Individuals With Patellofemoral Pain Have Impaired Self-Reported and Performance-Based Function: Systematic Review With Meta-Analysis and Meta-Regression

Search Strategy

We conducted the initial electronic search in MEDLINE, Embase, CINAHL, Web of Science, and SPORTDiscus between inception and February 2021 and updated it in January 2024. We combined the keywords and medical subject headings related to PFP and self-reported and performance-based measures of function with search filters to develop the search strategy. We created the primary search for MEDLINE and adapted it to the other databases through pilot searches (Supplemental Table 1). We did not search the gray literature.

Selection Criteria

One author (A.F.B.B.) imported identified studies into Covidence (Veritas Health Innovation), and duplicates were removed.¹⁸ Two authors (A.F.B.B. and J.C.P.S.) independently screened titles and abstracts for eligibility using the criteria presented in Table 1. We retrieved full-text articles of potentially relevant abstracts for further review. When the full text was not available, we requested it from the corresponding authors via email. If authors were unable to provide the full text, we excluded the study. When studies from the same author groups presented similar descriptive values of measures of function, we included only the first study published after confirmation with the corresponding author that both publications included the same cohort. Disagreements were resolved by consulting a senior author (R.V.B.).

Table 1.Inclusion and Exclusion Criteria for Studies

View Fullscreen

Data Extraction

One author (R.V.B.) extracted study and participant characteristics (eg, lead author, year of publication, sample size, sex, and participant age), self-reported and performance-based measures of function (eg, Anterior Knee Pain Scale [AKPS], Lower Extremity Functional Scale [LEFS], Knee injury and Osteoarthritis Outcome Score [KOOS], KOOS for PFP and osteoarthritis [KOOS-PF], forward step-down test [FSDT], and hop or balance tests), and predictors of interest to be included in the meta-regression. We selected physical (eg, body mass index [BMI] and strength) and nonphysical (eg, kinesiophobia and pain catastrophizing) predictors of interest based on recommended items from Reporting of Quantitative Patellofemoral Pain (REPORT-PFP), the biomechanical and psychological consensus of PFP.^6,19,20 A second author (J.C.P.S.) reviewed all extracted data. We extracted means, SDs, and sample sizes for all outcomes and used them for data analysis. When data were missing, we contacted corresponding authors for further information via email up to 3 times. If authors were unable to provide the data or did not respond to the requests and missing data could not be calculated using Review Manager 5.4 (The Cochrane Collaboration), we did not enter the study in the meta-analyses. For these studies, we only performed an individual study analysis by calculating the standardized mean differences (SMDs) and discussing them. We provide details on data-extraction management in Supplemental Material 2.

Methodologic Quality Assessment and Risk of Bias

We assessed internal and external validity of observational and nonrandomized interventional studies with a domain-based evaluation using the modified Downs and Black checklist, as performed by Hart et al.²¹ We assessed internal validity across the following 5 domains: performance bias (items 14, 15, and 19), reporting bias (item 16), detection bias (items 17, 18, and 20), selection bias (items 21–25), and attrition bias (item 26). We assessed external validity using items 11 through 13. Items were scored as yes, no, or unable to determine. Overall quality classification for each study was based on concerns across all applicable items and domains rather than the numeric summary score. Studies were classified across domains and external validity as low, moderate, or high quality based on item evaluation. A similar classification was performed for internal validity based on domain evaluation. We assessed the methodologic quality of the only randomized clinical trial²² included using the Physiotherapy Evidence Database (PEDro) scale²³ and its risk of bias using the Cochrane Risk of Bias for Randomized Trials 2 (RoB2) following Cochrane Handbook for Systematic Reviews of Interventions recommendations.²⁴ The 10-item PEDro scale consists of a score ranging from 0 to 10 and is used to rate trials according to the presence or absence of some methodologic quality criteria.²³ The score classifications are high quality (≥7 of 10), moderate quality (4–6 of 10), and low quality (≤3 of 10). The RoB2 comprises the following 5 domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported study. For each domain, the tool comprises a series of signaling questions scored as yes, probably yes, probably no, no, and no information. We classified each domain as low, high, or some concerns of risk of bias based on the tool’s algorithm.²⁵ We determined the overall risk of bias using the worst-score-counts method, which takes the lowest rating across all the domains.²⁶ Further details regarding study-quality assessment and risk of bias are provided in Supplemental Material 3 and Supplemental Tables 2 and 3. Two authors (A.F.B.B. and M.C.W.) independently assessed all studies, and disagreements were resolved by consulting a senior author (R.V.B.).

Data Synthesis and Analysis

One author (R.V.B.) performed meta-analyses using Review Manager 5.4 and random-effects models when ≥2 studies investigated the same outcome and comparator (pain-free individuals or pain-free limb of individuals with unilateral PFP). Another author (A.F.B.B.) reviewed all meta-analyses. We calculated SMDs with 95% CIs (Hedges g) once different scales or units of measurement were reported across studies, even in those using the same questionnaire or test (eg, Goharpey et al,²⁷ Peeler and Anderson,²⁸ and Guimaraes Araujo et al²⁹). We classified SMDs as small (≥0.2), moderate (0.5–0.79), and large (≥0.80) effects.³⁰ We quantified statistical heterogeneity for pooled results using the I² statistic and defined it as not important (<50%), moderate (50%–75%), or high (>75%).³¹ We estimated publication bias using the Egger regression test. For data that were not included in the meta-analyses, we calculated the SMDs with 95% CIs for individual comparisons and discussed them. Confidence intervals excluding zero were considered statistically significant.

We performed meta-regressions to identify predictors that could explain the SMDs (Hedges g) of function outcomes. Random-effects meta-regressions were performed using Comprehensive Meta-Analysis software (BioSTAT Consultants, Inc) when at least 10 studies included in a meta-analysis presented data for the same predictor.²⁵

Certainty of Evidence

Two authors (A.F.B.B. and M.C.W.) assessed the certainty of evidence using a modified Grading of Recommendations Assessment, Development, and Evaluation approach.^32,33 Given the observational nature of the research question of our systematic review, certainty of evidence started as high and was downgraded or upgraded according to Grading of Recommendations Assessment, Development, and Evaluation Handbook recommendations, which are described in Table 2.^32,34 We defined levels of certainty of evidence as follows: high when further research is very unlikely to change confidence in the estimate of the effect; moderate when further research is likely to have an important effect on confidence in the estimate of the effect and may change the estimate; low when further research is very likely to have an important effect on confidence in the estimate of the effect and is likely to change the estimate; and very low when there is very little confidence in the effect estimate.³²

Table 2.Outcome Level of Certainty of Meta-Analyses (GRADE Approach)

View Fullscreen

Anterior Knee Pain Scale

Moderate certainty of evidence from 40 studies (2414 individuals) showed that individuals with PFP have reduced self-reported function measured with the AKPS compared with pain-free individuals (large effect size, SMD = −3.45; 95% CI = −3.84, −3.06; I² = 88%; P < .001; Figure 2A). ^††

View Figure

• Download as Powerpoint
• Download Figure

Lower Extremity Function Scale

Moderate certainty of evidence from 9 studies (593 individuals) showed that individuals with PFP have reduced self-reported function measured with the LEFS compared with pain-free individuals (large effect size, SMD = −3.83; 95% CI = −5.10, −2.55; I² = 95%; P < .001; Figure 2B). ^{37,60,65,68–73}

Functional Index Questionnaire

Moderate certainty of evidence from 7 studies (337 individuals) showed that individuals with PFP have reduced self-reported function measured using the Functional Index Questionnaire compared with pain-free individuals (large effect size, SMD = −4.87; 95% CI = −6.97, −2.77; I² = 96%; P < .001; Figure 2C).^{28,74–77,105,106}

Knee injury and Osteoarthritis Outcome Score

High certainty of evidence from 5 studies (255 individuals) showed that individuals with PFP have reduced self-reported function measured using the KOOS compared with pain-free individuals (large effect size, SMD = −1.99; 95% CI = −2.41, −1.57; I² = 43%; P < .001; Figure 2D).^78–81,83

Activities of Daily Living Questionnaire

Moderate certainty of evidence from 5 studies (375 individuals) showed that individuals with PFP have reduced self-reported function measured using the Activities of Daily Living Questionnaire compared with pain-free individuals (large effect size, SMD = −2.79; 95% CI = −3.49, −2.08; I² = 83%; P < .001; Figure 2E).^29,84–87

KOOS for PFP and Osteoarthritis

Moderate certainty of evidence from 4 studies (124 individuals) showed that individuals with PFP have reduced self-reported function measured using the KOOS-PF compared with pain-free individuals (large effect size, SMD = −2.66; 95% CI = −3.47, −1.86; I² = 60%; P < .001; Figure 2F).^9,79,88,110

Lysholm Knee Scoring Scale

Moderate certainty of evidence from 3 studies (102 individuals) showed that individuals with PFP have reduced self-reported function measured using the Lysholm Knee Scoring Scale compared with pain-free individuals (large effect size, SMD = −2.23; 95% CI = −3.51, −0.96; I² = 82%; P < .001; Figure 2G).^60,85,90

Balance Tests

Low certainty of evidence from 12 studies (789 individuals) showed that individuals with PFP have reduced reach distance in balance tests compared with pain-free individuals (large effect size, SMD = −0.66; 95% CI = −1.12, −0.19; I² = 88%; P = .005; Figure 3).^‡‡ Low certainty of evidence from 2 studies (70 individuals) showed no differences between limbs for the YBT in individuals with unilateral PFP (small effect size, SMD = −0.20; 95% CI = −0.68, 0.27; I² = 0%; P = .39; Figure 4).^22,88

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Forward Step-Down Test

Moderate certainty of evidence from 9 studies (737 individuals) showed that individuals with PFP have fewer repetitions in the FSDT compared with pain-free individuals (large effect size, SMD = −0.80; 95% CI = −1.11, −0.50; I² = 68%; P < .001; Figure 3).^§§ Very low certainty of evidence from 2 studies (106 individuals) showed no differences between limbs for repetitions in the FSDT in individuals with unilateral PFP (small effect size, SMD = −0.36, 95% CI = −1.11, 0.38; I² = 73%; P = .34; Figure 4).^97,99

Single-Leg Hop Test

Moderate certainty of evidence from 7 studies (711 individuals) indicated a shorter distance in the SLHT for individuals with PFP than for pain-free individuals (small effect size, SMD = −0.42; 95% CI = −0.57, −0.27; I² = 0%; P < .001; Figure 3).^{9,42,45,61,62,93,101} Evidence from 1 high-quality study showed a shorter distance in the SLHT for the PFP limb compared with the pain-free limb of individuals with unilateral PFP, but this result was not confirmed via the calculated SMD and 95% CI (small effect size, SMD = −0.29; 95% CI = −0.86, 0.28; P = .32; Figure 4).⁹⁹

Single-Leg Triple-Hop Test

Moderate certainty of evidence from 2 studies (196 individuals) showed a shorter distance in the single-leg triple-hop test (SLTHT) in individuals with PFP than pain-free individuals (small effect size, SMD = −0.30, 95% CI = −0.58, −0.02; I² = 0%; P = .04; Figure 3).^29,42

Bilateral Squat Test

Moderate certainty of evidence from 2 pooled studies (70 individuals) showed no difference in the number of repetitions in the bilateral squat test between individuals with PFP and pain-free individuals (large effect size, SMD = −1.21; 95% CI = −2.71, 0.29; I² = 86%; P = .11; Figure 3).^27,97

Between-Limbs Comparisons for Other Performance-Based Measures

Evidence from 1 study showed fewer repetitions for the anteromedial lunge in the PFP limb compared with the pain-free limb of individuals with unilateral PFP (moderate effect size, SMD = −0.64; 95% CI = −1.17, −0.11; P = .02) (Figure 4).⁹⁷ The same study showed fewer repetitions for the balance-and-reach test and lower scores in the single-leg press test in the PFP limb compared with the pain-free limb of individuals with unilateral PFP; however, this result was not confirmed by the SMDs and 95% CIs (small effect sizes, SMD = −0.35; 95% CI = −0.87, 0.17; P = .19 and SMD = −0.49; 95% CI = −1.02, 0.03; P = .06, respectively) (Figure 4).⁹⁷

Self-Reported Function

We identified that individuals with PFP have impaired self-reported function compared with pain-free individuals. This finding is based on large effects across 7 different questionnaires/scales and supports previous evidence considering self-reported measures of function as determinants of treatment success or patient recovery after nonoperative management of PFP.^12,15 Impaired self-reported function as a consequence of having PFP not only affects individuals’ perception and perspectives about their physical function^5,111 but also can predict persistent or recurrent PFP (ie, poor prognosis) in the medium and long term.^11,112,113 Along with previous evidence, our results highlight the need to consider self-reported function as one of the primary condition-specific outcome measures during rehabilitation of individuals with PFP.

The assessment of self-reported function is clinician friendly and strongly recommended by the REPORT-PFP.¹⁹ Recommended questionnaires include the AKPS (also known as the Kujala scale) and the KOOS-PF.¹⁹ Although the AKPS is by far the most commonly used questionnaire (42 studies^‡‡‡ included in our review), the recently developed KOOS-PF¹¹⁴ seems to have better content validity, reliability, construct validity, and responsiveness.^115,116 The KOOS-PF is also more feasible for clinical use due to the its ease of administration and scoring, smaller number of items, and short time to complete.¹¹⁶ Changes in KOOS-PF ranging from 16 to 17.2 have been suggested to detect meaningful differences postintervention.¹¹⁶ However, only 5 studies included in our review used the KOOS-PF.^{9,79,88,89,110} More studies using the KOOS-PF to assess self-reported function of individuals with PFP are needed to further support the recommendation for using this tool instead of the AKPS.

Physical Performance

Performance-based measures of function can complement information from self-reported measures by objectively quantifying functional impairments using physical performance tests that are clinically accessible, low-cost, and time efficient.^117,118 We identified that individuals with PFP have impaired physical performance compared with pain-free individuals during balance tests FSDT, SLHT, and SLTHT. These tests are easily measured in clinical settings, and they represent aspects of daily function or sport and simulate common pain-provoking tasks (eg, stepping, jumping, and landing).^93,94,119 Given that self-reported function does not fully reflect the magnitude of performance deficits, we recommend using performance-based measures of function when assessing individuals with PFP. Whereas balance tests and FSDT may be more useful for evaluating sedentary or lower-functioning individuals due to their reduced physical demand, the SLHT may be used for athletic populations as it is more challenging. In a recent review, Berg et al also recommended the use of the SLHT for assessing performance deficits of youth and young adults, given its sufficient intrarater reliability, construct validity, and responsiveness.¹¹⁷

Although we observed that performance during balance tests FSDT, SLHT, and SLTHT was impaired when compared with that of controls, we found no differences between the painful and pain-free limbs of individuals with unilateral PFP. This suggested that the functional performance of the pain-free limb may be also compromised in individuals with PFP, as recently reported by Waiteman et al.¹²⁰ Individuals with PFP have reduced physical activity, which may result in reduced bilateral lower limb muscle strength and physical performance, regardless of pain laterality.^8,121 Reduced pain-free limb performance may also be a consequence of fear-avoidance belief, a commonly reported trait in this population.^7,93 Caution is warranted when comparing limbs using performance-based measures of function in clinical practice, as the pain-free limb of individuals with unilateral PFP may not be an accurate comparator.¹²² In the absence of reference values of performance-based measures of function for individuals with PFP, we encourage pretest and posttest bilateral assessments to aid clinicians in their decision-making processes because changes may occur bilaterally.^123,124 Future research is needed to provide reference values from age- and sex-matched pain-free samples, as the use of the contralateral limb does not seem appropriate.

Predictors of Self-Reported or Performance-Based Function

We performed meta-regressions to investigate physical and nonphysical factors that might explain poor self-reported and performance-based function in individuals with PFP. We observed no relationships between self-reported function and age, BMI, or duration of symptoms. Similarly, no relationship between performance-based function and age was observed. Although researchers have reported a direct relationship between these factors and clinical outcomes in PFP (eg, higher BMI was related to lower functional capacity), our findings showed that this relationship does not explain differences between groups.^125,126 This means that individuals with PFP have lower self-reported and performance-based function compared with pain-free controls regardless of age, BMI, or duration of symptoms. Another reason that we did not observe relationships between function and these variables is that, as commonly reported in PFP, most individuals from the studies included in this review had normal BMI (mean ± SD BMI of individuals with PFP = 23.04 ± 3.58 kg/m² and of pain-free individuals = 22.30 ± 3.05 kg/m²) and were young adults (mean ± SD age of individuals with PFP = 22.91 ± 6.55 years and of pain-free individuals = 23.24 ± 6.23 years).¹²⁵ This results in a constrained range of age and BMI across studies and may have influenced statistical analysis. Other factors, such as quadriceps strength, kinesiophobia, or both, may be more associated with impaired self-reported or performance-based function, as previously reported.^13,127,128 Quadriceps strength and kinesiophobia have been reported to be associated with pain intensity in individuals with PFP, which plays an important role in the perception of disability and function.^7,129,130 These uncontrolled factors may also be the source of potential heterogeneity between the studies. More studies are needed to better understand what physical and nonphysical factors might explain impaired self-reported and performance-based function in individuals with PFP. Furthermore, more longitudinal studies are necessary to investigate how function changes across time in individuals with PFP.

We also did not observe a relationship between self-reported function and self-reported pain. Glaviano et al reported high pain variability in individuals with PFP over 10 days, which explained almost 60% of the variance in self-reported function.¹³¹ One should assume that traditional methods of assessing pain (eg, worst levels of pain in the last month or week) used in the studies included in our review may not be sensitive to pain variation and are also more susceptible to recall bias.¹³² Longitudinal pain assessments of daily pain variability over a period may be better suited to investigate the relationship between self-reported pain and function in individuals with PFP versus isolated pain observations.

Limitations

The design of studies included in the review did not allow us to infer causality of self-reported and performance-based measures of function in individuals with PFP. We did not review the gray literature; thus, relevant but unpublished studies may have been excluded from our findings. In addition, the limited number of studies did not allow us to investigate whether other important physical and nonphysical predictors (eg, hip and knee strength, physical activity level, and psychological factors) may be more associated with self-reported and performance-based measures of function. More studies following the REPORT-PFP guidelines are needed to fill this gap in the literature.