This review used the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines and was registered on PROSPERO (CRD42020208212). Owing to limited guidance on conducting systematic reviews in this area, we used the UK MRC theoretical framework from which we derived a glossary of components to use as a checklist to extract and categorize qualitative and quantitative data (Table 1) [15]. This framework considers 3 linked dimensions in which to consider processes: implementation, mechanisms of impact, and contextual factors. Implementation involves an assessment of fidelity, dose delivered and received (ie, how often or effectively it was delivered), adaptations made, reach, content delivered, and method of content delivery [20,21]. Mechanisms of impact involve an assessment of participants’ satisfaction with the intervention, “unexpected consequences,” and “mediators” [18]. Contextual factors refer to the external barriers and facilitators, such as cultural or organizational factors, that may alter the implementation of an intervention [22].

The timeframe for the search was limited to between December 1, 2005, and December 16, 2020. This timeframe was selected because of the rapid increase in DPIs in addition to the fast-moving space of digital technology. The review aimed to capture the most recent DPIs. Five electronic bibliographic databases were used: Cochrane Library; Cochrane Collaboration Registry of Controlled Trials; Embase; PubMed; and MEDLINE. The main key search terms were GDM, RCT, and PE (see Multimedia Appendix 1 for a detailed list of search terms and strings and Multimedia Appendix 2 for key definitions). Reference lists of the included studies were also searched for additional eligible studies. Boolean search was used to combine the keywords with operators such as AND, NOT, and OR to further produce more relevant results, for example: ((GDM OR gestational diabetes OR pregnancy-induced diabetes OR diabetes in pregnancy) AND ((RCT OR controlled clinical trial OR pragmatic control trial OR clinical trial) AND (process evaluation OR program evaluation OR process assessment OR process acceptance OR outcome measures)); ((GDMs OR gestational diabetes OR pregnancy-induced diabetes)) AND ((RCT OR controlled clinical trial OR pragmatic clinical trial OR clinical trial) OR (process evaluation OR program assessment OR process acceptance OR outcome measures).

Inclusion criteria were published (peer-reviewed) RCTs of DPIs in women with a current diagnosis or history of GDM. Exclusion criteria were studies not published in English; studies where the target population was women who had a family history of T2D or women who were menopausal or postmenopausal; and gray literature, including abstracts in conference proceedings. For further details about the eligibility criteria, please see the study protocol [23].

After initial deduplication of the extracted data using Endnote (Clarivate), studies were uploaded to the platform Rayyan for screening. The 2 reviewers (IIMS and MB) independently screened the titles and abstracts of the studies. Disagreements were resolved via consensus or decided by a third reviewer (IPN) when consensus could not be reached. The full texts of potentially relevant studies were retrieved for further screening and independently appraised by 2 reviewers (IIMS and MB) for final inclusion.

Standardized data extraction forms were developed and piloted. Extracted data included a summary of study characteristics, evaluation of processes (if any), methods, and findings. Data were extracted by a single reviewer (IIMS) and verified by a second reviewer (NH); discrepancies were again resolved via consensus.

We conducted a narrative synthesis of the studies as we anticipated that the heterogeneity of the methods used to assess processes would preclude meta-analysis, as has been found previously [10].

The Mixed Method Appraisal Tool was used to assess the quality of included studies by 2 independent reviewers (IIMS) and (NH). The discrepancies were again resolved via consensus.

The PRISMA flow diagram for this review is shown in Figure 1. Of the 13,735 records initially identified, the full text of 378 full-text studies were screened for eligibility and 21 studies met the eligibility criteria. An additional 3 studies were added following citation searching. Therefore, 24 studies were included in this review. Most of these studies fulfilled at least 3 methodological quality criteria outlined by the Mixed Method Appraisal Tool for each study design (Multimedia Appendix 3 [3,24-46]).

The stated primary aims of all 24 DPI studies reviewed were to improve dietary and physical activity outcomes, increase self-efficacy levels and risk perception of T2DM, and decrease maternal postpartum weight and BMI. The 24 studies evaluated in-person (n=8), digital (n=7), and hybrid (n=9) interventions. Two of the studies were pilot studies of the feasibility and acceptability of DPIs. All studies were published between 2007 and 2020. Most studies (n=22, 92%) were conducted in high-income countries. Only 2 studies were conducted in low- to middle-income countries: Egypt (n=1) and Malaysia (n=1). The sample size ranged from 31 to 1180 participants. The intervention duration varied from 12 weeks to 6 years. See Multimedia Appendix 4 [3,24-46] for a description of study characteristics of the studies included in this review.

The most frequent PE components evaluated were content delivered and method of content delivery (n=24), dose delivered (n=20), reach (n=20), dose received (n=17), barriers (n=15), facilitators (n=8), and participants’ responses toward the intervention (n=8). Relatively few studies reported process data relating to adaptations (n=7), mediators (n=7), fidelity (n=5), or unexpected pathways or consequences (n=2). All PE components outlined by the MRC framework and reported in the reviewed DPIs are summarized in Multimedia Appendices 5 and 6 [3,24-46].

Five studies explicitly referred to a PE framework or PE quantitative or qualitative measures. Only 1 of these referred to process measures in the context of the MRC framework of complex interventions [24]. The others reported measures to ensure adherence to the study protocol [25], to measure “reach” (using the penetration, implementation, participation, and effectiveness metric) [26,27], or to assess participants’ satisfaction to the program using a survey [47].

We identified and reviewed 24 RCTS of DPIs that evaluated at least 1 PE component for women with previous or current GDM. Only 10 studies explicitly reported individual processes, and only 5 explicitly referred to a PE framework or PE quantitative or qualitative measures to assess processes. Overall, few studies reported and evaluated processes in relation to study outcomes, and although most DPIs linked PE components and DPI outcomes in some capacity, it was challenging to attribute any 1 process component to the effectiveness and success of the intervention.

The most complete of the PE components reported across all 24 studies was the combined component of content delivered and method of content delivery. Clear narratives were provided on how the interventions were delivered and their content. Most of the interventions also referred to broad behavior change techniques such as goal setting and MI, which are known to be effective in promoting lifestyle behavior change [56,57]. However, few studies linked these theories with mechanisms of action by providing a detailed psychological or behavior change theoretical framework on which their intervention was underpinned. Such frameworks are important to make sense of “how,” “why,” and “under what circumstances” intervention components work together to achieve the desired outcomes. Moving forward, we recommend that all RCTs of DPIs publish an overview of the theories they have used to design and evaluate the intervention for greater transparency.

Although most DPIs reported information on the dose delivered, PE measures on the dose received were less complete. Generally, increased frequency and longer sessions resulted in more engagement with the intervention and better outcomes [26-28,30,31,36]. More frequent sessions may also be preferred by the participants: for a study on the use of MI and hemoglobin A_1c outcomes, women said that they would like more sessions to improve their engagement with the intervention [58]. The timing of the intervention is also a consideration. Most of the DPIs that we reviewed delivered interventions during either the antenatal or postpartum periods. However, systematic review evidence suggests that optimal pregnancy outcomes can be achieved by delivering the intervention during both periods since engagement with a healthy lifestyle poses tends to be poorer during the postpartum stage [5,51]. Moreover, there was little information in the studies reviewed on how the dose received impacted the success of the intervention. The challenges in assessing adherence to the dose received are well recognized as people may have received a dose but choose not to take it [59]. This is also supported in the literature where a nurse-led psychological intervention for a T2D cluster RCT study identified a lower dose received than intended, but that this was not associated with the dose delivered [60]. Further research to explore how timing, frequency, and duration of the intervention affect outcomes among women with current or previous GDM is indicated.

Although all 24 studies included in this review sought to describe the reach of the intervention, few gave a detailed evaluation of this measure. Poor reach was attributed to generic reasons, most commonly low participation rates, loss to follow-up, and small sample sizes [26,27,29,36,39]. The increase in OGTT appointment attendance found in 1 study [41] is contrary to prior research linking loss to follow-up and withdrawal numbers to participants’ reluctance to attend OGTT appointments [49,54], which highlights the influence that different contexts have on outcomes in this population. Moreover, the studies that made adaptations to tailor their DPIs to personal or cultural needs [24,25,28,34,35,41,45] are likely to have increased reach by making the interventions more relevant. A clear definition of the target population and consideration of how the DPI can be adapted to make it as inclusive as possible seems important for its success.

Most of the studies reported on barriers to intervention engagement. Similar to previous systematic reviews [10,61], the barriers primarily related to lack of time, childcare commitments, and challenges to maintaining healthy lifestyles [24,29]. Focus groups with women suggest that these inherent barriers can be mitigated by combining direct contact with health care professionals with web-based interventions [48]. Moreover, it is worth noting that, despite the many barriers to participation identified, women were still in favor of the DPIs, and many resulted in positive outcomes. This aligns with previous research in T2D [62] and supports a need and willingness among women with previous or current GDM to engage with DPIs.

A particular concern identified by this review was the lack of reporting or evaluation of fidelity. Only 5 studies reported on this process component. Ensuring fidelity is a key aspect of any DPI because it safeguards against nonadherence to the study protocols and inadequate implementation delivery [47,63]. Ideally, fidelity should be measured prior to conducting full-scale implementation to distinguish between outcomes that are related to ineffectiveness from those related to protocol deviation [64,65]. In our review, methods to assess fidelity included audiotaping and patient registers of attendance [27,36]. Such methods have been tried and tested in the T2D setting [47,63] and are likely to be more robust than relying on the expertise of health care professionals as a fidelity measure [66]. To minimize research waste, fidelity assessments should be commenced at the design stage of a trial and incorporate standardized measures where possible.

Finally, we welcome the use of both quantitative and qualitative methods to evaluate the PE components in our studies because this enables a more in-depth understanding of the relationship between individual components and outcomes [52,67]. We reemphasize the importance of mixed methods when conducting thorough PE.

To the best of our knowledge, this is the first and only systematic review of PEs of T2DM prevention interventions for women with GDM. We used robust methodology, an established framework of PE for complex interventions [15], and restricted the review to RCTs, which are known to be the best design for evaluating complex health care interventions [15]. The glossary list of standardized DPI content and PE terms developed during the review process can be used to guide future RCTs in this area. However, while the MRC framework [18] is a valuable source for conducting systematic reviews on PEs, it has been criticized for lacking practical advice on how to conduct them [68]. The lack of information on PE frameworks in our review also made it difficult to categorize individual PE components against the framework. Moreover, there is a lack of consensus on how PE components should be defined and interpreted by researchers, which is only partially addressed in the MRC guidelines [15], which facilitate understanding of PE constructs but provide little clarity on what should be measured and how.

This systematic review has highlighted that there are important gaps in the reporting of PE metrics for RCTs of T2DM in women with GDM. We recommend rigorous, systematic, and in-depth PE guidance to facilitate reporting of these studies. Future research should focus on reaching consensus on the reporting of PE measures using established frameworks and evaluating PE in real-world health care settings to optimize the interpretation of study outcomes.