Executive Summary


This Future Research Needs (FRN) project is a followup to the draft Comparative Effectiveness Review (CER) “Physical Therapy Interventions for Knee Pain Secondary to Osteoarthritis.” The review was motivated by uncertainty around the effectiveness and comparative effectiveness of physical therapy (PT) treatments for adult patients with knee pain secondary to osteoarthritis (OA). The purpose of this FRN project is to identify and prioritize specific gaps in the current literature on PT for knee pain due to OA that would aid decisionmakers. We used a deliberative process to identify evidence gaps, translate gaps into researchable questions, and solicit stakeholder opinion on the importance of research questions. This report proposes specific research needs along with research design considerations that may be useful in advancing the field.

The analytic framework adapted from the original draft CER (Figure A) describes the process experienced by adults with knee pain secondary to OA once they are referred for PT. Important Key Questions (KQ) about the efficacy and effectiveness of these treatments (KQ 1), the relationship between intermediate and patient-centered outcomes and use of minimal clinically important differences (MCIDs) (KQ 2), and the potential harms of PT treatments (KQ 3) were addressed in the review.1

Figure A depicts the key questions within the context of the PICOTS. In general, the figure illustrates how different physical therapy interventions in adults with knee osteoarthritis may improve intermediate outcomes (e.g., joint movement, swelling, inflammation, or impairments in performance tests) and patient-centered outcomes (e.g., pain, activities of daily living, decreased disability; return to work/activities, or quality of life). Adverse events may occur at any point after treatment is received. Treatment effects may be modified by age, gender, race, baseline ADL/IADL disability, comorbidity, concomitant/prior treatments, activity level, or occupation. The figure also illustrates the questions about the validity, reliability, and minimal clinically important difference of the tests and measures to determine intermediate outcomes (e.g. manual muscle test, hand held dynamometer, isokinetic dynamometer). Dotted line points out the association between changes in intermediate outcomes with the changes in patient-centered outcomes.

Figure A

Analytic framework. KQ = Key Question; OA = osteoarthritis; PT = physical therapy

The authors of the draft CER found that the evidence for KQ 1 supported the use of various forms of exercise therapy and ultrasound. Exercise therapy was efficacious when supervised by a physical therapist and typically resulted in a clinically meaningful improvement in pain and disability outcomes. The evidence comparing various forms of exercise therapy demonstrated similar benefits in disability measures for aerobic, aquatic, and strengthening exercise. Adherence to exercise therapy was the key to efficacy. Diathermy, orthotics, and magnetic stimulation used as stand-alone treatments demonstrated no benefit. Evidence was insufficient to conclude the best treatment option among effective PT interventions or to conclude differences in effects by patient characteristics. No consistent associations between the duration of examined interventions or followup times and intermediate/patient-centered outcomes were found.

For KQ 2, the intermediate outcomes of gait, mobility restrictions, muscle strength, and range-of-motion measures were associated with patient-centered disability measures in individual studies. However, these intermediate measures could not adequately predict patient-centered outcomes. MCIDs were determined for several outcomes scales, but not used consistently.

For KQ 3, the authors found that adverse events were uncommon and not severe enough to deter participants from continuing treatment.

Study quality and heterogeneity in populations and treatments, including concomitant treatments, downgraded the strength of evidence to low or moderate in most cases. The authors also identified gaps in evidence limiting their ability to draw definitive conclusions. There were a limited number of comparative effectiveness studies and efficacy studies primarily addressed stand-alone therapies rather than combinations, common in current clinical practice. The CER did not address whether adjunct therapies were effective in regard to their intended goal of enabling patients to more fully participate in primary therapies. Which patients are likely to benefit from exercise therapy alone and who may need a broader treatment approach could not be addressed. Evidence was insufficient to draw conclusions about the most effective activities (aerobic, strength, etc.) or dosage (intensity, frequency, duration) within exercise therapy. Evidence about long-term effectiveness of PT interventions is limited. Another systematic review suggests that long-term effectiveness is enhanced when booster or followup PT sessions are employed.2


We used a deliberative process to identify and prioritize research questions relevant to the evidence gaps identified in the CER.1 Figure B illustrates the eight steps used to accomplish the objectives of this project.

Figure B describes the flow of the project in 8 steps. Step 1 is to identify the evidence gaps from the CER. Step 2 is to form and orient the stakeholder panel. Step 3 is to translate research gaps into researchable questions. Step 4 is to elicit stakeholder feedback on those questions via conference calls and emails. Step 5 is to revise the list of research gap questions based upon the stakeholder feedback. Step 6 involves having the stakeholders prioritize the revised set of research questions, higly prioritized questions are deemed research needs. Step 7 involves the EPC team identifying research design considerations for these research needs. And Step 8 is to develop the FRN report.

Figure B

Project flow. CER = Comparative Effectiveness Review; PICOTS = population, intervention, comparison, outcome, timing, and setting

First, research gaps identified in the CER were translated to research questions. Secondly, a diverse stakeholder panel with representation from various perspectives relevant to the topic was assembled. Research representatives were national experts familiar with evidence-based medicine and the obstacles faced in conducting well-designed research from the fields of rheumatology, orthopedics, and PT. Representatives from organizations supporting or conducting relevant research including the National Institute of Arthritis and Musculoskeletal and Skin Diseases, the National Institute on Aging, the American Physical Therapy Association as well as policy and payer representation from the Centers for Medicare and Medicaid Services and the Centers for Disease Control and Prevention participated on the stakeholder panel. Providers and consumers, including representation from the Arthritis Foundation, were also engaged because the decisional dilemmas faced by these groups are critical to identifying and prioritizing research questions.

We first held conference calls with stakeholders to refine the research gaps identified during the CER process. Based upon these conversations, we refined our initial list of research gap questions and categorized the questions by whether they were methodological, addressing issues necessary to enhance the usefulness of current research, or topical, addressing issues that have not been sufficiently addressed in the current literature. This list of research questions was sent to a select group of stakeholders for ranking. Stakeholders numerically ranked their top 3 methodological research questions from a total of 7 and their top 4 topical research questions from a total of 11.

Based upon the natural breakpoints in these rankings, we determined high, moderate, and low priority research gap questions. High priority questions were deemed research needs. We then identified and discussed research design considerations for research needs.


Prioritization Results

We analyzed weighted rankings for stakeholders participating in the Web-based prioritization process. From the 14 stakeholders invited to rank research questions, 12 ranked methodological questions and 11 ranked topical questions.

Methodological Research Needs

Natural breakpoints in weighted rankings revealed one high and four moderate priority methodological research questions Because only one methodological research question appeared as a high priority, we also considered the moderate priority research questions research needs. Addressing methodological research needs will enhance the utility and translation of current and future research on PT interventions for patients with knee pain secondary to OA.

  • Which patient-centered outcome measurement instruments should be used consistently by all relevant disciplines (e.g., PT, rheumatology, orthopedics)?
  • Which intermediate outcome measurement instruments should be used consistently by all relevant disciplines (e.g., PT, rheumatology, orthopedics)?
  • Should effectiveness research on PT treatments use MCIDs?
  • What confounding and effect modifying variables (e.g. OA severity, obesity, comorbidities, and concomitant therapies-including anti-inflammatory and analgesic medication) should be measured and reported in effectiveness research?
  • What minimum set of treatment factors (site, treatment components, frequency, duration, intensity, timing) should be reported consistently by all relevant disciplines (e.g., PT, rheumatology, orthopedics)?

Methodological research needs pertain to how effectiveness is measured and the consistency and completeness of research studies and reporting on interventions for knee pain secondary to OA. The draft CER emphasized that relatively few studies utilized MCIDs in evaluating efficacy and effectiveness. However, stakeholder discussions described problems in a reliance on MCIDs. While the concept of MCIDs offers a meaningful interpretation of scale scores, issues surrounding their calculation, reliability, and applicability to specific research populations, and the use of an average score to evaluate effectiveness of all patients deter their validity and utility.

Literature examined for the draft CER rarely provided adequate and consistent measurement and reporting of variables thought to confound or modify the effect of PT treatments for knee OA. Related to the reporting of confounding and effect modifying variables, stakeholders would like to see a consensus on the identification and measurement of specific intervention characteristics reported in studies.

Considerations for Potential Research Designs

Methodological research needs could be addressed through a consensus development process (i.e., consensus conference). Because knee OA is treated by more than one group of providers, a multidisciplinary approach to consensus development is ideal, including representation from clinical areas (PT, rheumatology, and orthopedics) and researchers with expertise in clinical outcomes, epidemiology, biostatistics, and health services research. Continuing consensus work, facilitated by the Osteoarthritis Research Society International and Outcome Measures in Rheumatology, on improving the reporting and measuring effectiveness in OA trials3 will offer valuable information to address this research need. Specific research needs, such as guidance in the use of MCIDs, may benefit from pre-work prior to the consensus development process. The information needs to facilitate a discussion on MCID could be identified, collected or generated, and distributed before discussion.

Topical Research Needs

A natural breakpoint in weighted rankings of topical research questions revealed four research needs. All topical research needs addressed the PICOTS (population, intervention, comparison, outcome, timing and setting) elements of populations and interventions. Addressing topical research needs will enhance understanding of efficacy and comparative effectiveness, which was limited in the draft CER. Current ongoing studies addressing specific hypothesis will not likely sufficiently answer the research questions. However, related ongoing studies should be watched and their contributions should be considered when future studies are planned.

First Topical Research Need

  • Which PT treatments work for which patients?

The draft CER, other reviews on the topic, current efficacy studies, and stakeholder discussions emphasized the need to address efficacy and comparative effectiveness for particular types of patients. While specific subgroups and interventions were not specified in this research need, subgroups can likely be defined by prevalent patient characteristics such as degree of symptoms, severity of disease, age, obesity and other characteristics that appear to have an effect on response to treatment.

Research Design Considerations

Topical research needs are best addressed with experimental designs. However, identifying specific patient subgroups (hypothesis generating research) may first be accomplished with less rigorous research designs. Review of previous systematic reviews, published trials including post hoc subgroup analyses, observational studies, and administrative databases could be used to extract hypothesized relationships between patient characteristics and specific therapies or multimodal treatments. The systematic review found very little evidence testing particular interventions for specific types of patients since very few studies reported the treatment outcomes for specific patient subpopulations. The systematic review focused on randomized controlled trials which can provide valid treatment estimates equally distributing patient characteristics and concomitant treatments among treatment groups. However, the review concluded that the results are applicable to the target population and much less to the subpopulations by age, gender, baseline OA severity, and response to pharmacological treatments. Therefore, future research is needed for hypotheses by garnering expert opinion about which patient subgroups may respond differently to specific therapies.

Once hypotheses are generated, they should be tested using rigorous experimental design. Randomized controlled trials (RCTs) are the best approach. Conducting RCTs on specific patient subgroups is feasible yet the systematic review found very weak evidence of treatment effects in patient subpopulations. The review concluded that the evidence from individual RCTs did not support robust conclusions about differences in PT effects by patient age, gender, baseline severity of knee OA and multijoint OA, or responses to prior PT and drug treatments. However, a more valuable study design would be a large scale RCT with representative samples of sufficient size (as determined by the appropriate power calculations) from various subgroups of patients identified a priori. In designing these trials, another important concern lies in defining the PT treatments. Treatment definition for the intervention and comparator should be sufficient to explain specific activities used in each PT session or a protocol that explains the sequence of therapies. Treatments compared should capture the full range of PT treatments that would be used in practice. Fidelity checks may be necessary to monitor compliance with protocols. Attention should be paid to other concomitant treats, especially anti-inflammatory drugs and analgesics.

Second Topical Research Need

  • How do the duration, intensity, and frequency of examined interventions affect sustained changes in patient-centered outcomes?

The CER found limited evidence to evaluate intervention characteristics. The duration of examined PT interventions was not consistently associated with better intermediate or patient-centered outcomes. Evidence regarding the association between the dose/intensity/frequency of examined interventions and outcomes was not available for the majority of comparisons. The effects of the treatments that significantly improved outcomes, including exercise (aerobic, aquatic, and strengthening) and ultrasound did not differ at shorter versus longer followup times. Moreover, electrical stimulation worsened pain at longer followup. Study risk of bias and heterogeneity in populations and treatments including concomitant treatments hampered strength of evidence to low or moderate in most cases. Stakeholder discussions confirmed that a better understanding of different intervention characteristics (especially dosage) and how they influence effectiveness would better inform decisionmaking.

Research Design Considerations

Processes similar to those mentioned above could be used to identify specific intervention characteristics that contribute to effectiveness. Again, experimental designs are likely the best approach to testing hypothesized relationships, yet very few RCTs examine the role of treatment intensity and duration on patient centered outcomes. The review found no high quality observational studies or administrative databases analyses suggesting significant improvement in patient centered outcomes with longer and more intense PT interventions in adult with knee OA. Design considerations for these experimental studies are also similar to those of this first research need. The approach might be implemented with trials testing the standard evidence-based treatment, exercise therapy. The most valid way to then address this research need would be with RCTs; however it may prove difficult to mount studies of adequate size. In that case quasi-experimental designs may be necessary. Prospective cohort studies with large samples may be preferred to small RCTs, yet no well designed prospective cohort analyzed the association between PT intensity and duration on pain, function, or disability in older adults with knee OA. In either case, investigators should be careful to appropriately define the PT treatment and document the intensity, duration, and frequency. Special attention should be paid to adherence among study participants. Studies should be sufficiently powered to detect differences between groups as determined by appropriate power calculation. A major concern is in powering the study adequately to test the effects of combinations of treatment variations. The cohort studies should pay additional attention to identifying and adjusting results for potentially confounding variables.

Third Topical Research Need

  • What is the comparative effectiveness of comprehensive multimodal PT treatments on patient-centered outcomes when compared with exercise alone?

The two remaining research needs have more focused hypotheses. Few studies comparing multimodal treatments to exercise alone are available, yet this question is particularly important to informing clinical practice. Current guidelines recommend that PT be delivered with a combination of modalities. Published research has focused instead on the marginal effects of individual PT interventions. The systematic review concluded that the studies overall had low applicability to the actual practice of PT because available studies focused on single modalities of PT rather than the combinations typically used in practice. In addition, many of the interventions were physical agents/modalities (i.e., orthotics, ultrasound, taping, etc.). This also contradicts the recommended practice of PT, in which physical agents/modalities are infrequently used in isolation, but rather combined with other more “active” interventions (i.e., exercises). The review found that few studies of combined PT modalities demonstrated no statistically significant benefit on the outcomes when compared with exercise alone.

Research Design Considerations

Given the specific hypothesis of this research need, an RCT is likely the best approach. Randomization eliminates concerns about inherent differences between the groups assigned to each intervention being responsible for differences in outcomes. An RCT will be resource intensive, requiring a large sample size because the marginal difference between the two active treatment arms is likely to be low and subgroups are particularly relevant in this question. Investigators should pay careful attention to defining the multimodal programs; only a limited number of combinations will be feasible.

Fourth Topical Research Need

  • In individuals who proceed to joint replacement surgery, do patients who underwent PT treatments prior to surgery fare better postoperatively?

The CER focused on community-dwelling adults with knee pain secondary to OA. While many patients with knee OA eventually undergo joint replacement surgery, postsurgical outcomes were beyond the scope of this review. Stakeholders brought up this question as a research gap. Benefits of pre-surgical PT treatments on patient outcomes after surgery remain unclear and this information would have important clinical implications.

Research Design Considerations

In first addressing this research need, investigators should examine previous literature to determine if studies that address this question are available. Once hypotheses are generated, more rigorous studies can be conducted. Due to the potentially long-term nature of this outcome and the difficulty in identifying group members a priori, an RCT or other prospective design may not be feasible. Therefore, testing the hypothesis that individuals receiving PT treatment fare better after knee replacement surgery might best be approached with case control studies. Large sample sizes and the identification, measurement, and appropriate adjustment for confounding variables with multivariate analysis would strengthen the internal validity of these studies. However, limited causal inference will be a limitation.


This FRNs project refined and prioritized research needs relevant to the KQs addressed in the draft CER, Physical Therapy Interventions for Knee Pain Secondary to Osteoarthritis.1 We conducted a deliberative process to refine and expand research gaps identified in the CER through conversations with stakeholders with various perspectives of expertise on the topic. This process identified 7 methodological and 11 topical research questions thought to address identified evidence gaps. We then had stakeholders rank research questions. The highly ranked questions were deemed research needs. Stakeholders prioritized five methodological and four topical research needs.

Addressing methodological research needs will enhance the utility and comparability of future studies of PT treatments for knee OA. A common set of patient-centered and intermediate outcomes—with guidance on interpreting changes in outcomes scale scores—will provide researchers with concrete approaches to collecting outcomes data and determining effectiveness. Guidance on how PT interventions should be defined in research studies and variables to report in studies as determined by a multidisciplinary panel will, when utilized, enhance the quality of research on the topic.

Topical research needs demonstrate the importance of understanding that all PT interventions may not be ideal for all patients. Advancement in the field needs to address which treatments are effective for which patients. Additionally, a better understanding of how PT treatments are defined is essential to understanding their effectiveness. Complete interventions definitions will enhance the internal validity of studies and allow replicability of effective treatments. Testing specific hypotheses will fill specific evidence gaps identified and prioritized by our stakeholders.

For the specific research design selected to study a particular population and intervention, future studies on PT interventions should pay close attention to reducing bias as much as possible for that particular design and conducting studies with adequate power to test hypothesized relationships, including among subgroups.

While a strength of this project is the multidisciplinary perspective brought by broad stakeholder participation, our inability to collect a representative perspective from a larger sample of stakeholders is also a limitation. The stakeholders participating in this project represented various perspectives on knee OA and PT. However, the prioritized research needs reflect the opinions of these stakeholders and may not be generalizable to the population of stakeholders on this topic.


Addressing research needs identified in this FRN project will help to create a broader and stronger evidence base in which clinical decisions can be made. Future research addressing specific research questions is likely to establish a preliminary research agenda on this topic:

  • Which patient-centered outcome measurement instruments should be used consistently by all relevant disciplines (e.g., PT, rheumatology, orthopedics)?
  • Which intermediate outcome measurement instruments should be used consistently by all relevant disciplines (e.g., PT, rheumatology, orthopedics)?
  • Should effectiveness research on PT treatments use MCID?
  • What confounding and effect modifying variables (e.g., OA severity, obesity, comorbidities, and concomitant therapies-including anti-inflammatory and analgesic medication) should be measured and reported in effectiveness research?
  • What minimum set of treatment factors (site, treatment components, frequency, duration, intensity, timing) should be reported consistently by all relevant disciplines (e.g., PT, rheumatology, orthopedics)?
  • Which PT treatments work for which patients?
  • How do the duration, intensity, and frequency of examined interventions affect sustained changes in patient-centered outcomes?
  • What is the comparative effectiveness of comprehensive multimodal PT treatments on patient-centered outcomes when compared with exercise alone?
  • In individuals who proceed to joint replacement surgery, do patients who underwent PT treatments prior to surgery fare better postoperatively?


Shamliyan TA, Wang S-Y, Olson-Kellogg B, Kane RL. Physical Therapy Interventions for Knee Pain Secondary to Osteoarthritis. Comparative Effectiveness Review No. 77. Rockville, MD: Agency for Healthcare Research and Quality; Nov, 2012. (Prepared by the Minnesota Evidence-based Practice Center under Contract No. 290-2007-10064-I.) AHRQ Publication No. 12(13)-EHC115-EF. www​.effectivehealthcare​.ahrq.gov/reports/final.cfm.
Pisters MF, Veenhof C, van Meeteren NLU, et al. Long-term effectiveness of exercise therapy in patients with osteoarthritis of the hip or knee: a systematic review. Arthritis & Rheumatism. 2007 Oct 15;57(7):1245–53. [PubMed: 17907210]
Riddle DL, Stratford PW, Singh JA, et al. Variation in outcome measures in hip and knee arthroplasty clinical trials: a proposed approach to achieving consensus. Journal of Rheumatology. 2009 Sep;36(9):2050–6. [PMC free article: PMC3882025] [PubMed: 19738212]