Predictors of Pain,Function,and Change in Patellofemoral Pain

Abstract

Background:

Identification of factors predictive of outcome and change is important to improve treatment for patellofemoral pain (PFP). Few studies have examined the predictive value of psychological factors in PFP, although they have been reported to be important predictors in other musculoskeletal pain conditions.

Purpose:

To evaluate predictors of pain, function, and change 1 year after an exercise-based intervention in PFP.

Study Design:

Cohort study; Level of evidence, 3.

Methods:

In sum, 112 patients were recruited to a randomized controlled trial; 98 attended 1-year follow-up. There were no between-group differences in the trial; thus, the material was analyzed as 1 cohort. Nine baseline factors—sex, bilateral pain, worst pain, pain duration, Anterior Knee Pain Scale (AKPS), kinesiophobia, anxiety and depression, self-efficacy, and number of pain sites throughout the body—were investigated for their predictive ability on outcome at 1 year (AKPS, worst pain) and for change at 1 year (global change score, change in AKPS, and change in worst pain). Multivariable linear regression models with stepwise backward removal method were used to find predictors of poor outcome.

Results:

Number of pain sites at baseline was a significant predictor of worse outcome for AKPS (B = −2.7; 95% CI, –4.0 to −1.3; P < .01), worst pain (B = 0.5; 95% CI, 0.2-0.8; P < .01), global change (B = −0.8; 95% CI, –1.2 to −0.5; P < .01), change in AKPS (B = −2.7; 95% CI, –4.0 to −1.3; P < .01), and change in worst pain (B = 0.5, 95% CI, 0.2-0.8; P < .01) at 1 year. Baseline scores for AKPS and worst pain predicted respective 1-year levels and change scores (P < .01). Lower self-efficacy and male sex predicted less global change (P < .01). Longer pain duration predicted final score and change score for worst pain (P < .01). The predictive models had reasonable fit with adjusted R² from 0.22 to 0.35.

Conclusion:

Higher number of pain sites throughout the body was a consistent predictor of poor outcome and less change at 1 year. Baseline levels for AKPS and worst pain predicted respective final scores and change scores.

Registration:

NCT02114294 (ClinicalTrials.gov identifier).

Keywords

knee physiotherapy patellofemoral pain prognosis widespread pain

Patellofemoral pain (PFP) is defined as pain around or behind the patella that is provoked by activities loading the patellofemoral joint in the absence of other distinct pathology.¹⁴ PFP is common, with annual prevalence estimates of 23% in the general population and 29% in adolescents.⁴⁰ Women are affected approximately twice as commonly as men.⁴⁰ Newer surveys suggest higher degrees of chronicity than previously thought, with the majority of patients still reporting pain at 2- to 8-year follow-up.^30,39,40

Previous studies have identified some factors predictive of poor outcome in patients with PFP. A recent systematic review found that longer pain duration (>4 months) was the most reported prognostic factor of a poor outcome.³⁴ In addition, more intense self-reported symptoms at baseline, as evidenced by low Anterior Knee Pain Scale (AKPS) score, was a consistent predictor of poor outcome in a recent long-term analysis.³⁰ Other factors have been shown to predict poorer outcome, but these predictors are not consistent among studies.^34,38,39 Generally, strength of the evidence is reduced by methodological limitations in original studies (most commonly related to small samples).³⁴

Psychological factors have been found to predict poorer outcomes in other musculoskeletal conditions.^{1,2,18,20,22,32,48,49} Existing studies suggest that psychological factors such as anxiety, depression, catastrophizing, and kinesiophobia may be important in PFP.³³ Two studies reported correlations between concurrent improvements in catastrophizing and kinesiophobia and improvements in pain and function,^16,37 but to our knowledge, the prognostic value of baseline psychological measures on outcomes of pain and function has not previously been examined in PFP.

The aim of the present study was to evaluate predictors of outcomes for function (AKPS), pain, and measures of change after 1 year for patients with PFP who participated in a randomized controlled trial (RCT) comparing 3 exercise-based interventions.

Methods

Study Design

A post hoc analysis was performed with data from a previously reported 1-year follow-up of a single-blind, single-center randomized controlled trial. The trial has been reported, including the protocol.^24,25 Patients (N = 112) were randomly assigned to a 6-week intervention consisting of patient education combined with isolated hip exercise (n = 39), traditional knee-based exercise (n = 37), or free training (n = 36); 98 patients completed 1-year follow-up. The trial was registered with the ClinicalTrials.gov database (NCT02114294). The Ethics Committee Health Region South-East, Norway, granted ethical approval (2013/1860/REKsør-øst). Patients provided informed consent.

Participants

Patients were recruited from referrals by primary care physicians or other medical specialists (eg, orthopaedic surgeons, rheumatologists) to the Department of Physical Medicine and Rehabilitation, Sørlandet Hospital, between September 2014 and September 2017.

Patients were eligible for inclusion if they were 16 to 40 years of age with a minimum 3-month history of PFP (pain ≥3 of 10) reproduced by at least 2 activities (stair ascent or descent, hopping, running, prolonged sitting, squatting, or kneeling) and present on at least 1 clinical test (compression of the patella, palpation of the patellar facets). In patients with bilateral pain, the worst knee was included.

Exclusion criteria were as follows: (1) clinical, radiographic, or magnetic resonance imaging findings indicative of other significant specific pathology, including meniscal, ligament, or cartilage injury, as well as osteoarthritis, apophysitis, significant knee joint effusion, or recurrent patellar subluxation or dislocation; (2) significant pain from hip or back hindering the ability to perform the prescribed exercises; (3) previous surgery to the knee joint; (4) nonsteroidal anti-inflammatory drug or cortisone use over an extended period; (5) previous trauma to the knee joint with an effect on the presenting clinical condition; (6) physiotherapy or other similar exercises for PFP syndrome within the previous 3 months.

Interventions

All patients received the same patient education and were randomized to isolated hip exercise, traditional knee-based exercise, or free training for 6 weeks.²⁴ The patients in the hip and knee groups performed exercises 3 times weekly (1 session under guidance from the physiotherapist and 2 home sessions).

Outcome and Predictor Variables

Data measured at baseline and 1 year were used for this post hoc analysis. Participants filled out the self-report questionnaires without assistance. Observers blinded to group assignment collected all outcomes.

Demographic measures were collected for all patients, including age, sex, body mass index, education, work/sick leave, pain duration, and presence of unilateral or bilateral pain.

The main outcome of the study was the AKPS. This is a self-report questionnaire consisting of 13 questions pertaining to symptoms of anterior knee pain.²⁸ The total score ranges from 0 to 100 (no symptoms).

Secondary outcomes were pain intensity and global change. Pain intensity was measured by a visual analog scale (range, 0-10 [most pain]). Usual pain and worst pain during the past week were measured. A global change score consisting of an 18-point Likert scale from −9 (maximum deterioration) to +9 (maximum improvement) was used to determine patients’ global assessment of change from baseline.⁵

Psychological predictors included kinesiophobia, self-efficacy, and anxiety and depression. The Tampa Scale for Kinesiophobia (TSK) is a 13-item, self-report questionnaire used to assess fear of movement/reinjury (range, 13-52 [most]).²³ The Knee Self-Efficacy Scale (K-SES) was used to assess self-efficacy (range, 0-10 [highest]).⁴⁴ Questions pertain to how certain respondents feel about performing various activities now and in the future. The 10-question Hopkins Symptom Checklist (HSCL-10) assesses symptoms of anxiety and depression (range, 1-4 [most symptoms]).^15,42

Number of pain sites during the past week was assessed with an adaptation of the Standardized Nordic Questionnaire. Patients record presence of pain (yes/no) in 9 body areas (head, neck, shoulder, elbow, hand, upper back, hip, knee, and ankle).^27,29 Only 1 point is scored per bodily area; thus, the score does not differentiate between bilateral and unilateral pain.

Statistical Analysis

All statistical analyses were performed with SPSS (v 23; IBM Corporation). All data were checked for normality and outliers via stem-and-leaf plot and Q-Q plot. Scatter plots were used to inspect correlation data for outliers.

Missing Values

Missing values for AKPS, HSCL-10, K-SES, and TSK were handled as follows: If <25% of items were missing, the values were substituted with the arithmetic mean of values from the available items.⁸ If ≥25% of items were missing, the outcome was regarded as missing for the patient. If an item was missing for the Nordic Questionnaire, the total score was regarded as missing.

Predictive Models

Outcomes were AKPS, worst pain, and global change score at 1 year. At 1 year, there were no significant between-group differences in either primary or secondary outcomes measures in the RCT (data not shown). Thus, for the current study, the data were analyzed as a single cohort.³⁵ Selection of potential predictors was based on findings in previous studies and clinical judgment.⁴³ We limited the number of variables in the model to 9 based on our sample size of 98.²¹ On the basis of this, we chose the following 9 baseline measures as potential predictors of outcome, which were entered in each model: sex, bilateral pain, worst pain, pain duration (categories: 3-6, >6-12, >12-24, >24 months), AKPS, TSK, HSCL-10, K-SES, and number of pain sites. Treatment group was not included as a variable, as there were no between-group differences in the outcomes in question.³⁵ Multivariable models were also tested without the baseline variable for AKPS and worst pain for their 1-year outcomes.

Multivariable linear regression models were then constructed with the outcome in question as the dependent variable and potential predictors as independent variables.⁴¹ A backward stepwise elimination (P≤ .10) was used to identify a group of factors that were associated with outcome at 1 year. The final model is presented with the constant (Y-intercept), the unstandardized regression coefficients (B), and their 95% CIs. P < .05 was used to define statistical significance in the final model. The amount of variance accounted for by the final model is expressed as adjusted R². Models were checked for collinearity through the variance inflation factor.

Results

Of the 317 patients evaluated for inclusion, 47 did not meet inclusion criteria; 87 met exclusion criteria; and 71 declined to participate. Ultimately, 112 patients were randomized to 1 of 3 arms of the intervention, and 98 (88%) completed 1-year follow-up and were included in the present analysis (see flowchart; Figure 1). All groups showed statistically significant improvement from baseline to 1 year for pain and function. The intention-to-treat analysis in the main study did not find any between-group differences for any outcomes at 1 year. Thus, for the current post hoc analysis, we treated the whole group of patients as 1 cohort. Baseline values for the cohort are presented in Table 1. There were few missing data, except for the Nordic pain questionnaire (number of pain sites), in which 6 (5%) total scores were incomplete and treated as missing.

Figure 1.

CONSORT (Consolidated Standards of Reporting Trials) flowchart. VAS, visual analog scale.

Table 1

Baseline Characteristics^a (N = 112)

	Mean ± SD or n (%)
Age, y	27.6 ± 7.3
Sex
Female	73 (65)
Male	39 (35)
Body mass index, kg/m²
<25	51 (46)
25-30	39 (35)
>30	22 (20)
Pain
Unilateral	31 (28)
Bilateral	81 (72)
Symptom duration, mo
3-6	8 (7)
>6-12	23 (21)
>12-24	24 (21)
>24	57 (51)
Education >13 y	33 (29)
Sick listed	15 (13)
Regular use of analgesics	19 (17)
Use of insoles	16 (14)
Use of knee support or brace	19 (17)
VAS, 0-10
Usual pain	4.1 ± 1.9
Worst pain	6.1 ± 2.1
AKPS, 0-100	65.7 ± 12.1
Knee self-efficacy, 0-10	6.1 ± 1.8
Kinesiophobia
Normal to mild, 13-32	88 (78)
Moderate to high, 33-52	24 (21)
No. of pain sites
0-1	29 (26)
2-3	40 (36)
≥4	37 (33)
Missing	6 (5)
Emotionally distressed, HSCL-10 >1.8	30 (27)

AKPS, Anterior Knee Pain Scale; HSCL-10, 10-question Hopkins Symptom Checklist; VAS, visual analog scale.

Anterior Knee Pain Scale

Lower AKPS baseline and higher number of pain sites were significant predictors of lower AKPS at 1 year, accounting for 35% of the variance (Table 2).

Table 2

Multivariable Linear Regression for the Dependent Variable AKPS at 1 Year^a

Predictor	B	95% CI	P Value	R ²
Constant^b	43.7	27.8 to 59.6	<.01	0.35
Sex^c	5.1	−0.3 to 10.5	.06
AKPS baseline^d	0.6	0.4 to 0.8	<.01
No. of pain sites^e	−2.7	−4.0 to −1.3	<.01

Model includes AKPS at baseline. Excluded predictor variables after stepwise elimination: pain duration (0, 3-6 months; 1, >6-12 months; 2, >12-24 months; 3, >2 years); self-efficacy (0-10 [highest self-efficacy]); worst pain, visual analog scale for worst pain in past week (0-10); kinesiophobia (13-52 [most kinesiophobia]); bilateral pain (0, no; 1, yes); Hopkins Symptom Checklist (1-4 [most symptoms]).

Constant: Y-intercept of the model.

Sex: 0, male; 1, female.

AKPS, Anterior Knee Pain Scale: 0-100 (least symptoms).

Number of pain sites throughout body during past week: 0-9.

Worst Pain

Higher baseline pain levels, longer pain duration, and higher number of pain sites predicted higher levels of pain at 1 year, accounting for 26% of the variance (Table 3).

Table 3

Multivariable Linear Regression for the Dependent Variable Worst Pain at 1 Year^a

Predictor	B	95% CI	P Value	R ²
Constant^b	−1.1	−2.9 to 0.7	.24	0.26
Worst pain at baseline^c	0.4	0.1 to 0.6	<.01
Pain duration^d	0.7	0.2 to 1.2	<.01
No. of pain sites^e	0.5	0.2 to 0.8	<.01

Model includes worst pain at baseline. Excluded predictor variables after stepwise elimination: sex (0, male; 1, female); Anterior Knee Pain Scale (0-100 [least symptoms]); self-efficacy (0-10 [highest self-efficacy]); kinesiophobia (13-52 [most kinesiophobia]); bilateral pain (0, no; 1, yes); Hopkins Symptom Checklist (1-4 [most symptoms]).

Constant: Y-intercept of the model.

Worst pain: visual analog scale for worst pain in past week (0-10).

Pain duration: 0, 3-6 months; 1, >6-12 months; 2, >12-24 months; 3, >2 years.

Number of pain sites throughout body during past week: 0-9.

Change Scores

For the outcome global change, the variables female sex, lower AKPS at baseline, higher self-efficacy, and lower number of pain sites were significant predictors of higher change scores at 1 year. The model accounted for 28% of the variance. Table 4 shows results for multivariable linear regressions for global change scores at 1 year.

Table 4

Multivariable Linear Regressions for Global Change Scores at 1 Year^a

Predictor	B	95% CI	P Value	R ²
Constant^b	6.7	2.0 to 11.4	<.01	0.28
Sex^c	2.2	0.9 to 3.6	<.01
Pain duration^d	−0.6	−1.2 to 0.0	.06
AKPS baseline^e	−0.1	−0.2 to 0.0	.02
Self-efficacy^f	0.7	0.3 to 1.2	<.01
No. of pain sites^g	−0.8	−1.2 to −0.5	<.01

Constant: Y-intercept of the model.

Sex: 0, male; 1, female.

Pain duration: 0, 3-6 months; 1, >6-12 months; 2, >12-24 months; 3, >2 years.

AKPS, Anterior Knee Pain Scale: 0-100 (least symptoms).

Self-efficacy: 0-10 (highest self-efficacy).

Number of pain sites throughout body during past week: 0-9.

For the outcome change in AKPS at 1 year, the variables lower AKPS at baseline and lower number of pain sites were significant predictors of higher change scores in AKPS at 1 year. This model accounted for 22% of the variance (Table 5).

Table 5

Multivariable Linear Regressions for Change in AKPS at 1 Year^a

Predictor	B	95% CI	P Value	R ²
Constant^b	43.7	27.8 to 59.6	<.01	0.22
Sex^c	5.1	−0.3 to 10.5	.06
AKPS^d	−0.4	−0.6 to −0.2	<.01
No. of pain sites^e	−2.7	−4.0 to −1.3	<.01

Excluded predictor variables after stepwise elimination: worst pain, visual analog scale for worst pain in past week (0-10); kinesiophobia (13-52 [most kinesiophobia]); bilateral pain (0, no; 1, yes); pain duration (0, 3-6 months; 1, >6-12 months; 2, >12-24 months; 3, >2 years); Hopkins Symptom Checklist (1-4 [most symptoms]); self-efficacy (0-10 [highest self-efficacy]).

Constant: Y-intercept of the model.

Sex: 0, male; 1, female.

AKPS, Anterior Knee Pain Scale: 0-100 (least symptoms).

Number of pain sites throughout body during past week: 0-9.

For the outcome change in worst pain at 1 year, the variables higher baseline pain levels, shorter pain duration, and lower number of pain sites predicted largest reduction in worst pain at 1 year. This model accounted for 30% of the variance (Table 6).

Table 6

Multivariable Linear Regressions for Change in Worst Pain at 1 Year^a

Predictor	B	95% CI	P Value	R ²
Constant^b	−1.1	−2.9 to 0.7	.24	0.30
Worst pain^c	−0.6	−0.9 to −0.4	<.01
Pain duration^d	0.7	0.2 to 1.2	<.01
No. of pain sites^e	0.5	0.2 to 0.8	<.01

Excluded predictor variables after stepwise elimination: bilateral pain (0, no; 1, yes); Hopkins Symptom Checklist (1-4 [most symptoms]); self-efficacy (0-10 [highest self-efficacy]); kinesiophobia (13-52 [most kinesiophobia]); Anterior Knee Pain Scale (0-100 [least symptoms]); sex (0, male; 1, female).

Constant: Y-intercept of the model.

Worst pain: visual analog scale for worst pain in past week (0-10).

Pain duration: 0, 3-6 months; 1, >6-12 months; 2, >12-24 months; 3, >2 years.

Number of pain sites throughout body during past week: 0-9.

Alternative Models

We also tested predictive models for AKPS and worst pain at 1 year that excluded their baseline values. For the AKPS model with the baseline AKPS value removed, the variables longer pain duration, lower self-efficacy, and higher number of pain sites were significant predictors for lower AKPS at 1 year, accounting for 31% of the variance. For the worst pain model excluding baseline levels of worst pain, we found that longer pain duration, lower baseline AKPS, and higher number of pain sites predicted higher pain levels at 1 year, accounting for 24% of the variance.

Predictive models in which number of pain sites was excluded as a variable found HSCL-10 replacing it, although as a weaker predictor than number of pain sites, with a lower R² for all models (for HSCL-10 in all models,P≥ .04 and R² = 0.11-0.26).

Discussion

The most consistent predictor of outcome at 1 year in patients with PFP was number of pain sites at baseline. Higher number of pain sites was a significant predictor of poorer outcomes in all multivariable models. For the outcomes AKPS and worst pain, their baseline values were significant predictors of outcome at 1 year. Low self-efficacy, longer pain duration, and male sex were also significant predictors of poorer outcomes at 1 year, although these results were found in only 1 analysis.

Comparison With Previous Studies

Number of Pain Sites

Previous studies identified widespread pain as a predictor of poor prognosis in other musculoskeletal conditions^1,20 and in knee pain in general practice.⁴⁶ Large epidemiologic surveys have found number of pain sites to be associated with lower function,²⁷ number of nonmuskuloskeletal symptoms,⁴⁵ and a predictor of long-term work disability.²⁶ Fibromyalgia survey criteria, in which number of pain sites is a central factor, have been identified as a strong predictor of poor outcome after total hip or knee joint replacement⁷ and for postsurgical opiate use after total hip or knee joint replacement,⁶ regardless of whether the patient fulfilled diagnostic criteria for fibromyalgia. Our finding that higher number of pain sites predicted poorer outcome for pain and function and poorer response to treatment (lower change scores) is consistent with these previous studies.

Pain Duration

Lankhorst et al,³⁰ Collins et al,^11,12 and Rathleff et al³⁹ found longer pain duration to be predictive of poorer outcome in the long and short term in patients with PFP. The recent systematic review by Matthews et al³⁴ found longer duration of PFP symptoms (>4 months) to be the most reported prognostic factor of a poor outcome. Our results partially support this finding, with pain duration being a significant predictor for worst pain and change in worst pain. It is important to point out that none of the patients in our cohort reported pain duration <4 months, which might affect the predictive value of pain duration.

Psychological Factors

We found self-efficacy to be one of several predictors of global change and of AKPS when baseline AKPS levels were excluded from the model. Self-efficacy has not previously been examined as a predictor of PFP outcomes. A recent systematic review found that catastrophic thinking and poor coping strategies predicted higher pain levels after total knee replacement but kinesiophobia did not.² This is in keeping with our results, in which self-efficacy, but not kinesiophobia, was a predictor for global change score.

Domenech et al¹⁶ and Piva et al³⁷ found that changes in kinesiophobia were strongly associated with changes in pain and function after treatment for PFP, but neither reported on the predictive value of baseline levels of kinesiophobia on final outcomes or change, as examined in this study. In the current study, baseline kinesiophobia was not predictive of any of the outcomes examined. We did not examine associations between concurrent changes in the various measures.

In the current study, alternative multivariable models were also constructed in which number of pain sites was excluded as an independent variable. Within these models, we found that HSCL-10 replaced number of pain sites in the final models, although as a weaker predictor than number of pain sites and with a lower R² for all models. Although this finding has not previously been reported in PFP, it is consistent with previous findings indicating that anxiety and depression are predictors of poorer outcomes in back pain and in general musculoskeletal pain^1,22 and are associated with widespread pain.^9,13

Sex

Sex is a variable that has been examined in previous PFP studies. Given the sex differences in PFP prevalence, with women being more commonly affected than men, it has been hypothesized that female sex might predict worse outcome. Previous studies have not reported sex as predictive for outcome,³⁴ although Lankhorst et al³¹ found a positive trend for females responding better to exercise therapy. Our analysis found female sex to be a significant predictor of higher global change score.

Directionality of Baseline Values on Final Outcome vs Change

Our findings that baseline measures of pain and function (AKPS) predict final outcomes are in agreement with previous studies.^12,30 It should be noted that the directionality of predictive value of baseline levels for final outcome was opposite to that of predictive value for change. That is, having more symptoms at baseline predicted more symptoms at follow-up but also higher change scores (more improvement) for the outcomes.

The differences in directionality of predictors on either final outcomes or change in scores are intuitive from a statistical standpoint but still important to point out. It means that the predictive value of baseline levels of scores depends on whether we wish to predict final outcome or change scores.

In contrast, higher number of pain sites, longer pain duration, and lower self-efficacy were predictors for worse final outcome and less change.

Strengths and Weaknesses of the Study

Although this study is larger than most comparable predictor studies within PFP,³⁴ the limited sample size may affect the robustness of our results. The data extraction for this predictive analysis is based on an RCT in which the original study was not designed for prognostic analyses, which is a weakness when compared with prespecified analyses in a larger study designed for the purpose of prognosis. Data were analyzed as a single cohort, as there were no between-group differences in the 1-year outcomes of the RCT.³⁵ This is advantageous in increasing the sample size but allows for no analysis regarding which factors predict response to a specific treatment. Our cohort did not include any patients with pain duration <4 months, which may limit generalizability of the results to populations with shorter duration at baseline. Furthermore, the relatively strict inclusion criteria used in the main RCT may limit the generalizability of the results. The inherent clinimetric limitations of self-report measures should also be pointed out. Transformations of pain, feelings, or thoughts to numbers are unlikely to be linear as the scales imply, and results based on such analyses should be interpreted with caution.⁴

Implications for Future Research and Clinical Practice

The most consistent predictor of pain, function, and change in PFP at 1 year was number of pain sites. This is consistent with previous studies within other musculoskeletal pain conditions^1,6,7,22,46 but has not been examined in PFP previously. The meaning of multiple pain sites is unclear. Widespread pain is closely associated with various psychological factors^9,13 and has been suggested as an indicator of central sensitization.^6,7,19,36 Psychological factors, widespread pain, and sensitization are predictive of outcomes in other musculoskeletal pain conditions.^{1,2,6,7,20,22,26,46} Our findings suggest that they may be important to consider within PFP, as with other chronic musculoskeletal pain conditions. The recognition of multiple pain sites and self-efficacy as prognostic factors in a clinical setting suggests that we should broaden our perspective. The current recommendations^3,10 are mainly based on biomechanical considerations. As such, we recommend the use of a biopsychosocial model in approaching PFP clinically and in future research.^17,33,47

Our findings that baseline levels of worst pain, function (AKPS), and pain duration predict outcomes are consistent with previous studies.³⁴ The recognition of these risk factors for poorer outcome may be useful in a clinical setting—for example, in informing patients regarding realistic outcome expectations. The finding of sex as a predictor of outcome is not consistent with previous studies and should be examined further.

Conclusion

Number of pain sites was the most consistent predictor of pain, function, and change scores 1 year after an education- and exercise-based intervention for PFP. This finding has not previously been reported in PFP but is consistent with previous research for other musculoskeletal conditions. Baseline levels for AKPS and worst pain predicted respective final scores and change scores.

Footnotes

Acknowledgements

The authors thank the physiotherapists at the Department of Physical Medicine and Rehabilitation, Sørlandet Hospital Kristiansand, for performing the interventions and outcome measures in this trial.

One or more of the authors has declared the following potential conflict of interest or source of funding: This study was funded by the Research Department of Sørlandet Hospital. AOSSM checks author disclosures against the Open Payments Database (OPD). AOSSM has not conducted an independent investigation on the OPD and disclaims any liability or responsibility relating thereto.

Submitted May 16, 2019; accepted September 30, 2019.

References

Artus

Campbell

Mallen

Dunn

van der Windt

. Generic prognostic factors for musculoskeletal pain in primary care: a systematic review. BMJ Open. 2017;7(1):e012901.

Baert

Lluch

Mulder

Nijs

Noten

Meeus

. Does pre-surgical central modulation of pain influence outcome after total knee replacement? A systematic review. Osteoarthritis Cartilage. 2016;24(2):213-223.

Barton

Lack

Hemmings

Tufail

Morrissey

. The “Best Practice Guide to Conservative Management of Patellofemoral Pain”: incorporating level 1 evidence with expert clinical reasoning. Br J Sports Med. 2015;49(14):923-934.

Brox

. The Fear Avoidance Beliefs Questionnaire—the FABQ—for the benefit of another 70 million potential pain patients. Scand J Pain. 2019;19(1):1-2.

Brox

Gjengedal

Uppheim

, et al. Arthroscopic surgery versus supervised exercises in patients with rotator cuff disease (stage II impingement syndrome): a prospective, randomized, controlled study in 125 patients with a 2 1/2-year follow-up. J Shoulder Elbow Surg. 1999;8(2):102-111.

Brummett

Janda

Schueller

, et al. Survey criteria for fibromyalgia independently predict increased postoperative opioid consumption after lower-extremity joint arthroplasty: a prospective, observational cohort study. Anesthesiology. 2013;119(6):1434-1443.

Brummett

Urquhart

Hassett

, et al. Characteristics of fibromyalgia independently predict poorer long-term analgesic outcomes following total knee and hip arthroplasty. Arthritis Rheumatol (Hoboken, NJ). 2015;67(5):1386-1394.

Chavance

. Handling missing items in quality of life studies. Commun Stat Theory Methods. 2004;33(6):1371-1383.

Christensen

Johansen

Knardahl

. Psychological predictors of change in the number of musculoskeletal pain sites among Norwegian employees: a prospective study. BMC Musculoskelet Disord. 2017;18(1):140.

10.

Collins

Barton

van Middelkoop

, et al. 2018 Consensus statement on exercise therapy and physical interventions (orthoses, taping and manual therapy) to treat patellofemoral pain: recommendations from the 5th International Patellofemoral Pain Research Retreat, Gold Coast, Australia, 2017. Br J Sports Med. 2018;52(18):1170-1178.

11.

Collins

Bierma-Zeinstra

Crossley

van Linschoten

Vicenzino

van Middelkoop

. Prognostic factors for patellofemoral pain: a multicentre observational analysis. Br J Sports Med. 2013;47(4):227-233.

12.

Collins

Crossley

Darnell

Vicenzino

. Predictors of short and long term outcome in patellofemoral pain syndrome: a prospective longitudinal study. BMC Musculoskelet Disord. 2010;11:11.

13.

Croft

Schollum

Silman

. Population study of tender point counts and pain as evidence of fibromyalgia. BMJ. 1994;309(6956):696-699.

14.

Crossley

Stefanik

Selfe

, et al. 2016 patellofemoral pain consensus statement from the 4th International Patellofemoral Pain Research Retreat, Manchester. Part 1: terminology, definitions, clinical examination, natural history, patellofemoral osteoarthritis and patient-reported outcome measures. Br J Sports Med. 2016;50(14):839-843.

15.

Derogatis

Lipman

Rickels

Uhlenhuth

Covi

. The Hopkins Symptom Checklist (HSCL): a self-report symptom inventory. Behav Sci. 1974;19(1):1-15.

16.

Domenech

Sanchis-Alfonso

Espejo

. Changes in catastrophizing and kinesiophobia are predictive of changes in disability and pain after treatment in patients with anterior knee pain. Knee Surg Sports Traumatol Arthrosc. 2014;22(10):2295-2300.

17.

Engel

. The need for a new medical model: a challenge for biomedicine. Science. 1977;196(4286):129-136.

18.

Garcia

Park

, et al. Depression symptomatology and anterior cruciate ligament injury: incidence and effect on functional outcome—a prospective cohort study. Am J Sports Med. 2016;44(3):572-579.

19.

Graven-Nielsen

Arendt-Nielsen

. Assessment of mechanisms in localized and widespread musculoskeletal pain. Nat Rev Rheumatol. 2010;6(10):599-606.

20.

Grotle

Foster

Dunn

Croft

. Are prognostic indicators for poor outcome different for acute and chronic low back pain consulters in primary care? Pain. 2010;151(3):790-797.

21.

Harrell

Jr Lee

Mark

. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15(4):361-387.

22.

Hartvigsen

Hancock

Kongsted

, et al. What low back pain is and why we need to pay attention. Lancet. 2018;391(10137):2356-2367.

23.

Haugen

Grovle

Keller

Grotle

. Cross-cultural adaptation and validation of the Norwegian version of the Tampa Scale for Kinesiophobia. Spine. 2008;33(17):e595-e601.

24.

Hott

Brox

Pripp

Juel

Paulsen

Liavaag

. Effectiveness of isolated hip exercise, knee exercise, or free physical activity for patellofemoral pain: a randomized controlled trial. Am J Sports Med. 2019;47(6):1312-1322.

25.

Hott

Liavaag

Juel

Brox

. Study protocol: a randomised controlled trial comparing the long term effects of isolated hip strengthening, quadriceps-based training and free physical activity for patellofemoral pain syndrome (anterior knee pain). BMC Musculoskelet Disord. 2015;16:40.

26.

Kamaleri

Natvig

Ihlebaek

Bruusgaard

. Does the number of musculoskeletal pain sites predict work disability? A 14-year prospective study. Eur J Pain. 2009;13(4):426-430.

27.

Kamaleri

Natvig

Ihlebaek

Bruusgaard

. Localized or widespread musculoskeletal pain: does it matter? Pain. 2008;138(1):41-46.

28.

Kujala

Jaakkola

Koskinen

Taimela

Hurme

Nelimarkka

. Scoring of patellofemoral disorders. Arthroscopy. 1993;9(2):159-163.

29.

Kuorinka

Jonsson

Kilbom

, et al. Standardised Nordic questionnaires for the analysis of musculoskeletal symptoms. Appl Ergon. 1987;18(3):233-237.

30.

Lankhorst

van Middelkoop

Crossley

, et al. Factors that predict a poor outcome 5-8 years after the diagnosis of patellofemoral pain: a multicentre observational analysis. Br J Sports Med. 2016;50(14):881-886.

31.

Lankhorst

van Middelkoop

van Trier

, et al. Can we predict which patients with patellofemoral pain are more likely to benefit from exercise therapy? A secondary exploratory analysis of a randomized controlled trial. J Orthop Sports Phys Ther. 2015;45(3):183-189.

32.

Lentz

Zeppieri

Jr George

, et al. Comparison of physical impairment, functional, and psychosocial measures based on fear of reinjury/lack of confidence and return-to-sport status after ACL reconstruction. Am J Sports Med. 2015;43(2):345-353.

33.

Maclachlan

Collins

Matthews

MLG

Hodges

Vicenzino

. The psychological features of patellofemoral pain: a systematic review. Br J Sports Med. 2017;51(9):732-742.

34.

Matthews

Rathleff

Claus

, et al. Can we predict the outcome for people with patellofemoral pain? A systematic review on prognostic factors and treatment effect modifiers. Br J Sports Med. 2017;51(23):1650-1660.

35.

Moons

Royston

Vergouwe

Grobbee

Altman

. Prognosis and prognostic research: what, why, and how? BMJ. 2009;338:B375.

36.

Nijs

Torres-Cueco

van Wilgen

, et al. Applying modern pain neuroscience in clinical practice: criteria for the classification of central sensitization pain. Pain Physician. 2014;17(5):447-457.

37.

Piva

Fitzgerald

Wisniewski

Delitto

. Predictors of pain and function outcome after rehabilitation in patients with patellofemoral pain syndrome. J Rehabil Med. 2009;41(8):604-612.

38.

Rathleff

Graven-Nielsen

Holmich

, et al. Activity modification and load management of adolescents with patellofemoral pain: a prospective intervention study including 151 adolescents. Am J Sports Med. 2019;47(7):1629-1637.

39.

Rathleff

Olesen

Rasmussen

Roos

. Is knee pain during adolescence a self-limiting condition? Prognosis of patellofemoral pain and other types of knee pain. Am J Sports Med. 2016;44(5):1165-1171.

40.

Smith

Selfe

Thacker

, et al. Incidence and prevalence of patellofemoral pain: a systematic review and meta-analysis. PLoS One. 2018;13(1):e0190892.

41.

Steyerberg

. Clinical Prediction Models: A Practical Approach to Development, Validation, and Updating. 2nd ed. Berlin, Germany: Springer; 2019.

42.

Strand

Dalgard

Tambs

Rognerud

. Measuring the mental health status of the Norwegian population: a comparison of the instruments SCL-25, SCL-10, SCL-5 and MHI-5 (SF-36). Nord J Psychiatry. 2003;57(2):113-118.

43.

Sun

Shook

Kay

. Inappropriate use of bivariable analysis to screen risk factors for use in multivariable analysis. J Clin Epidemiol. 1996;49(8):907-916.

44.

Thomee

Wahrborg

Borjesson

Thomee

Eriksson

Karlsson

. A new instrument for measuring self-efficacy in patients with an anterior cruciate ligament injury. Scand J Med Sci Sports. 2006;16(3):181-187.

45.

Tschudi-Madsen

Kjeldsberg

Natvig

, et al. A strong association between non-musculoskeletal symptoms and musculoskeletal pain symptoms: results from a population study. BMC Musculoskelet Disord. 2011;12:285.

46.

van der Waal

Bot

Terwee

, et al. Course and prognosis of knee complaints in general practice. Arthritis Rheum. 2005;53(6):920-930.

47.

Vicenzino

Maclachlan

Rathleff

. Taking the pain out of the patellofemoral joint: articulating a bone of contention. Br J Sports Med. 2019;53(5):268-269.

48.

Wertli

Eugster

Held

Steurer

Kofmehl

Weiser

. Catastrophizing—a prognostic factor for outcome in patients with low back pain: a systematic review. Spine J. 2014;14(11):2639-2657.

49.

Wertli

Rasmussen-Barr

Held

Weiser

Bachmann

Brunner

. Fear-avoidance beliefs—a moderator of treatment efficacy in patients with low back pain: a systematic review. Spine J. 2014;14(11):2658-2678.