knee Pain
Physical Therapy Interventions for Knee Pain Secondary to Osteoarthritis: A Systematic Review FREE
Abstract
Background: Osteoarthritis is a leading cause of disability. Nonsurgical treatment is a key first step.
Purpose: Systematic literature review of physical therapy (PT) interventions for community-dwelling adults with knee osteoarthritis.
Data Sources: MEDLINE, the Cochrane Library, the Physiotherapy Evidence Database, Scirus, Allied and Complementary Medicine, and the Health and Psychosocial Instruments bibliography database.
Study Selection: 193 randomized, controlled trials (RCTs) published in English from 1970 to 29 February 2012.
Data Extraction: Means of outcomes, PT interventions, and risk of bias were extracted to pool standardized mean differences. Disagreements between reviewers abstracting and checking data were resolved through discussion.
Data Synthesis: Meta-analyses of 84 RCTs provided evidence for 13 PT interventions on pain (58 RCTs), physical function (36 RCTs), and disability (29 RCTs). Meta-analyses provided low-strength evidence that aerobic (11 RCTs) and aquatic (3 RCTs) exercise improved disability and that aerobic exercise (19 RCTs), strengthening exercise (17 RCTs), and ultrasonography (6 RCTs) reduced pain and improved function. Several individual RCTs demonstrated clinically important improvements in pain and disability with aerobic exercise. Other PT interventions demonstrated no sustained benefit. Individual RCTs showed similar benefits with aerobic, aquatic, and strengthening exercise. Adverse events were uncommon and did not deter participants from continuing treatment.
Limitation: Variability in PT interventions and outcomes measures hampered synthesis of evidence.
Conclusion: Low-strength evidence suggested that only a few PT interventions were effective. Future studies should compare combined PT interventions (which is how PT is generally administered for pain associated with knee osteoarthritis).
Primary Funding Source: Agency for Healthcare Research and Quality.
Osteoarthritis (OA) is a progressive joint disorder (1–2). Knee OA affects 28% of adults older than 45 years and 37% of adults older than 65 years in the United States (2–3). Osteoarthritis is a leading cause of disability among noninstitutionalized adults (2). Its prevalence and health impact are expected to increase as the population ages (4).
Osteoarthritis treatments aim to reduce or control pain, improve physical function, prevent disability, and enhance quality of life (5). Nonsurgical OA management combines pharmacologic treatments with physical therapy (PT) interventions (6–9). Guidelines recommend exercise as the core treatment of OA (6–7, 10). The marginal effects of specific exercise types (aerobic, aquatic, strength, and proprioception) have not been systematically reviewed.
This review evaluates the efficacy and comparative effectiveness of available PT interventions for adult patients with knee OA (11).
We searched MEDLINE, the Cochrane Library, the Physiotherapy Evidence Database, Scirus, Allied and Complementary Medicine, and the Health and Psychosocial Instruments bibliography database from 1970 to February 29 2012. We manually searched reference lists from systematic reviews and eligible studies. We used relevant Medical Subject Headings (MeSH) terms and text words, including osteoarthritis knee,physical therapy modalities, pain measurement, activities of daily living, and quality of life (Supplement 1). We searched ClinicalTrials.gov for completed trials related to the key questions. We did not contact primary investigators, but we did request additional information from sponsors of ongoing trials.
At least 2 investigators determined study eligibility (14). We included original publications of randomized, controlled trials (RCTs) published in English. Eligible trials enrolled community-dwelling adults with knee OA and reported pain as an inclusion criterion or outcome. Disagreements about the appropriateness of an article were resolved through discussion.
Eligible interventions were those within the scope of PT practice (regardless of whether the articles clearly described the involvement of a physical therapist or physical therapist assistant) (Appendix Table 1) (15). Eligible comparators included sham stimulation, usual care, and no active treatment for analyses of efficacy and PT interventions for the analysis of comparative effectiveness. Eligible patient-centered outcomes included knee pain, disability, quality of life, perceived health status, and global assessments of treatment effectiveness. Eligible intermediate outcomes included gait function, strength, transfers, joint function, or a composite measure of functional performance.
We excluded studies involving children, adolescents, hospitalized patients, or patients in long-term care facilities. We also excluded studies of surgical or pharmacologic treatments for knee OA and those that examined PT delivered in rehabilitation programs for adults with knee OA who had knee arthroplasty within 6 months before the study. Because the effects of PT may differ between hip and knee OA, we synthesized the results from studies that enrolled patients with knee or hip OA only if they reported the outcomes separately.
To assess harms of treatments, we included the findings of nonrandomized clinical trials, case series, and observational cohort or case–control studies. Possible adverse events included injuries related to exercise, back or foot pain, falls, blisters related to orthotics, or skin burns related to diathermy and electrical stimulation. We defined harms as a totality of all possible adverse consequences of discontinuing an intervention or treatment because of adverse events (16). We included all evidence of adverse events with eligible interventions regardless of how authors perceived causality of treatments (16).
Two researchers used standardized forms to extract data (17). One reviewer abstracted an article, and a second reviewer checked the data for accuracy. Discrepancies were documented, discussed, and resolved by consensus.
For categorical variables, we abstracted the number of participants randomly assigned to each treatment group and follow-up duration after randomization. For continuous variables, we abstracted means and SDs and the follow-up duration after randomization. For crossover trials, we abstracted the outcome levels after randomization if reported by the authors. We abstracted mean age; mean body mass index; proportions of women, minorities, and participants with disabilities; severity of knee OA; comorbid conditions; multijoint OA; baseline physical activity level; occupation; and concomitant drug and PT interventions. We abstracted settings; supervision of treatments by physical therapists; and dose, length, and intensity of the interventions when reported by the authors.
We assessed risk of bias in RCTs by using predefined criteria from the Cochrane Risk of Bias tool that included adequacy of randomization, allocation concealment, and intention-to-treat principles (14). We assigned studies as having medium risk of bias if at least 1 criterion was not met and high risk of bias if 2 or more criteria were not met. We abstracted masking information, but we did not include it in the assessment of risk of bias. We abstracted loss of follow-up and used the number of randomly assigned participants in calculating mean differences in measurement of pain, disability, and other outcomes.
We assessed strength of evidence according to guidelines from the Evidence-based Practice Center Program at the Agency for Healthcare Research and Quality (AHRQ) (Table 1) (18). We focused on direct evidence from head-to-head RCTs. We judged the strength of evidence for each major outcome following the GRADE (Grading of Recommendations Assessment, Development, and Evaluation) criteria according to risk of bias, consistency, precision, and when appropriate strength of association (18). We defined treatment effect estimates as precise when pooled estimates had reasonably narrow 95% CIs and pooled sample size was greater than 400 (19). We defined strength of association by using the Cohen criteria of large effect corresponding to standardized mean differences (SMDs) greater than 0.8 (20).
We focused on patient-centered outcomes, including pain, disability, and quality of life. We categorized intermediate outcomes as measurements of gait function, strength, transfers, or joint function or a composite measure of functional performance. Follow-up duration was categorized as less than 6 weeks, 6 to 13 weeks, 14 to 26 weeks, or more than 26 weeks.
We categorized eligible interventions in accordance with the definitions and hierarchy of interventions found in the Guide to Physical Therapist Practice (15) (Appendix Table 1).
Exercise interventions in eligible RCTs are described in Appendix Table 2. When we found more than 1 study from a particular trial, we used the results from the latest published paper.
Pooling criteria required that definitions of PT interventions and outcomes be the same. We categorized instruments according to similar domains with respect to pain, disability, quality of life, and composite function (Appendix Table 3). We used SMDs to combine effect estimates obtained from quality-of-life instruments or pain scales (14). Negative direction in absolute values of measured pain, disability, and other outcomes corresponded to improvement in the outcomes.
We prioritized pooled analyses over nonpooled results and longer versus shorter follow-up. We used the Hedges methods to calculate SMDs for different continuous measures of the same outcome (21). We evaluated the consistency of findings by comparing the direction and strength of the effect (18) along with the degree of statistical heterogeneity (based on the chi-square and I2 statistics) in effects across studies (22–23). We used random-effects models to pool results to account for inevitable variation in patient populations, concomitant treatments, and specific components of PT interventions (24). We used meta-regression and subgroup analyses to evaluate the effects of a priori–defined clinical and study characteristics on pain and physical function.
When heterogeneity was significant, we explored the effects of clinical diversity (age, sex, race, baseline activities of daily living, instrumental activities of daily living, comorbid conditions, inclusion of adults with previous knee arthroplasty, and obesity); exercise type, dose, and duration; specific study quality criteria; and physical therapist supervision. We investigated heterogeneity with individual quality criteria and crossover design rather than global quality scores (24–26). We did not use statistical tests for publication bias (14, 27–29).
We back-transformed SMDs to mean differences with several measures. For disability, we used EuroQol-5D (EQ-5D) (14), a multiattribute, preference-based health status measuring instrument (30). For quality of life, we used the 36-Item Short Form Health Survey (SF-36) (31). For pain, we used the visual analogue scale (VAS) (32). For composite function, we used the Western Ontario and McMaster Universities (WOMAC) Osteoarthritis Index function score (33). For gait function, we used walking speed (32). Pooled SDs of these measures were derived from large population-based studies of noninstitutionalized adults (30–33).
We multiplied the SMDs by the pooled SDs mentioned previously to yield an estimate of the difference in mean outcome scores (with vs. without intervention) on EQ-5D score (SD, 0.38), SF-36 score (SD, 10.9), VAS score (SD, 22 on a scale of 0 to 100), WOMAC physical function score (SD, 18.5), and walking speed (SD, 0.2 m/s) (30–33). We categorized treatment effects by the clinical importance of differences in intermediate outcomes. We used definitions of minimum clinically important differences from published studies and evidence-based reports (Appendix Table 4) (34). To assess the clinical importance of pain reduction with interventions, we did subgroup analyses with a subset of the studies that used the same VAS instrument for pain measures. We then compared mean reduction in pain with the cutoff for minimum clinically important differences in VAS scores, as reported in observational studies (Appendix Table 4).
We used Stata, version 11 (StataCorp, College Station, Texas), for all analyses (35).
The study was funded by AHRQ. The questions were developed with stakeholder input as part of AHRQ's Effective Health Care Program. The AHRQ provided copyright release for this manuscript but had no role in the literature search, data analysis, conduct of the study, preparation of the review, or interpretation of the results. It reviewed and approved the submitted manuscript without revisions.
The 4266 retrieved reports yielded 212 eligible articles (from 193 RCTs) that contributed to the evidence synthesis (Figure 1). We excluded 2085 references at screening; 1605 because of ineligible target population, interventions, and outcomes or because they were guidelines or reviews; and 154 of the observational nontherapeutic studies that examined the association between intermediate and patient-centered outcomes. No sponsors of ongoing trials responded to our requests for scientific information packages.
Most RCTs demonstrated adequacy of randomization (138 of 193 RCTs). Adequacy of allocation concealment was unclear in most studies (129 of 193 RCTs) (Appendix Figure 1). The most common reason for increased risk of bias was no planned intention-to-treat analyses (118 of 193 RCTs). One third of RCTs (68 of 193) reported open-label design with no masking of the outcome assessment. The mean attrition rate was 10.3% (SD, 10.6%).
Overall, RCTs had good applicability to our target population because they primarily recruited older adults with knee OA. More than 70% of the participants were women. Most participants were overweight; body mass index of participants averaged approximately 29 kg/m2. In 100 RCTs (52%), participants were taking anti-inflammatory drugs or pain relievers. One half of the studies provided no information about exact pharmacologic treatments. Few studies specified that they excluded patients with previous knee surgery. Most studies did not report participants' occupation, knee injury, comorbid conditions, or duration of condition or the proportion of participants who had baseline disability or surgery.
Only 84 RCTs met pooling criteria and provided evidence for 13 PT interventions on pain (58 RCTs), physical function (36 RCTs), and disability (29 RCTs). Evidence on long-term PT effects was available for 7 intervention–outcome pairs. We found few statistically significant differences in outcomes between active and control treatments (Supplement 2). We downgraded strength of evidence because of risk of bias and low precision of the estimated treatment effects. Individual RCTs reported the same direction of the effect, but variations in statistical significance and strength of the effects contributed to statistically significant heterogeneity in some cases, which we explored according to a priori–defined characteristics of participants and treatments.
Meta-analyses provided low-strength evidence that aerobic exercise (11 RCTs) and aquatic exercise (3 RCTs) improved disability; aerobic exercise (19 RCTs), strengthening exercise (17 RCTs), and ultrasonography (6 RCTs) reduced pain and improved function; and at short- but not long-term follow-up proprioception exercise reduced pain and Tai Chi improved function. Nonpooled results from individual RCTs did not show consistent statistically significant, strong, or clinically important changes in outcomes. Despite differences in interventions and outcome measures, the results from RCTs consistently demonstrated no benefits from specific education programs, diathermy, orthotics, or magnetic stimulation.
The effects of proprioception exercise were examined in 4 RCTs (247 participants) (39–42). Proprioception exercise improved pain (SMD, −0.71 [95% CI, −1.31 to −0.11], corresponding to a back-transformed VAS score difference of −15.6 on a scale of 0 to 100 [CI, −28.8 to −2.4]), but not composite function or gait function. The improvement was clinically important. Strength of evidence was low due to a high risk of bias in included trials.
Aerobic exercise effects were analyzed in 11 RCTs (1553 participants) (36–38, 43–52). Aerobic exercise led to statistically significant improvements in long-term pain (> 26 weeks) (SMD, −0.21 [CI, −0.35 to −0.08], corresponding to a back-transformed VAS score difference of −4.6 [CI, −7.7 to −1.8]) (36–37,46–49), and disability (SMD, −0.21 [CI, −0.37 to −0.04], corresponding to a back-transformed EQ-5D score difference of −0.08 [CI, −0.14 to −0.02]) (38, 46–48), but not psychological disability (43–45, 51) or health perception (38, 47–48). Within 3 months, aerobic exercise also improved composite function (WOMAC function score difference, −15.4 [CI, −24.8 to −5.92]) (45, 49, 53) and gait function (walking speed difference, −0.11 m/s [CI, −0.15 to −0.08 m/s]) (43, 45, 51, 53–56). At 12 months, the benefits of aerobic exercise continued for gait function (walking speed difference, −0.11 m/s [CI, −0.17 to −0.05 m/s]) (46, 52) but not for composite function (37, 46, 49). The pooled results showed that improvement in disability (but not in pain) was clinically important. Strength of evidence was low due to high risk of bias of the included trials. Several individual RCTs demonstrated clinically important improvements in pain and disability with aerobic exercise.
Effect estimates were statistically homogeneous. A meta-regression analysis exploring pain relief after approximately 3 months of aerobic exercise compared with placebo found no factor that could have consistently modified PT effects. Pain relief at approximately 3 months was consistent in RCTs that examined aerobic exercise supervised by a physical therapist. However, improvement in composite function at 3 months with aerobic exercise was greater in RCTs that reported no supervision of a physical therapist. Subgroup analyses with a subset of the studies that used the VAS instrument for pain measures found that the effect size of aerobic exercise at 3 months exceeded the minimum clinically importance difference; however, the result was not statistically significant.
Three RCTs (348 participants) studied the effectiveness of aquatic exercise (57–59). Aquatic exercise reduced disability (SMD, −0.28 [CI, −0.51 to −0.05], corresponding to a back-transformed EQ-5D score difference of −0.11 [CI, −0.19 to −0.02]), but had no statistically significant effects on pain relief or quality of life (57–58).
The effects of strengthening exercise were examined in 9 RCTs (1982 participants) (39, 46, 58, 60–65). Strengthening exercise had no statistically significant effect on disability or quality of life (46, 58, 60, 62). However, we saw a persistent improvement in pain relief (SMD, −0.68 [CI, −1.23 to −0.14], corresponding to a back-transformed VAS score difference of −15.0 [CI, −27.1 to −3.1]); composite function (SMD, −1.00 [CI, −1.95 to −0.05], corresponding to a back-transformed WOMAC function score difference of −18.5 [CI, −36.1 to −0.93]); and gait function (SMD, −0.39 [CI, −0.59 to −0.20], corresponding to a back-transformed walking speed difference of −0.08 m/s [CI, −0.12 to −0.04 m/s]) at 3 to 12 months of follow-up (39–40, 46,58, 60–64, 66–73). The improvements in pain and composite function were clinically important. Strength of evidence was low due to medium risk of bias and significant heterogeneity.
Magnitude of the effect differed across the studies, with the I2 statistic greater than 0.64 and the chi-square P values less than 0.004. We used meta-regression to explore heterogeneity in gait function or composite function at 3 months after strengthening exercise compared with placebo, and we found no factor that could explain the heterogeneity. However, meta-regression exploring heterogeneity in pain relief approximately 3 months after strengthening exercise indicated that younger participants had significantly better outcomes (P = 0.020). In contrast with aerobic exercises, the involvement of a physical therapist in strengthening exercise did not demonstrate consistent association with outcomes. For example, in comparison with studies without PT involvement, studies with PT involvement demonstrated smaller effect size for gait function 3 months after strengthening exercise, yet the same studies showed greater effect size in long-term pain relief after strengthening exercise.
Subgroup analyses with a subset of the studies using the VAS instrument for pain measures found that strengthening exercise resulted in statistically and clinically significant long-term pain reduction (transformed mean difference, −12.8 [CI, −22.9 to −2.7], which exceeded the cutoff for minimum clinically important difference).
Tai Chi effects were analyzed in 3 RCTs (167 participants) (74–76). Tai Chi improved composite function measures (SMD, −0.44 [CI, −0.88 to 0], corresponding to a back-transformed WOMAC function score difference of −8.14 [CI, −16.3 to 0]), at approximately 3 months but had no statistically significant effects on pain or disability. The improvement in composite function was not clinically important. Strength of evidence was low due to medium risk of bias and imprecision.
Three RCTs (162 participants) studied the effectiveness of massage (77–79). Massage improved composite function (SMD, −0.55 [CI, −0.93 to −0.18], corresponding to a back-transformed WOMAC function score difference of −10.2 [CI, −17.3 to −3.33]) (77–78). The improvement was clinically important. Strength of evidence was low due to high risk of bias and imprecision.
The effects of orthotics were examined in 7 RCTs (364 participants) (80–86). Orthotics had no effect on short-term outcomes of composite function or gait function. Three Japanese studies offered low strength of evidence that elastic subtalar strapping improved composite function at approximately 3 months (SMD, −0.27 [CI, −0.53 to −0.02], corresponding to a back-transformed WOMAC function score difference of −5.00 [CI, −9.81 to −0.37]) (87–89). The improvement was not clinically important. Strength of evidence was low due to high risk of bias and imprecision.
Two RCTs (105 participants) studied the effectiveness of taping (90–91). Low-strength evidence suggested that taping did not improve pain, disability, composite function, or gait function (90–91). Different reporting formats precluded pooled analyses. Individual studies suggested that taping may provide short-term pain relief (90–92).
Electrical stimulation effects were analyzed in 7 RCTs (390 participants) (69, 93–98). Electrical stimulation led to statistically significant short-term improvements in pain (SMD, −0.71 [CI, −0.98 to −0.43], corresponding to a back-transformed VAS score difference of −15.6 [CI, −21.6 to −9.5]) (68, 96, 99–103). However, electrical stimulation worsened pain at 6 months (SMD, 0.57 [CI, 0.09 to 1.06], corresponding to a back-transformed VAS score difference of 12.5 [CI, 2.0 to 23.3]) (Appendix Figure 2) (93, 97). Low-strength evidence showed statistically significant improvements from electrical stimulation at 3 months for global assessment (95–96) and muscle strength (measured at 60-degree extension) (69, 94). Pooled analyses provided moderate-strength evidence of no improvement in disability or joint function and low-strength evidence of no improvement in composite or gait functional measures (69, 93–96, 102–105).
Subgroup analyses with a subset of the studies using the VAS instrument for pain measures found that electrical stimulation resulted in clinically significant short-term pain reduction (mean reduction, −17.2 [CI, −23.1 to −11.4], which exceeded the cutoff for minimum clinically important difference). At 3-month follow-up, however, electrical stimulation tended to worsen pain measured with VAS (effect size, 0.1 [CI, −6.2 to 6.3]).
Six RCTs (387 participants) studied the effectiveness of ultrasonography (94, 110–114). Low-strength evidence demonstrated that ultrasonography led to statistically and clinically significant reductions in pain (SMD, −0.74 [CI, −0.95 to −0.53], corresponding to a back-transformed VAS score difference of −16.3 [CI, −20.9 to −11.7]). Ultrasonography also resulted in statistically and clinically significant improvements in composite function (SMD, −1.14 [CI, −1.85 to −0.42], corresponding to a back-transformed WOMAC function score difference of −21.2 [CI, −29.8 to −12.8]) and gait function (SMD, −1.48 [CI, −2.08 to −0.89], corresponding to back-transformed walking speed difference of −0.30 m/s [CI, −0.42 to −0.18 m/s]) (94,110–111). Ultrasonography did not improve disability (112–113). Strength of evidence was low due to medium risk of bias and an imprecise estimate of the treatment effect.
Subgroup analyses with a subset of the studies by using the VAS instrument for pain measures found that ultrasonography resulted in statistically and clinically significant short-term pain reduction (transformed mean reduction, −10.5 [CI, −18.6 to −2.4], which exceeded the cutoff for minimum clinically important difference). At 3-month follow-up, however, the effect size of −6.9 (CI, −11.7 to −2.0) no longer exceeded the minimum clinically importance difference despite being statistically significant.
Diathermy effects were analyzed in 5 RCTs (382 participants) (94, 115–118). Diathermy led to statistically significant pain reduction at 1 month (SMD, −0.53 [CI, −0.96 to −0.10], corresponding to a back-transformed VAS score difference of −11.7 [CI, −21.1 to −2.2]) (116–119), but the effect was statistically insignificant at 3 months (94, 117–118). Diathermy had no effect on disability, composite function, joint function, or gait function (94, 116–119). Strength of evidence was low due to medium risk of bias and an imprecise estimate of the treatment effect. Subgroup analyses with a subset of the studies using the VAS instrument for pain measures found that diathermy resulted in clinically significant short-term pain reduction (mean reduction, −18.4 [CI, −28.0 to −8.8], which exceeded the cutoff for minimum clinically important difference). At 3-month follow-up, however, the effect size of −0.7 (CI, −8.2 to 6.8]) no longer exceeded the minimum clinically important difference.
In individual RCTs, PT interventions demonstrated similar effects on patient-centered outcomes (Table 2). Aerobic and aquatic exercise had the same benefits for disability and pain relief (126, 128), a finding consistent with the similar effect sizes demonstrated by these interventions in efficacy studies. One study demonstrated that Tai Chi was better than stretching exercise for physical disability, psychological disability, global assessment, and transfer function (129).
Moderate-strength evidence suggested that with exercise (aerobic and strengthening), high adherence (defined as the percentage of classes attended) was associated with better outcomes (Figure 2) (57,130–133). The higher exercise adherence subgroup had the lowest risk for incident disability in activities of daily living (130), a lower average depression score (131), a higher mean Quality of Well-Being Scale score (57), and greater improvements in 6-minute walking distance and disability (131). These results highlight the importance of exercise adherence.
Evidence was inconclusive for the treatment-modifying effects of patient age (46, 107, 134), body mass index (46, 135), race (46), knee malalignment (72, 134), and comorbid conditions (131, 136). Evidence from 5 RCTs showed no statistically significant differences in effects between men and women (46, 81, 89,137–138). Baseline OA severity may modify the effects of PT interventions on clinical outcomes, but the effect varied depending on the treatments, outcomes, or OA severity definitions (84–85, 134–135, 137,139).
Adverse events were uncommon and varied across interventions (Table 3). Skin irritation was reported with braces, insoles, taping, and electrical stimulation; swelling with braces, diathermy, and exercise; muscle soreness with electrical stimulation; throbbing sensation with diathermy, electrical stimulation, and pulsed electromagnetic fields; increased pain with diathermy, exercise, insoles, and pulsed electromagnetic fields; falls with insoles; and need for surgery with diathermy. Adverse events were not severe enough to deter participants from continuing treatment.
Our comprehensive analysis of patient-centered outcomes with PT interventions has implications for clinical practice. Our findings reflect previously published guidelines (7, 10) and systematic reviews (145–147) about core PT interventions. Specifically, interventions that empower patients to actively self-manage knee OA (such as aerobic, strength, and proprioception exercise) improved patient-centered outcomes. No single intervention, however, improved all outcomes. Pooled analyses provided low-strength evidence of no benefits from diathermy, orthotics, and magnetic stimulation.
Because of variability in the definitions of outcomes, we had to calculate SMDs. Statistically significant differences in SMDs do not necessarily reflect the clinical importance of improvement in outcomes. We evaluated clinical importance of improvement in outcomes on the basis of back-transformation of pooled SMDs and whether the results from individual RCTs met established criteria for clinically important changes. Aerobic, strength, and proprioception exercise demonstrated consistent effects for pain and disability. Yet, we must use caution when inferring clinically important benefits of PT interventions in clinical settings.
Our analyses further indicate a possible association between high adherence to exercise and improvement in knee pain and function. Thus, therapeutic exercise programs should focus on achieving higher adherence rather than increasing the amount or intensity of exercise.
Our review was complicated by the discrepancy between the recommended practice of PT and the design of studies that examined the interventions. Current practice guidelines recommend that PT be delivered with a combination of interventions (15). Most notably, current PT practice would not administer taping or bracing alone, but rather in combination with therapeutic exercise and possibly other interventions. Published RCTs have focused on the effects of combining several PT interventions; however, the varied components in these studies preclude meta-analyses (94, 111, 121, 148).
In addition, clinical care for adults with knee OA includes pharmacologic interventions (6–9, 147, 149–150). Randomized trials equally distribute concomitant treatments among treatment groups and, thus, can provide valid estimates of effects of the examined PT interventions. Examined trials rarely reported other treatments that patients may have received, which impeded meta-regression analyses, nor did they analyze outcomes separately in patient subgroups by concomitant treatments.
In most cases, strength of evidence was low due to imprecision and risk of bias. Sample sizes were small; total study participants reached 400 with only a few interventions (education and aerobic and strengthening exercise). Most trials had moderate risk of bias because they frequently excluded patients from the analyses. Many trials did not sufficiently describe the nature and intensity of interventions or the involvement of physical therapists, which further impeded our ability to conduct meta-analyses (151–152).
Inconsistent definitions and measurements of the outcomes hampered synthesis of evidence. Validated measurements of functional impairments relevant to PT practice are listed in the Guide to Physical Therapist Practice (15); however, the guide recommends neither clinically important thresholds for such measures nor monitoring treatment effects according to patient-centered outcomes. The Osteoarthritis Research Society International has recommended evaluating treatment success according to patient-centered outcomes and clinically important differences in the WOMAC scale (153–154). However, most trials reported outcomes as average scores for all patients in each treatment group without evaluating how many patients had clinically meaningful improvement in pain, function, or quality of life.
Our review has limitations. We included only studies published in English in journals or reported in ClinicalTrials.gov. Despite a comprehensive review, we do not know how many funded but unregistered studies we may have missed. We assumed publication bias without conducting formal statistical tests for publication bias (29). We did not contact the principal investigators of the completed registered studies for unpublished data. We relied on published information about methods and treatment adherence and did not contact the authors for clarifications of poorly reported information (155–156). We focused on interventions applicable to PT practice and did not analyze the benefits of weight loss in adults who were obese and had knee OA, which is associated with statistically significant reductions in self-reported disability (157).
Our report has implications for future research. First, consensus is needed about methods to judge the benefits of PT interventions (158). Benefits should be defined as rates of clinically important improvements in pain, independence in daily activities, and quality of life. Through meta-analyses of individual-patient data from previously conducted RCTs, researchers could categorize patients according to the clinical importance of any changes they experienced. Such meta-analyses may also provide good estimates of treatment effects in patient subpopulations by age, comorbid condition, severity of knee OA, and concomitant treatments. Fully powered trials should examine comprehensive and multimodal interventions that more closely resemble PT practice.
In conclusion, our analysis suggests that only a few PT interventions were effective, specifically exercise (aerobic, aquatic, strengthening, and proprioception) and ultrasonography. Limited direct evidence of comparative effectiveness demonstrated similar benefits in disability measures with aerobic, aquatic, and strengthening exercise. Pain relief was consistent with aerobic exercise supervised by physical therapists. No single PT intervention improved all outcomes, and some interventions, specifically diathermy, orthotics, and magnetic stimulation, demonstrated no benefit. Adverse events were uncommon and did not deter participants from continuing treatment. Most studies tested single interventions; evidence is sparse on the effectiveness of PT interventions delivered in combination, as is typical in practice.
Nenhum comentário:
Postar um comentário