Abstract
Objective
To identify prognostic factors for health outcomes following an 8‐week supervised exercise therapy and education program for individuals with knee and hip osteoarthritis (OA) alone or with concomitant hypertension, heart or respiratory disease, diabetes mellitus, or depression.
Methods
We included individuals with knee and/or hip OA from the Good Life With OsteoArthritis in Denmark (GLA:D) program. GLA:D consists of 2 patient education sessions and 12 supervised exercise therapy sessions. Before GLA:D, participants self‐reported any comorbidities and were categorized into 8 comorbidity groups. Twenty‐one potential prognostic factors (demographic information, clinical data, and performance‐based physical function) gathered from participants and clinicians before the program were included. Outcomes were physical function using the 40‐meter Fast‐Paced Walk Test (FPWT), health‐related quality of life using the 5‐level EuroQol 5‐domain (EQ‐5D‐5L) index score, and pain intensity using a visual analog scale (VAS; range 0–100) assessed before and immediately after the GLA:D program. Within each comorbidity group, associations of prognostic factors with outcomes were estimated using multivariable linear regression.
Results
Data from 35,496 (40‐meter FPWT) and 37,576 (EQ‐5D‐5L and VAS) participants were included in the analyses. Clinically relevant associations were demonstrated between age, self‐efficacy, self‐rated health, and pain intensity and change in 40‐meter FPWT, EQ‐5D‐5L, or VAS scores across comorbidity groups. Furthermore, anxiety, education, physical function, and smoking were associated with outcomes among subgroups having depression or diabetes mellitus in addition to OA.
Conclusion
Age, self‐efficacy, self‐rated health, and pain intensity may be prognostic of change in health outcomes following an 8‐week exercise therapy and patient education program for individuals with OA, irrespective of comorbidities.
INTRODUCTION
International treatment guidelines recommend exercise therapy and patient education as first‐line treatment of knee or hip osteoarthritis (OA) (1, 2). This is supported by several randomized controlled trials showing that exercise therapy is effective in reducing pain intensity and improving physical function and quality of life (3, 4).
SIGNIFICANCE & INNOVATIONS.
To our knowledge, this is the first study that investigates potential prognostic factors for different health outcomes following an exercise and education program in individuals with knee and hip osteoarthritis (OA) stratified for concomitant comorbidities.
Irrespective of comorbidities, higher age and pain intensity were associated with less improvement, while higher self‐efficacy and self‐rated health were associated with larger improvement after the exercise therapy and education program.
Concomitant diabetes mellitus and/or depression in individuals with OA may have an effect‐modifying effect on the association between prognostic factors such as anxiety, education, physical function, and smoking and outcomes after the exercise therapy and education program.
Although the findings of this exploratory analysis need to be confirmed in future studies, the identified prognostic factors, and possible effect‐modifying nature of some of the comorbidities, may be used clinically to better target individuals with OA and comorbidities to improve outcomes from an exercise therapy and education program.
Still, substantial heterogeneity is observed in treatment effects between individuals, with ~50% reported to fail in responding to the treatment (5). The reason for the difference in outcomes after exercise therapy is largely unknown, although factors such as age, sex, level of physical activity, and anxiety may be associated with outcomes. However, findings are inconsistent across studies (6, 7, 8, 9, 10).
In addition to their knee or hip problems, individuals with OA often have other long‐term conditions (11, 12), with heart failure, coronary heart disease, hypertension, diabetes mellitus, chronic obstructive pulmonary disease, and depression being some of the most common (11, 12, 13, 14). These conditions share several similarities with OA. They are among the leading causes of global disability affecting millions of people in the world (15), and exercise therapy is an effective and recommended treatment for all the conditions (16, 17, 18, 19, 20, 21). Furthermore, some of the same factors as for OA (e.g., age and anxiety) have been reported to be associated with outcomes after exercise therapy in individuals with heart failure, coronary heart disease, or diabetes mellitus (22, 23). For most of the conditions, however, knowledge of potential prognostic factors is sparse. Importantly, whether prognostic factors for the outcomes after exercise therapy in individuals with OA and specific comorbidities are the same as in those without comorbidities is currently unknown.
A deeper understanding of what factors might be prognostic of the outcomes after exercise therapy and education in individuals with knee or hip OA with or without comorbidities may improve the identification of individuals who benefit from exercise and education and be used to tailor treatment based on individual characteristics to potentially optimize outcomes (24). Thus, we aimed to identify potential prognostic factors for health outcomes (function, quality of life, and pain) following an 8‐week exercise therapy and patient education program in individuals with knee or hip OA alone or in combination with heart or respiratory disease, hypertension, diabetes mellitus, or depression.
PATIENTS AND METHODS
Study design
We used registry data from the Good Life With OsteoArthritis in Denmark (GLA:D) program in an explorative cohort design to investigate potential prognostic factors for health outcomes following an exercise therapy and education program. The GLA:D program is an ongoing nationwide treatment program for individuals with knee and hip OA and consists of 2 patient education sessions followed by twelve 60‐minute sessions of neuromuscular exercise (twice weekly; 6 weeks) supervised by physical therapists certified to deliver the treatment program and evaluate the effects using predefined and validated outcomes prior to the program, immediately after the program (~3 months) and at 12 months. GLA:D has previously been described in detail, including participant characteristics, treatment, and outcomes (25, 26). The Danish Data Protection Agency has previously approved the GLA:D registry, and all participants have consented to submitting their data to the registry, whereas ethics approval of GLA:D is unneeded according to the local ethics committee of the North Denmark Region.
The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (27), with minor adaptations from the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) guideline (28), was followed to report the study (Table 1).
Table 1.
REMARK profile (overview of included participants, outcome measures, prognostic factors, and analyses conducted)*
| No. | Remarks | Factors | Results | |
|---|---|---|---|---|
| Participants, outcome measures, and prognostic factors | – | – | ||
| Participants | ||||
| Assessed for eligibility | 44,755 | Started GLA:D program within the period July 1, 2014 to October 31, 2019 | ||
| Excluded | ||||
| No reply | 6,391 | Did not reply to baseline questionnaire | ||
| Missing data | 126 | Missing data on most affected joint | ||
| Missing data | 37 | Missing data on comorbidities of interest | ||
| Other | 567 | Other comorbidity combinations than the included | ||
| Missing outcomes | 2,138/58 | Missing baseline outcomes (for 40‐meter FPWT/for EQ‐5D and VAS) | ||
| Included | 35,496/37,576 | For 40‐meter FPWT/for EQ‐5D and VAS | ||
| Outcome measures | – | – | ||
| 40‐meter FPWT | 34% (12,218/35,496) missing at follow‐up | |||
| EQ‐5D | 26% (9,864/37,576) missing at follow‐up | |||
| VAS | 26% (9,878/37,576) missing at follow‐up | |||
| Factors | ||||
| Factors available with complete data | f1 = age (con.), f2 = sex (dic.), f3 = pain in other joints (cat.), f4 = pain medicine (dic.), f5 = prior knee/hip surgery (dic.) | |||
| Factors available with missing data | f6 = anxiety (cat.), f7 = VAS (con.), f8 = 40‐m FPWT (con.), f9 = bodily pain (cat.), f10 = BMI (con.), f11 = education (cat.), f12 = born in Denmark (dic.), f13 = civil status (dic.), f14 = employment (cat.), f15 = activity level (cat.), f16 = no. of comorbidities (cat.), f17 = self‐rated health (con.), f18 = self‐efficacy (con.), f19 = duration of symptoms (cat.), f20 = previous knee/hip injury (dic.), f21 = smoking (dic.) | |||
| No. of missing | f6 = 9; f7 = 32; f8 = 2,130; f9 = 10,371; f10 = 151; f11 = 13; f12 = 7; f13 = 6, f14 = 5; f15 = 25; f16 = 45; f17 = 16; f18 = 36; f19 = 4,117; f20 = 777; f21 = 2 | f6 and f7 only missing for the analysis of 40‐meter FPWT; f8 only missing for the analyses of EQ‐5D and VAS; f9 fully or partly missing in period from April 9, 2018 to October 31, 2019 due to technical problems in data collection; f19 missing for 6 months due to technical problems in data collection; f20 missing for 2 months as the variable was first introduced on October 9, 2014 | ||
| Statistical analyses | ||||
| Univariable | ||||
| 40‐meter FPWT | 35,496 | Based on 25 and 35 imputed data sets† | f1 to f7, f9 to f21 | Supplementary Tables 2–4 |
| EQ‐5D and VAS | 37,576 | Based on 25 and 35 imputed data sets† | f1 to f21‡ | Supplementary Tables 2–4 |
| Multivariable | ||||
| 40‐meter FPWT | 35,496 | Based on 25 and 35 imputed data sets† | f1 to f7, f9 to f21 | Figures 1, 2, 3 and Supplementary Tables 16, 17, and 18 |
| EQ‐5D and VAS | 37,576 | Based on 25 and 35 imputed data sets† | f1 to f21‡ | Figures 1, 2, 3 and Supplementary Tables 16, 17, and 18 |
| Sensitivity 1§ | 35,496/37,576¶ | Based on 25 and 35 imputed data sets† | f1 to f21 | Supplementary Tables 7, 10, and 13 |
| Sensitivity 2# | 35,496/37,576¶ | Based on 25 and 35 imputed data sets† | f1 to f21 | Supplementary Tables 8, 11, and 14 |
| Sensitivity 3** | 23,278/27,698/27,712†† | f1 to f21 | Supplementary Tables 9, 12, and 15 |
Cat. = categorical; con. = continuous; dic. = dichotomous; EQ‐5D = EuroQol 5‐domain (questionnaire); f = factor; FPWT = Fast‐Paced Walk Test; GLA:D = Good Life With OsteoArthritis in Denmark; REMARK = Reporting Recommendations for Tumor Marker Prognostic Studies; VAS = visual analog scale.
N = 25 data sets (for model with VAS as dependent variable and for model with EQ‐5D as dependent variable); n = 35 data sets (for model with 40‐meter FPWT as dependent variable).
f7 not included as prognostic factor in the model with VAS as the dependent variable.
Only included participants with knee as the most affected joint.
N = 35,496 (for model with EQ‐5D as dependent variable); n = 37,576 (for model with VAS as dependent variable).
Adjusted for hypertension.
Complete case analyses.
N = 23,278 (for model with 40‐meter FPWT as dependent variable); n = 27,698 (for model with VAS as dependent variable); n = 27,712 (for model with EQ‐5D as dependent variable).
Participants
Individuals with knee or hip joint pain or functional impairments that result in contact with the health care system and not meeting any of the following exclusion criteria are eligible for GLA:D: 1) another reason for the joint symptoms than OA as evaluated by a physical therapist, e.g., inflammatory joint disease or patellar tendinopathy; 2) other symptoms that are more pronounced than the OA symptoms, e.g., chronic generalized pain or fibromyalgia; and 3) unable to understand Danish.
For the present study, participants had to have complete baseline data for all outcomes and the comorbidities of interest to be included. Only data from participants enrolled in the GLA:D program in the period from July 1, 2014 to October 31, 2019 were used, as no information on comorbidities were collected before this period.
Comorbidity groups
Prior to the GLA:D program, participants were asked to indicate if they had any of the following comorbidities (i.e., the comorbidities of interest in the present study): hypertension, heart diseases, respiratory diseases, diabetes mellitus (type 1 or 2), and depression. These conditions have previously been reported as some of the most prevalent somatic and psychiatric conditions in the GLA:D register (12), and exercise therapy is part of the recommended treatment (16, 17, 18, 19, 20, 21). Participants were grouped by comorbidity pattern. Because of a total of 32 comorbidity patterns, where the majority only contained few participants, we only included the 7 most prevalent comorbidity patterns in this study and additionally collapsed patterns that included hypertension with the same pattern that did not include hypertension (e.g., diabetes mellitus plus hypertension was collapsed with diabetes mellitus alone). As a result, 8 groups constituting 98% of the participants from the GLA:D registry were defined: 1) knee or hip OA only, and hip or knee OA plus the following: 2) hypertension; 3) heart diseases; 4) respiratory diseases; 5) diabetes mellitus; 6) depression; 7) heart plus respiratory diseases; or 8) heart diseases plus diabetes mellitus. Groups 3–8 included participants with or without concomitant hypertension.
Outcomes
The following 3 outcomes were used: objectively assessed physical function; self‐reported health‐related quality of life; and self‐reported mean pain intensity during the last month in the most affected joint. All were assessed at baseline and immediately after the 8 weeks of supervised exercise therapy and education (~3 months). Physical function was assessed by the physical therapist using the 40‐meter Fast‐Paced Walk Test (FPWT, meters/second), which is recommended by the Osteoarthritis Research Society International as a performance‐based function test for knee and hip OA (29). Health‐related quality of life was reported using the 5‐level EuroQol 5‐domain (EQ‐5D‐5L) questionnaire (30), which was scored using the Danish crosswalk value set (ranges from –0.624 to 1.000, corresponding to worst to best health), while mean pain intensity during the last month was reported by participants on a 100‐mm visual analog scale (VAS, range 0 [no pain] to 100 [maximum pain]) (31).
Prognostic factors
Potential prognostic factors were gathered from participants and physical therapists as part of the baseline questionnaires in GLA:D. Among factors available, we included 21 factors covering demographic information, clinical data, and performance‐based physical function (Table 1). The inclusion of factors was based on clinical reasoning and/or published literature suggesting an association with outcomes for the individual conditions, as no previous studies explicitly investigating prognostic factors in individuals with multiple conditions have been identified. A detailed description of factors is available in Supplementary Table 1, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722, while response categories are shown in Table 2.
Table 2.
Baseline characteristics of included participants with and without comorbidities (hypertension, heart and respiratory diseases, diabetes mellitus, and depression) in the analyses for EQ‐5D/VAS and 40‐meter FPWT, respectively*
| Characteristic | EQ‐5D and VAS | 40‐meter FPWT | ||
|---|---|---|---|---|
| OA only (n = 19,927) | OA + comorbidities (n = 17,649) | OA only (n = 18,793) | OA + comorbidities (n = 16,703) | |
| Age, mean ± SD years | 63.2 ± 10.0 | 67.4 ± 8.8 | 63.3 ± 10.0 | 67.4 ± 8.7 |
| Female | 14,759 (74) | 12,024 (68) | 13,910 (74) | 11,373 (68) |
| BMI, mean ± SD kg/m2 | 27.4 ± 5.0 | 29.4 ± 5.5 | 27,4 ± 5.0 | 29.4 ± 5.5 |
| Educational level | ||||
| Primary school | 2,954 (15) | 3,784 (21) | 2,785 (15) | 3,598 (22) |
| Secondary school | 2,168 (11) | 2,025 (11) | 2,040 (11) | 1,926 (12) |
| Short‐term education† | 3,978 (20) | 3,446 (20) | 3,754 (20) | 3,248 (19) |
| Middle‐term education‡ | 8,318 (42) | 6,578 (37) | 7,853 (42) | 6,224 (37) |
| Long‐term education§ | 2,505 (13) | 1,809 (10) | 2,357 (13) | 1,700 (10) |
| Born in Denmark | 19,109 (96) | 17,011 (96) | 18,023 (96) | 16,089 (96) |
| Living alone | 4,583 (23) | 5,106 (29) | 4,323 (23) | 4,843 (29) |
| Employment status | ||||
| Employed/student | 7,814 (39) | 3,801 (22) | 7,333 (39) | 3,577 (21) |
| On sick leave (full or part time) | 908 (5) | 847 (5) | 860 (5) | 796 (5) |
| Unemployed | 395 (2) | 363 (2) | 374 (2) | 343 (2) |
| Retired (including if due to inability to work) | 10,808 (54) | 12,637 (72) | 10,223 (54) | 11,985 (72) |
| Smoking | 1,843 (9) | 1,592 (9) | 1,760 (9) | 1,506 (9) |
| Physical activity level | ||||
| Inactive | 405 (2) | 604 (3) | 377 (2) | 554 (3) |
| Low (e.g., walking and limited housework) | 5,051 (25) | 5,843 (33) | 4,766 (25) | 5,531 (33) |
| Moderate (e.g., swimming and unlimited housework) | 7,391 (37) | 6,425 (36) | 6,994 (37) | 6,068 (36) |
| High (e.g., prolonged biking and fitness) | 5,587 (28) | 3,937 (22) | 5,258 (28) | 3,754 (22) |
| Very high (e.g., running, tennis, and skiing) | 1,486 (7) | 831 (5) | 1,389 (7) | 784 (5) |
| Pain in other knee or hip joints | ||||
| Bilateral joint (from most affected joint) | 6,979 (35) | 6,180 (35) | 6,576 (35) | 5,868 (35) |
| Hip or knee¶ | 3,334 (17) | 3,268 (19) | 3,147 (17) | 3,092 (19) |
| Bilateral joint and hip or knee¶ | 827 (4) | 882 (5) | 787 (4) | 839 (5) |
| Duration of symptoms in most affected joint | ||||
| 0–6 months | 5,173 (29) | 4,649 (29) | 4,855 (29) | 4,364 (29) |
| 7–12 months | 3,544 (20) | 2,976 (19) | 3,317 (20) | 2,788 (19) |
| 13–48 months | 5,160 (29) | 4,636 (29) | 4,890 (29) | 4,426 (30) |
| >48 months | 3,828 (22) | 3,504 (22) | 3,620 (22) | 3,331 (22) |
| Prior injury in most affected joint | 9,163 (47) | 7,998 (46) | 8,643 (47) | 7,553 (46) |
| Prior surgery in most affected joint | 4,603 (23) | 3,475 (20) | 4,353 (23) | 3,294 (20) |
| No. of bodily pain areas | ||||
| 1–2 | 6,351 (44) | 5,189 (41) | 5,980 (44) | 4,920 (41) |
| 3–4 | 3,844 (26) | 3,468 (27) | 3,645 (27) | 3,31 (28) |
| 5–6 | 1,793 (12) | 1,611 (13) | 1,694 (12) | 1,533 (13) |
| ≥7 | 2,132 (15) | 2,045 (16) | 2,013 (15) | 1,922 (16) |
| No. of other comorbidities# | ||||
| 1 | 4,749 (24) | 4,864 (28) | 4,468 (24) | 4,612 (28) |
| 2 | 788 (4) | 1,100 (6) | 738 (4) | 1,043 (6) |
| ≥3 | 136 (1) | 234 (1) | 132 (1) | 216 (1) |
| Use of analgesics within recent 3 months** | 12,102 (61) | 11,917 (68) | 11.452 (61) | 11.295 (68) |
| Anxious/depressed | ||||
| Slightly | 3,212 (16) | 3,081 (17) | 3,034 (16) | 2,904 (17) |
| Moderately | 784 (4) | 1,110 (6) | 731 (4) | 1,047 (6) |
| Severely/extremely | 120 (1) | 292 (2) | 114 (1) | 276 (2) |
| Self‐efficacy (ASES), mean ± SD | 69.4 ± 17.6 | 65.7 ± 18.0 | 69.4 ± 17.6 | 65.7 ± 18.0 |
| Self‐rated health (EQ‐5D VAS), mean ± SD | 71.2 ± 18.6 | 67.1 ± 19.0 | 71.2 ± 18.5 | 67.2 ± 19.0 |
| 40‐meter FPWT, mean ± SD meters/second | 1.55 ± 0.33 | 1.41 ± 0.32 | 1.55 ± 0.33 | 1.41 ± 0.32 |
| Quality of life (EQ‐5D), mean ± SD (range –0.624 to 1.000) | 0.719 ± 0.110 | 0.697 ± 0.120 | 0.719 ± 0.110 | 0.698 ± 0.120 |
| Pain intensity (VAS), mean ± SD (range 0–100 mm) | 46.1 ± 22.0 | 49.2 ± 21.9 | 46.1 ± 22.0 | 49.2 ± 21.9 |
Values are the number (%) unless indicated otherwise. ASES = Arthritis Self‐Efficacy Scale; BMI = body mass index; EQ‐5D = EuroQol 5‐domain (questionnaire); FPWT = Fast‐Paced Walk Test; OA = osteoarthritis; VAS = visual analog scale.
Less than 3 years after secondary school.
Three to 4 years after secondary school.
At least 5 years after secondary school.
Hip if most affected joint is knee and vice versa.
Other comorbidities than hypertension, heart and respiratory diseases, diabetes mellitus, and depression (i.e., ulcer or another bowel disease, kidney or liver disease, anemia or blood disease, cancer, rheumatoid arthritis, neurologic disease, other medical diseases).
At least 1 of the following: acetaminophen, oral or topical nonsteroidal antiinflammatory drug, morphine, tramadol, or codeine.
Statistical analysis
Multiple imputation
Missing values for the prognostic factors at baseline and the outcomes at follow‐up were imputed for each comorbidity group separately using multiple imputation with chained equations (32) under the assumption of data being missing at random (33). The multiple imputation models included all outcomes (i.e., 40‐meter FPWT, EQ‐5D‐5L, and VAS) at baseline and follow‐up, all prognostic factors, and as auxiliary variable (i.e., not included in the primary models) the most affected joint variable was included. The assumption of linearity for continuous and ordered categorical factors was visually examined using scatter plots with lowess smoothing line (34), and if unfulfilled, they were included as categorical factors. Continuous variables were imputed using linear regression, semicontinuous variables (i.e., variables in which a substantial fraction of values are equal, e.g., 1 or 0; in this study, the variables EQ‐5D‐5L and VAS scores) using predictive mean matching, ordered categorical factors using ordered logistic regression, nominal categorical factors using multinomial logistic regression, and binary variables using logistic regression. The MI package in Stata, version 16.1, was used to generate 35 and 25 imputed data sets approximately equal to the proportion of missing data for FPWT, EQ‐5D‐5L, and VAS scores, respectively (Table 1) (32).
Main analyses
In the main analyses, potential prognostic factors’ association with outcome immediately after the GLA:D program were investigated using linear regression. For each outcome (i.e., the dependent variable) a linear regression model was fitted in the group with OA only and additionally in each comorbidity group separately. Both univariable (i.e., including each prognostic factor separately; see Supplementary Tables 2–4, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722) and multivariable regression models (i.e., including all factors) were fitted (Table 1). Continuous and ordered categorical factors were handled as continuous in the models after visual inspection using scatter plots revealed no serious nonlinearity, except for bodily pain, which was included as categorical. Certain categories in categorical factors were collapsed due to sparse data. All models were adjusted for baseline outcome scores and fitted across the imputed data sets with coefficients and 95% confidence intervals (95% CIs) estimated using Rubin's rules (35). The models that included individuals with OA only were considered as primary, while any differences in prognostic effects of factors between these and the specific comorbidity groups were tested using interaction terms between the specific factors and groups (no comorbidities versus comorbidities), with P values less than or equal to 0.05 considered statistically significant.
In addition to the main analyses, 3 sensitivity analyses were carried out. First, the impact of combining individuals with knee and hip OA was explored by repeating all analyses excluding those with hip OA. Secondly, the impact of combining patterns with and without the coexistence of hypertension was explored by repeating analyses conditioning for hypertension; and last, a complete case analysis using only participants with complete data at all time points was conducted to explore the impact of missing data (Table 1).
Prior to all analyses, exposure variables (i.e., prognostic factors) were investigated for collinearity by calculating variance inflation factors (VIFs). The level of collinearity was considered unproblematic if the mean VIF and individual VIFs were ≤5 and ≤10, respectively (36). Model assumptions (i.e., residuals being normally distributed, linearity, and homoscedasticity) were assessed using quantile–quantile plots and scatter plots of residuals and predicted values.
All analyses were conducted using Stata, version 16.1. Before commencing the analyses, the statistical analysis plan was uploaded to the Open Science Framework, where the statistical syntaxes used for the main analyses are also available (https://osf.io/ua5rp/).
RESULTS
Data from a total of 35,496 (40‐meter FPWT) and 37,576 (EQ‐5D‐5L and VAS) participants were included in the analyses, with missing data for the outcomes at follow‐up ranging between 26% and 34%, whereas missingness for most prognostic factors was <1% (Table 1). Participants having the comorbidities of interest were on average a little older than the OA‐only participants and had slightly poorer baseline scores in all 3 outcome measures (Table 2). Characteristics of each comorbidity group separately are available in Supplementary Tables 5 and 6, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722.
Physical function (40‐meter FPWT)
The average improvement in the 40‐meter FPWT for all included participants from before to after the program was 0.12 meters/second (95% CI 0.12, 0.13). Of the 21 potential prognostic factors, older age and female sex were associated with less improvement (–0.04 [95% CI –0.04, –0.03] and –0.04 [95% CI –0.05, –0.03], respectively), which did not statistically differ in the comorbidity groups (P > 0.05), apart from the hypertension and heart disease groups, where the association for female sex was weaker than in the OA‐alone group, and for the hypertension group alone, in which the association of older age was stronger than in the OA‐alone group (Figure 1). Being anxious or depressed was associated with greater improvement in the diabetes mellitus group but not in any of the other comorbidity groups, while estimates for prior injury, self‐rated health, and bodily pain areas indicated some association in the heart plus respiratory disease and heart disease plus diabetes mellitus groups, accompanied, however, by wide confidence intervals, indicating large uncertainties concerning the real effect sizes (Figure 1). Other factors were not associated with change in 40‐meter FPWT scores in the OA‐alone group, with coefficients close to 0.00, and for the majority, statistically insignificant (P > 0.05) (Figure 1). Results from sensitivity analyses did not alter the interpretation of the results (see Supplementary Tables 7–9, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722).
Figure 1.
Results from multivariable linear regression for associations with change in function (40‐meter Fast‐Paced Walk Test [FPWT]) based on 35 imputed data sets (complete data available in Supplementary Table 16, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722). As an example, the figure shows that female patients on average improve (i.e., increase) their walking speed 0.04 meters/second less than male patients among those having osteoarthritis (OA) alone. Error bars indicate 95% confidence intervals; circles represent beta. a = difference per category on a 5‐point scale ranging from primary school to long‐term education; b = includes full‐ or part‐time sick leave; c = includes if retired due to inability to work; d = difference per point on a 5‐point scale ranging from inactive to very high activity level (i.e., running, tennis, etc.); e = hip if most affected joint is knee and vice versa; f = difference per point on a 4‐point scale ranging 0–6 to >48 months; g = self‐reported in most affected joint; h = difference per point on a 4‐point scale ranging 0 to ≥3 (other comorbidities than the groups); i = at least 1 of the following: acetaminophen, oral or topical nonsteroidal antiinflammatory drug, morphine, tramadol, or codeine; j = difference per point on a 4‐point scale ranging from not at all to severely/extremely. Test for interaction between OA‐alone and comorbidity group: k = P ≤ 0.05; l = P ≤ 0.01; m = P ≤ 0.001. ASES = Arthritis Self‐Efficacy Scale (pain and other symptoms); Bilat. = bilateral; BMI = body mass index; DM = diabetes mellitus; EQ‐5D = EuroQol 5‐domain (questionnaire); HD = heart disease; HT = hypertension; RD = respiratory disease; VAS = visual analog scale.
Health‐related quality of life (EQ‐5D‐5L)
Participants improved on average (0.038 [95% CI 0.036, 0.040]) on the EQ‐5D‐5L from before to after the treatment program. In the OA‐alone group, 14 of the 21 factors showed a weak (but statistically significant) association with change in outcome, with female sex (0.017 [95% CI 0.013, 0.021]), self‐efficacy (0.009 [95% CI 0.008, 0.010]), self‐rated health (0.004 [95% CI 0.003, 0.005]), and pain intensity (–0.004 [95% CI –0.005, –0.003]) having the strongest associations with change (Figure 2). The results were similar in most comorbidity groups (no statistically significant differences from the OA‐alone group), however, there were a few exceptions. In contrast to the OA‐alone group, higher age and smoking were associated with larger improvements in the depression group and diabetes mellitus group, respectively. Further, higher education and function (40‐meter FPWT) showed a stronger association with improvement in the diabetes mellitus and depression groups, while pain in other joints and comorbidities were associated with less improvement in the heart disease and heart plus respiratory disease groups, respectively, than in the OA‐alone group. Other differences were either minor or included imprecise estimates (Figure 2). In the complete case analysis, the observed association of higher age with change in the depression group nearly vanished, while other results in all 3 sensitivity analyses were as in the main analyses (see Supplementary Tables 10–12, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722).
Figure 2.
Results from multivariable linear regression for associations with change in quality of life (EQ‐5D) based on 25 imputed data sets (complete data available in Supplementary Table 17, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722). As an example, the figure shows that female patients on average improve their quality of life 0.017 points more than male patients among those having osteoarthritis (OA). Error bars indicate 95% confidence intervals; circles represent beta. a = difference per category on a 5‐point scale ranging from primary school to long‐term education; b = includes full‐ or part‐time sick leave; c = includes if retired due to inability to work; d = difference per point on a 5‐point scale ranging from inactive to very high activity level (i.e., running, tennis, etc.); e = hip if most affected joint is knee and vice versa; f = difference per point on a 4‐point scale ranging 0–6 to >48 months; g = self‐reported in most affected joint; h = difference per point on a 4‐point scale ranging 0 to ≥3 (other comorbidities than the groups); i = at least 1 of the following: acetaminophen, oral or topical nonsteroidal antiinflammatory drug, morphine, tramadol, or codeine; j = difference per point on a 4‐point scale ranging from not at all to severely/extremely. Test for interaction between OA‐alone and comorbidity group: k = P ≤ 0.05; l = P ≤ 0.01; m = P ≤ 0.001 (see Figure 1 for other definitions).
Pain intensity (VAS)
Pain intensity decreased on average –12.2 mm (95% CI –12.6, –11.9) on the VAS scale from before to after the GLA:D program. About one‐half of the factors were statistically significantly associated with change in the OA‐alone group (Figure 3). The strongest associations were observed for higher age and longer duration of symptoms, which were associated with less pain relief (0.68 [95% CI 0.11, 1.26] and 1.32 [95% CI 1.01, 1.64], respectively), whereas higher self‐efficacy (1.50 [95% CI –1.72, –1.28]), self‐rated health (–0.63 [95% CI –0.84, –0.42]), and physical function (–3.73 [95% CI –4.94, –2.52]) were associated with greater reduction in pain intensity. In a few comorbidity groups, some associations differed from the OA‐alone group. In the diabetes mellitus group, the association of smoking with change was more pronounced and reversed to greater reduction in pain intensity than in the OA‐alone group. Also, associations of pain in other joints, education, and body mass index (BMI) differed statistically significantly in the respiratory disease, heart plus respiratory disease, and heart disease plus diabetes mellitus groups, respectively; however, estimates for these included wide confidence intervals, limiting clear interpretations of the real associations (Figure 3). Results from sensitivity analyses did not alter the interpretation of the results (see Supplementary Tables 13–15, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722).
Figure 3.
Results from multivariable linear regression for associations with change in pain (VAS) based on 25 imputed data sets (complete data available in Supplementary Table 18, available on the Arthritis Care & Research website at http://onlinelibrary.wiley.com/doi/10.1002/acr.24722). As an example, the figure shows that those having had prior surgery on average improve (i.e., decrease) their pain intensity 1.38 mm less than those not having had prior surgery among those having osteoarthritis (OA). Error bars indicate 95% confidence intervals; circles represent beta. a = difference per category on a 5‐point scale ranging from primary school to long‐term education; b = includes full‐ or part‐time sick leave; c = includes if retired due to inability to work; d = difference per point on a 5‐point scale ranging from inactive to very high activity level (i.e., running, tennis, etc.); e = hip if most affected joint is knee and vice versa; f = difference per point on a 4‐point scale ranging 0–6 to >48 months; g = self‐reported in most affected joint; h = difference per point on a 4‐point scale ranging 0 to ≥3 (other comorbidities than the groups); i = at least 1 of the following: acetaminophen, oral or topical nonsteroidal antiinflammatory drug, morphine, tramadol, or codeine; j = difference per point on a 4‐point scale ranging from not at all to severely/extremely. Test for interaction between OA‐alone and comorbidity group: k = P ≤ 0.05; l = P ≤ 0.01; m = P ≤ 0.001 (see Figure 1 for other definitions).
DISCUSSION
Among 21 potential prognostic factors for health outcomes in individuals with knee or hip OA with or without comorbidities, the number of factors statistically significantly associated with outcomes after exercise therapy and education varied greatly, and most associations were weak. This was generally observed across all groups irrespective of coexisting comorbidities. The most notable exceptions were higher age and pain intensity, which were associated with less improvement, while higher self‐efficacy and self‐rated health were associated with larger improvement after exercise therapy and education.
This is the first investigation of prognostic factors for health outcomes following an exercise therapy and education program in knee or hip OA stratified for specific concomitant chronic conditions. Previous studies have assessed individuals with OA with and without comorbidities together assuming prognostic effects of factors being equal across all OA comorbidity groups (6, 7, 8, 9, 10). That might be why a recent study using data from the GLA:D registry failed to find individualized predictions of changes in different health outcomes based on baseline characteristics better than predictions based on group average improvements (37).
Few factors were associated with change in performance‐based physical function (40‐meter FPWT), which may partly be explained by the overall relatively small observed improvement leaving less room for difference in change within the different factors. The factors age, sex, and anxiety/depression were the ones having a potentially relevant association (i.e., not close to null), however, the latter only in individuals with OA and concurrent diabetes mellitus. Similar findings for age and sex were found in a previous study on individuals with heart conditions (22), whereas no previous studies on prognostic factors for performance‐based outcomes in OA have been identified.
Several factors were statistically significantly associated with change in quality of life (EQ‐5D‐5L), including self‐efficacy, self‐rated health, and pain intensity in individuals with OA alone and across all comorbidity groups, and higher education level and performance‐based physical function in the groups with coexisting depression and diabetes mellitus, respectively. We were unable to identify any studies investigating prognostic factors for quality of life after exercise therapy in individuals with OA; however, a previous study in individuals with diabetes mellitus found that higher BMI was associated with poorer quality of life (23), which was unsupported by our study.
Previous studies on individuals with knee or hip OA have found female sex, higher education and physical activity level, fewer comorbidities, anxiety, younger age, better physical performance, shorter duration of symptoms, living with a partner, no previous surgery, no use of analgesics, and no obesity to be associated with greater pain relief after a treatment including exercise therapy (7, 8, 9, 10). Associations of the latter 7 are supported by our findings. Contrary to our study, participants in 2 of the 4 previous studies also received other treatment modalities than exercise therapy, and these 2 studies also used univariable analyses for screening of factors (7, 8), which is generally not recommended and known to produce selection bias (38). Supporting this, the 5 first factors previously found associated with greater pain relief were all significantly associated with pain relief in our univariable analyses but not in the main multivariable analyses. Furthermore, in 3 of the studies, most prognostic factors were dichotomized (7, 8, 9), which increases the risk of spurious findings (39, 40). Taken together, this may explain some of the discrepancies between previous studies and our findings. Adding to previous findings, we found that higher self‐efficacy, self‐rated health, and performance‐based physical function were associated with greater pain relief, and that being a smoker was associated with less pain relief, apart from the diabetic group, where smoking had the opposite effect.
Although many factors were statistically significantly associated with improvements in health outcomes in the present study, the number varied greatly between the different health outcomes, and the clinical importance of most associations may be limited. For the 40‐meter FPWT, taking the range of possible values of age, sex, and anxiety/depression in the present study into account, only younger age and anxiety/depression may be associated with clinically important improvements in physical function, as other associations were far from the 0.20–0.30 meters/second previously reported as clinically important in individuals with hip OA (41). The minimal clinically important worsening and improvement in EQ‐5D‐5L score has been suggested to be 0.05 and 0.07, respectively, in individuals with knee and hip OA (30), which means that only associations found for self‐efficacy, self‐rated health, and pain intensity may be clinically important prognostic factors for quality of life in individuals with OA alone and across comorbidity groups. Additionally, higher education level and performance‐based physical function were associated with a clinically relevant improvement in quality of life in the groups with coexisting depression and diabetes mellitus, respectively, but not in any of the other comorbidity groups or the OA‐alone group. For pain intensity, changes between 7 mm and 37 mm (range 0–100 mm), depending on the baseline level of pain, may be considered clinically relevant (42), meaning that self‐efficacy may be associated with clinically relevant changes irrespective of comorbidities. The same is the case for smoking, although only in individuals with coexisting diabetes mellitus.
Improvements following exercise therapy may be less affected by individuals’ characteristics and more a result of the treatment itself and/or contextual factors (43), thus observed differences in treatment response may just be due to random variance. Nevertheless, based on our results, a focus on nonmodifiable factors such as age, education, etc., may help to identify those individuals in need of more support than others, and including education aimed at improving self‐efficacy targeted those with low self‐efficacy at baseline may potentially optimize the use of resources and improve outcomes following exercise and education in individuals with OA with and without comorbidities. Furthermore, our findings suggest that the presence of certain comorbidities has an effect‐modifying effect on the association between some prognostic factors and outcomes, which should be considered when attempting to predict a treatment response. Yet, given the explorative nature of the present study, and because it is one of the first studies in the area, results need to be confirmed. Factors not considered in the present study, such as kinesiophobia and pain catastrophizing, which have shown prognostic abilities in other populations (44, 45), should ideally be included in future studies.
Several limitations warrant consideration in the present study. First, misclassification of prognostic factors may have occurred and potentially affected any association. However, the degree of misclassification is likely low, as most factors were self‐reported, and furthermore, they are unlikely to be associated with outcomes (i.e., nondifferential). Second, the accuracy of comorbidities being self‐reported may be debated, and we were unable to distinguish between different heart and respiratory diseases and types 1 and 2 diabetes mellitus; thus, it is unknown what specific diseases these categories mainly comprise. Third, as we included many factors without any predefined hypotheses, the risk of chance findings is inherently increased. However, because all factors were included and retained regardless of their P value, the significance level of 0.05 for each factor was maintained, and data‐driven approaches that are known to produce selection bias (38) were reduced. However, caution should be taken when interpreting the single estimates. Fourth, as this is a single‐arm study, it is unknown if the observed prognostic factors are prognostic regardless of the specific treatment or may be predictive of differential treatment response. Furthermore, as all participants in the present study received neuromuscular exercises therapy, it is unknown whether the prognostic ability of factors differs in other exercise programs.
The strengths of this study include the large sample size, the inclusion of a large number of potential prognostic factors that were mutually adjusted for in the analyses, and handling the majority of factors in their original format as opposed to categorization, which has been shown to be suboptimal (39, 40). Furthermore, the fact that the data are from a nationwide clinical registry including patients from urban and rural settings supports the generalizability of the findings to clinical practice.
In conclusion, higher age and pain intensity displayed clinically relevant associations with less improvement, while higher self‐efficacy and self‐rated health displayed clinically relevant associations with larger improvement in different health outcomes following an 8‐week exercise therapy and patient education program in individuals with knee or hip OA alone and across different comorbidity groups. Additionally, anxiety/depression, education, physical function, and smoking seem to be prognostic in subgroups with diabetes mellitus and depression. The identified prognostic factors may be used clinically to better target individuals with OA who gain most from an exercise and education program.
AUTHOR CONTRIBUTIONS
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. Dr. Pihl 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
Pihl, Roos, Taylor, Skou.
Acquisition of data
Roos, Grønne, Skou.
Analysis and interpretation of data
Pihl, Taylor, Skou.
Supporting information
Disclosure Form
Table S1 Supplementary Tables
Supported by the European Research Council under the European Union's Horizon 2020 Research and Innovation Programme (project MOBILIZE; grant agreement no. 801790), Næstved, Slagelse, and Ringsted Hospitals’ Research Fund, and the Association of Danish Physiotherapists Research Fund. The start‐up phase of GLA:D was partly funded by the Danish Physiotherapy Association's Fund for Research, Education and Practice Development, the Danish Rheumatism Association, and the Physiotherapy Practice Foundation. Dr. Skou is recipient of a program grant from Region Zealand (Exercise First) and an ongoing grant from the European Research Council under the European Union's Horizon 2020 Research and Innovation Programme (project ESCAPE; grant agreement no. 945377).
Author disclosures are available at https://onlinelibrary.wiley.com/action/downloadSupplement?doi=10.1002%2Facr.24722&file=acr24722‐sup‐0001‐Disclosureform.pdf.
REFERENCES
- 1. Bannuru RR, Osani MC, Vaysbrot EE, Arden NK, Bennell K, Bierma‐Zeinstra SM, et al. OARSI guidelines for the non‐surgical management of knee, hip, and polyarticular osteoarthritis. Osteoarthritis Cartilage 2019;27:1578–89. [DOI] [PubMed] [Google Scholar]
- 2. Fernandes L, Hagen KB, Bijlsma JW, Andreassen O, Christensen P, Conaghan PG, et al. EULAR recommendations for the non‐pharmacological core management of hip and knee osteoarthritis. Ann Rheum Dis 2013;72:1125–35. [DOI] [PubMed] [Google Scholar]
- 3. Fransen M, McConnell S, Hernandez‐Molina G, Reichenbach S. Exercise for osteoarthritis of the hip. Cochrane Database Syst Rev 2014:Cd007912. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Fransen M, McConnell S, Harmer AR, van der Esch M, Simic M, Bennell KL. Exercise for osteoarthritis of the knee: a Cochrane systematic review. Br J Sports Med 2015;49:1554–7. [DOI] [PubMed] [Google Scholar]
- 5. Hay E, Dziedzic K, Foster N, Peat G, van der Windt D, Bartlam B, et al. Programme grants for applied research. In: Optimal primary care management of clinical osteoarthritis and joint pain in older people: a mixed‐methods programme of systematic reviews, observational and qualitative studies, and randomised controlled trials Southampton (UK): NIHR Journals Library 2018:1–260. [PubMed] [Google Scholar]
- 6. French HP, Galvin R, Cusack T, McCarthy GM. Predictors of short‐term outcome to exercise and manual therapy for people with hip osteoarthritis. Phys Ther 2014;94:31–9. [DOI] [PubMed] [Google Scholar]
- 7. Weigl M, Angst F, Aeschlimann A, Lehmann S, Stucki G. Predictors for response to rehabilitation in patients with hip or knee osteoarthritis: a comparison of logistic regression models with three different definitions of responder. Osteoarthritis Cartilage 2006;14:641–51. [DOI] [PubMed] [Google Scholar]
- 8. Wright AA, Cook CE, Flynn TW, Baxter GD, Abbott JH. Predictors of response to physical therapy intervention in patients with primary hip osteoarthritis. Phys Ther 2011;91:510–24. [DOI] [PubMed] [Google Scholar]
- 9. Legha A, Burke DL, Foster NE, van der Windt DA, Quicke JG, Healey EL, et al. Do comorbidities predict pain and function in knee osteoarthritis following an exercise intervention, and do they moderate the effect of exercise? Analyses of data from three randomized controlled trials. Musculoskeletal Care 2020;18:3–11. [DOI] [PubMed] [Google Scholar]
- 10. Dell'Isola A, Jönsson T, Nero H, Eek F, Dahlberg L. Factors associated with the outcome of a first‐line intervention for patients with hip or knee osteoarthritis or both: data from the BOA register. Phys Ther 2020;100:1771–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Wesseling J, Welsing PM, Bierma‐Zeinstra SM, Dekker J, Gorter KJ, Kloppenburg M, et al. Impact of self‐reported comorbidity on physical and mental health status in early symptomatic osteoarthritis: the CHECK (Cohort Hip and Cohort Knee) study. Rheumatology (Oxford) 2013;52:180–8. [DOI] [PubMed] [Google Scholar]
- 12. Muckelt PE, Roos E, Stokes M, McDonough S, Grønne D, Ewings S, et al. Comorbidities and their link with individual health status: a cross‐sectional analysis of 23,892 people with knee and hip osteoarthritis from primary care. J Comorb 2020;10:2235042X20920456. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Kadam UT, Jordan K, Croft PR. Clinical comorbidity in patients with osteoarthritis: a case‐control study of general practice consulters in England and Wales. Ann Rheum Dis 2004;63:408–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Schafer I, Kaduszkiewicz H, Wagner HO, Schon G, Scherer M, van den Bussche H. Reducing complexity: a visualisation of multimorbidity by combining disease clusters and triads. BMC Public Health 2014;14:1285. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. GBD 2016 Disease and Injury Incidence and Prevalence Collaborators . Global, regional, and national incidence, prevalence, and years lived with disability for 328 diseases and injuries for 195 countries, 1990–2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet 2017;390:1211–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Pedersen BK, Saltin B. Exercise as medicine: evidence for prescribing exercise as therapy in 26 different chronic diseases. Scand J Med Sci Sports 2015;25 Suppl 3:1–72. [DOI] [PubMed] [Google Scholar]
- 17. Anderson L, Thompson DR, Oldridge N, Zwisler AD, Rees K, Martin N, et al. Exercise‐based cardiac rehabilitation for coronary heart disease. Cochrane Database Syst Rev 2016:Cd001800. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Huang G, Shi X, Gibson CA, Huang SC, Coudret NA, Ehlman MC. Controlled aerobic exercise training reduces resting blood pressure in sedentary older adults. Blood Press 2013;22:386–94. [DOI] [PubMed] [Google Scholar]
- 19. Thomas DE, Elliott EJ, Naughton GA. Exercise for type 2 diabetes mellitus. Cochrane Database Syst Rev 2006:Cd002968. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. McCarthy B, Casey D, Devane D, Murphy K, Murphy E, Lacasse Y. Pulmonary rehabilitation for chronic obstructive pulmonary disease. Cochrane Database Syst Rev 2015:Cd003793. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Cooney GM, Dwan K, Greig CA, Lawlor DA, Rimer J, Waugh FR, et al. Exercise for depression. Cochrane Database Syst Rev 2013:Cd004366. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Uddin J, Zwisler AD, Lewinter C, Moniruzzaman M, Lund K, Tang LH, et al. Predictors of exercise capacity following exercise‐based rehabilitation in patients with coronary heart disease and heart failure: a meta‐regression analysis. Eur J Prev Cardiol 2016;23:683–93. [DOI] [PubMed] [Google Scholar]
- 23. Dixit S, Maiya A, Shastry B. Effect of aerobic exercise on quality of life in population with diabetic peripheral neuropathy in type 2 diabetes: a single blind, randomized controlled trial. Qual Life Res 2014;23:1629–40. [DOI] [PubMed] [Google Scholar]
- 24. De Rooij M, van der Leeden M, Cheung J, van der Esch M, Häkkinen A, Haverkamp D, et al. Efficacy of tailored exercise therapy on physical functioning in patients with knee osteoarthritis and comorbidity: a randomized controlled trial. Arthritis Care Res (Hoboken) 2017;69:807–16. [DOI] [PubMed] [Google Scholar]
- 25. Skou ST, Roos EM. Good Life with osteoArthritis in Denmark (GLA:D): evidence‐based education and supervised neuromuscular exercise delivered by certified physiotherapists nationwide. BMC Musculoskelet Disord 2017;18:72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Ageberg E, Link A, Roos EM. Feasibility of neuromuscular training in patients with severe hip or knee OA: the individualized goal‐based NEMEX‐TJR training program. BMC Musculoskelet Disord 2010;11:126. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Vandenbroucke JP, von Elm E, Altman DG, Gotzsche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Int J Surg 2014;12:1500–24. [DOI] [PubMed] [Google Scholar]
- 28. Altman DG, McShane LM, Sauerbrei W, Taube SE. Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK): explanation and elaboration. PLoS Med 2012;9:e1001216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Dobson F, Hinman RS, Roos EM, Abbott JH, Stratford P, Davis AM, et al. OARSI recommended performance‐based tests to assess physical function in people diagnosed with hip or knee osteoarthritis. Osteoarthritis Cartilage 2013;21:1042–52. [DOI] [PubMed] [Google Scholar]
- 30. Bilbao A, Garcia‐Perez L, Arenaza JC, Garcia I, Ariza‐Cardiel G, Trujillo‐Martin E, et al. Psychometric properties of the EQ‐5D‐5L in patients with hip or knee osteoarthritis: reliability, validity and responsiveness. Qual Life Res 2018;27:2897–908. [DOI] [PubMed] [Google Scholar]
- 31. Hawker GA, Mian S, Kendzerska T, French M. Measures of adult pain: visual analog scale for pain (VAS pain), numeric rating scale for pain (NRS pain), McGill Pain Questionnaire (MPQ), Short‐Form McGill Pain Questionnaire (SF‐MPQ), Chronic Pain Grade Scale (CPGS), Short Form‐36 bodily pain scale (SF‐36 BPS), and Measure of Intermittent and Constant Osteoarthritis Pain (ICOAP). Arthritis Care Res (Hoboken) 2011;63 Suppl 11:S240–52. [DOI] [PubMed] [Google Scholar]
- 32. White IR, Royston P, Wood AM. Multiple imputation using chained equations: issues and guidance for practice. Stat Med 2011;30:377–99. [DOI] [PubMed] [Google Scholar]
- 33. Little RJ, Rubin DB. Statistical analysis with missing data. 2nd edition. Hoboken (NJ): Wiley; 2002. [Google Scholar]
- 34. William SC. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 1979;74:829–36. [Google Scholar]
- 35. Rubin DB. Multiple imputation for nonresponse in surveys. New York: J. Wiley; 1987. [Google Scholar]
- 36. O'Brien RM. A caution regarding rules of thumb for variance inflation factors. Qual Quant 2007;41:673–90. [Google Scholar]
- 37. Baumbach L, List M, Grønne DT, Skou ST, Roos EM. Individualized predictions of changes in knee pain, quality of life and walking speed following patient education and exercise therapy in patients with knee osteoarthritis ‐ a prognostic model study. Osteoarthritis Cartilage 2020;28:1191–201. [DOI] [PubMed] [Google Scholar]
- 38. Royston P, Moons KG, Altman DG, Vergouwe Y. Prognosis and prognostic research: developing a prognostic model. BMJ 2009;338:b604. [DOI] [PubMed] [Google Scholar]
- 39. Altman DG, Royston P. The cost of dichotomising continuous variables. BMJ 2006;332:1080. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Royston P, Altman DG, Sauerbrei W. Dichotomizing continuous predictors in multiple regression: a bad idea. Stat Med 2006;25:127–41. [DOI] [PubMed] [Google Scholar]
- 41. Bennell K, Dobson F, Hinman R. Measures of physical performance assessments: Self‐Paced Walk Test (SPWT), Stair Climb Test (SCT), Six‐Minute Walk Test (6MWT), Chair Stand Test (CST), Timed Up & Go (TUG), Sock Test, Lift and Carry Test (LCT), and Car Task. Arthritis Care Res (Hoboken) 2011;63 Suppl 11:S350–70. [DOI] [PubMed] [Google Scholar]
- 42. Tubach F, Ravaud P, Baron G, Falissard B, Logeart I, Bellamy N, et al. Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Ann Rheum Dis 2005;64:29–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Zou K, Wong J, Abdullah N, Chen X, Smith T, Doherty M, et al. Examination of overall treatment effect and the proportion attributable to contextual effect in osteoarthritis: meta‐analysis of randomised controlled trials. Ann Rheum Dis 2016;75:1964–970. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Ang DC, Bair MJ, Damush TM, Wu J, Tu W, Kroenke K. Predictors of pain outcomes in patients with chronic musculoskeletal pain co‐morbid with depression: results from a randomized controlled trial. Pain Med 2010;11:482–91. [DOI] [PubMed] [Google Scholar]
- 45. Helminen EE, Arokoski JP, Selander TA, Sinikallio SH. Multiple psychological factors predict pain and disability among community‐dwelling knee osteoarthritis patients: a five‐year prospective study. Clin Rehabil 2020:269215519900533. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Disclosure Form
Table S1 Supplementary Tables