   
   Not anxious/depressed470 (79.9)Reference 
   Anxious/depressed107 (18.2)1.28 (0.82, 2.01) 
  VAS for health (range 0–100)76.5 (70.0–85.0)0.97 (0.96, 0.99)0.99 (0.97, 1.01)
 Social support (31) (range 12–60)16 (12–22)1.02 (1.00, 1.04)1.01 (0.99, 1.03)
 Pain Coping Inventory (32)   
  Distraction (range 5–20)11 (9–13)1.05 (1.00, 1.11)1.02 (0.95, 1.09)
  Transformation (range 4–16)8 (7–10)1.08 (1.02, 1.15)1.05 (0.97, 1.13)
  Reducing demands (range 3–12)6 (5–7)1.05 (0.95, 1.15) 
  Resting (range 5–20)9 (8–11)1.08 (1.01, 1.16)0.99 (0.91, 1.08)
  Retreating (range 7–28)10 (8–13)1.01 (0.96, 1.06) 
  Worrying (range 9–36)14 (11–16)1.03 (0.98, 1.07) 

Categorical variables were coded as dummy variables. Combinations of categories were based on frequency distributions (if categories were relatively small, they were combined [18]) or on combinations used in comparable studies.

For the physical and radiographic examination variables, we used the clinical signs and symptoms relating to the index knee or the index hip (i.e., the most affected knee or hip) in the analyses. Of the 832 and 588 participants with knee and hip symptoms, 44.5% and 64.6% reported unilateral symptoms, respectively. For participants with bilateral symptoms, we defined an index joint based on the following decision tree: 1) highest Kellgren/Lawrence score (19), 2) lowest degree of active knee flexion (index knee) or active hip internal rotation (index hip), 3) highest pain during knee flexion (index knee) or highest pain during hip internal rotation (index hip), and 4) crepitus during knee flexion (index knee) or highest pain during hip flexion (index hip). In participants for whom we could not define an index knee or hip based on these signs, we randomly assigned an index joint.

Statistical analysis.

Changes in activity limitations between baseline and 2-year followup measurements were analyzed using paired-samples t-tests. Logistic and linear regression models were built to predict the 2-year course of activity limitations. Dependent variables were the dichotomous and continuous outcome measures of activity limitations.

Independent variables (the prognostic factors described in the previous paragraph) were selected in 4 steps. First, the individual relationship of each prognostic factor with the outcome was studied. Second, after controlling for colinearity, the prognostic factors identified at step 1 with a value of P < 0.20 were entered into a multivariable regression model. This was done separately for each of the 7 groups of related factors (see previous paragraph). Third, the prognostic factors identified at step 2 (the multivariable regression analyses per group) with a value of P < 0.20 were entered into a backward stepwise regression model (P removal 0.10). Variables were entered in 7 consecutive blocks (Tables 1 and 2). Adding a next block of prognostic factors to the model may cause loss of significance of the prognostic factors added in a previous block. Therefore, to remove all prognostic factors with a value of P > 0.10 from the final model, at step 4 the prognostic factors identified in step 3 were entered in 1 block into a backward stepwise regression model (P removal 0.10).

A potential nonlinear behavior of continuous factors with the outcome was examined by using restricted cubic spline functions and spline plots (18). Multicategory dummy variables went to the next step of the model building process when ≥1 of the set of dummy variables was significantly related to the outcome.

For the paired-samples t-tests, we used data from the complete cases only. For the regression analyses, missing values were estimated using the Multiple Imputation by Chained Equation (MICE) procedure (20, 21). Before missing values were imputed, we studied the missing data mechanisms (18). Missing values were at least partly missing at random; therefore, imputation of the missing values may reduce bias (22). Missing values were imputed in 4 groups of related variables, whereby 10 multiple imputed data sets were generated. The outcome variable was included in all of the imputation models (23).

The regression coefficients (B) and 95% confidence intervals (95% CIs) of the final regression models were estimated over the 10 imputed data sets according to Rubin's rules (24). Furthermore, for the logistic regression models, the odds ratios (ORs), area under the receiver operating characteristic curve (AUC), Hosmer-Lemeshow goodness-of-fit test (25), and Nagelkerke's R2 statistic were calculated (26), and for the linear regression models the R2 was calculated.

Paired-samples t-tests were performed with SPSS, version 15.0 (SPSS). The MICE procedure and regression analyses on the imputed data sets were performed with R software, version 2.8.1 (R foundation for Statistical Computing).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Study population.

Of the 1,002 participants included at baseline, 926 (92.4%) completed the WOMAC-PF after 2 years. The knee and hip strata consisted of 832 and 588 participants, respectively. The baseline characteristics of the study population are presented in Tables 1 and 2.

In the knee and hip strata, missing outcome values were imputed for 77 (9.3%) and 57 (9.7%) participants, respectively. The mean number of missing values per independent variable was 21 (median 21 [2.5%], range 0–105 [0–12.6%]) for the knee stratum, and 11 (median 12 [2.0%], range 0–31 [0–5.3%]) for the hip stratum.

Two-year course of activity limitations.

Using the dichotomous outcome measure of activity limitations, 405 (48.7%) of the participants with knee symptoms were classified as experiencing a poor 2-year outcome. In participants with hip symptoms, 292 (49.7%) were classified as experiencing a poor 2-year outcome.

The median baseline WOMAC-PF scores for the knee and hip strata were 14 (interquartile range [IQR] 7–23) and 16 (IQR 8–25), respectively. In both participants with knee symptoms (mean ± SD WOMAC-PF change score −0.7 ± 9.8; P = 0.05) and participants with hip symptoms (−0.8 ± 10.4; P = 0.09), a small overall improvement in activities was observed between baseline and 2-year followup. The correlation between activity limitations at baseline and after 2 years was 0.65 (P = 0.00) for participants with knee symptoms and 0.62 (P = 0.00) for participants with hip symptoms. An overview of the mean WOMAC-PF change scores grouped by the dichotomous outcome subgroups is given in Table 3.

Table 3. Mean WOMAC-PF 2-year change scores of participants with knee symptoms and participants with hip symptoms, grouped by the dichotomous outcome definition: poor versus good outcome on activity limitations*
 Stratum knee (n = 832)Stratum hip (n = 588)
 No. (%)WOMAC-PF changeNo. (%)WOMAC-PF change
  • *

    Values are the mean ± SD unless otherwise indicated. Mean values were calculated over the 10 imputed data sets. WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; PF = physical function scale.

Poor outcome405 (48.7)5.45 ± 6.38292 (49.7)5.89 ± 6.15
 Stayed in low function score group188 (22.6)0.21 ± 5.87127 (21.6)0.85 ± 5.04
 Moved to worse group217 (26.1)9.99 ± 6.82165 (28.1)9.77 ± 7.00
Good outcome427 (51.3)−6.57 ± 5.10296 (50.3)−7.43 ± 5.69
 Stayed in high function score group163 (19.6)−0.11 ± 2.14108 (18.4)−0.02 ± 2.45
 Moved to better group264 (31.7)−10.56 ± 6.93188 (32.0)−11.68 ± 7.55
Total832 (100.0)−0.72 ± 9.84588 (100.0)−0.82 ± 10.41

Prognostic factors in participants with knee symptoms.

Examination of a potential nonlinear behavior of continuous factors with each outcome measure showed that for all factors, a linear fit was appropriate. The prognostic factors, analyzed in the univariable and multivariable regression analyses per group, and their association with the dichotomous outcome measure of activity limitations in participants with knee symptoms are given in Table 1.

The final multivariable logistic regression model identified the following 10 baseline prognostic factors for a poor 2-year outcome on activity limitations: young age, non-Western ethnicity, bilateral hip pain, morning stiffness in the knee, high comorbidity count, high body mass index (BMI), high bodily pain, poor general health perception, frequent use of the pain coping strategy distraction, and infrequent use of the pain coping strategy worrying (AUC 0.69 [95% CI 0.66, 0.73], Nagelkerke's R2 = 0.15, Hosmer-Lemeshow statistic = 13.49; P = 0.14). ORs and 95% CIs are given in Table 4. The AUC of the univariable logistic regression model with only the baseline WOMAC-PF score as an independent variable was 0.61 (95% CI 0.57, 0.65).

Table 4. Multivariable logistic regression models of prognostic factors associated with a poor 2-year outcome on activity limitations in patients with knee symptoms and patients with hip symptoms*
VariableKnee stratum (n = 832)Hip stratum (n = 588)
 OR (95% CI)POR (95% CI)P
  • *

    For the multivariable regression analyses, prognostic factors were selected according to variable selection steps 3 and 4 as described in the Statistical analysis section. T0 = baseline; PCI = Pain Coping Inventory; AUC = area under the receiver operating characteristic curve. See Table 1 for additional definitions.

  • Categories are based on the definition of ethnic minorities of Statistics Netherlands, whereby Western immigrants of Indonesian and Japanese origin are analyzed as a separate group.

Intercept3.36 (0.32, 35.91)0.3218.10 (1.83, 179.11)0.01
WOMAC function T0  0.98 (0.97, 1.00)0.07
Age, years0.97 (0.94, 1.00)0.05  
Ethnicity    
 Native or immigrant of Western originReference   
 Immigrant of non-Western origin4.03 (1.06, 15.38)0.04  
 Immigrant of Indonesian origin1.80 (0.68, 4.79)0.24  
Hip pain    
 No painReference   
 Unilateral1.26 (0.90, 1.78)0.18Reference 
 Bilateral with index hip1.60 (0.98, 2.59)0.061.16 (0.77, 1.75)0.48
 Bilateral with equal symptoms2.76 (1.33, 5.74)0.012.22 (1.11, 4.45)0.02
Morning stiffness knee <30 minutes1.37 (0.99, 1.88)0.051.73 (1.20, 2.50)0.00
Comorbidity count    
 0Reference Reference 
 11.01 (0.67, 1.53)0.951.04 (0.64, 1.70)0.88
 21.21 (0.77, 1.89)0.411.10 (0.63, 1.93)0.73
 ≥31.53 (0.93, 2.53)0.091.82 (1.05, 3.14)0.03
Body mass index, kg/m21.06 (1.02, 1.11)0.00  
Hip flexion, degrees  0.98 (0.96, 0.99)0.01
SF-36 subscale for bodily pain0.98 (0.97, 1.00)0.00  
SF-36 subscale for general health0.99 (0.98, 1.00)0.060.98 (0.97, 0.99)0.00
PCI subscale for distraction1.26 (0.98, 1.62)0.07  
PCI subscale for transformation  1.07 (0.98, 1.14)0.06
PCI subscale for worrying0.63 (0.42, 0.95)0.03  
AUC0.69 (0.66, 0.73) 0.69 (0.64, 0.73) 
Nagelkerke's R20.15 0.14 
Hosmer-Lemeshow statistic13.490.1410.180.34

The final multivariable linear regression model for participants with knee symptoms identified the following 5 baseline prognostic factors for 2-year worsening on the continuous outcome measure of activity limitations: low WOMAC-PF score (few activity limitations at baseline, B = −0.43; 95% CI −0.50, −0.36), non-Western ethnicity (B = 5.12 [95% CI 0.24, 10.01]) or Indonesian ethnicity (B = 4.42 [95% CI 0.46, 8.38]), pain on palpation of the joint line (B = 2.56 [95% CI 1.27, 3.86]), no intraarticular fluid (positive refill test: B = −2.20 [95% CI −4.58, 0.18]), and low Short Form 36 (SF-36) bodily pain score (higher pain at baseline, B = −0.10 [95% CI −0.15, −0.06]). The R2 of the model was 0.20, which is an increase in explained variance of 6% compared with the univariable regression model with only the baseline WOMAC-PF score as an independent variable (R2 = 0.14).

Prognostic factors in participants with hip symptoms.

The prognostic factors analyzed in the univariable and multivariable regression analyses per group, and their association with the dichotomous outcome measure of activity limitations in participants with hip symptoms, are given in Table 2.

The final multivariable logistic regression model identified the following 7 baseline prognostic factors for a poor 2-year outcome on activity limitations: low WOMAC-PF score (few activity limitations at baseline), bilateral hip pain, morning stiffness in the knee, high comorbidity count, reduced active hip flexion, poor general health perception, and frequent use of the pain coping strategy transformation (AUC 0.69 [95% CI 0.64, 0.73], Nagelkerke's R2 = 0.14, Hosmer-Lemeshow statistic = 10.18; P = 0.34). ORs and 95% CIs are given in Table 4. The AUC of the univariable logistic regression model with only the baseline WOMAC-PF score as an independent variable was 0.55 (95% CI 0.50, 0.60).

The final multivariable linear regression model for participants with hip symptoms identified the following 2 baseline prognostic factors for 2-year worsening on the continuous outcome measure of activity limitations: low WOMAC-PF score (few activity limitations at baseline, B = −0.46 [95% CI −0.55, −0.37]) and a low SF-36 bodily pain score (higher pain at baseline, B = −0.09 [95% CI −0.15, −0.04]). The R2 of the model was 0.20, which is an increase in explained variation of 1% compared with the univariable regression model with only the baseline WOMAC-PF score as an independent variable, for which the R2 was 0.19.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

The present study investigated predictors of the 2-year course of activity limitations in 1,002 participants with early OA of the knee and/or hip. Overall, half of the participants were classified as experiencing a poor 2-year outcome on the dichotomous measure of activity limitations. On the continuous measure, a slight improvement in activities was observed after 2 years of followup. The correlation between activity limitations at baseline and after 2 years was moderate.

Our observations are in line with recent studies in which the course of activity limitations in knee and hip OA was found to deteriorate slowly over time (8). The slight decrease that we observed in activity limitations might be explained by the natural course of OA. This course varies considerably over time, with fluctuating severity of symptoms (2, 8). CHECK is an inception cohort: participants were included based on knee and/or hip symptoms. Therefore, it is likely that there was a relatively high prevalence of exacerbation of symptoms at the time of inclusion. It is plausible that after 2 years, some of these symptoms had improved (regression toward the mean [RTM]). However, the alternative to an inception cohort is an open population study, which is less efficient due to the inclusion of nonsymptomatic patients. Moreover, the observed decrease in activity limitations was very small and only statistically significant in participants with knee symptoms due to the large sample size.

In accordance with studies in established OA, large interindividual differences in the course of activity limitations were observed (8). Apparently, already at an early stage of the disease the course of activity limitations is variable. Identification of prognostic factors is therefore highly relevant. This enables physicians to identify patients at risk for development of activity limitations and to select patients eligible for treatment.

One of the strongest prognostic factors for a poor outcome on activity limitations in participants with knee symptoms was concomitant hip pain. In participants with hip symptoms, we found that morning stiffness of the knee was a prognostic factor for a poor outcome on activity limitations. Apparently, patients with both knee and hip symptoms have a relatively high risk at developing activity limitations. An explanation for this finding could be that having both early knee and hip symptoms is a sign of the development of generalized OA. Because generalized OA is characterized by the involvement of ≥3 joints, it is likely that this subset of OA leads to more activity limitations. In addition, patients with both knee and hip symptoms are probably less able to compensate for limitations in their symptomatic knee or hip by using their ipsilateral hip or knee joint in comparison with patients with only knee or only hip symptoms.

Comorbidity count was identified as a prognostic factor for a poor outcome on activity limitations in both participants with knee symptoms and participants with hip symptoms. OA is a disease with one of the highest rates of comorbidity (33), and a relationship with activity limitations has frequently been shown (8). Therefore, future research should address how comorbidity influences activity limitations, and how therapeutic interventions aimed at preventing activity limitations in OA can be adjusted/optimized for OA patients with comorbidities.

The predictors of activity limitations identified in the present study are similar to predictors found in studies in established OA. Associations between activity limitations in knee and/or hip OA and ethnicity (11, 34, 35), multiple-site joint pain (36, 37), morning stiffness (38), comorbidity count (5, 39), BMI (8), impaired hip flexion (33), bodily pain (8), general health perception (6), and the pain coping strategy transformation (40) have been identified previously in established OA. Furthermore, the results of the present study confirm earlier observations that radiographic status is not closely associated with physical function in patients with knee or hip OA (41–43). Apparently, in both early and established OA the same underlying processes of an increase in activity limitations are playing a role.

Few activity limitations at baseline was found to be a prognostic factor for an increase in activity limitations. This finding is contrary to expectation and previous observations (5). The contrasting results might be explained by a floor effect of the WOMAC, which may have caused RTM. In the present study, participants with few activity limitations at baseline (a WOMAC-PF baseline score of 0–5 in the knee stratum or of 0–6 in the hip stratum) had a relatively high risk of declining. They could only remain stable or increase in activity limitations over time, whereas participants with more activity limitations at baseline could both improve and deteriorate on the WOMAC-PF scale. Therefore, in future studies of activity limitations in patients with early OA, the use of an outcome instrument with an extended measurement range for few activity limitations is recommended.

In the multivariable logistic model for participants with knee symptoms, younger age was found to be a prognostic factor for an increase in activity limitations. This association is contrary to the associations found in other longitudinal studies of activity limitations in patients with OA (6, 7). The contrasting observations might be explained by the fact that our study population was younger (mean ± SD age of 56 ± 5 years) than the other study populations (mean ± SD ages of 65 ± 9 and 69 ± 10 years, respectively) (6, 7). OA-related symptoms at a relatively young age might be indicative of a more severe and progressive course of the disease, with an associated poor functional prognosis.

In the present study, we investigated the predictive values of the candidate prognostic factors by means of logistic and linear regression analyses. The linear regression models identified fewer prognostic factors for the course of activity limitations than the logistic regression models, possibly because the results of the linear regression analyses were more influenced by a floor effect of the WOMAC-PF. Therefore, to avoid bias due to a possible floor effect, we conducted additional linear regression analyses in which we adjusted for the WOMAC-PF baseline score in all steps of the model-building process. The additional linear regression analyses with WOMAC-PF baseline adjustment identified almost the same prognostic factors as the logistic regression models. In both participants with knee symptoms and participants with hip symptoms, 6 prognostic factors identified in the logistic analyses were also identified in the linear analyses with WOMAC-PF baseline adjustment.

Some comments have to be made about the applicability of the present study's results in clinical practice. First, plotting the WOMAC-PF change score against the WOMAC-PF baseline score identified that a floor effect of the WOMAC-PF had caused RTM (44). The degree in which participants could improve in activities was limited by this floor effect. Therefore, it is likely that the slight decrease in activity limitations we found was an underestimation of the true 2-year decrease in activity limitations.

Second, the discriminative ability and the variance in outcome accounted for by the presented regression models are relatively low. Because we did not correct for overfitting the presented regression coefficients, the explained variance and discriminative ability of the models may have been overoptimistic. Our models need validation and further evaluation before they can be used clinically in actual patient care (45). It is expected that a longer followup time is necessary for constructing reliable prediction rules for use in clinical practice. Thus, further research on predictors for activity limitations using a longer followup time is recommended.

Finally, the overlap in predictors for a poor outcome on activity limitations in participants with knee symptoms and participants with hip symptoms was possibly caused by the overlap in data used in both the knee and hip analyses. Nevertheless, many older people with joint pain in the lower limb have more than 1 joint affected (37). Therefore, our results seem to be applicable to the typical knee and hip OA population.

In conclusion, participants overall reported little change in activity limitations after 2 years. However, large between-subject differences were observed in the course of activity limitations. The course of activity limitations is to some extent predictable already at an early stage of knee and hip OA.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  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 submitted for publication. Ms Holla 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. Holla, Steultjens, Roorda, ten Wolde, Dekker.

Acquisition of data. Steultjens, Roorda, ten Wolde.

Analysis and interpretation of data. Holla, Steultjens, Roorda, Heymans, ten Wolde, Dekker.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

The authors thank the steering committee of the Dutch Arthritis Association, which is chaired by Professor J. W. J. Bijlsma and coordinated by J. Wesseling, MSc. Involved are the Academic Hospital Maastricht, Erasmus Medical Center Rotterdam, Jan van Breemen Institute/VU Medical Center Amsterdam, Kennemer Gasthuis Haarlem, Martini Hospital Groningen/Allied Health Care Center for Rheumatology and Rehabilitation Groningen, Medical Spectrum Twente Enschede/Twenteborg Hospital Almelo, St. Maartenskliniek Nijmegen, Leiden University Medical Center, University Medical Center Utrecht, and Wilhelmina Hospital Assen.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  • 1
    Michaud CM, McKenna MT, Begg S, Tomijima N, Majmudar M, Bulzacchelli MT, et al. The burden of disease and injury in the United States 1996. Popul Health Metr 2006; 4: 11.
  • 2
    Arden N, Nevitt MC. Osteoarthritis: epidemiology. Best Pract Res Clin Rheumatol 2006; 20: 325.
  • 3
    Lawrence RC, Felson DT, Helmick CG, Arnold LM, Choi H, Deyo RA, et al, for the National Arthritis Data Workgroup. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States: part II. Arthritis Rheum 2008; 58: 2635.
  • 4
    World Health Organization. International classification of functioning, disability and health: ICF. Geneva: WHO; 2001.
  • 5
    Ayis S, Dieppe P. The natural history of disability and its determinants in adults with lower limb musculoskeletal pain. J Rheumatol 2009; 36: 58391.
  • 6
    Mallen CD, Peat G, Thomas E, Lacey R, Croft P. Predicting poor functional outcome in community-dwelling older adults with knee pain: prognostic value of generic indicators. Ann Rheum Dis 2007; 66: 145661.
  • 7
    Sharma L, Cahue S, Song J, Hayes K, Pai YC, Dunlop D. Physical functioning over three years in knee osteoarthritis: role of psychosocial, local mechanical, and neuromuscular factors. Arthritis Rheum 2003; 48: 335970.
  • 8
    Dekker J, van Dijk GM, Veenhof C. Risk factors for functional decline in osteoarthritis of the hip or knee [review]. Curr Opin Rheumatol 2009; 21: 5204.
  • 9
    Marra CA, Cibere J, Tsuyuki RT, Soon JA, Esdaile JM, Gastonguay L, et al. Improving osteoarthritis detection in the community: pharmacist identification of new, diagnostically confirmed osteoarthritis. Arthritis Rheum 2007; 57: 123844.
  • 10
    Rosemann T, Wensing M, Joest K, Backenstrass M, Mahler C, Szecsenyi J. Problems and needs for improving primary care of osteoarthritis patients: the views of patients, general practitioners and practice nurses. BMC Musculoskelet Disord 2006; 7: 48.
  • 11
    Felson DT, Zhang Y. An update on the epidemiology of knee and hip osteoarthritis with a view to prevention. Arthritis Rheum 1998; 41: 134355.
  • 12
    Wesseling J, Dekker J, van den Berg WB, Bierma-Zeinstra SM, Boers M, Cats HA, et al. CHECK (Cohort Hip and Cohort Knee): similarities and differences with the Osteoarthritis Initiative. Ann Rheum Dis 2009; 68: 14139.
  • 13
    Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Development of criteria for the classification and reporting of osteoarthritis: classification of osteoarthritis of the knee. Arthritis Rheum 1986; 29: 103949.
  • 14
    Altman R, Alarcon G, Appelrouth D, Bloch D, Borenstein D, Brandt K, et al. The American College of Rheumatology criteria for the classification and reporting of osteoarthritis of the hip. Arthritis Rheum 1991; 34: 50514.
  • 15
    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.
  • 16
    Roorda LD, Jones CA, Waltz M, Lankhorst GJ, Bouter LM, van der Eijken JW, et al. Satisfactory cross cultural equivalence of the Dutch WOMAC in patients with hip osteoarthritis waiting for arthroplasty. Ann Rheum Dis 2004; 63: 3642.
  • 17
    Veenhof C, Bijlsma JW, van den Ende CH, van Dijk GM, Pisters MF, Dekker J. Psychometric evaluation of osteoarthritis questionnaires: a systematic review of the literature. Arthritis Rheum 2006; 55: 48092.
  • 18
    Steyerberg EW. Clinical prediction models: a practical approach to development, validation, and updating. New York: Springer Science & Business Media; 2009.
  • 19
    Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis. Ann Rheum Dis 1957; 16: 494502.
  • 20
    Clark TG, Altman DG. Developing a prognostic model in the presence of missing data: an ovarian cancer case study. J Clin Epidemiol 2003; 56: 2837.
  • 21
    Van Buuren S, Boshuizen HC, Knook DL. Multiple imputation of missing blood pressure covariates in survival analysis. Stat Med 1999; 18: 68194.
  • 22
    Vergouwe Y, Royston P, Moons KG, Altman DG. Development and validation of a prediction model with missing predictor data: a practical approach. J Clin Epidemiol 2010; 63: 20514.
  • 23
    Moons KG, Donders RA, Stijnen T, Harrell FE Jr. Using the outcome for imputation of missing predictor values was preferred. J Clin Epidemiol 2006; 59: 1092101.
  • 24
    Rubin DB. Multiple imputation for nonresponse in surveys. New York: Wiley & Sons; 1987.
  • 25
    Hosmer DW, Lemeshow S. Applied logistic regression. New York: Wiley & Sons; 1989.
  • 26
    Nagelkerke NJ. A note on a general definition of the coefficient of determination. Biometrika 1991; 78: 6912.
  • 27
    Aaronson NK, Muller M, Cohen PD, Essink-Bot ML, Fekkes M, Sanderman R, et al. Translation, validation, and norming of the Dutch language version of the SF-36 health survey in community and chronic disease populations. J Clin Epidemiol 1998; 51: 105568.
  • 28
    Ware JE Jr, Sherbourne CD. The MOS 36-item Short-Form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992; 30: 47383.
  • 29
    The EuroQol Group. EuroQol: a new facility for the measurement of health-related quality of life. Health Policy 1990; 16: 199208.
  • 30
    Hurst NP, Kind P, Ruta D, Hunter M, Stubbings A. Measuring health-related quality of life in rheumatoid arthritis: validity, responsiveness and reliability of EuroQol (EQ-5D). Br J Rheumatol 1997; 36: 5519.
  • 31
    Fink P, Jensen J, Borgquist L, Brevik JI, Dalgard OS, Sandager I, et al. Psychiatric morbidity in primary public health care: a Nordic multicentre investigation. Part I: method and prevalence of psychiatric morbidity. Acta Psychiatr Scand 1995; 92: 40918.
  • 32
    Kraaimaat FW, Evers AW. Pain-coping strategies in chronic pain patients: psychometric characteristics of the Pain-Coping Inventory (PCI). Int J Behav Med 2003; 10: 34363.
  • 33
    Van Dijk GM, Veenhof C, Lankhorst G, Dekker J. Limitations in activities in patients with osteoarthritis of the hip or knee: the relationship with body functions, comorbidity and cognitive functioning. Disabil Rehabil 2009; 31: 168591.
  • 34
    Dunlop DD, Semanik P, Song J, Manheim LM, Shih V, Chang RW. Risk factors for functional decline in older adults with arthritis. Arthritis Rheum 2005; 52: 127482.
  • 35
    Jordan JM. Effect of race and ethnicity on outcomes in arthritis and rheumatic conditions. Curr Opin Rheumatol 1999; 11: 98103.
  • 36
    Belo JN, Berger MY, Koes BW, Bierma-Zeinstra SM. Prognostic factors in adults with knee pain in general practice. Arthritis Rheum 2009; 61: 14351.
  • 37
    Peat G, Thomas E, Wilkie R, Croft P. Multiple joint pain and lower extremity disability in middle and old age. Disabil Rehabil 2006; 28: 15439.
  • 38
    Thomas E, Peat G, Mallen C, Wood L, Lacey R, Duncan R, et al. Predicting the course of functional limitation among older adults with knee pain: do local signs, symptoms and radiographs add anything to general indicators? Ann Rheum Dis 2008; 67: 13908.
  • 39
    Van Dijk GM, Veenhof C, Schellevis F, Hulsmans H, Bakker JP, Arwert H, et al. Comorbidity, limitations in activities and pain in patients with osteoarthritis of the hip or knee. BMC Musculoskelet Disord 2008; 9: 95.
  • 40
    Steultjens MP, Dekker J, Bijlsma JW. Coping, pain, and disability in osteoarthritis: a longitudinal study. J Rheumatol 2001; 28: 106872.
  • 41
    Belo JN, Berger MY, Reijman M, Koes BW, Bierma-Zeinstra SM. Prognostic factors of progression of osteoarthritis of the knee: a systematic review of observational studies. Arthritis Rheum 2007; 57: 1326.
  • 42
    Juhakoski R, Tenhonen S, Anttonen T, Kauppinen T, Arokoski JP. Factors affecting self-reported pain and physical function in patients with hip osteoarthritis. Arch Phys Med Rehabil 2008; 89: 106673.
  • 43
    Paradowski PT, Englund M, Lohmander LS, Roos EM. The effect of patient characteristics on variability in pain and function over two years in early knee osteoarthritis. Health Qual Life Outcomes 2005; 3: 59.
  • 44
    Barnett AG, van der Pols JC, Dobson AJ. Regression to the mean: what it is and how to deal with it. Int J Epidemiol 2005; 34: 21520.
  • 45
    Reilly BM, Evans AT. Translating clinical research into clinical practice: impact of using prediction rules to make decisions. Ann Intern Med 2006; 144: 2019.