Objectives The objective of this 30-week randomised crossover trial was to determine whether a multi-modal realignment treatmentwould be successful in relieving pain and improving function among persons with medial tibiofemoral osteoarthritis (OA).
Methods The authors conducted a double-blind randomised crossover trial of a multi-modal realignment treatment for medial tibiofemoral OA. Trial participants met American College of Rheumatology criteria for OA, with knee pain, aching or stiffness on most days of the past month and radiographic evidence of a definite osteophyte with predominant medial tibiofemoral OA. The authors tested two different treatments: (A) control treatment consisting of a neutral knee brace (no valgus angulation), flat unsupportive foot orthoses and shoes with a flexible mid-sole; and (B) active treatment consisting of a valgus knee brace, customised neutral foot orthoses and shoes designed for motion control. For each subject, the trial lasted 30 weeks, including 12 weeks each of active treatment and control treatment separated by a 6-week washout period. The primary outcome of the linear regression model was change in knee pain and function, as assessed by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC).
Results 80 participants with medial tibiofemoral OA were randomised. Their mean age was 62 years, their mean body mass index was 34 kg/m2 and their mean WOMAC Pain score was 9.2 (0–20 scale). There was no evidence of a carryover effect. The regression model demonstrated that the mean difference in pain between the active treatment and the control treatment was −1.82 units (95% CI −3.05 to −0.60; p=0.004) on the WOMAC Pain scale, indicating a small but statistically significant decrease in pain with the multi-modal active treatment. For WOMAC Function, the realignment intervention had a non-significant effect on function, with a −2.90 unit decrease (95% CI −6.60 to 0.79) compared with the control condition (p=0.12).
Conclusion Multi-modal realignment treatment decreases pain in persons with medial tibiofemoral OA.
Statistics from Altmetric.com
The etiopathogenesis and progression of symptomatic knee osteoarthritis (OA) are driven by mechanical factors.1 A number of biomechanical studies have demonstrated improvements in certain aspects of gait and biomechanics with valgus bracing among persons with knee OA.2,–,4 In addition, numerous randomised controlled trials (RCTs) have investigated the clinical effects of valgus bracing on improving pain and function outcomes among persons with symptomatic knee OA.3 ,5,–,15 While the findings have generally been supportive, several study design limitations, including the studies being uncontrolled5,–,10 and/or underpowered,11,–,15 have called the results of these few RCTs into question.16 ,17
The two largest studies to date have suggested positive effects on symptoms. Kirkley et al18 found significant improvements in a braced treatment group compared with an unbraced control group. However, this trial may not have been appropriately controlled since an active brace intervention was compared to no intervention at all.18 Given the profound effects of placebo on OA19 and the potential for symptomatic improvements with neoprene sleeve interventions alone,20 the findings of Kirkley et al require confirmation. A more recent RCT by Brouwer et al21 found that valgus knee bracing resulted in improved knee function but in no significant improvements in knee pain compared with no bracing. However, many participants in the Brouwer et al trial did not fully adhere to brace treatment as a result of skin irritation and poor fit.
Given that the findings of previous trials have been somewhat equivocal, disease management guidelines have not advocated for the use of braces and have recommended that further research is needed.17 Echoing this need, recent systematic reviews of unloader braces for knee OA found modest evidence for their effectiveness and also recommended that further research be conducted,3 ,16 hence the need for an appropriately powered and well-controlled trial on the effect of valgus knee bracing on persons with medial knee OA.
Unfortunately, biomechanical studies demonstrate that even with appropriate valgus bracing, large mechanical stresses on the knee can persist, suggesting that the addition of other interventions to further improve limb alignment may be of therapeutic value.9 Multiple orthotic modalities to decrease forces across a knee may be necessary in order to fully unload an osteoarthritic medial compartment and to bring about significant improvements in knee pain and function. Recognising the relationship between knee and foot biomechanics,22 the combination of a valgus knee brace with motion control shoes and neutral foot orthoses has been proposed as a promising multi-modal strategy.3
The overall objective of this study was to determine whether the provision of a multi-modal realignment treatment relieves knee pain and improves function among patients with medial tibiofemoral OA in a 30-week randomised crossover clinical trial. We tested the hypothesis that, compared to control treatment, a multi-modal realignment treatment that includes a valgus knee brace, motion control shoes and neutral foot orthoses (customised shoe inserts) is effective in reducing pain and improving function among persons with medial tibiofemoral OA.
Materials and methods
This study was a double-blind (participant and assessor) randomised crossover trial of a realignment treatment for medial tibiofemoral OA, with the primary outcome being knee pain and function as assessed by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC; VAS version)23. The trial was prospectively registered with the NIH Clinical Trials Registry (NCT00124462), and approval was provided by the Boston University institutional review board and the New England Baptist Hospital institutional review board. All participants provided written informed consent. The study conformed to the CONSORT (CONsolidated Standards Of Reporting Trials) requirements for RCTs. Study design and protocol development began in January 2005. Recruitment and enrolment began in June 2005. Follow-up of all subjects was completed in September 2008.
Given the possibility that even braces with no realigning capabilities can provide sensory input and improve symptoms on that basis alone, we compared the effects of active brace treatment to the effects of a control treatment consisting of comparable braces without realigning capabilities:
Control treatment (A): a neutral knee brace that does not have any varus/valgus angulation was given along with flat unsupportive foot orthoses and shoes with a flexible mid-sole.
Active treatment (B): a valgus knee brace was given along with customised neutral foot orthoses and motion control shoes.
A run-in design was used in order to maximise the likelihood of recruiting subjects who would remain in the trial. Subjects were randomised to receive either brace treatment A or brace treatment B for the initial 12 weeks. After 12 weeks, we removed the assigned brace, and participants received no treatment for 6 weeks. Following this 6-week washout period, the alternative brace treatment was assigned for the final 12 weeks. For each subject, the trial lasted for a total of 30 weeks.
The active treatment consisted of a DonJoy OAdjuster knee brace (DonJoy Braces, Coconut Creek, Florida, USA) with bilateral customised semirigid functional foot orthoses, which were crafted according to methods previously described24 using heat-mouldable Fastech (Fastech Labs, Troy, Michigan, USA) shells and medium-density Nickleplast-S (Alimed, Dedham, Massachusetts, USA) inners to support a neutral foot position (figure 1).The customised foot orthoses replaced the normal insoles of a New Balance 830 motion control shoe (New Balance, Brighton, Massachusetts, USA). The control brace consisted of the DonJoy Montana brace with a loosened screw at the hinge allowing varus/valgus laxity, a flat unmoulded 1/16 Fastech orthotic blank of a material (Poron) identical to that in the active treatment but without the heat-mouldable core and a New Balance court shoe (model 505) with a low-density mid-sole and a flexible upper soleto minimise motion control. Shoes and orthoses were worn on both feet during each treatment period.
Trial participants met American College of Rheumatology criteria for OA, with knee pain, aching or stiffness on most of the past 30 days and evidence of a definite osteophyte on radiograph. In addition, because we were interested in persons with predominantly medial tibiofemoral OA, participants had radiographic evidence of disease in the medial tibiofemoral compartment without predominant lateral tibiofemoral or patellofemoral involvement. Medial tibiofemoral disease25 required definite radiographic OA with at least grade 1 medial joint space narrowing (0–3 scale) using the Osteoarthritis Research Society International atlas.26 Individuals with clinical evidence of patellofemoral disease or knee pathology (other than medial compartment OA) that was likely to be contributing to knee pain (such as pes anserine bursitis) were excluded. Participants had to be ambulatory and limited in usual activities due to knee pain. The anatomical axis was measured from short films using previously validated methods.27
Exclusion criteria included the following: (1) individuals who usually used an ambulation aid (such as a cane, crutch, walker or wheelchair) to walk; (2) amputation of a foot or a previous major trauma to a foot that would raise concerns about whether an orthosis might worsen foot pain; (3) known neuropathy due to diabetes or other causes; (4) history of deep vein thrombosis; (5) pain emanating more from the back or hip than from the knee, as determined by a screening questionnaire and clinical exam; (6) planning to move from the area within 9 months of study screening; (7) symptomatic comorbid disease that limits walking more than knee pain; (8) receiving corticosteriod injections in the knee in the month prior to starting the trial; (9) receiving a stable dose of glucosamine and/or chondroitin and/or nonsteroidal anti-inflammatory drug for at least 2 months before beginning the trial and committing to not starting any new treatments during the trial (for participants receiving glucosamine and/or chondroitin and/or nonsteroidal anti-inflammatory drug); (10) bilateral total knee replacement or planning for total knee replacement of the index knee in the next 6 months; (11) other known causes of arthritis, including rheumatoid arthritis, systemic lupus erythematosus, gout, psoriatic arthritis and pseudogout; (12) failure to pass the 2-week run-in test; (13) height of 5 ft 0 inches or less due to incompatibility with brace fitting; or (14) past use of prescription brace or customised orthotic.
Additionally, persons with low WOMAC Pain scores at the time of prerandomisation screening were excluded. In order to properly evaluate response to treatment, we required that patients have a minimal score of at least 2 of 5 on at least 2 of 5 WOMAC questions, or a total of >6 of 20 on the WOMAC Pain scale in the index knee during both a prerandomisation phone call and a screening visit. In the event that both knees met all eligibility criteria, the most symptomatic knee served as the index knee.
Participants were recruited from among patients in the rheumatology and orthopaedic sections of Boston Medical Center and New England Baptist Hospital and from persons already recruited for other clinical trials, using various forms of advertising in local public media.
The efficacy of a clinical trial is maximised if subjects comply with assigned treatment and come to scheduled visits. One way to increase the likelihood of subject compliance with a trial is to perform a run-in test—a period of observation prior to randomisation during which subjects experience major components of the study protocol. Those subjects who have trouble complying with the protocol are excluded before being randomised. Participants had a 2-week placebo run-in28 with administration of the control shoe and control foot orthosis. Blindness was reassessed at the beginning of the study, and equal numbers of participants considered the control shoe and orthosis active and placebo interventions, respectively.
We preserved allocation concealment by having the randomisation codes held by a biostatistician at the Boston University School of Public Health, which is external to the Clinical Epidemiology and Research Training Unit where the trial examiners were located. We stratified patients into those with end-stage OA (Kellgren–Lawrence grade 4) and those with disease that was mild or moderate yet still predominantly affecting the medial joint. We then performed computer-generated blocked stratified randomisation in order to ensure that roughly equal numbers of patients with severe and mild/moderate diseases were randomised into each of the two treatment arms. The enrolment of participants and their assignment to intervention were conducted by separate research coordinators.
Participants were told only that we were comparing two types of treatment for their knee arthritis. We did not specify which treatment constituted active treatment. A single blinded examiner administered the knee pain and function outcome measures, while a second investigator with experience in foot orthotic customisation (KDG) fitted the shoe and customised the foot orthoses.
We applied a number of methods to monitor and improve adherence. Educational messages on the potential benefits of non-pharmacological treatment and skills training on donning the alignment treatment were discussed and provided to participants to increase confidence and motivation to comply.29,–,31 At each visit, we inquired about adherence in detailed fashion. To assess adherence, we called the subjects every week during the active phases of the trial (phase 1: 0–12 weeks (visits occurred at 0, 1, 3, 6 and 12 weeks) and phase 2: 18–30 weeks (visits occurred at 18, 19, 21, 24 and 30 weeks)). In addition, subjects kept a diary recording of their daily use of the combination of brace, shoes and insert during the course of the trial, and a pamphlet addressing common concerns that might interfere with adherence was provided. Participants were encouraged to wear the interventions for a minimum of 4 h/day.
The primary outcomes were the WOMAC Pain and Function subscales.23 The WOMAC, which has been extensively validated and is recommended by the Osteoarthritis Research Society International for use in OA clinical trials, has three subscales: the pain (5 items) subscale, the stiffness (2 items) subscale and the physical function (17 items) subscale. In this study, results on the pain and physical function subscales were analysed separately. All of the subscales have high test–retest reliability, and validation studies have shown high correlations with other indices probing similar constructs.23 In terms of responsiveness to change, the WOMAC has been compared to other measures of patient status in OA, including the Doyle Index, the Lequesne Index, walk time and range of motion,32,–,35 and has generally been found to be more responsive than these other instruments.
Analysis of this trial focused on the primary outcome measure, which was change in WOMAC Pain during treatment. The primary study question was whether the change in pain during the active treatment period differed from the change in pain during the control treatment period. We made similar comparisons of changes in WOMAC Function scores during the active treatment period and the control treatment period.
We performed a regression analysis of the relationship between treatment type (active or control) and WOMAC Pain and Function scores during the treatment period. We used a generalised estimating equation correction to account for repeated measures within individual participants. These methods were similar to those used in a prior crossover trial in OA.36 Additional regression models were constructed using an interaction term of Treatment×Period as predictors. In the two-period crossover trial, the Treatment×Period interaction was equivalent to the carryover effect, so this provided a method for assessing carryover effect. The washout period of 6 weeks between treatment periods was intended to reduce the likelihood of a carryover effect. This analysis used an intent-to-treat approach, with the last observation brought forward for missing values.
All statistical analyses were performed using version 9 of the SAS package. The GENMOD procedure was used for regression analyses with generalised estimating equation corrections for repeated measures.
Statistical power estimates
Based on the results of Horlick and Loomer15 and Kirkley et al18 on valgus knee bracing, we anticipated a conservative treatment effect difference of 30% in pain and function. We estimated a correlation of 0.6 in the primary outcome measure (WOMAC Pain) for measures taken within a subject. With 80 participants, we had 80% power (with an α of 0.05 in a two-sided test) to detect a treatment effect of a 30% reduction in WOMAC Pain compared with the control condition.
Of the 860 potential participants who were contacted by phone, 229 were eligible for a screening visit. The most common reason for ineligibility was insufficient pain. Of the 150 participants who had a screening visit, 80 were found to be eligible and were randomised. Of the 80 participants enrolled, 56 completed the study (figure 2). The main reasons for early termination were loss to follow-up, non-compliance and scheduled joint replacement.
There were no statistically significant differences in age, gender, body mass index or radiographic severity between those who were randomly assigned to receive the active treatment first and those who were randomly assigned to receive the control treatment first (table 1). Baseline WOMAC Pain score was slightly higher in the group randomised to the active treatment first.
When analysing the crossover trial findings for pain, we first tested for treatment period and differential carryover effects (table 2 and figure 3). The treatment period effect was a 0.43 unit difference (p=0.66) in WOMAC Pain change score, and the differential carryover effect was only a −0.10 point difference in the WOMAC Pain change score (p=0.95), indicating that treatment effectiveness did not depend on whether a participant was randomised to the active treatment or the control treatment first. Therefore, we removed the carryover term from the model. In model 2, excluding the differential carryover effect, the realignment intervention had a significant effect on pain, with a −1.82 unit decrease in pain (95% CI −3.05 to −0.6) compared with the control condition (p=0.004).
For WOMAC Function, the treatment period effect was a 0.58 unit difference (p=0.84) (table 3 and figure 4), and the differential carryover was a 0.99 unit difference on the 68 unit WOMAC Function scale (p=0.82), indicating once again that treatment effectiveness did not depend on whether a participant was randomised to the active treatment or the control treatment first. Therefore, we removed the carryover term from the model. In model 2, excluding the differential carryover effect, the realignment intervention had a non-significant effect on function with a −2.90 unit decrease (95% CI −6.60 to 0.79) compared with the control condition (p=0.12).
To facilitate an understanding of phase-specific effects (consistent with figures 3, 4), we have inserted additional information about pain and function by phase (table 4). This incorporates the carryover effect test and the estimated treatment effect from model 2 (without carryover effect) from tables 2, 3. The results are consistent with a favourable effect for realignment compared to placebo in each treatment phase.
The adherence of study participants and the reported side effects are detailed in table 5. On average, the participants wore the interventions for more than 3 h/day. The most frequent side effect reported among persons wearing the realignment treatment compared to the control was problems with brace positioning or slipping (16 participants compared to 4 participants in the placebo group). Participants also more frequently reported pain from poorly fitting shoes during the active treatment period (7 participants compared to 1 participant during the control period).
The need to develop efficacious, conservative, non-pharmacological treatment approaches that are capable of ameliorating the symptoms of knee OA is an important research objective.37 Despite insufficient attention from researchers and clinicians, treatments that are capable of targeting the pathomechanics of OA are likely to be efficacious.38 We have demonstrated that the application of realignment treatment consisting of a DonJoy OAdjuster knee brace, customised foot orthoses and New Balance 830 motion control shoes leads to a statistically significant improvement in knee pain compared to a placebo intervention.
The direction of effect (a 1.8 unit reduction in WOMAC Pain) is consistent with the positive effects seen in other randomised trials of knee braces.15 ,18 ,21 However, the magnitude of effect (~20% reduction in pain from baseline) is slightly less than that seen in previous brace trials. This discrepancy may result from differences in the study design and in the particular interventions tested. First, our study intervention involved a brace, a motion control shoe and a foot orthosis. Two previous trials demonstrated that wearing a valgus knee brace alone can result in substantial improvements in the pain and function of patients with medial knee OA.15 ,18 However, in contrast to these two previous trials, we also employed a rigorous control intervention, making the demonstration of a sizeable treatment effect more challenging. A third trial by Brouwer et al21 failed to demonstrate the efficacy of unloader bracing (either varus or valgus) within a study sample that included persons with both lateral and medial tibiofemoral OA. Among the possible reasons for this failure to demonstrate treatment efficacy were problems with adherence to brace treatment mainly because of skin irritation and poor fit. A subgroup analysis of persons with medial knee OA did find significant improvements in function among persons who were braced. In addition, we have conducted a placebo controlled trial where the control group received a similar intervention and equal attention as the intervention group. Hence, the specific treatment effect is distinct from the placebo effect. The placebo effects for self-reported outcomes, such as pain and function, in OA trials are substantial.39 This is a challenging aspect of trials, particularly of non-pharmacological treatments such as this trial.40
The medial compartment of the knee absorbs 60–70% of the force across the joint during weight bearing.41 ,42 The overwhelming majority of treatments available for OA involve drugs and/or surgery. Despite strong evidence for the potent effect of mechanics (in particular the external knee adduction moment) on disease progression and symptoms,38 ,43,–,47 there are few interventions that effectively reduce mechanical load. Our study results provide further evidence that an intervention targeted to the reduction of mechanical load can have a durable therapeutic benefit with non-serious side effects that are largely managed with training or other minor adjustments.
There are a number of limitations of this study that warrant discussion. First, there was no control group without an intervention because of concerns about unblinding. Prior research has highlighted the difficulty of disentangling the placebo effect in non-pharmacological clinical trials.40 ,48 ,49 Consistent with this prior research, participants in the current crossover trial experienced symptomatic improvement during both the active intervention period and the control intervention period, making it more challenging to detect relative improvement during the active intervention. There were also a number of dropouts in the trial. Most dropouts occurred during the second treatment condition, and efforts to reduce the impact of these censored data using an intent-to-treat analysis were applied.50 Lastly, the daily duration of optimal treatment is unknown for this type of intervention; therefore, the prescribed 4 h of daily use in this study was largely arbitrary. In the clinic, braces are often prescribed for use only during aggravating weight-bearing activities.
There are also a number of important strengths of this study that merit discussion. The crossover design facilitates efficient recruitment and protection against confounding by patient-related factors. Similarly, we were adequately powered to detect a clinically meaningful treatment effect. The effect we found, while statistically significant, approximates that of clinical significance. Reports of the minimum clinically important difference in pain using the WOMAC suggest that values reporting an improvement of 10–20% are of clinical importance.51 ,52 It is possible that treatment effects may differ according to baseline alignment or radiographic severity. We are currently undertaking further research on this sample to determine if that is the case.
Multi-modal realignment treatment has significant effects on pain in persons with medial tibiofemoral OA. Further studies of this intervention are warranted to corroborate our findings. In addition, biomechanical studies that include clinical outcome measures will be helpful in determining whether clinical improvement is a function of mechanical alterations.
The authors would like to thank the participants involved in the study, without whom this trial would not have been possible. The authors would also like to thank the staff responsible for coordinating the study. The trial was jointly supported by the National Institute on Disability and Rehabilitation Research (grant H133G040201) and DonJoy. The authors would like to thank DonJoy for providing the knee braces used in this trial and New Balance for providing the shoes. Neither company was involved in the study design and played a role in writing or editing this manuscript. Dr Hunter was funded by an Australian Research Council Future Fellowship.
Funding National Institute on Disability and Rehabilitation Research.
Competing interests None.
Patient consent Obtained.
Ethics approval Boston University institutional review board and the New England Baptist Hospital institutional review board.
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.