Preventing Type 2 Diabetes after Gestational Diabetes: A Systematic Review Mapping Physical Activity Components using the Socio-Ecological Model

Objectives Gestational diabetes commonly occurs during pregnancy and increases lifetime risk of type 2 diabetes following pregnancy. Engaging in physical activity postnatally can reduce this subsequent risk. Interventions aiming to increase physical activity after gestational diabetes may not address the wide range of post-pregnancy barriers. A socio-ecological approach highlights the need to include multi-level factors such as social, community and organisational factors. The aim of the review was to map intervention components to prevent type 2 diabetes after gestational diabetes using the socio-ecological model as a framework and investigate how physical activity changes align with different intervention components utilised. Methods Eligible studies included any study type within 5 years of a gestational diabetes diagnosis and targeted physical activity. A systematic search of MEDLINE, Cochrane Library, Web of Science, CINAHL Complete, and Scopus was conducted in October 2022. Results were categorised based on whether findings demonstrated no increases, non-statistically significant increases or statistically significant increases in physical activity. Results Forty-eight studies were included (37 different interventions). Thirty-eight studies were assessed as “adequate” quality, only two studies were “good” quality, and the remaining were limited quality. Mixed physical activity outcomes were observed across components used at the intrapersonal level, with components across other levels of the socio-ecological model showing more increases in physical activity. Intervention components within the social and organisational levels, for example childcare provision, providing group-based sessions and offering remote delivery, were more often present in interventions with physical activity increases. Conclusions for Practice Future interventions targeting physical activity after gestational diabetes should aim to include social and organisational-level components in their intervention design. This systematic review was registered in PROSPERO (ID: CRD42021272044).


Significance
What is Already Known on this Subject?Physical activity can reduce risk of type 2 diabetes after gestational diabetes.However, interventions in this population are not sustainable, have low attendance and high dropout rates.What this Study Adds?This review has highlighted intervention components across levels of the socio-ecological model which could have important implications in uptake and maintenance of physical activity after gestational diabetes.Findings should be considered in intervention planning and design, ensuring a combination of multi-level approaches are purposefully included.Examples include providing childcare, facilitating social support through group-based sessions and offering increased flexibility through remote delivery.

Introduction
Gestational Diabetes Mellitus (GDM) is a common complication in pregnancy, resulting in short-and long-term complications in both women and their infants (Metzger, 2010).One potential long-term complication is the development of Type 2 Diabetes Mellitus (T2DM), where subsequent risk is ten-fold that of women with a normoglycemic pregnancy (Vounzoulaki et al., 2020).Preventing T2DM after GDM is currently recognised as one of the top-ten research priorities, according to literature, Health Care Professionals (HCP) and women who have had GDM (Ayman et al., 2021).The National Institute for Health and Care Excellence (NICE) recommends educating about lifestyle change after a GDM pregnancy to reduce T2DM risk (NICE, 2020).
Lifestyle changes, including physical activity (PA) and dietary changes, have been shown to reduce onset of T2DM by over 50% (Knowler et al., 2002;Pan et al., 1997;Tuomilehto et al., 2001).When these are adopted by women with previous GDM, T2DM development can also be effectively prevented (Bentley-Lewis et al., 2008;Chasan-Taber, 2015).PA alone may independently reduce risk of future T2DM, however is not effectively encouraged after GDM (Bao et al., 2014;Jones et al., 2017aJones et al., , 2017b)).Despite interventions improving dietary behaviour and resulting in weight-loss, challenges in PA uptake remain (Jones et al., 2017a(Jones et al., , 2017b)).Furthermore, the UK National Diabetes Prevention Programme, "Healthier You", has struggled to engage people under the age of 65 (NHS, 2019).Taken together, this could be as interventions and diabetes prevention programs may not address the unique barriers faced by women of reproductive age (Lim et al., 2021), such as balancing family demands, adjusting to a new role as a mother, lack of childcare and support (Dennison et al., 2019).These barriers are not exclusively within an individual's power to overcome and change (Ioannou et al., 2024).Further understanding regarding effective intervention components, and their potential impacts on PA, is needed.
An integrated system-wide approach could be more effective than single-level interventions to overcome barriers to health behaviours and improve health outcomes (Rutter et al., 2017).This is because individual behaviours do not happen in isolation, with cultural, social and other contextual factors largely determining health behaviours (McGlashan et al., 2018).The Socio-Ecological Model (SEM) focuses on the relationships between individuals and their surrounding social, physical and policy environments (Stokols, 1996).Identifying and targeting multiple levels of the SEM could therefore result in longer-term sustained behaviour change (Mcleroy et al., 1988).Figure 1 displays an adapted version of the SEM used as an a priori framework for the present study, highlighting the five levels of influence on individual behaviour.
Previous systematic reviews have examined the effectiveness of lifestyle interventions in women with previous GDM.These have evolved over the past five years, with some of the first reviews examining whether lifestyle interventions can reduce risk of T2DM in women with previous GDM (Chasan-Taber, 2015).More recently, reviews have focused on the cost-effectiveness of lifestyle interventions and the benefits and timing of lifestyle interventions (Goveia et al., 2018;Hewage et al., 2020).Only two reviews have specifically looked at intervention design.Peacock et al., (2014) highlighted that most interventions may not be translatable into real life settings.Jones et al., (2017aJones et al., ( , 2017b) ) summarised knowledge and practices around tailoring multi-modal lifestyle interventions.Neither review analysed intervention components based on existing theory.Buelo et al., (2019) examined both the effectiveness of interventions and the extent to which factors influencing intervention effectiveness were addressed.Their mixed method synthesis evaluated to what extent barriers identified have been addressed in lifestyle interventions.They grouped their qualitative themes based on Dahlgren and Whitehead's determinants of health model (Dahlgren & Whitehead, 1991), but did not analyse the intervention components according to the model.
Reviews in other topics have analysed interventions according to the SEM, and subsequent PA outcomes (Mehtälä et al., 2014).This approach has not been used before for interventions aiming to promote activity in women with previous GDM.Evaluating to what extent these interventions have incorporated a socio-ecological approach in their design, and understanding what effect specific components within each level may have on PA outcomes, can inform future intervention design in this area and subsequent policy decisions.

Aim
The aim is to investigate the extent interventions to prevent T2DM after GDM have integrated a socio-ecological approach, and the impact on PA outcomes.The questions the review addressed included:

Methods
Methods fully comply with the PRISMA 2020 checklist (Page et al., 2021).A protocol was prepared for registration in PROSPERO (ID: CRD42021272044) but was not published elsewhere.

Eligibility
Table 1 includes a summary of the inclusion criteria.Studies had to include women with a GDM diagnosis in the previous 5 years and have any type of PA component.Interventions initiated in pregnancy, with the aim of changing postpartum or life-long behaviours were also included.PA did not have to be the sole focus of the intervention, meaning interventions including diet or weight-loss targets were still included.
If weight loss was targeted and if there were dietary components to the intervention, it was still included.To map intervention components using the SEM, any study type, e.g., protocol papers, were included.

Search Strategy
A literature search was carried out in October 2022.The search was conducted in 5 databases: MEDLINE (via EBSCO), Cochrane library, Web of Science (via Clarivate analytics), CINAHL Complete (via EBSCO), and Scopus.Search terms and keywords from previous reviews of similar themes were included or modified for the purpose for the present review (Buelo et al., 2019;Goveia et al., 2018;Hewage et al., 2020).Where it was possible to include limits, results were filtered to only include publications in English language.Date limits were applied, excluding papers published before January 2000, as done by Peacock et al., (2014).A breakdown of the themes used, and search terms is displayed in Table 2.

Data Collection Process
Data was extracted using a standardised data extraction excel sheet piloted on three papers by EI.Published papers were grouped together when they were related to a singular intervention or study.For example, where the data needed to answer the review questions spanned across a protocol and a results paper, these were grouped by the intervention name, with data items recorded as one entry and relevant information extracted from all linked publications.A maximum of two attempts were made to contact a study's author where data was unavailable.

Quality Assessment
The included studies were evaluated for risk of bias using the empirically grounded quality assessment tool, QualSyst, by Kmet et al.,(2004).QualSyst provides a systematic, reproducible, and quantitative means of assessing the quality of research from different study types (Kmet et al., 2004).The present study included different study types, providing valuable information to answer the research questions which would otherwise not have been considered (Clarke & Oxman, 2003;Hawker et al., 2002).Therefore, Qual-Syst, a more generic quality assessment tool, was suitable for assessing risk of bias in included, variable study design types.Lee et al., (2008) defined the quality of the paper based on QualSyst summary scores as strong (> 0.80), good (0.71-0.79), adequate (0.50-0.70) and limited (< 0.50).These boundaries were used in the present review to narratively synthesise the differences between findings for higher or lower quality studies (Booth et al., 2012).Studies were scored 'N/A', '2' for 'YES', '1' for 'PARTIAL' and '0' for 'NO' on 14 different items.The total possible score was double the number of 'N/As' subtracted from 28.A summary score was calculated by summing the total score and dividing by the total possible score.The 14 data items scored included (extracted directly from Kmet et al., 2004): 1. Question / objective sufficiently described?2. Study design evident and appropriate?3. Method of subject/comparison group selection or source of information/input variables described and appropriate?4. Subject (and comparison group, if applicable) characteristics sufficiently described? 5.If interventional and random allocation was possible, was it described?6.If interventional and blinding of investigators was possible, was it reported?7. If interventional and blinding of subjects was possible, was it reported?8. Outcome and (if applicable) exposure measure(s) well defined and robust to measurement/misclassification bias?Means of assessment reported?9. Sample size appropriate?10.Analytic methods described/justified and appropriate?11.Some estimate of variance is reported for the main results?12. Controlled for confounding?13. Results reported in sufficient detail?14. Conclusions supported by the results?

Synthesis Methods
An adapted SEM was used as an a priori framework to classify intervention components (Fig. 1).Each circle represents a 'level', and each level is labelled e.g., interpersonal.For this study, the intrapersonal level was defined as intervention components targeting psychological factors e.g., behaviour change strategies and/or educational components.The interpersonal, or social, level included components related to other individuals surrounding women who have had GDM e.g., their partners, or intervention delivery staff.The organisational level was defined as where components were targeted at or based in organisations.For example, out of healthcare settings or remotely, or the inclusion of childcare.The community level was used to represent interventions making use of community or locally based resources, while the policy level was taken to represent guidelines utilised in interventions.
PA outcomes as reported by each study were categorised as 'U' if no outcomes were available e.g., if the paper was a protocol paper, 'N' if there were no changes in PA, 'Y' if PA outcomes increased and 'Y*' if these were significant increases.These were narratively synthesised to better understand commonly utilised intervention components within and across levels of the SEM, according to PA outcomes.Interventions were labelled alphabetically.These letters were used in tables to group interventions under one label and to visually depict patterns of intervention components by SEM level.

Results
After duplicates were excluded, a total of 3603 publications were retrieved from the database searches and reference lists.After screening the titles and abstracts, 77 publications were sought and further assessed for eligibility.At full-text screening, papers were excluded because they did not include relevant (n = 5), or enough information (n = 1).Some studies were also excluded due to a sole weight-loss focus, with no measures of PA behaviour change (n = 17) or because the target for intervention timing was outside of the 5-year postpartum period after GDM (n = 7).After full-text screening, 47 publications were included in the final review (comprising 36 different interventions) (Fig. 2).A summary of characteristics of included papers is displayed in Table 3.

Quality and Study Type
Table 4 displays study quality by study type.Twenty-four studies were RCTs and 16 were protocol studies.Most studies were "adequate" quality (n = 38), with only one study falling in the "good" quality range and the remaining limited quality (n = 8).The only "good" quality study was an RCT and saw significant PA increases (see Table 3, study (j) 'Estudio PARTO' by Burkart et al., (2020) for details).

Levels of Influence
Intervention components at the intrapersonal and social levels appeared in every intervention (n = 36), with the second most commonly appearing level being Policy (n = 24).Community level components appeared the least often (n = 8), and organisation level components the second least (n = 22).Five interventions utilised all five levels of the SEM (14%), 14 had four levels (41%) and 11 had three levels (30%).More interventions with components at 4 or 5 levels of the SEM saw significant PA increases.None of the interventions had only one level of influence (Fig. 3).m, q, t, G, j l, m, w i, w b, w l, q, t, u, w i i, l, t j, m, t, w, G i, l, q, u, w i, l, t, w j, m, q, u i, j, m, q, t, u, w

Social (Interpersonal)
Five different intervention components were identified under the social (interpersonal) level.The most common social component was using HCPs including exercise physiologists, dietitians, midwives to deliver the intervention.The second most common social component was using groupbased sessions.All but one intervention including this component saw PA increases.Of the few interventions delivered by laypeople, three saw significant PA increases, and the rest had yet to publish results.One-fifth of the interventions allowed participants to bring their family or partner to sessions or actively included them in the interventions.The final, and least used intervention component at the social level was using forums for connecting women to each other, to ask questions or share their experiences and tips.

Organisational
Four different intervention components were identified at the organisational level.Remote delivery of the intervention was the most utilised intervention component, and all papers with published results reported PA increases.The next most used intervention component was being based out of hospitals where women were cared for during pregnancy.Implementing exercise during the session and providing childcare during sessions were the least used overall, yet most used in interventions with significant PA increases.Aside from these main four components, one intervention also provided healthy food and drink at sessions and provided transportation (Guo et al., 2018).One other intervention removed the issue of cost by providing access, for example, to pool and PA services for free (Rönö et al., 2014).Finally, two interventions provided a gift card or form of monetary incentive for participating in the intervention (Lipscombe et al., 2019;McManus et al., 2018).

Community
Only two intervention components were identified at the community level.These components included basing interventions in local settings or involving local communities as part of the intervention.For example, the 'MAGDA' study and 'Dulce mothers' carried out sessions in community health centres and both saw significant PA increases (O'Reilly et al., 2016;Philis-Tsimikas et al., 2014;Shih et al., 2014).The 'Families Defeating Diabetes' intervention had walking groups taking place in local malls and also saw significant increases in PA (McManus et al., 2018).

Policy
Fifty-six percent of the interventions implemented PA guidelines (n = 20).This included country-specific guidelines, such as the Chief Medical Officer's guidelines in the UK, or more generally the WHO PA guidelines (Davies et al., 2019;WHO, 2016).This was the only component stated at a policy level.Use of PA guidelines was similarly spread across different PA results.

Discussion
The aim of the present review was to: (a) map PA intervention components using the SEM, (b) understand how many levels and (c) determine what intervention components across the SEM are commonly utilised in interventions seeing PA increases.Overall, significant PA increases mostly occurred when four or all five levels of the SEM were utilised.Intervention components which had more increases in PA were remote delivery of the intervention, providing childcare, and having group-based sessions.
In addition to the 16 protocol papers, expanding the inclusion criteria for any study type resulted in an additional eight papers included in the present review, mainly as pilot, feasibility, or pre-post studies.Mixed PA outcomes were observed from these eight papers, therefore, it was not the case that non-RCTs were more likely to show meaningful PA increases.It is possible that, due to standards for publishing RCTs, these tend to include more explicit information regarding study design (Kmet et al., 2004;Schulz et al., 2010).However, QualSyst's performance as a quality assessment tool seemed evenly spread over the different study types, therefore it is likely that quality was adequately assessed, and that studies generally were not reported well.
All interventions included both the intrapersonal and social levels.These levels also included the greatest variety of intervention components.This is important, as these levels of influence are theorised as having the strongest influences on an individual (Kilanowski, 2017).Use of behaviour change strategies to reduce T2DM risk after GDM has also been determined as important, at the very least for reducing energy intake (Lim et al., 2020).While interventions should include the intrapersonal level due to the influences on individual behaviour, it is likely that ability to increase PA is constrained by wider factors across the other levels.This is evidenced in the present review as significant PA increases occurred mostly when four or five levels of the SEM were utilised.Additionally, the present review identified that intervention components used across the intrapersonal level, including use of behaviour change techniques, showed mixed PA outcomes, with components across other levels showing greater variation and more definitive.Therefore, while the intrapersonal level matters, wider levels of the SEM may act as constraints for increasing activity in women with previous GDM and therefore need to be included in interventions.
Despite the number of and type of level potentially impacting PA outcomes, findings of the present review indicate that intervention components within each level are also important.More specifically, distinctive patterns across intervention components from the social and organisational levels were seen.For example, providing childcare (organisational) was a key component that appeared most in interventions seeing significant PA increases, and did not appear in interventions with no or non-significant PA increases.This result is in agreement with literature which has identified childcare (or lack of), as a barrier to participation, given women's identified "role as a mother" (Dennison et al., 2019a).In terms of the SEM, childcare as a barrier is not wholly within an individual's capability to overcome (Ioannou et al., 2024).It is a structural barrier, which from a practical perspective, to overcome, would need to be addressed by the non-intrapersonal levels of the SEM (Ioannou et al., 2023).To increase activity, it is therefore important that interventions targeting women after GDM not only target behaviour change strategies, but also address barriers at either the organisational or community level, for example, by addressing social norms around the role of a mother, and/or providing childcare.
Group-based (social) and remote delivery (organisational) were also most seen in interventions with significant and non-significant PA increases.This may seem conflicting, however, a blended approach could improve PA outcomes in future interventions.Tang et al., (2015) highlighted that PA done at home could better engage women after GDM, as lack of time and flexibility were key barriers.However, groupbased sessions could be effective for managing chronic conditions (Harden et al., 2015).Specifically, women after GDM value connections made with other women who have shared a similar experience (Kelly-Whyte et al., 2021).While it could be expected that forums would provide a similar sort of comfort, the present review did not find this to be a particularly beneficial intervention component.In part, this could be because forums are less personal, and could not be providing the type of social support women with previous GDM are looking for.A recent study by Dennison et al., (2022) highlighted that women after GDM want more support, including connecting and meeting with other mums who have had GDM.Therefore, connecting women with previous GDM e.g., through group-based sessions, where there is flexibility to incorporate PA at home and in own time, could be useful to improve PA outcomes in interventions trying to reduce T2DM risk.

Limitations
The present review categorised intervention components based on where they sit within a system, however, SEM levels refer to systems changes (Mcleroy et al., 1988).For example, using HCPs to deliver the intervention was categorised at the social level.While this is a social interaction, the interventions were not actively targeting HCP behaviour, knowledge, or attitudes to help or benefit women.While the present review used the SEM to map intervention components and design, interventions should focus on at least two different levels of influence (Stokols, 1996).Using the example above, this could involve targeting beliefs HCPs hold that may be unhelpful, to enable them to most effectively provide the support that women after GDM have indicated they would like to receive (Dennison et al., 2022).
Another limitation of the present review was how PA outcomes were quantified and interpreted.PA targets and measures in the identified interventions were heterogeneous.To accommodate for this, and enable meaningful synthesis, an intervention was categorised as having successful PA outcomes, based on whatever PA outcomes were used within the individual study.However, it is important to note that reporting of PA outcomes was greatly varied.How interventions themselves classified significance also varied.To accommodate for this, this review considered and looked for patterns across interventions seeing changes in PA, whether these were or were not significant.While this method was useful for synthesising and understanding the results of the review, it is limited.However, results of the review were consistent with other literature discussed above, providing a degree of confidence.

Conclusions
While interventions to prevent T2DM after GDM do incorporate multiple levels of the SEM, those which included components at the organisational levels, targeting structural barriers like providing childcare, had a greater number of significant PA increases.Future interventions targeting this population should, at the very least, address childcare barriers in their intervention design.They should also consider how to encourage social support between women who have had GDM, for example, through group-based sessions, and consider how the offer of remote delivery can provide increased flexibility for participation.

Table 1
Inclusion criteria summaryTable provides a summary for the eligibility criteria of the present review.Each inclusion criterion was separated by the review questions being addressed.GDM Gestational Diabetes Mellitus; T1DM Type 1 Diabetes Mellitus; T2DM Type 2 Diabetes Mellitus; PA Physical Activity

Table 4
Summary of study types included in the review and analogous study quality RCT Randomised Control Trial; N total number of studies; n sub-number of studies

Table 5
How intervention strategies align with the adopted a-priori socio-ecological model Behaviour change strategies include goal setting, motivational interviewing, action plans, self-monitoring (logbook), reminders etc. Education refers to DPP content, risk perception, healthy lifestyle advice, printed materials etc. Exercises given refers to a program given or instructional/how to do exercise.Monitor refers to use of accelerometer or pedometer.App or web refers to either a program or component if the intervention including technology (website, application etc.).Reminders/messages refer to any automated messages, postcards or feedback given HCP refers to who delivered the intervention (e.g., midwife, dietitian).Family, partner etc. refers to being able or these being included in intervention.Forum refers to forums or online ways of connecting.Group refers to in-person group-based sessions.Community refers to being in community contexts or community driven.Local refers to local settings or taking place locally.Remote delivery refers to intervention delivery via phone or calls for motivation.Exercise during session refers to implementing group walks or any intervention where exercise was completed during (usually face-to-face) sessions.Childcare refers to some form of childcare or reimbursement provided.Hospitals refers to the intervention taking place, in part of fully, at the hospital in which GDM was diagnosed or cared for Table5summarises the intervention components according to the a priori framework used for analysis.