Maintaining long-term physical activity after cancer: A conceptual framework to inform intervention development

Purpose: This paper describes a conceptual framework of maintenance of physical activity and its application to future intervention design. Methods: Evidence from systematic literature reviews and in-depth (N=27) qualitative interviews with individuals with cancer were used to develop a conceptual framework of long-term physical activity behaviour. Determinants of long-term PA were listed and linked with domains of the Theoretical Domains Framework which in turn were linked to associated behaviour change techniques (BCTs) and �nally to proposed mechanisms of action (MoA). Results: The conceptual framework is presented within the context of non-modiable contextual factors (such as demographic and material resources) and in the presence of learnt and adapted behavioural determinants of skills, competence and autonomous motivation that must be established as part of the initiation of physical activity behaviour. An inventory of 8 determinants of engagement in long-term PA after cancer was developed. Clusters of BCTs are presented along with proposed MoA which can be tested using mediation analysis in future trials. Conclusion: Understanding the processes of PA maintenance after cancer and presentation of implementable and testable intervention components and mechanisms of action to promote continued PA can inform future intervention development. Implications for cancer survivors. This resource can act as a starting point for selection of intervention components for those developing future interventions. This will facilitate effective support of individuals affected by cancer to maintain PA for the long term.


Introduction
Each year more than 17 million people are diagnosed with cancer worldwide (1). Cancer and its treatment can result in numerous adverse physical and psychological consequences, some of which can persist for years after treatment completion. Individuals with a history of cancer are also at risk of cancer recurrence and developing other chronic conditions such as heart disease. Engaging in regular physical activity (PA) can mitigate many of these adverse effects including fatigue, anxiety, depression, physical functioning and health-related quality of life (2) and reduce cancer recurrence and improve survival (3).
Despite this around 70% of people with cancer are not meeting physical activity guidelines (4)(5)(6)(7). Systematic reviews and metaanalysis have found clinically signi cant effects of interventions to support initiation of physical activity behaviour change with substantial increases in activity levels from baseline to end of intervention (8-10). However, to sustain the health bene ts of physical activity individuals must be habitually physical active for the long term. There is some evidence of modest maintenance effects (11) in people with cancer but activity levels typically regress towards baseline as the time from end of intervention increases (12). There are scant examples of interventions developed with an explicit aim to support sustained increases in physical activity in people affected by cancer and further research is required.
Key to developing effective interventions is to identify determinants of the target behaviour in the population in question as well as potential processes of change. Intervention components that in uence those determinants can then be identi ed. This can be achieved by reviewing existing literature and consultation with the intervention development team. It is also increasingly recognised that empirical research with the speci c population is imperative to fully understand the 'problem' and identify potential solutions. These are core principles of intervention development frameworks such as Intervention Mapping and the MRC Guidance for development of complex interventions (13).
Such development frameworks also recommended the use of theory of behaviour/behaviour change to guide the identi cation of pathways of change and appropriate intervention components. The Theory of Planned Behaviour, Social Cognitive Theory and the Transtheoretical Model have received much attention within the PA and cancer literature (14)(15)(16). However empirical evidence suggests no one theory is superior and that theory-based interventions are as effective as those without explicit theoretical underpinning (17,18). This may be because both types of intervention include similar behaviour change techniques (the main catalysts of intervention effects) (13,17) and existing theories have multiple overlapping constructs. To address the latter, issue the Theoretical Domains Framework (TDF) was developed. It brings together 33 models of behaviour/behaviour change, including 128 separate constructs and is increasingly used by researcher developing complex interventions that target health behaviour change (19). The TDF has 14 theoretical domains from which researchers can draw on to support identi cation of pathways of behaviour change.
Furthermore, for optimal transparency of the intervention development process, and to advance our understanding of not just what works, but how it works, researchers are encouraged to identify the proposed mechanisms of action (MoA) through which interventions are hypothesised to exert their effects. MoA are 'the processes through which behaviour change techniques affect behaviour' (20). For example, a barrier to engagement in physical activity might be a lack of belief in one's capability to perform the behaviour. Intervention developers would select BCTs believed to impact the MoA 'belief in capability', for example, verbal persuasion and focus on past success. Subsequent evaluation of the intervention would include measuring change in this MoA and conducting mediation analysis to determine its impact on behaviour (21). Central to such endeavours is an agreed matrix of BCTs and hypothesised MoA which can act as a standardised resource, enabling synthesis of data. This has been achieved through a series of studies of literature synthesis (22) and expert consensus (23), triangulated to develop a Theory and Technique tool linking BCTs and their MoA (24).
To date there is a lack of empirical data, conceptual or theoretical understanding of the process of engaging in sustained physical activity that can inform the identi cation of behavioural determinants. Indeed, in concluding remarks following their recent metaanalysis of sustaining physical activity behaviour after intervention completion, McEwan et al. (2021) recommend 'future efforts to develop and test theoretical frameworks that specially focus on maintenance of health behaviour could help optimise interventions that are concerned with supporting long-term physical activity adherence"(p8). This paper describes the development of a conceptual framework of sustained physical activity engagement in people with cancer through meta-analysis and primary qualitative research. This led to the generation of an inventory of determinants of this behaviour and a corresponding matrix BCTs and associated MoA that can be used as a basis for developing future interventions.

Methods
A modi ed version of French et al's (25) approach to developing theory informed behaviour change interventions was used and involved four phases: 1: Who needs to do what, differently?
A systematic review and meta-analysis of long-term physical activity behaviour change in cancer survivors was conducted by our group (11). We explored factors associated with success (or lack thereof) including context, population characteristics and behaviour change techniques.

What barriers, enablers and processes need to be addressed?
To afford an in-depth understanding of the barriers, enablers and processes involved in long-term PA behaviour change a qualitative study of 27 cancer patients who had taken part in a previous PA programme in the UK was conducted (26). Inductive thematic analysis was conducted, and ndings were combined with those from the aforementioned systematic review and meta-analysis (15) and evidence from relevant qualitative meta-syntheses and reviews (12,(27)(28)(29), selecting and structuring factors to create constructs that informed the development of an inventory of key determinants of sustained PA behaviour. These data also informed the development of a conceptual framework describing the processes of sustained PA engagement.

Using a theoretical framework, what intervention components could address these barriers and enablers?
In consultation with co-authors, barriers and enablers were mapped to the associated domains of the Theoretical Domains Framework.
Using published expert consensus linking BCTs to the TDF domains (30) BCTs were identi ed that might address these barriers and enablers. In addition, we reviewed the recently published compendium of 'self-enacted techniques' (31) and selected additional intervention components hypothesised to impact on the target determinants.

How can the behaviour change be understood?
Using the Theory and Techniques Tool we identi ed key MoA associated with each behaviour change technique which could be used to assess intervention causal pathways (mediating behaviour change).
This was an iterative process involving regular meetings and revisions at all stages with co-authors.

Who needs to do what differently?
The systematic review and meta-analysis, including 19 studies, concluded that existing interventions with a long-term follow-up were successful in achieving moderate improvements in sustained behaviour change (11). Older adults, those with existing functional limitations and who had fewer contacts with those delivering the PA programme were less likely to sustain PA increases. Furthermore, PA programmes included in the review with poorer long-term behaviour outcomes were less likely to include the BCTs of action planning, graded tasks, and social support (unspeci ed) (see (11) for full details). These ndings were triangulated with the qualitative data (26) and relevant reviews (12,(27)(28)(29) generating hypotheses of the processes at play and informing step 2; generation of an inventory of barriers and enablers and conceptual framework.

What barriers and enablers need to be addressed?
Findings from the qualitative study and systematic reviews were selected and structured to identify an inventory of 8 determinants of engagement in long-term PA behaviour after cancer and a conceptual framework illustrating the processes at play was developed (see Central to the conceptual framework are the founding factors related to initiation of behaviour required before long-term maintenance can be achieved (depicted in the outer segment of Fig. 1). Such factors are vital for consideration during the initiation of PA behaviour before maintenance can be achieved. Contextual factors are key and include socioeconomic status, demographics (including age, sex, education level) material resources and environment (i.e. access to facilities and/or appealing outdoor space). Appropriate attention must be paid to these factors when identifying when, where, and how an individual will engage in a physically active behaviour. This will ensure the chosen activity is appropriate to the individual's personal context and resources. Furthermore, the initial intervention must result in motivation to initiate change. The participant must develop the necessary skills to engage with their new activity and that activity must be appropriate to their pathophysiological status.
A case study: A participant in the qualitative study described how the practice nurse at her GP surgery repeatedly told her she needed to do more exercise and suggested walking. No support or consultation was provided. The participant reported walking on a handful of occasions but then stopped. Reasons included pain on walking, concerns about breathlessness and di culties accessing suitable walking routes. On taking part in the Move More intervention she engaged in a conversation with the practitioner regarding her history of PA, likes, and dislikes, her priorities, and commitments as well as existing health condition, which included obesity, mobility issues/joint pain and COPD. A local weekly yoga class was identi ed which, on the date of interview, the participant had been attending for more than 2 years.
Once PA behaviour is initiated, the individual experiences the consequences of that behaviour, including an impact on their affective state and physiological outcomes. When engaging in PA with others this will also include social interaction. If the behaviour results in desired/positive outcomes that behaviour is reinforced and is more likely to be sustained. Inevitably at some point the behaviour will be disrupted. This may be due to affective factors, such as low mood or boredom, or practical, e.g. discontinuation of an exercise class.
Alternatively, a life stressor event such as ill health or caring responsibilities can disrupt engagement, as can change to the environment/ resources. Adaptation to this disruption then ensues. This will include a prioritisation process of physical and psychological resources. It may also involve problem solving, assessment and/or change regarding accessibility to local amenities and /or personal resources. Pros and cons of re-engaging with the discontinued activity including re ection on the consequences of behaviour (experienced outcomes) will be in uenced by the degree of intrinsic motivation and self-e cacy (con dence) the individual has for that activity. For some social support will play a role here with some individuals requiring practical or emotional support to reengage. Consequently, the behaviour remains ceased/reduced or resumption occurs. This is a cyclical process.
The inventory of barriers and enablers are set out in Table 1. Finding pleasure/enjoyment in the chosen activities was a key enabler of maintenance identi ed in the qualitative data and wider literature. Those individuals who were habitually active talked about the fun they had and enjoyment they felt from being physically active. Individuals engaging regularly in physical activity describe feelings of empowerment as a result, seeing it as evidence that they have overcome the physical challenges of cancer and its treatment and regaining a sense of ownership and control of their bodies.
This is linked to a feeling of con dence in their ability to engage in these activities. An individual's sense of perceived value/experienced outcomes is also integral to motivation to continue to engage in PA participation. Such outcomes included extrinsic factors such as the importance of PA in maintaining health, function, and independence, as well as intrinsic factors such as a sense of wellbeing. For those who participate in group activities social interaction is an important facilitator to engagement. Disruption to PA engagement can typically be divided into affective factors, such as low mood, anxiety and boredom and external factors including terminated exercise classes, ill health of life stressors. The perceived appropriateness of PA to an individual's age is also important with individuals less inclined to engage if social comparison or personal identity are not aligned with the activities. Finally monitoring of engagement by self or others is an important element of participation. This process enables individuals to track progress and identify reductions in PA that can then be addressed. compendium of self-enacted techniques has several overlapping techniques with Michie's taxonomy but also includes some important additions which we hypothesize would impact some of the barriers and enablers identi ed here. For example, boredom was a barrier to continued PA engagement. This might be overcome with the use of the self-enactment technique "add challenge (stops behaviour becoming boring)". Enjoyment was an important enabler to continued PA and the technique of "task crafting (enjoyment)" and "focus on enjoyment (pleasant aspects)" are likely to facilitate this.

How can the behaviour change be understood?
Mechanisms of Action associated with the BCTs were identi ed using the Theory and Techniques tool (33). See Table 1. Measuring them in future interventions would enable calculation of mediating mechanisms of behaviour change.

Discussion
This paper presents a novel conceptual model of long-term engagement in physical activity after a cancer diagnosis, with accompanying inventory of determinants and suggested behaviour change techniques. This resource can act as a starting point for selection of intervention components for those developing future interventions. It is essential that it is used alongside a robust development process, using qualitative and co-design approaches to understand the local context and speci c participant group, working with multidisciplinary teams to choose methods and modes of delivery that are locally feasible.
This conceptual framework is a holistic appreciation of the problem of sustained PA participation after a cancer diagnosis. The outer elements of the framework, including contextual factors such as environment, socioeconomic status, material resources are key, shaping the possibility of initial PA engagement necessary before sustained action can be achieved. Historically, consideration of these contextual factors has often been absent from intervention development. A personalised approach, assessing pathophysiological state and evaluating these contextual factors at an individual level is key to establishing initial PA engagement that has the potential to be sustained. This perspective is in keeping with the recently updated MRC Framework for the Development and Evaluation of Complex Interventions (13). A key addition to which was the inclusion of context in the de nition of complex interventions as well as considerations for the systems within which the intervention sits.
Comparisons can be drawn between the conceptual model, inventory of behavioural determinants and associated intervention components presented here and Kwasnicka et al's review of the theoretical explanations for maintenance of physical activity behaviour (34). Kwasnicka and colleagues coded constructs of 100 theories of behaviour and set out ve theoretical themes they deem relevant to PA maintenance, four of which are also captured in our work. These include 'maintenance motives'; regular grati cation is more likely to lead to sustained engagement, 'self-regulation'; including the need for high coping self-e cacy, 'resources'; psychological and physical assets, and environmental and social in uences.
In recent years the concept of behavioural maintenance in physical activity has received considerable attention with thought paid to the way maintenance is conceptualised and operationalised. The conceptual model presented here aligns with the evolution in thinking from maintenance of a behaviour as a speci c state, to consideration of the underlying mechanisms of action that determine behaviour. This includes shifts back and forth between re ective and reactive processes (35). Rhodes & Sui (35) present a new, working de nition of physical activity maintenance as 'a dynamic development of mechanisms of action that engender greater perceived behavioural enactment e ciency that partially supplant prior mechanisms of action that required greater perceived cognitive recourses to enact physical activity". They argue that some constructs that were critical to the initiation of a behaviour will still be important for engagement in the longer-term. This is re ected in the conceptual model and associated BCTs presented here where constructs and techniques are essential to embed in the process of initial behaviour enactment to support longer term maintenance. Whilst it has been argued that maintained behaviours are characterised by intrinsic motivations and self-determined actions, self-regulatory skills are still required over time as inevitable disruptions occur and more effortful regulation is required. Hence the suggested inclusion of BCTs that consolidate self-regulatory processes and skills in the initiation of behaviour change, as well as considering long-term behaviour change where these skills may need to be revisited to manage a period of disruption. Furthermore, some individuals may spend longer in a phase of effortful self-regulation than others and need to revisit these skills more frequently. This is also supported by evidence from a review of determinants of physical activity behaviour in older adults which found that self-regulatory strategies such as action planning and coping planning were positively associated with both activity initiation and maintenance (36). Development of a typology of sustained physical activity that proceeded and informed this model argues that individuals fall into three 'types' (26). Those who, after initial support to engage successful maintain increased levels of PA through planning and prioritisation. Those who are 'intermittently active' with cycles of action and inaction with frequent periods of effortful self-regulation, and those with consistently low levels of PA with minimal engagement in PA during or after intervention participation. Reviewing individuals after 12 weeks of intervention participation to ascertain which of these 'types' they most align with could help personalise the intervention components/BCTs that need to be emphasised to support maintenance of behaviour.
An important consideration when developing interventions including evidence derived BCTs is to have con dence that these strategies are being delivered and utilised by recipients as intended. Knittle et al (31) recently published the compendium of self-enactable techniques v1.0 to change and self-manage motivation and behaviour. The focus of this taxonomy is that of techniques which individuals can enact themselves, rather than those which are delivered or enacted by the intervention providers. They point to evidence that suggests maintenance of behaviour change following interventions is dependent on the extent to which individuals can enact the BCTs involved (37)(38)(39). They also argue that existing taxonomies have insu cient focus on the way techniques are delivered, received (comprehended and understood) and enacted in everyday lives. This new classi cation also includes techniques from additional behavioural domains that may have utility in affecting behaviour and/or it's determinants, from sport and occupational psychology.  (40) who used qualitative methods to explore participant's understanding of the speci c BCTs included in the Diabetes Prevention Program. The focus was self-regulatory strategies such as problem solving, action planning, self-monitoring of behaviour and goal setting. Some techniques, including selfmonitoring were well understood and participants accurately described their use. Others, including action planning and problem solving were harder to understand and additional support was needed to enable participants to operationalise these intervention components.
It is therefore imperative that those designing and delivering interventions to support long-term PA behaviour change provide appropriate and accessible explanations of the fundamental BCTs to ensure they can be enacted. Assessing this during and after intervention delivery is also important.
In addition to the need for evidence informed intervention components it is important to understand how these components exert their effects on behaviour. This paper identi es potentially effective intervention components/BCTs linked to the constructs of the conceptual model and goes on to state the hypothesised MoA through which these BCTs/groups of BCTs exert their effect on behaviour. This was achieved by using a publicly available database where links between BCTs and proposed MoA have been collated based on expert-veri ed consensus (ref). It has been widely acknowledged that research exploring potential MoA as mediators of behaviour change in speci c context with speci c populations is necessary to advance the eld of behavioural science. Haggar et al (41) describe a process model, including the type of data and analysis needed to contribute to the evidence base of testing proposed MoA. Such contributions will provide data that can be synthesised and increase our collective knowledge of the MoA of interventions. This is important work given the current lack of such evidence, as demonstrated in a recent series of meta-reviews (42).

Conclusion
This conceptual model and accompanying inventory of potential intervention components is intended to inform those developing novel interventions to promote long term engagement in physical activity after a cancer diagnosis. The model and its proposed MoA could be tested in future, using methods outlined by Haggar and colleagues. Use of this model alongside participatory approaches, consulting with end users and key stakeholders to ensure relevance and appropriateness are encouraged. Competing interests: The authors declare that they have no competing interests.

Abbreviations
Authors contributions: Chloe Grimmett was responsible for the conception, design, data interpretation and drafting of the manuscript. Claire Foster and Bernardine Pinto made substantial contributions to the conception of the manuscript. Carl May, Teresa Corbet, Kate Morton and Katherine Bradbury made substantial contributions to the conception and data interpretation of the manuscript. All authors read, revised, and approved the nal manuscript.
Data availability: The data generated during the current study are available from the corresponding author on reasonable request.
Ethics approval and consent to participate: Not applicable to this paper. Ethics approvals were given for the qualitative work proceeding this publication.

Consent: Not applicable
Availability of data and materials: Not applicable Acknowledgements: Not applicable Figure 1 Conceptual framework of long-term engagement in PA after cancer