This review focuses on face-to-face interventions that promote physical activity (PA). It is part of a suite of three complementary reviews that provide both an update and a progression for the previously completed Cochrane review assessing "Interventions for promoting physical activity" (Foster 2005a). The titles of the other reviews in this suite are:
- Face-to-face versus remote and web 2.0 interventions for promoting physical activity;
- Remote and web 2.0 interventions for promoting physical activity.
The evidence base for both face-to-face PA promotion and remote and web 2.0 interventions is rapidly growing and diverging. Consequently, we have divided this Cochrane update into two separate reviews that focus on comparisons of each of these delivery methods to true control groups. The third review allows a head-to-head comparison between these intervention approaches for promoting PA. In all of the reviews we also consider how the effectiveness of PA interventions is influenced by implementing the intervention in a group or individually. The paradigm through which we will approach these different methods of PA intervention delivery in this suite of reviews is summarised below (Figure 1).
|Figure 1. Delivery of PA interventions described according to interaction with implementer and other participants|
Description of the condition
The health benefits of adequate levels of PA have been well documented (WHO 2010a). Previous systematic reviews and meta-analyses of observational studies have demonstrated the role of PA in the prevention and treatment of coronary heart disease, hypertension, stroke, type II diabetes, obesity, metabolic syndrome, breast cancer, colon cancer, osteoporosis, falls, depression, anxiety and negative self-concept (Haskell 2007; Janssen 2010; Kesaniemi 2001; Nelson 2007; Strong 2005; Warburton 2006; Williams 2001). It is estimated that in 2008 physical inactivity caused 9% of premature mortality and 5.3 million deaths worldwide (Lee 2012). This included between 6 - 10% of all deaths from major non-communicable diseases globally and their burden is increasing rapidly in low- and middle-income countries (Lee 2012; WHO 2010b).
The World Health Organisation (WHO) recommends that adults should accumulate at least 150 minutes of moderate intensity or 75 minutes of vigorous intensity or an equivalent combination of aerobic PA throughout the week (WHO 2010a). This should be achieved in bouts of at least 10 minutes duration (WHO 2010a). Muscle strengthening activities involving the major muscle groups are also recommended on at least two days per week (WHO 2010a). Investigations into the dose-response relationship indicate that PA at levels higher than the minimum recommendations confer greater health benefit (Kesaniemi 2001; WHO 2010a)
The available data suggest that 31.1% of the worlds adult population are not meeting the minimum recommendations for PA (Hallal 2012). The direct economic burden of physical inactivity is 1.5 - 3.0% of health care system costs and it is an emerging expense in low-and middle-income countries (Oldridge 2008). It has been estimated that increasing the number of people that achieve the WHO PA recommendations by 10% or 25% would prevent more than 553,000 and 1.3 million deaths, respectively, globally each year (Lee 2012).
Description of the intervention
It has long been accepted that various interventions can promote PA participation and improve health (Dishman 1996). This has prompted growing global interest and investment in PA interventions by different stakeholders using a variety of methods (Heath 2012). It is evident that there are opportunities to influence personal, social and environmental determinants of PA in different contexts and populations (Bauman 2012). A previous Cochrane review found that PA interventions had a moderate effect on participation levels (Foster 2005a). However, conclusions could not be drawn about the effectiveness of isolated components for achieving and maintaining recommended levels of PA in the population.
For the purposes of this review an intervention is any deliberate attempt to increase the PA levels of the participants. It may be delivered using various methods and implemented through a broad range of professions (for example health professional, exercise specialist, PA researchers). This is consistent with the principles of the previous versions of this review (Foster 2005a). The critical additional component of face-to-face interventions is that the interaction with the implementer occurs in person. These interventions can be delivered to groups or individually and several examples are presented below ( Table 1).
How the intervention might work
One of the earliest reviews of determinants of PA stated that few PA interventions or adherence studies were based on any theoretical or psychological models (Dishman 1990). However, when the review was repeated four years later the authors noted a marked increase in the use of theories in studies and interventions (Dishman 1994). It is now accepted that well designed PA interventions are based upon behavioural theories (Bartholomew 2001), but understanding how these are translated into practical strategies needs further evaluation (Foster 2005a). Behavioural theories provide a foundation for an intervention that can explain the drivers of PA behaviour and potential pathways for change (Foster 2005b). They also inform the planning, development and implementation of PA interventions and the majority of studies have adopted social psychology theories (Biddle 2011).
Most of the theory previously applied to PA interventions has stemmed from face-to-face approaches. Epidemiological studies have tried to explain the variation of PA behaviour by examining the impact of different factors or correlates. These have been described as operating at two levels: the intrinsic and extrinsic level (Sallis 1997). Correlates like social class, personality, cognition, attitudes and beliefs operate at an intrinsic level. Extrinsic factors are divided into incentive structures like sports facilities or access to green spaces, and legal restrictions like prohibitive laws (Sallis 1999). Intrinsic factors have received greater attention than the external factors in attempts to explain behavioural choices influenced by face-to-face interventions.
Four common theories were cited in the reviews conducted by Dishman (Dishman 1990; Dishman 1994). These were the Health Belief Model (HBM) (Becker 1974); the Theory of Reasoned Action/Planned Behaviour (TRA) (Fishbein 1975); Social Cognitive Theory (SCT) (Bandura 1986); and the Transtheoretical Model of behaviour change (TTM) (Prochaska 1982). These models have "conceptual convergence" (Biddle 2001) and share two common constructs (Rodgers 1991):
- outcome expectancy – the belief that the behaviour will lead to a specific outcome;
- social norms - the influence of expected behaviours within a social group.
More recently other theories have emerged that include: the Health Action Process Model (Schwarzer 2008); the Common Sense Model of Illness Perceptions (Hagger 2003); and the Behavioural Choice Theory (Epstein 2001). These models tend to encompass the connection between intrinsic and extrinsic drivers of behaviour.
There are several common strategies used in face-to-face interventions for changing PA behaviour that can be grounded in each of these theoretical models and some examples of applying these were described previously ( Table 1). Firstly, stimulus control is the manipulation of factors that prompt and improve accessibility to PA opportunities while reducing the desirability and appeal of sedentary behaviour (Foster 2005b). The provision of exercise classes and walking groups in local communities are examples of this that are typically delivered face-to-face. Secondly, drawing the attention of the participant to the immediate benefits of PA reinforces the behaviour and may increase the likelihood that it will be repeated (Hillsdon 2002). This may include the face-to-face provision of rewards, progress reviews and praise from practitioners or peers. Finally, higher levels of PA self monitoring are associated with better behaviour outcomes (King 1988). Examples of this include the participant keeping a PA diary or using a pedometer for reflection. Each of these strategies can contribute to building social support and developing self-efficacy for engaging in PA (McAuley 1994). However, the key component of these face-to-face approaches may be the elicitation of empathy as the implementer (Miller 2012; Rogers 1969):
- understands the participants feelings;
- responds to what has been said in a way that reflects the participants mood;
- conveys the ability to share the participants feelings.
Why it is important to do this review
Although it is known that the behaviour of individuals can be influenced by PA interventions, the most effective delivery method is not clear (Foster 2005a). In recent times there has been an emergence of remotely delivered interventions that has accelerated with the advent of web 2.0 technology (van den Berg 2007). This impetus has displaced several more traditional face-to-face methods of implementing PA interventions (van den Berg 2007).
It is intended that this review will provide an up-to-date indication of the effectiveness of face-to-face PA interventions. Understanding the effectiveness of these more traditional approaches to implementation should influence PA policy makers and professionals. This review will also describe the importance of individual versus group delivery. Identifying the most effective implementation methods is integral for optimising health-related outcomes related to the promotion of PA participation.
To compare the effectiveness of face-to-face interventions for PA promotion in community dwelling adults (aged 16 years and above) with a control exposed to placebo, no and/or minimal intervention. The influence of delivering the intervention to a group versus individually versus mixed (combined group and individually) will also be assessed.
If sufficient data exists, the following secondary objectives will be explored:
- to assess how the professional delivering the intervention (for example health professional, exercise specialist) influences the effectiveness in changing PA;
- to assess how the intensity of the intervention delivery (for example frequency, duration of contact) influences the effectiveness in changing PA;
- to assess how the prescription of the intervention (for example duration, frequency, intensity) influences the effectiveness in changing PA.
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) comparing face-to-face PA interventions for community dwelling adults with a placebo, no and/or minimal intervention control group. We will include studies if the principle component of the intervention is delivered using face-to-face methods. To assess behavioural change over time the included studies must have a minimum of six months follow-up from the start of the intervention to the final results. We will exclude studies that have more than a 20% loss to follow-up if they do not apply an intention to treat analysis.
Types of participants
Community dwelling adults, age 16 years to any age, free from pre-existing medical conditions or with no more than 10% of subjects with pre-existing medical conditions that may limit participation in PA.
We will exclude interventions on trained athletes or sports students.
We will only include studies that measure PA at an individual level.
Types of interventions
Face-to-face PA interventions can be delivered using supervised methods (for example exercise class) and/or an unsupervised approach (for example home exercise programme). The interventions can be delivered to groups or individuals. They can involve one-off or ongoing interactions between the implementer and the participants that include:
- counselling and/or advice;
- self-directed and/or prescribed exercise;
- home-based and/or facility-based exercise;
- written education and/or motivational support material.
We will exclude mass media and multiple risk factor interventions.
The comparison will be with a control group exposed to placebo, no and/or minimal intervention.
Types of outcome measures
The primary outcomes of this review aim to inform the impact of face-to-face interventions on PA levels. This includes data that assesses change between baseline and follow-up for:
- PA levels expressed as an estimate of total energy expenditure (kcal/kg/week or kcal/week), total minutes completed at a moderate or vigorous intensity, proportion reaching a pre-determined threshold level (for example meeting current public health recommendations) and/or frequency of participation. PA can be assessed using objective methods (for example accelerometers, pedometers) and/or more subjective tools (for example PA diary, survey).
- cardiorespiratory fitness (CF), which is often used as a marker for PA and demonstrates similar associations with health-related outcomes (Blair 2001). It is either estimated from a sub-maximal fitness test or recorded directly from a maximal fitness test. CF is typically expressed as a VO
2max score, which is an abbreviation for maximal oxygen uptake (ml/kg/min or ml/min).
We will also collect information about whether the intervention was delivered to a group versus individually versus mixed (combined group and individually).
The secondary outcomes of this review aim to inform the differential effects of various intervention components. This includes data relevant to:
- quality of life (for example quality-adjusted life years (QALYs)];
- cost (for example cost-benefit, cost-utility);
- adverse events (for example musculoskeletal injury, cardiovascular event).
Search methods for identification of studies
We will search CENTRAL, MEDLINE, EMBASE, CINAHL, AMED, PsycINFO, SPORTdiscus, OpenGrey, SCISearch, ACM Digital Library and IEEE Xplore Digital Library.
We will amend the search strategy used for MEDLINE where necessary to search the other databases listed (Appendix 1).
We will apply no language restriction to the searches.
Searching other resources
We will conduct hand searching for the International Journal of Behavioural Nutrition and Physical Activity. The reference lists of all relevant articles identified during the search will be checked by the authors. We will also use published systematic reviews of PA interventions as a source for identifying RCTs.
We will communicate directly with authors to identify and request unpublished studies and data. A comprehensive list of relevant articles along with the inclusion criteria for the review will be sent to the first author of each paper that meets the inclusion criteria, asking if they know of any additional published or unpublished studies which might be relevant.
Data collection and analysis
Selection of studies
We expect to identify several thousand citations with the initial search strategy. Two authors (CF, JR) will independently screen these titles manually to exclude those which are obviously outside the scope of the review. The authors will be conservative at this stage and where disagreement occurs we will include the citation for abstract review. Two authors (CF, JR) will independently review the abstracts of all citations that pass the initial title screening. They will apply the following inclusion criteria to determine if the full paper will be needed for further scrutiny.
Does the study:
- aim to examine the effectiveness of a PA/CF promotion strategy to increase PA/CF levels;
- use principally face-to-face methods to promote PA to the intervention group;
- allocate participants in to the intervention or control group using a method of randomisation;
- have a control group that is exposed to placebo, no and/or minimal intervention;
- include adults of 16 years and older;
- recruit community dwelling adults that are free of chronic disease or with no more than 10% of subjects with pre-existing medical conditions that may limit participation in PA;
- have a follow-up period of at least six months between commencing the intervention and measuring the outcomes;
- analyse the results by intention-to-treat or, failing that, is there less than 20% loss to follow up.
The authors will be conservative at this stage and where disagreement occurs we will include the citation for full text review. Two authors (CF, JR) will review the full text of all studies that pass the abstract screening using the inclusion criteria described above to identify a final set of eligible studies. The studies included in the previous "Interventions for promoting physical activity" Cochrane review will also be allocated within the new suite of reviews by two authors (CF, JR) (Foster 2005a). When there is persisting disagreement at this stage it will be resolved by consensus after a third author (MT) has been asked to review the study in question. We will link publications and reports that utilise the same data to avoid replication in the analysis.
Data extraction and management
The data extraction form has been piloted independently by two authors (CF, JR) and subsequently adjusted to ensure it captures the relevant data. Two authors (AK, KW) will independently extract the data from all of the selected studies using the standard form. When there is disagreement a third author will review the study and a consensus will be reached (TW). We will extract data from multiple publications of the same study separately and then combine them to avoid replication. Any missing or ambiguous data will be clarified with the study first author via email.
Assessment of risk of bias in included studies
We will only assess and report in a risk of bias table studies that meet the inclusion criteria (Higgins 2011).
Two authors (AK, KW) will assess the risk of bias. Where there is disagreement between review authors in risk of bias assessment, a third author (MH) will be asked to appraise the study independently and discrepancies will be resolved by consensus between all three authors.
We will assess the studies for the five general domains of bias: selection, performance, attrition, detection, and reporting. When sufficient information is available, we will designate each domain as "high" or "low" risk of bias. When there is a lack of information or uncertainty over the potential for bias, we will describe the domain as "unclear". We will judge the studies overall as "low", "medium", or "high" risk of bias given consideration of the study design and size, and the potential impact of the identified weakness noted in the table for each study.
The authors have determined a priori that the best evidence is likely to come from cluster RCTs and CBA studies. Although this differs from the usual evidence hierarchy, it is a better approach than the problematic application of the usual criteria when appraising the evidence for social and public health interventions (Petticrew 2003).
Measures of treatment effect
For each study with dichotomous outcomes, we will express the effect size using an odds ratio (OR). For each study with continuous outcomes, we will express the effect size using the standardised mean difference (SMD) between the post-intervention values of the randomised groups. We will complete a narrative summary of the study results. When there is suitable intervention homogeneity and sufficient data we will also complete a formal meta-analysis of the included studies.
Unit of analysis issues
If possible, we will analyse the studies using the mean and standard deviation (SD) and visualised using forest plots. Alternatively, we will report only the point estimate with confidence intervals (CIs) and P values.
If a study has more than one intervention arm that uses a face-to-face delivery approach, then we will examine the overall effects of the intervention versus control by splitting the control group (means and SDs). We will weight the values according to the overall numbers within each arm. This approach is more appropriate than comparing the effects of each intervention arm against the full control group within a meta-analysis, because it avoids including participants twice in the comparison and effect calculations. For each study with dichotomous outcomes we will calculate an OR and 95% CIs. We will use the number of participants in each arm that are reported as an event (that is categorised as active at a pre-determined level) or no event (for example not active). Where appropriate, we will calculate individual study effects and then the pooled effect sizes as ORs with 95% CIs using a random-effects model. We will calculate any missing 95% CIs using established approaches (Higgins 2008).
If possible, we will re-analyse studies which randomise or allocate by clusters but do not account for clustering during analysis. Where the population reporting attainment of a PA level is stated as a percentage of the population meeting a specified attainment level the analysis will be considered as being at the same level as allocation for each cluster. Alternatively, if appropriate, we will employ statistical methods that allow analysis at the level of the individual while accounting for the clustering in the data. If successful, effect estimates and their standard errors from correct analyses of cluster-randomised trials may be meta-analysed using the generic inverse-variance method in RevMan.
Dealing with missing data
We will exclude studies that are found to have a high degree of incomplete data for assessment (that is less than 40% of data) during the risk of bias assessment or present evidence that missing data is likely to be associated with the reported intervention effect. We will contact the authors of potentially included studies if missing data are unclear or data have not been fully reported. Missing data will be captured in the data extraction form and reported in the risk of bias table.
Assessment of heterogeneity
We will quantify and evaluate the amount of heterogeneity to determine whether the observed variation in the study results is compatible with the variation expected by chance alone (Deeks 2011). Heterogeneity will be assessed through examination of the forest plots and quantified using the I
Assessment of reporting biases
We will plot trial effect against standard error and present it as funnel plots (Sterne 2011). Given asymmetry could be caused by a relationship between effect size and sample size or by publication bias, we will examine any observed effect for clinical heterogeneity and additional sensitivity tests may be carried out (Sterne 2011).
When possible we will report all continuous outcomes on the original scale. If the outcomes are to be combined from different scales we will standardise them as required for the analysis. We will only undertake a meta-analysis when the data are clinically homogeneous and we will follow established Cochrane methods (Deeks 2011). If data are available, sufficiently similar, and of adequate quality, we will use the Cochrane Collaboration's statistical software, Review Manager 2012, to perform the statistical analyses. We will use a random-effects model as the default to incorporate heterogeneity between studies. We will not combine evidence from differing study designs and outcome types in the same forest plot (Christinsen 2009).
Subgroup analysis and investigation of heterogeneity
We will perform a subgroup analysis to compare interventions that were delivered to a group versus individually versus mixed (combined group and individually).
When possible we will categorise the available data according to established exposure characteristics or appropriate quantiles to perform additional subgroup analyses that assesses the relationship between the interventions and outcomes by:
- delivering professional (for example health professional, exercise specialist);
- intensity of intervention delivery (for example frequency, duration of contact);
- prescription (for example duration, frequency, intensity).
We will conduct a sensitivity analysis for studies with low risk of bias as reported within the "Risk of bias" tables.
The authors wish to acknowledge Karen Rees and the Cochrane Heart Group for their contributions and support during the planning of this suite of reviews.
Appendix 1. Search strategies
We will base the initial search strategy on the previous methods used for the "Interventions for promoting physical activity" Cochrane review (Foster 2005a). We will combine all searches with an appropriate RCT filter except CENTRAL, SCI Search, ACM Digital Library and IEEE Xplore Digital Library. MEDLINE, EMBASE, CINAHL and PsycINFO have validated RCT filters. There is currently no validated RCT filter available for SportDISCUS, AMED and OpenGrey, so we will use appropriate text words. The search strategy for MEDLINE is outlined below.
1 Physical Exertion/
2 Physical fitness/
3 exp "Physical education and training"/
4 exp Sports/
5 exp Dancing/
6 exp Exercise therapy/
7 exp Exercise/
8 (physical$ adj5 (fit$ or train$ or activ$ or endur$ or exertion$)).tw.
9 (exercis$ adj5 (train$ or physical$ or activ$)).tw.
13 ((exercise$ adj3 aerobic$) or aerobics)).tw.
14 (("lifestyle" or life-style) adj5 activ$).tw.
15 (("lifestyle" or life-style) adj5 physical$).tw.
17 Health education/
18 Patient education/
19 Primary prevention/
20 Health promotion/
21 Behaviour therapy/
22 Cognitive therapy/
23 Primary health care/
39 16 and 28
Above search combined with RCT filters [the Cochrane Highly Sensitive Search Strategy for identifying randomized trials in MEDLINE: sensitivity-maximizing version (2008 revision)]
1. randomized controlled trial.pt.
2. controlled clinical trial.pt.
5. drug therapy.fs.
9. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8
10. exp animals/ not humans.sh.
11. 9 not 10
We will combine all studies retrieved and remove duplicates. The titles and abstracts will then be manually screened to identify face-to-face versus remote and web 2.0 interventions for allocation into the relevant review in this suite (see "Selection of studies"). We will report the total number of initially retrieved studies and the number of studies allocated to each review during the subsequent screening process.
Contributions of authors
All named authors contributed to the conceptual planning of this suite of reviews to update the previously completed Cochrane review assessing "Interventions for promoting physical activity" (Foster 2005a). This protocol was initially drafted by Justin Richards and Charlie Foster. Editorial contributions were received from Margaret Thorogood and Melvyn Hillsdon.
Declarations of interest
Melvyn Hillsdon has received a research council grant to investigate the feasibility of a primary care physical activity intervention. This was not a study of outcomes and therefore had no bearing on this review. He has been a member of a NICE programme development group on walking and cycling and was paid for travel expenses.
Sources of support
- British Heart Foundation Core Grant (021/P&C/Core/2010/HPRG), UK.
- NIHR Cochrane Incentive Scheme 2012, UK.