Objective To develop a clinical prediction rule to identify patients with patellofemoral pain (PFP) who are more likely to benefit from foot orthoses.
Design Posthoc analysis of one treatment arm of a randomised clinical trial.
Setting Single-centre trial in a community setting in Brisbane, Australia.
Participants 42 participants (mean age 27.9 years) with a clinical diagnosis of PFP (median duration 36 months).
Interventions Foot orthoses fitted by a physiotherapist.
Main Outcome Measures Five-point global improvement scale at 12-week follow-up, dichotomised with marked improvement equalling success.
Results Potential predictor variables identified by univariate analyses were age, height, pain severity, anterior knee pain scale score, functional index questionnaire score, foot morphometry (arch height ratio, mid-foot width difference from non-weight bearing to weight bearing) and overall orthoses comfort. Parsimonious fitting of these variables to a model that explained success with orthoses identified the following: age (>25 years), height (<165 cm), worst pain visual analogue scale (<53.25 mm) and a difference in mid-foot width from non-weight bearing to weight bearing (>10.96 mm). The pretest success rate of 40% increased to 86% if the patient exhibited three of these variables (positive likelihood ratio 8.8; 95% CI 1.2 to 66.9).
Conclusion Post-hoc analysis identified age, height, pain severity and mid-foot morphometry as possible predictors of successful treatment of PFP with foot orthoses, thereby providing practitioners with information for prescribing foot orthoses in PFP and stimulating further research.
Statistics from Altmetric.com
Foot orthoses are frequently used by practitioners in the management of patellofemoral pain (PFP) despite a dearth of research-based evidence for their clinical efficacy. Our recent randomised clinical trial (RCT) found that those who wore foot orthoses were 66% more likely to experience success than those who wore flat inserts at 6 weeks on a dichotomised measure of success.1 2 Although this effect was not evident at 12 and 52 weeks, the relatively wide confidence interval (eg, number need to treat four; 99% CI 2 to 511 suggest a heterogeneous response to foot orthoses. This is consistent with evidence of non-systematic effects between individuals participating in laboratory-based research on the biomechanical effects of orthoses.3,–,5 Unlike previous RCT on foot orthoses for PFP,6 7 we did not use specific criteria, such as excessive pronation, as inclusion or stratification criteria. Point estimates of the effect from our RCT thus may well underestimate the effect for some patients, similar to previous reports for other body regions, such as low back pain.8,–,10
Only one published study has investigated possible predictors of success with foot orthoses in the treatment of patients with PFP. Sutlive et al11 collected an extensive range of lower limb measures of the thigh, knee and leg (eg, Craig's test, Q angle, tibial varum/valgum) and the foot (eg, great toe extension, forefoot and rearfoot posture, ankle dorsiflexion), as well as characteristics of the individual patient (eg, age, sex) and PFP (eg, pain severity, duration). They found that the best individual predictors of improvement with foot orthoses, indicated by a reduction in pain or disability greater than 50%, were forefoot valgus alignment of at least 2°, great toe extension less than or equal to 78° and navicular drop less than or equal to 3 mm. At best, these predictors increased the success rate from 60% to 86%. However, the reliability of these measures is questionable (intraclass correlation coefficient; ICC (SEM)): orefoot alignment 0.25 (7.0°); great toe extension 0.55 (7.8°); navicular drop 0.51 (2.8 mm).11 The authors also reported poor to moderate reliability for rearfoot measures, relaxed calcaneal stance and tibial varum/valgum. These findings highlight the need for a more reliable approach to lower limb morphometry if such measures are to be considered as possible predictors of outcome. We employed a recently developed approach12 that uses digital callipers to measure arch height and mid-foot width in weight bearing and non-weight bearing. A pilot study demonstrated high intrarater (ICC 0.97–0.99) and interrater (0.90–0.98) reliability associated with these measures.12
In their review of foot orthosis research, Nigg et al13 concluded that the traditional notion of skeletal re-alignment being the major function of foot orthoses was questionable, and highlighted the importance of comfort. They proposed that effective orthoses would be perceived as comfortable if they supported the skeleton's preferred path during a given movement task, possibly acting as a surrogate to optimal muscle activity and energy expenditure as well as minimising fatigue while improving function. On this basis, we posit that comfort may be a potential predictor of successful outcome with foot orthoses.
Practitioners and patients would benefit from the ability to identify a priori those with PFP who are likely to respond favourably to foot orthoses. Clinical prediction rules are tools used to improve decision-making by identifying a parsimonious group of factors from the patient history and physical examination to maximise accuracy in predicting response to treatment.14
The aim of this post-hoc analysis was to identify possible predictors of successful treatment of PFP with foot orthoses; in particular, to determine if patient characteristics (eg, age, height, weight, gender, foot morphometry), perceived comfort of orthoses or pain characteristics (pain severity and duration) measured at the outset could predict success at 12 weeks.
We conducted a post-hoc analysis of data from a previously reported single-blind, single-centre RCT of foot orthoses, flat inserts, physiotherapy and foot orthoses plus physiotherapy in PFP.1 2 Only the group randomly assigned to receive foot orthoses was used in this analysis. We limited analysis to 12-week data to minimise the number of comparisons and the consequential risk of inflated type I error, and to curtail spurious associations through post-hoc data dredging.15 We also considered 12 weeks a likely time point of interest to patients (ie, several weeks after discharge). Ethical approval was conferred by the University of Queensland's Medical Research Ethics Committee.
Participants responded to calls for volunteers in the southeast corner of Queensland between May 2004 and May 2006. Inclusion criteria were: (1) age 18–40 years; (2) anterior or retropatellar knee pain of non-traumatic origin that was greater than 6 weeks' duration and provoked by at least two predefined activities (prolonged sitting or kneeling, squatting, jogging or running, hopping, jumping, or stair walking); (3) pain on palpation of the patellar facets, or with step down from a 25-cm step, or double leg squat; and (4) pain over the previous week of at least 30 mm on a 100 mm visual analogue scale. Exclusion criteria were concomitant injury or pathology of other knee structures, previous knee surgery, patellofemoral instability (history of sublaxation or dislocation; positive apprehension test), knee joint effusion, Osgood–Schlatter's or Sinding–Larsen–Johanssen disease, any foot condition that precluded the use of foot orthoses, hip or lumbar spine pain (local or referred), physiotherapy within the previous year, previous foot orthoses treatment or the use of anti-inflammatory drugs or corticosteroids.
Participants attended six appointments (30 min) over a 6-week period with one of 17 trained registered physiotherapists. These sessions were used to fit and modify the orthoses to optimise effectiveness, attempting to replicate clinical practice and to encourage wearing of the orthoses for the duration of the study.
The foot orthoses and fitting protocol used have been described elsewhere.2 In brief, participants were provided with four pairs (to fit a range of shoes) of prefabricated ethylene-vinyl acetate foot orthoses (Vasyli International, Labrador, Australia), which have a manufacturer-specified 6° varus wedge and arch support inbuilt. Physiotherapists selected from a range of ethylene-vinyl acetate densities (shore A ratings of 75°, 60°, 52°) and modification techniques (heat moulding, wedge and heel raise additions) to suit the individual participant, and followed a standard protocol for orthosis fitting and modification that emphasised fitting first to optimise comfort and second to improve pain-free function.2 16
Participants rated their perceived effect of the orthoses on a 5-point Likert scale (marked improvement, moderate improvement, same, moderate worsening, marked worsening). Patients were dichotomised as having experienced a successful outcome (those who scored ‘marked improvement’) or non-success (any other score on the scale).
Predictor variables (independent variables)
We collected data pertaining to participant characteristics, orthosis fit and the musculoskeletal condition (PFP) at baseline and the first intervention appointment. Participant characteristics covered the usual demographic features of age, sex, height, weight and body mass index, as well as morphometric foot measures.12 We took (weight bearing, equal weight on each foot) and non-weight bearing measures of the mid-foot (recorded at 50% of total foot length) using a digital calliper (Mitutoyo, Kanon, Japan). Weight-bearing arch height was also expressed as a proportion of the truncated foot length (posterior heel to first metatarsophalangeal joint line) and termed the ‘arch height ratio’.16,–,18 We also calculated the change in mid-foot width from non-weight bearing to weight bearing.12 We conducted a pilot study on test–retest reliability (n=10 participants, three raters) before conducting the RCT and found sound reliability (weight-bearing arch height ICC (SEM mm) 0.98 (0.4); weight-bearing mid-foot width 0.92 (1.1)).
At the commencement of each physiotherapy appointment, participants rated the overall comfort of the foot orthoses on a visual analogue scale (0 mm = too uncomfortable to wear; 100 mm = no discomfort). They completed an identical scale following any modifications made to the orthoses during the appointment. Mündermann et al19 found moderate reliability for a similar scale (ICC 0.799).
PFP was characterised by way of the duration of knee pain, severity of worst and usual pain over the the preceding week (visual analogue scale, 0 mm = no pain; 100 mm = worst pain imaginable),20 the anterior knee pain scale21 and the functional index questionnaire.22 The reliability of these measures in individuals with PFP has been established as fair to substantial23 (worst pain ICC (SEM) 0.76 (0.6); usual pain 0.56 (0.6); anterior knee pain scale 0.81 (3.1); functional index questionnaire 0.49 (1.2)).20
We used SPSS version 15.0 statistical software package to determine whether any potential prognostic variables identified patients who benefited from orthoses. The reference criterion for success was a score of ‘marked improvement’ on the global improvement scale at the 12-week follow-up. We tested individual variables from self-report measures, history and foot morphometry for a univariate relationship with success by using independent sample t tests for continuous variables and χ2 tests for categorical variables. Those with a significance level of p<0.20 were retained as potential prediction variables.24 This liberal significance level was selected to increase the likelihood that no potential predictor variables would be overlooked.
For continuous variables with a significant univariate relationship, sensitivity and specificity values were calculated for all possible cut-off points, and plotted as receiver operator characteristic curves.25 The point on the curve nearest the upper left-hand corner represented the value with the best diagnostic accuracy, and was selected as the cut-off defining a positive test.25 Sensitivity, specificity and positive likelihood ratios were calculated for each potential predictor variable. To determine the most accurate set of variables for the prediction of treatment success, potential predictor variables were entered into a stepwise logistic regression model. A significance level greater than 0.10 was required for removal from the equation to minimise the likelihood of excluding potentially useful variables.24 Variables retained in the regression model were included in the clinical prediction rule as the most parsimonious subset of predictors for identifying success with foot orthoses.
At 12 weeks, data were available for 42 participants (table 1). A successful outcome was recorded in 17 (40%) participants, with the remainder (25, 60%) rated as non-successes. Analyses revealed nine potential predictor variables (table 2) that exhibited a significance level of less than 0.20; these were subsequently entered into the logistic regression. The cut-off values determined by the receiver operator characteristic curves and diagnostic accuracy statistics for all nine variables are reported in table 2. The positive likelihood ratios ranged from 1.5 to 4.4. Four variables were retained in the final regression model: age over 25 years, height less than 165 cm, worst pain less than 53.25 mm, mid-foot width difference greater than 10.96 mm (table 3). These four variables were used to form the most parsimonious combination of predictors for identifying patients with PFP likely to have a dramatic improvement with orthoses. The pretest probability for the likelihood of success with orthoses was 40%. If the patient exhibited three of the four variables the positive likelihood ratio was 8.8 (95% CI 1.2 to 66.9) and the post-test probability of success increased to 86%. If the patient was positive on two of the four variables the positive likelihood ratio was 2.2 (1.1 to 4.2) with a 61% post-test probability of success. In real participant numbers, of the 33 who were positive on at least one variable, 17 experienced a successful outcome; of the 21 who were positive on at least two variable, 12 were in the success group; and of the seven who were positive on three variables, six were in the success group. There were no patients who satisfied all four criteria.
The probability of success when using foot orthoses indiscriminately (ie, without any predetermined stratification on foot posture or any other patient characteristic) for the treatment of PFP was 40%. This can potentially be increased more than twofold (ie, to 86%) if the orthoses are supplied to patients who have three of the following features: worst pain severity less than 53 mm; an increase in mid-foot width of more than 11 mm when the foot is loaded; older than 25 years, or shorter than 165 cm. Although these are preliminary findings from post-hoc analyses, it is readily apparent that careful selection or stratification of those receiving a specific intervention such as foot orthoses is likely to improve the probability of success substantially.
There are three key differences between our findings and those of the only other published study on this topic.11 First, a predictor set cluster emerged in our study (positive likelihood ratio 8.8; success rate improved from 40% to 86%) but not in that of Sutlive et al11 (best positive likelihood ratio 4; success rate improved from 60% to 86%). Second, our clinical prediction rule included pain severity along with patient characteristics (age, height and foot morphometry), whereas Sutlive et al11 found only foot measures of forefoot alignment, navicular drop and great toe extension. Third, the foot measures used in our study exhibited high reliability, whereas those used by Sutlive et al11 may be sufficiently unreliable to compromise the validity of their testing. Notwithstanding these differences, there is a strong similarity in the nature of the foot measures identified in that they all represent surrogate clinical measures of foot mobility. In summary, future research of predictors of success would be best served by not only including reliable measures of foot morphometry, but also pain severity and patient characteristics (eg, age and height).
In addition to the cluster of predictor variables, a number of variables had significant univariate relationships with successful orthoses outcome. These included measures of less severe PFP (lower usual pain severity, and higher functional index questionnaire and anterior knee pain scale scores), another foot posture measure (lower arch height ratio, indicating increased pronation) and an orthosis-related measure of comfort (lower overall comfort). The latter is a particularly interesting finding considering that we used comfort as the primary criterion for fitting the orthoses, and the work of Nigg et al.13 Although this finding does contrast our hypothesis of success having an association with greater comfort, the small increase in probability of success should be kept in mind (40% to 55.6%). Nevertheless, this finding does highlight the need to investigate further the role of perceived comfort in clinical prediction rules for foot orthoses.
This post-hoc analysis was conducted as a preliminary investigation into a clinical prediction rule for success with foot orthoses. The nature of the data source is that the sample size was powered to detect between-group differences within a large RCT. As such, it is possible that the small sample size, as well as the inclusion of a number of variables in the logistic regression model, may have resulted in over-fitting of the model, which can lead to spurious findings. However, in the initial stages of identifying predictor variables it is important and necessary to include all potential predictors. Any variable that may have fallen out as a predictor also needs to be re-examined in future studies, especially given the likelihood of type II error in post-hoc analyses such as this. Nevertheless, this study provides preliminary findings that highlight the need for the conduct of larger-scale investigations into clinical prediction rules for foot orthoses success in PFP.
This preliminary investigation of the potential predictors of success with foot orthoses treatment in PFP has identified that a combination of three of the following (age, height, pain severity and foot width mobility) increases the probability of success from 40% to 86%. This post-hoc analysis serves as a basis for further studies and may be used by practitioners as a guide to assist clinical decisions in managing PFP with foot orthoses.
The authors wish to thank Vasyli International for donating the foot orthoses, and the project physiotherapists and their practice staff for delivering the interventions.
Funding The RCT was funded primarily by the National Health and Medical Research Council of Australia (Primary Health Care Project Grant #301037). NC was the recipient of a National Health and Medical Research Council of Australia Public Health Scholarship (#351663). Vasyli International donated the foot orthoses. The funding organisations had no involvement in the study design, data collection, analysis and interpretation, writing of the report, or decision to submit the paper for publication.
Competing interests BV has been reimbursed by Vasyli International for seminar presentations not directly related to this study and has also received research funding from this company. Vasyli International is the manufacturer and donator of the foot orthoses in this study. All other authors declare that they have no competing interests.
Ethics approval Ethics approval was conferred by the University of Queensland's Medical Research Ethics Committee.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.