Exercise interventions and patient beliefs for people with chronic hip and knee pain: a mixed methods review

  • Protocol
  • Intervention



This is the protocol for a review and there is no abstract. The objectives are as follows:

Overarching objective

To improve our understanding of the complex inter-relationship between pain, psychosocial effects, physical function and exercise.

Specific aims and objectives:

To systematically review the evidence on the impact of physical exercise on patients' pain, physical and psychosocial functioning including:

  • identifying the most effective formats for delivering exercise advice;

  • explaining why some exercise interventions may be more effective than others;

  • recommending exercise formats and content by constructing a "toolbox" which describes the most effective exercise interventions for healthcare providers and patients to use.

These will be achieved by conducting:

  1. a synthesis of quantitative data on the benefits and harm of exercise interventions for improving pain, physical and psychosocial functioning;

  2. a synthesis of qualitative data on participant's experiences, opinions and preferences of physical exercise;

  3. a cross-studies synthesis of the quantitative and qualitative data to assess the extent to which existing evaluated interventions address the needs and concerns of people living with osteoarthritis


Description of the condition

Severe peripheral joint pain, often labelled as osteoarthritis (OA), is extremely prevalent worldwide (Bedson 2004; Woolf 2003), and a major cause of disability and healthcare expenditure (Gupta 2005; Leardini 2004; March 1997; Vos 2012). In the UK, nearly 20% of people over 50 years have severe disabling knee or hip pain (Jinks 2004; Peat 2001), also labelled as OA (Bedson 2004), which slowly worsens over time, compromising quality of life and independence (Dawson 2005). The economic burden of joint pain/OA is significant (Gupta 2005). Annually 15% of people over 50 years consult their general practitioners for knee pain (Jinks 2004). The societal cost is about 1% of gross national product (GNP) from 36 million lost working days costing £3.2 billion (NICE 2008). The personal suffering and psycho-socio-economic consequences of chronic joint pain will increase as people live longer, adopt sedentary lifestyles and obesity rises (Underwood 2004). By 2020, OA is projected to be the 4th leading cause of disability across the world (Woolf 2003).

Description of the intervention

Exercise is recommended to reduce joint pain and improve physical function (Fransen 2008; Fransen 2009; NICE 2008; Zhang 2008). In addition, successful completion of a challenging exercise programme highlights to people their capability, challenges inappropriate health beliefs, disrupts detrimental behaviour (fear-avoidance) and teaches people exercise is a safe, beneficial, active coping strategy they can use to improve self-efficacy (confidence in one’s ability to perform a specific health behaviour or task), self-reliance, helplessness and disability (Hurley 2010; Keefe 1996a; Penninx 2002). Unfortunately, as there is no summary of the evidence describing the reciprocity between pain, physical and psychosocial function and the utility of exercise on addressing these problems, the importance of these inter-relationships remains under-appreciated, and potential treatment options under-utilised.

Information and advice about the role of exercise in the management of joint pain form part of most self-management (Miles 2011; Newman 2004) and physiotherapy programmes (Walsh 2009). The aim is to effect behavioural change, i.e. encourage people to exercise regularly, but the most effective way to deliver exercise advice that will bring about this behavioural change and get people exercising regularly is unclear (Hurley 2009). Didactic programmes explain the benefits of exercise for joint pain management using verbal or written information may enlighten people, but they do not detail how to start exercising, what (not) to do, when, how or how much, and fail to convince people who have suffered many years of activity-related pain that exercise is harmless. Consequently, didactic programmes may have limited ability to improve health beliefs, self-confidence, self-efficacy, coping and affect behavioural change. Exercise remains a burdensome, time consuming, effortful concept that causes pain. Programmes that include a participatory exercise component may encourage regular exercise more effectively (Griffiths 2007). On these programmes participants gain first-hand experience of what exercises to do, how, that exercise is not harmful, how exercise can be used to reduce pain, and improves their physical function, health beliefs, anxiety and depression, and potentially their general quality of life, overall (Hurley 2007; Hurley 2010). Again, without a systematic review of the evidence on the effectiveness of exercise education delivery, the best way to bring about participation in regular exercise is unclear, wasting time, effort and resources and potentially missing effective treatment options.

How the intervention might work

Conceptual framework

Relationship between chronic pain, physical and psychosocial function

Chronic joint pain and disability are the commonest symptoms of OA, and attract most attention. Because OA and joint pain is often regarded as a benign, untreatable, inevitable consequence of ageing the psychosocial sequelae (anxiety, depression, health beliefs, behaviours, quality of life, participation and dependency) are often underestimated by healthcare professionals and lay people. However, this overlooks the complex, reciprocal relationship between pain, physical and psychosocial functioning - each affect and are affected by the others (Hurley 2003; Figure 1). For example, chronic joint pain is bewildering and distressing because it has no obvious cause, increases insidiously and is unaccountably episodic. People's reactions to pain are highly variable and influenced by the beliefs, meanings and explanations they attach to it.

Figure 1.

Complex reciprocal inter-relationship between pain, physical and psychcosocial function and exercise intervention

Relationship between health beliefs and psycho-social outcomes

Beliefs about the cause, prognosis and effectiveness of treatment are key determinants of illness behaviour and response to treatment (Main 2002; Turk 1996). People commonly believe joint pain is the inevitable, incurable consequence of ageing, caused or exacerbated by activity, that evoke feelings of helplessness, anxiety, depression and "fear-avoidance" behaviour (Figure 2), when people avoid physical activity for fear of causing additional pain and damage (Keefe 1996a). However, avoiding activity results in greater muscle weakness, joint instability and stiffness, exacerbated pain, disability and dependency (Dekker 1992). Challenging these erroneous health beliefs is vital for successful pain management. Inappropriate health beliefs and behaviours can be altered by positive experiences that show people how active coping strategies such as exercise can reduce pain, improve physical functioning, self-efficacy, anxiety, helplessness, catastrophising and depression (Keefe 1996b; Main 2002; Turk 1996). Appreciating the complex inter-relationship between clinical symptoms and psychosocial effects of joint pain could provide additional strategies for better joint pain management.

Figure 2.

Effect of erroneous health beliefs

Better appreciation of the complex reciprocal relationship between pain, psychosocial effects and physical functioning would help us understand better the consequences of joint pain, identify the most effective ways of teaching the value of exercise, and more efficient models of care for people suffering chronic joint pain. This is best achieved by a systematic review of the relevant literature to establish what interventions are most effective and quantify the size of the treatment effect produced. However, the complex reciprocity between joint pain, psychosocial impact, physical functioning and exercise will be influenced by many factors that are difficult to measure, depend on nebulous, labile, personal beliefs, experiences, emotions, preferences and prejudices. A systematic review asking questions on effectiveness and synthesising outcome evaluations only would miss important facets and cannot accommodate information from qualitative studies better placed to assess pain, the psychosocial effects of pain and the benefits of exercise. These are best captured using recently developed methods that synthesise quantitative (systematic reviews) with qualitative studies of people's views and experiences (Lorenc 2008; Oliver 2008; Rees 2006; Thomas 2004). Appreciating the views, beliefs, experiences and preferences of target populations for an intervention provides greater insight into how an intervention achieves its effects, why it may not be as effective as anticipated and expose gaps in our understanding. This enables us to adapt existing, or develop new, healthcare interventions that best address people’s needs (Harden 2004; Oliver 2008; Rees 2006).

Why it is important to do this review

This review focuses on exercise-based rehabilitation programmes, defined as programmes that have an active participatory exercise component (for management of OA and the psychosocial variables affected by the condition). Establishing the effect of exercise-based rehabilitation programmes on the psychosocial impact of chronic joint pain will increase our understanding about how and why these interventions are effective and identify the effective elements of exercise programmes.

To meet the aims the review will answer the following questions:

  1. what are the effects of exercise-based rehabilitation programmes on physical and psychosocial functioning for patients with chronic knee and/or hip pain?

  2. what are patient's experiences, opinions and preferences of exercise-based rehabilitation programmes and the advice they receive about exercise?

  3. what implications can be drawn, from the qualitative synthesis of people's views, to inform the appropriateness and acceptability of exercise-based rehabilitation programmes for people living with osteoarthritis?


Overarching objective

To improve our understanding of the complex inter-relationship between pain, psychosocial effects, physical function and exercise.

Specific aims and objectives:

To systematically review the evidence on the impact of physical exercise on patients' pain, physical and psychosocial functioning including:

  • identifying the most effective formats for delivering exercise advice;

  • explaining why some exercise interventions may be more effective than others;

  • recommending exercise formats and content by constructing a "toolbox" which describes the most effective exercise interventions for healthcare providers and patients to use.

These will be achieved by conducting:

  1. a synthesis of quantitative data on the benefits and harm of exercise interventions for improving pain, physical and psychosocial functioning;

  2. a synthesis of qualitative data on participant's experiences, opinions and preferences of physical exercise;

  3. a cross-studies synthesis of the quantitative and qualitative data to assess the extent to which existing evaluated interventions address the needs and concerns of people living with osteoarthritis


Criteria for considering studies for this review

Types of studies

Quantitative clinical trials or qualitative studies reporting the views and opinions of people who have participated on exercised-based rehabilitation programmes of studies published after 1985.

Quantitative clinical trials of exercise programmes must:

  • use random allocation or;

  • use a comparison group design;

  • report pre-and post-intervention data.

Trials meeting the above criteria which also have a qualitative study nested within them will also be considered for inclusion in the review.

Qualitative studies reporting the views and opinions of participants of exercise-based programmes must:

  • report methods of data collection and data analysis;

  • report people's perspectives, beliefs, feelings, understanding, experiences or behaviour about exercise or advice on exercise;

  • present views as data (e.g. direct quotes from participants or description of findings);

While no limits will be placed on location or language, we will only include studies that were published after 1985. We set this limit because before then there was a paucity of well-designed and well-reported randomised controlled trials (RCTs) of community-based interventions or studies of people's views specifically relevant to this topic.

Types of participants

Males or females aged 45 years or older, with a medical diagnosis of osteoarthritis (as defined by the study) or self-reporting chronic hip and/or knee pain (defined as more than 6 months duration).

Types of interventions

  • Exercise-based rehabilitation programmes

    • Any type of land based or aquatic exercise programme aiming to improve OA symptoms delivered in hospitals or the community. Programmes could vary in their content (e.g. range of motion, aerobics, tai chi) and their delivery mode, e.g. length, frequency or intensity

  • Control group

    • No treatment or waiting list group

    • Any non-exercise intervention (e.g. medication, lifestyle/diet changes, information on OA)

Types of outcome measures

Major outcomes
  1. Pain

  2. Physical function

  3. Self-efficacy

  4. Depression

  5. Anxiety

  6. Quality of life

  7. Adverse effects of exercise

These outcomes will be included in the 'Summary of findings' table. For quantitative synthesis, randomised trials will need to have measured at least one of the following outcomes:

  • pain;

  • physical function;

  • self-efficacy;

  • depression.

Minor outcomes
  • Anxiety;

  • Quality of life;

  • Adverse effects of exercise.

For the qualitative synthesis studies will have reported people's opinions and experiences of exercise, e.g. their views and beliefs about the utility of exercise in the management of chronic pain, barriers to adherence to exercise advice. We will use these in the cross-synthesis to attempt to explain some of the findings of the quantitative synthesis.

Search methods for identification of studies

Electronic searches

Our electronic database searching will involve comprehensives strings of thesaurus and free-text terms for the key features of research relevant for this review:

  • condition: chronic knee and hip pain: e.g. 'osteoarthritis/chronic joint pain'

  • intervention: 'exercise', physical activity, aerobic, walking, tai chi, physiotherapy

We will combine the two strings to identify reports that contain terms for all two features (population AND topic). In order to make the search results more precise, we will exclude terms that have surgery or arthroplasty in the title.

We will apply these search strategies for a comprehensive search of the following clinical, public health, psychology, social care databases from 1966 to present:

  • Allied and Complementary Medicine (AMED);

  • Applied Social Sciences Index and Abstracts (ASSIA);

  • BiblioMap (EPPI-Centre database of health promotion research);

  • British Nursing Index (BNI);

  • Cochrane Database of Systematic Reviews (CDSR);

  • Cochrane Central Register of Controlled Trials (CENTRAL);

  • Cumulative Index to Nursing and Allied Health Literature (CINAHL);

  • Database of Abstracts of Reviews of Effects (DARE);

  • Database of Promoting Health Effectiveness Reviews (DoPHER);


  • Health Management Information Consortium (HMIC);

  • Health Technology Assessment (HTA);

  • International Bibliography of the Social Sciences (IBSS);


  • NHS Economic Evaluation Database (NHS EED);

  • Physiotherapy Evidence Database (PEDro) (systematic reviews and appraised randomised trials in physiotherapy);

  • PsycINFO;

  • PubMed;

  • Sociological Abstracts (SOCABS);

  • Social policy and Practice (SPP);

  • Social Services Abstracts;

  • Trials Register of Promoting Health Interventions (TRoPHI);

  • Web of Knowledge.

Searching other resources

We will hand search the following resources:

  • Arthritis Australia

  • Arthritis Foundation (USA)

  • Arthritis New Zealand

  • Arthritis Society (Canada)

  • Cochrane Collaboration Behavioral Medicine Field Register archive

  • British library Integrated Catalogue (BLIC)

  • British Index to Theses

  • Dart Europe

  • British Library Electronic Theses Online Service (ETHOS)

  • Economic and Social Research Council (ESRC)

  • Evidence Database on Ageing Care

  • Fade: The North West Grey Literature Service (UK)

  • Google

  • Google Scholar

  • New York Academy of Medicine Library Catalogue

  • Arthritis Care UK

  • Healthtalkonline.org

  • NHS Evidence

  • NHS Health Scotland Library

  • Opengrey.eu

  • Rehabdata

  • Social Science Research Network

  • Alerts from Rehab database

  • Northern Ireland Qualitative Archive on Ageism

  • Arthitis Research Centre of Canada

We will also check references of included studies by

  • Checking where included studies have been cited, using Google Scholar;

  • Checking references of selected reviews in the topic area that the research team were aware of from a systematic review of reviews on adult social care; outcomes concurrently being undertaken at the EPPI-Centre;

  • Asking key experts/authors of included studies.

An example of the thesaurus and free-text strings applied to MEDLINE are outlined in Appendix 1.

We will engage with experts from the research, advocacy and policy sectors in the field of osteoarthritis rehabilitation. They will inform key stages in the review including: advising on the scope, informing the search strategy, reviewing the final report and disseminating the research findings.

Data collection and analysis

Selection of studies

We will apply inclusion and exclusion criteria successively to titles, abstracts and full reports. We will obtain the full-text reports for studies that appear to meet the criteria or where we have insufficient information to decide. We will enter information of these full-text reports into a separate database, and reapply the inclusion and exclusion criteria and studies that meet these criteria will be included in the review.

All review authors involved in study screening (KD, HH, MH, NW) will take part in a moderation exercise where results are discussed to ensure consistency in applying the review exclusion criteria. For the initial title and abstract screening, we will carry out "double screening" on 200 papers before continuing with independent screening. For the second round of screening on full reports, the criteria will be independently applied by a second screener to 10% of the reports. A 90% agreement rate will be required before proceeding to independent screening. The remaining sample of potential studies will be screened independently by review authors (KD, HH, MH, NW). Where a review author (for example HH) is unable to reach a decision, consensus will be reached through discussion with a second review author (for example KD), and when required a third review author (MH or NW).

Data extraction and management

To manage the review we will use the 'EPPI-Reviewer' software (Thomas 2010). This programme uses a single web location to house the documents, record details of each study included in the review and monitor progress of the review, including screening decisions, descriptive and evaluative codes, text and the data used and produced when reviewing studies in-depth.

Four review authors (KD, HH, MH, NW) will extract descriptive details from full reports using a specially designed data collection form, which will be piloted to ensure relevant detail is consistently collected from the studies. If one of the review authors is an author or a co-author of one of the included studies, that person will not be involved in any decisions regarding data extraction from the study to avoid bias.

We will extract the following information from all studies:

  • aims and focus of the research;

  • study design;

  • details about the intervention including:

    • format – written, didactic, non-participatory/participatory, lay/professional led, individual/group therapy, etc;

    • content – type, frequency, intensity, etc;

    • setting – hospital / outpatient / community / home-based, etc.

  • details about the study populations and settings as per the PROGRESS-Plus framework (Kavanagh 2008):

    • broad social determinants of health and well-being (e.g. ethnicity, occupation, gender, education, socioeconomic status;

    • characteristics that impinge on health and well-being by attracting discrimination, such as age;

    • other contextual features pertinent to the experiences of living with knee and hip pain, such as housing.

Quantitative outcome data

For outcome measurements, whenever possible, we will extract raw scores. Where trials report pain or function using more than one outcome measure tool, we will extract data according to the following hierarchy: Western Ontario McMaster Universities Osteoarthritis Index (WOMAC), Arthritis Impact Measurement Scale (AIMS), Visual Analogue Scale (VAS), other. Similarly, when articles present psycho-social outcomes using different scales, preference will be given to outcomes reported using a standardised measurement tool.

Qualitative "views" data

For studies which include "views" data, whenever possible, we will extract participant's quotes first, followed by and distinguished from authors description and analysis of participants views. We will follow the conceptual framework in order to support the identification of factors potentially impacting on participation in and experiences of exercise (Figure 2). Further themes and capturing of data will occur through the coding of data as described in the synthesis section. We will endeavour to record and discuss our reflections on the data extracted from individual papers, generating cumulative knowledge about the potential links and connections between the data and further analytical themes.

Assessment of risk of bias in included studies

In pairs, four review authors (HH, KD, MH, NW) will independently assess the risk of bias for all included studies using the Cochrane Collaboration's tool for assessing risk of bias addressing the following criteria (Higgins 2011a): sequence generation, allocation concealment, blinding, incomplete outcome data, selective reporting, and any 'other concerns' that may be identified. We will judge each study as 'low risk' (risk of bias avoided and/or "addressed"), 'high risk' (risk of bias not adequately addressed) or 'unclear risk' (uncertain risk) of bias. Any disagreements will be resolved by discussion and if necessary by consulting a fifth author (SO).

Sequence generation:

'Low risk' of bias if a random component in the sequence generation process is described (for example, referring to a random number table). 'High risk' of bias when the authors describe a non-random component in the sequence generation process (e.g. sequence is generated by hospital or clinic record number). 'Unclear risk' of bias if not specified in the paper

Allocation sequence concealment:

'Low risk' of bias if the personnel allocated to enrol participants could not foresee assignment by using an appropriate method used to conceal allocation (e.g. central allocation including telephone, web-based and pharmacy-controlled randomisation, sequentially numbered, opaque, sealed envelopes. 'High risk' of bias if concealment allocation was not guaranteed. 'Unclear risk' of bias if not specified in the paper


Blinding of participants and personnel

It is not possible to truly blind the intervention provider or recipient in RCTs evaluating exercise programs versus non-exercise programs. However, we will describe for each included study the methods used, if any, to blind study participants and personnel from knowledge of which intervention a participant received.

Blinding of outcome assessment

'Low risk' of bias if the authors state explicitly that the primary outcome variables were assessed blindly. 'High risk' of bias if the outcomes were not assessed blindly and this likely to affect results. 'Unclear risk' of bias" if not specified in the paper.

Completeness of outcome data:

'Low risk' of bias if there is no missing data or missing outcome data is balanced in numbers across intervention groups, with similar reasons for that missing data and unlikely to alter the results in the study). 'High risk' of bias if missing outcome data was likely to bias the results. 'Unclear risk' of bias if not specified in the paper.

Reporting bias:

'Low risk' of bias if there is no evidence that outcomes were selectively reported (for example, all relevant outcomes in the methods section are reported in the results section). 'High risk' of bias if some important outcomes are subsequently omitted from the results. 'Unclear risk' of bias if not specified in the paper.

'Other' bias:

'Low risk' of bias if there is no evidence of other risk of biases, and 'High risk' of bias if there are concerns.

Assessment of rigour in qualitative studies

We will assess the quality of "views studies" using a tool developed at the EPPI-Centre (Harden 2004). This tool addresses methodological rigour of studies and considers whether the findings are grounded in the data and reflect people's views. The development of the criteria was originally informed by those engaged in ensuring increased transparency and explicit methods for assessing the quality of qualitative research (Boulton 1996; Cobb 1987; Mays 1995; Popay 1998), and is in accordance with the Cochrane Qualitative Research Methods Group guidance on adopting a quality appraisal framework (Hannes 2011).

We will assess each study according to the extent to which they provide:

  • clearly stated aims and objectives

  • clear description of methodology, including systematic data collection methods

  • clear description of the sample

  • clear description of context, including detail on factors important for interpreting the results

  • clear description of the data analysis to establish reliability and validity

Two authors (KD, HH) will make judgements on the 'reliability' (e.g. how rigorously the study was conducted) and 'usefulness' of the study findings (e.g. how rich and complex the description and analysis of people's views is in the study's results and whether or not the data elucidated people's own explanations and views about exercise. Any disagreements will be resolved by discussion and if necessary by consulting a third review author (SO).


"High" (low risk of bias/error) on reliability if a thorough attempt has been taken to increase rigour in the sampling, the data collected and in the analysis of the data; "medium" if some steps were taken and "low" if minimal steps were taken or it was unclear what attempts study authors have made to avoid methodological bias and error in conducting the study.


"High" on usefulness if the findings are judged to be well grounded/supported by the data, contribute either depth or breadth of findings (in relation to their ability to answer the review question) and privilege the perspectives and experiences of people living with osteoarthritis. "Medium" usefulness will be assigned to those studies which meet the same criteria as high quality studies, but are only fairly well grounded in the data. "Low" usefulness will be assigned to studies which are found to be "limited" on any of the above criteria.

As one of the aims of the review is to synthesis patient's experiences and preferences in relation to exercise to better understand the factors which might contribute to the success of exercise-rehabilitations programs, we will not exclude studies failing to meet a minimum quality threshold (i.e., those scoring low for both reliability and usefulness). Instead, we will use the quality assessment to assess the contribution of each study to the development of explanations and relationships.

Measures of treatment effect

Continuous data

We will calculate the mean differences (MDs) when the continuous data were measured by the same scale or unit. When similar outcomes were measured by different scale or units, we will use standardised mean differences (SMDs). We will report 95% confidence intervals (CIs).

We will present any highly skewed continuous data in a table.

Dichotomous data

For dichotomous (binary) data, we will calculate the risk ratios (RRs) with 95% CIs, or convert odds ratios (OR) to SMDs, using the Cox-Snell formula, and where appropriate we will combine results from different trials.

Unit of analysis issues

We will identify the level at which randomisation occurred (e.g. cluster-randomised trials; cross-over trials, and repeated measures) at the data extraction stage and address the following issues if they arise:

  • cluster-randomised trials. With the help of ICC estimates, design effect will be calculated and the variance inflated accordingly, for use in the review. If ICC estimates are not available, then these trials will be analysed using imputed ICC estimates from similar trials and a sensitivity analysis will be conducted for the purpose. Authors may have reported some form of multilevel model to separate causal treatment effect from organisation clustering. If not we will seek more detail from the authors to enable clustering to be accounted for in the meta-analysis. If such information cannot be obtained, and we are concerned that clustering may cause serious bias, this will be documented and that study may be omitted from the meta-analysis, as for any other source of bias.

  • crossover trials. We do not anticipate cross-over trials.

  • multiple interventions per patient. We will analyse studies that compared the effect of two or more types of land-based or aquatic exercise-based rehabilitation programmes with a control condition.

  • multiple follow-up. For trials that report outcomes at multiple time points within these time frames, we will select the longest follow-up.

Dealing with missing data

If data is missing in trials, we will record the amount of missing data, reasons and pattern of missing data (missing completely at random, missing at random, missing not at random) and how the missing data was handled (ignored, last observation carried forward, statistical modelling, etc) (Carpenter 2008). We will perform sensitivity analyses to assess the impact of the missing data on the findings of the study and the implications to the findings of the review will be considered in the discussion.

An equivalent of missing data, as it is understood for quantitative studies, will not be applied to qualitative studies of people's views. However, we will report the number of studies contributing to each theme/sub-theme, to asses the extent to which certain studies may be contributing "more" or "less" to that theme/sub-theme.

Assessment of heterogeneity

We will compute the I2 statistic to assess the percentage of total variation caused by the heterogeneity of trials (Higgins 2003). We will assess statistical heterogeneity across studies by visual inspection of the forest plot and using the Chi2 test with a significance level of P < 0.10, and the I2 test and tentatively assign I2 statistic value of < 49% as "low" heterogeneity, 50 to 74% "moderate" and 75 to 100% "high" heterogeneity ((Higgins 2003). If there is moderate heterogeneity (Chi2 P < 0.10 and I2 value > 50 to 74%), we will use a random-effects model. If there is no clinical and no important statistical heterogeneity (I2 < 50%), we will combine results using a fixed-effect model. We consider the potential cause of heterogeneity by conducting subgroup and sensitivity analyses as described below.

Assessing differences in the "views" studies with regard to the study setting and sample (e.g. gender, age, type and severity of OA/chronic pain) will be noted and inform the qualitative synthesis, when reported and deemed appropriate in potentially explaining variation in the studies findings.

Assessment of reporting biases

We will use funnel plots (effect size versus standard error) to assess publication bias if sufficient trials are found. Where possible, we will compare the outcomes and comparisons reported in the papers against trial protocols to detect unreported results that may indicate reporting bias. These will contribute to the overall assessment of potential for bias.

We will not apply the assessment of reporting biases to qualitative studies.

Data synthesis

The methods or approaches to synthesis used will be driven by the research question, the types of studies/data that are included in the review, the detail and quality of reporting in these studies, and their heterogeneity. The synthesis of study findings will be informed and shaped by the conceptual framework and the type of interventions identified. If a wide variety of approaches and patient populations are found in the studies identified, we will use the random-effects model for meta-analyses.

Quantitative synthesis

Where it is appropriate, we will use standard methods for statistical meta-analysis to synthesise data, using the Cochrane Collaboration's statistical software, Review Manager 2013. We will use a fixed-effect meta-analysis for combining data where it is reasonable to assume that studies are estimating the same underlying treatment effect: i.e. where trials are examining the same intervention, and the trials' populations and methods are judged to be sufficiently similar. If substantial statistical heterogeneity is detected, we will use a random-effects model. If we perform random-effects analyses, the results will be presented as the average treatment effect with 95% confidence intervals. Where meta-analysis is not appropriate, we will present a narrative description of study results in line with the final conceptual framework, as with previous EPPI-Centre reviews.

Qualitative synthesis

We will synthesise studies of patients' views using framework analysis used in previous EPPI-Centre reviews (Lorenc 2008; Oliver 2008). A framework synthesis accommodates a range of different types of studies and can be conducted relatively quickly by a team of review authors. We shall read the abstracts from search outputs, then the full texts as they become available, to confirm, extend or refute as key issues and recurrent themes the concepts in our initial framework as illustrated in Figure 1. We will code the included studies according to the framework, revisiting studies as the framework is refined. The codes will then act as an index to navigate the data and allow the literature to be sub-divided into sections, albeit overlapping, that are more manageable for in-depth analysis. If necessary, we shall work through a series of frameworks that will accommodate more data with increasing coherence. Subsequently, each element of the framework will be interrogated in turn, tabulating the data under key themes in order to present distilled summaries. In the last stage, we will draw together what can be learnt from the tables and summaries and finding associations between themes and providing explanations for those findings across the included studies to illuminate patients' responses to aspects of arthritis and approaches to self management. This approach will provide a clear path from the original research data, to individual study author's descriptions and analyses to the findings of the qualitative review synthesis. This will then be used to inform the final cross-study synthesis.

Cross study synthesis

Cross study synthesis of clinical trials and "views studies" comprises a matrix that juxtaposes the components of the (in)effective interventions that correspond with patients' views on exercise and psychosocial health to expose (in)effective elements of exercise intervention and research gaps (Candy 2011). The cross-synthesis will be guided by questions such as:

  • which components of (in)effective interventions correspond with views expressed by patients?

  • does this match suggest why or how the intervention (does not) work?

  • what components appear in effective interventions but not in ineffective interventions?

  • does the "views" synthesis suggest these components are significant from a patient perspective?

  • does addressing the psychosocial effects of joint pain improve pain and physical functioning?

Summary of evidence

We will prepare summary tables of the findings for the quantitative and qualitative evidence synthesis. For trials, we shall use the methods and recommendations described in the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2011). We will use the GRADEpro software to produce a 'Summary of findings' table (GRADEpro 2008). Similarly, we will create "Summary of qualitative evidence" tables to summarise the key findings and be informed by the assessment of rigour, detailing the extent to which the findings are trustworthy, based on their reliability and usefulness, in answering the review question.

Subgroup analysis and investigation of heterogeneity

We will investigate whether sub-populations respond differently to an exercise-based rehabilitation programme by comparing the responses of different sub-groups to the exercise programme. Potentially participants with chronic hip pain only respond differently to exercise-based rehabilitation programmes than people with knee pain only and both respond differently to people with hip and knee pain. Other a priori planned sub-group analyses include age, gender and severity of symptoms as defined by the studies included in the review.

For trials, we will conduct a statistical test for heterogeneity across subgroup results and compute an I2 statistic. We will use the random-effects models to analyse the variation in the mean effects in the different subgroups using meta-regression techniques that reduce false-positive results when comparing subgroups in a fixed-effect model (Higgins 2011b). We will consider any post-hoc subgroups analyses as exploratory analyses. We will not apply a qualitative equivalent.

Sensitivity analysis

If aspects of a trial (e.g. atypical intervention, methodology, missing information) appears to unduly influence the findings of the review, we will conduct sensitivity analyses to assess the influence of that study and reported in a summary table). For qualitative studies we will consider whether certain individual and contextual factors can explain variation in the type of views identified.


The search protocol was designed, refined and conducted by Ms Claire Stansfield, Information Officer, EPPI-Centre, Social Science Research Unit, Institute of Education, London.


Appendix 1. Example of search strategy


Controlled vocabulary:




Osteoarthritis, Hip

Osteoarthritis, Knee



Free text, title and abstract:






"degenerative joint"


Terms used in conjunction – both columns ANDed:

Controlled vocabulary

· Musculoskeletal Diseases

· Chronic Pain

· Pain

· Joint Diseases

OR freetext in title and abstract

· "joint pain"

· "chronic joint"

· arthriti*

· "Chronic pain"

Controlled vocabulary

· Knee Joint

· hip joint

· knee

· hip

OR freetext in title and abstract

· Knee

· Knees

· Hip

· hips

Exclusion of surgery terms.

Exclude in title:

· Arthroplasty

· Surgery

· Surgical

· "hip replacement"

· "knee replacement"

· osteotomy

· arthroscopic

· postoperative

· "post operative"



Controlled vocabulary

  • Exercise therapy

  • Exercise

  • Muscle Stretching Exercises

  • Physiotherapy

  • Rehabilitation

  • "Tai Ji"

  • Walking

  • Yoga


Free text – title and abstract

Aerobic OR Aerobics OR aquarobics OR balneotherapy OR classes OR exercise OR exercises OR hydrotherapy OR "muscle strength" OR "muscle strengthening" OR "Muscle training" OR "nonpharmacological" OR "non-pharmacological" OR "nonsurgical" OR "non-surgical" OR "physical activity" OR "physical therapy" OR "physical training" OR "pool therapy" OR physiotherap* OR pilates OR "resistance training" OR "strength training" OR "T ai Chi" OR "Tai Chi" OR "taichi" OR Walk OR walking OR walks OR yoga OR manipulative therap* OR stretch OR stretches OR stretching

Example of PubMed search

Where mh = MeSH terms, and tiab = title and abstract terms

#25Search #23 AND #24
#24Search ("1985"[Date - Publication] : "3000"[Date - Publication])
#23Search #19 NOT #22
#22Search #21 NOT #20
#21Search animals[mh]
#20Search humans[mh] AND animals[mh]
#19Search #18 AND #14
#18Search #15 OR #16 OR #17
#17Search "Exercise therapy"[mh] OR Exercise[mh] OR "Muscle Stretching Exercises"[mh] OR Rehabilitation[mh:noexp] OR "Tai Ji"[mh] OR Walking[mh] OR Yoga[mh]
#16Search (stretch[tiab] OR stretches[tiab] OR stretching[tiab]) AND (muscle[tiab] OR muscles[tiab] OR physical[tiab] OR physically[tiab])
#15Search Aerobic[tiab] OR aerobics[tiab] OR aquarobics[tiab] OR balneotherapy[tiab] OR classes[tiab] OR exercise[tiab] OR exercises[tiab] OR hydrotherapy[tiab] OR "muscle strength"[tiab] OR "muscle strengthening"[tiab] OR "Muscle training"[tiab] OR "nonpharmacological"[tiab] OR "non-pharmacological"[tiab] OR "nonsurgical"[tiab] OR "non-surgical"[tiab] OR "physical activity"[tiab] OR "physical activities"[tiab] OR "physical therapy"[tiab] OR "physical therapies"[tiab] OR "physical training"[tiab] OR "pool therapy"[tiab] OR physiotherap*[tiab] OR pilates[tiab] OR "resistance training"[tiab] OR "strength training"[tiab] OR "T ai Chi"[tiab] OR "Tai Chi"[tiab] OR "taichi"[tiab] OR Walk[tiab] OR walking[tiab] OR walks[tiab] OR yoga[tiab] OR "manipulative therapy"[tiab] OR "manipulative therapies"[tiab] OR "water therapy"[tiab] OR "water therapies"[tiab]
#14Search #13 NOT #12
#13Search #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #11
#12Search arthroplasty[ti] OR surgery[ti] OR surgical[ti] OR "hip replacement"[ti] OR "knee replacement"[ti] OR osteotomy[ti] OR arthroscopic[ti] OR postoperative[ti] OR "post operative"[ti]
#11Search #9 AND #10
#10Search Knee Joint[mh] OR hip joint[mh] OR knee[mh] OR hip[mh] OR knee[tiab] OR knees[tiab] OR hip[tiab] OR hips[tiab]
#9Search Musculoskeletal Diseases[mh:noexp] OR "joint pain"[tiab] OR "chronic joint"[tiab] OR arthriti*[tiab] OR "Chronic pain"[tiab] OR “Chronic Pain"[mh] OR Pain[mh:noexp] OR Joint Diseases[mh:noexp]
#7Search Osteoarthr*[tiab] OR Arthralgia*[tiab] OR Arthrosis[tiab] OR "degenerative joint"[tiab]
#6Search Arthralgia[mh:noexp] NOT Arthralgia/surgery[mh:noexp]
#5Search Osteoarthritis[mh:noexp] NOT Osteoarthritis/surgery[mh:noexp]
#4Search Arthritis[mh:noexp] NOT Arthritis/surgery[mh:noexp]
#3Search Coxarthros*[tiab] OR Gonarthros*[tiab]
#2Search Osteoarthritis, Hip[mh] NOT Osteoarthritis, Hip/surgery[mh]
#1Search Osteoarthritis, Knee[mh] NOT Osteoarthritis, Knee/surgery[mh]

Contributions of authors

Roles and responsibilities
Draft the protocolHurley, Dickson, Walsh, Oliver
Develop a search strategyStansfield, Dickson
Search for trials (usually 2 people)Hauari, Dickson
Obtain copies of trialsHauari, Dickson
Select which trials to include (2 + 1 arbiter)Dickson, Hauari, Hurley, Walsh
Extract data from trials (2 people)Dickson, Hauari, Walsh, Hurley
Enter data into RevManDickson, Hurley, Hauari
Carry out the analysisHauari, Dickson, Grant
Interpret the analysisHauari, Dickson, Grant, Hurley, Walsh, Oliver
Draft the final reviewDickson, Hauari, Hurley, Oliver, Walsh, Grant
Update the reviewHurley, Walsh

Declarations of interest

None of the authors have any conflicting interests to declare.

Sources of support

Internal sources

  • No sources of support supplied

External sources

  • Arthritis Research UK, UK.

    Educational Grant, Number 20163