Appendix 2. Detailed description of the included studies

Authors, year / Design and population / Independent variable measurement / Dependent variable measurement / Analysis and results
Avoidance of activities – Muscle weakness – Activity limitations
Knee OA / Cross-sectional
Holla et al. (2012a) / N = 151 (CHECK)
Early symptomatic knee OA. All participants had pain and/or stiffness in the knee. 78.1% fulfilled the ACR clinical criteria for knee OA. / Independent variable:
PCI resting subscale.
Mediator:
Maximal isokinetic muscle strength (average of knee extension and flexion) around the most affected knee (Nm/kg). / WOMAC-PF score and stair-climbing time were combined into one latent variable for activity limitations. / Structural equation model including pain, psychological distress, avoidance, muscle strength and activity limitations.
Test for mediation: bootstrapping
The indirect association between resting and activity limitations via muscle weakness was statistically significant (β = 0.04, P = 0.01). Of the total association between resting and activity limitations, 18.1% was mediated by muscle weakness.
Steultjens et al. (2002) / N = 107
Diagnosis of knee OA according to the ACR clinical criteria. / Independent variable:
PCI resting subscale.
Mediator:
Maximal isometric muscle strength (sum of knee extension and flexion) around the most affected knee (N/kg). / Scoring of videotaped performances on a number of standardized tasks including walking, sitting down, reclining onto a bed, and bending. Based on five items (i.e. 5-m walking, stand-to-sit, stand-to-recline, level of guarding, and level of rigidity) an overall score for activity limitations was computed. / Regression analyses.
Test for mediation: criteria of Baron and Kenny + Sobel test.
The indirect effect of muscle weakness on the relationship between avoidance and activity limitations was 0.059 (P = 0.01).
Knee OA / Longitudinal
Pisters et al. (2014) / N = 216 (CARPA)
Follow-up: 1, 2, 3 and 5 years
Diagnosis of knee OA by a medical specialist according to the ACR radiographic or clinical criteria. / Independent variable:
PCI resting subscale.
Mediator:
Maximal isometric knee extension strength. Sum of the left and right leg (N/kg). / Self-reported activity limitations: WOMAC-PF.
Performance-based activity limitations: 10 meter timed walking test. / GEE analyses.
Test for mediation: criteria of Baron and Kenny + Sobel test.
The indirect effect of muscle weakness on the relationship between avoidance of activities and self-reported and performance-based activity limitations was 0.636 (P = 0.03) and 0.188 (P = 0.02), respectively.
Comparable results were found after adjustment for age, gender, duration of symptoms, BMI, educational level and comorbidity.
Hip OA / Longitudinal
Pisters et al. (2014) / N = 149 (CARPA)
Follow-up: 1, 2, 3 and 5 years
Diagnosis of hip OA by a medical specialist according to the ACR radiographic or clinical criteria. / Independent variable:
PCI resting subscale.
Mediator:
Maximal isometric hip abduction strength. Sum of the left and right leg (N/kg). / Self-reported activity limitations: WOMAC-PF.
Performance-based activity limitations: 10 meter timed walking test. / GEE analyses.
Test for mediation: criteria of Baron and Kenny + Sobel test.
Multivariable analyses:
* Adjustment for age, gender, duration of symptoms, BMI, educational level and comorbidity.
The indirect effect of reduced muscle strength on the relationship between avoidance of activities and performance-based activity limitations was 0.115. This mediation effect was statistically significant (P < 0.05).
Muscle weakness – Activity limitations
Knee OA / Cross-sectional
Amin et al. (2009) / N = 265
Diagnosis of knee OA according to the ACR clinical criteria and osteophytosis. / Maximal isokinetic knee extension strength around the most affected knee. Sex-specific tertiles (Nm/kg). / WOMAC-PF. / Univariable analysis: -
Multivariable analysis:
* Adjustment for 5 covariates.
Participants in the lowest sex-specific tertile of quadriceps strength reported higher WOMAC-PF scores (mean: 30.2) than participants in the moderate (mean: 23.9) and highest (20.3) sex-specific tertile of quadriceps strength (P < 0.001).
Brown et al. (2009) / N = 82
Participants who were scheduled for total knee replacement with symptomatic and radiographic knee OA, based on a physician (i.e. orthopedic surgeon) diagnosis using MRI, radiographs, signs and symptoms. / Maximal isokinetic knee extension and flexion strength around the surgical knee. (Nm/kg). / 6-minute walk test, the number of chair rises in 30 seconds and the time to ascend and descend 2 flights of 11 stairs. / Univariable analyses:
Extension and flexion strength of the surgical leg were significantly (P < 0.05) associated with the 6-minute walk distance (extension: r = 0.56, flexion: r = 0.62), number of chair rises (extension: r = 0.55, flexion: r = 0.63), and time to ascend (extension: r = -0.47, flexion: r = -0.48) or descend (extension: r = -0.49, flexion: r = -0.53) a flight of 11 stairs.
Multivariable analyses:
* After univariable preselection, knee extension strength and knee flexion strength were submitted to stepwise regression together with 4-6 covariates.
Flexion strength of the surgical leg was associated with the number of chair rises (β = 0.60, P < 0.01).
Chun et al. (2013) / N = 329
Koreans aged 65 years or older with KL-grade 2-3 (N = 258) or 4 (N = 71). / Maximal isokinetic knee extension strength. Average of both legs (Nm/kg). / SPPB: scores ≤ 9 = poor outcome; scores 9-12 = good outcome. / Univariable analyses:
Knee extension strength was associated with the poor outcome group in both the minimal to moderate (rs = -0.38, p < 0.01) and the severe (rs = -0.43, p < 0.01) OA stratum.
Multivariable analyses:
* After univariable preselection knee extension strength was subjected to multivariable analysis together with 5-7 covariates.
Knee extension strength was associated with the poor outcome group in both the minimal to moderate (OR [95% CI] = 0.81 [0.73 to 0.90]) and the severe (OR [95% CI] = 0.72 [0.58 to 0.89]) OA stratum.
Fitzgerald et al. (2004) / N = 105
Knee OA according to the ACR clinical and radiographic criteria. / Maximal isometric knee extension strength around the most affected knee (torque / BMI). / WOMAC-PF and a timed get up and go (GUG) test. A principal component analysis was performed to combine the WOMAC scores (pain, stiffness, and physical function) and GUG-score into a single principal component score of physical function (PCPF). / Univariable analyses:
Knee extension strength was significantly (P < 0.01) associated with the GUG time (r = -0.45), WOMAC-PF (r = -0.33) and PCPF (r = 0.36).
Multivariable analyses:
* Adjustment for 2 covariates:
Knee extension strength was found to be independently associated with the PCPF. Inclusion of knee extension strength changed the R2 with 10% (P = 0.001).
Goncalves et al. (2012) / N = 136
Diagnosis of uni- or bilateral knee OA according to the ACR clinical and radiographic criteria. All participants had a KL-grade of 2 or 3 and were referred for physical therapy. / Maximal isometric knee extension and flexion strength around the most affected knee (N). / Step test (ST), Timed up and go test (TUGT), 20-meter walk test (20MWT), 6-minute walk test (6MWT), KOOS subscales for function in daily living (DL) and function in sport and recreation (SR) , Physical function subscale of the SF-36 (SF-36 PF). / Univariable analyses:
Knee extension strength was significantly (P < 0.05) associated with the KOOS-DL (r = 0.35), KOOS-SR (r = 0.23), SF-36 PF (r = 0.29), ST (r = 0.35), TUGT (r = -0.21), 20MWT (r = 0.28) and 6MWT (r = 0.39).
Knee flexion strength was significantly associated with the KOOS-DL (r = 0.49), KOOS-SR (r = 0.38), SF-36 PF (r = 0.47), ST (r = 0.37), TUGT (r = -0.40), 20MWT (r = 0.51) and 6MWT (r = 0.56).
Multivariable analyses (self-report measures only):
* After univariable preselection knee extension strength and knee flexion strength were subjected to multivariable regression together with 7 covariates.
Knee flexion strength was significantly associated with the KOOS-DL (β = 0.26, P = 003), the KOOS-SR (β = 0.29, P < 0.01), and the SF-36 PF (β = 0.22, P = 0.01).
Hurley et al. (1997) / N = 103
Rheumatologist diagnosed knee OA based on clinical history and physical examination. All patients fulfilled the ACR clinical criteria for knee OA. / Maximal isometric knee extension strength around the most affected knee (N). / Aggregate functional performance time (AFPT = sum 50 ft walk time + get up and go time + stairs ascending time + stairs descending time.
Lequesne index for knee OA. The Lequesne index was arbitrarily categorized into: mild/moderate limitations (<7); severe/very severe limitations (7-14); and extremely severe limitations(>14). / Univariable analyses:
Knee extension strength (r = -0.402, P < 0.001) and the Lequesne index (r = -0.385, P < 0.001) were associated with the AFPT.
Multivariable analyses (Lequesne index only):
* Adjustment for 6 covariates.
Quadriceps strength significantly influenced the Lequesne index. Together with AFPT quadriceps strength accounted for 45% of the variance in the Lequesne index.
Kauppila et al. (2009) / N = 88
Patients with a diagnosis of primary knee OA who were scheduled for unilateral total knee replacement. All patients had radiographic knee OA (KL-grade ≥ 2). / Maximal isometric knee extension and flexion strength around the affected knee (Nm/kg). / WOMAC-PF and physical function (PF) subscale of the RAND-36. / Univariable analysis: extension (r = -0.45, P < 0.001) and flexion (r = -0.39, P = 0.001) strength were associated with the WOMAC-PF. extension (r = 0.38, P<0.001) and flexion (r = 0.29, P = 0.001) strength were associated with the RAND-36 PF.
Multivariable analyses:
* After univariable preselection flexion and extension strength were subjected to stepwise regression together with maximal 4 covariates.
No statistically significant (P < 0.05) associations.
Liikavainio et al. (2008) / N = 54
Patients with uni- or bilateral knee OA according to the ACR clinical criteria. / Maximal isometric knee extension and flexion strength around the most affected knee (Nm/kg). / WOMAC-PF, lift test (number of lifts in 1 minute), repeated sit-to-stand time, stair ascending and descending velocity (m/s), 10-m walk time, timed up and go (TUG) time, 20-m walk velocity (m/s), five-minute walk distance (m). / Univariable analyses:
Knee extension strength was significantly (P < 0.001) associated with the WOMAC-PF (r = -0.499), number of lifts in one minute (r = 0.555), repeated sit-to-stand time (r = -0.620), stair ascending velocity (r = 0.560), stair descending velocity (r = 0.598), 10-m walk time (r = -0.552), TUG time (r = -0.693), 20-m walk velocity (r = 0.617) and 5-min walk distance (r = 0.720).
Knee flexion strength was significantly (P < 0.001) associated with the WOMAC-PF (r = -0.513), number of lifts in one minute (r = 0.542), repeated sit-to-stand time (r = -0.638), stair ascending velocity (r = 0.561), stair descending velocity (r = 0.551), 10-m walk time (r = -0.524), TUG time (r = -0.649), 20-m walk velocity (r = 0.677) and 5-min walk distance (r = 0.704).
Multivariable analyses: -
Maly et al. (2005) / N = 54
Patients with physician diagnosed medial compartment knee OA (confirmed by radiographs). Unilateral knee OA: N = 22, bilateral knee OA: N = 26. / Maximal isokinetic knee extension and flexion strength around the most affected knee (Nm). / Six-minute walk (SMW) distance (m), stair climbing (STR) time (s), timed up and go (TUG) time. / Univarable analyses:
Knee extension strength was significantly (P < 0.001) associated with SMW distance (r = 0.47), STR time (r = -0.50) and TUG time (r = -0.49).
Knee flexion strength was significantly (P < 0.001) associated with SMW distance (r = 0.47), STR time (r = -0.52) and TUG time (r = -0.51).
Multivariable analyses:
* Knee extension and flexion strength were subjected to stepwise regression together with 6 covariates.
Knee extension strength was significantly (P < 0.001) associated with SMW distance and TUG time. Knee flexion strength was significantly (P < 0.001) associated with STR time.
Maly et al. (2006) / N = 54
Patients with physician diagnosed medial compartment knee OA (confirmed by radiographs). Unilateral knee OA: N = 22, bilateral knee OA: N = 26. Radiographs were taken at the beginning of the study to confirm the presence of OA in the medial compartment. / Maximal isokinetic knee extension and flexion strength around the most affected knee (Nm). / WOMAC-PF. / Univariable analyses: no statistically significant (P < 0.05) associations .
Multivariable analyses:
* Knee extension and flexion strength were subjected to stepwise regression together with 7 covariates.
Knee extension strength was significantly (P = 0.001) associated with the WOMAC-PF.
McAlindon et al. (1993) / N = 98
Radiographic OA based on JSN (0-2), osteophytosis (0-2), and sclerosis (0-1) scored in the medial, lateral, and patellofemoral compartments of each knee. A score greater than 0 was defined as radiographic OA. / Maximal isometric knee extension strength. Average of both legs (KgF). / Stanford Health Assessment Questionnaire. Three parts of the questionnaire (sections 2, 4, and 8) are related to lower limb function and were taken together as a lower limb score (range 0-9). This measure was dichotomized into: disability vs. no disability. / Univariable analyses: -
Multivariable analyses:
* Adjustment for 4 covariates.