Realist synthesis is a theory-driven approach to evidence synthesis based on the philosophical principles of realism. Realism is defined as a broad logic of inquiry that sits between positivism and constructivism and agrees that social reality cannot be measured directly [26]. However, it can be understood by examining the relationship between context and outcome, underlined by causal forces (mechanisms) in which events occur and the outcomes produced [26]. A realist review is concerned with answering how an intervention works, whom it works for, and in what circumstances it works [27]. It differs from the traditional systematic review in that it emphasises the importance of contextual factors in shaping the effectiveness of interventions and their associated mechanisms. A realist review involves an iterative process of theory building and testing — which includes building an initial programme theory based on preliminary data and then refining this theory based on additional qualitative and quantitative data.

Gender transformative approaches are both ontologically and epistemologically complex [28]. GTAs work by changing participants’ decision-making process and altering the resources and opportunities available to them within a specific context by eliciting certain mechanism(s). Therefore, it is essential to unpack ‘what works’ to explain these interactions. This review aims to achieve this by actively seeking out the contextual (C) factors that are hypothesised to have triggered the relevant mechanism (M) to generate the outcome (O) of interest [26]. In a realist review, secondary data is used to develop these CMO configurations (CMOCs), categorise them into theory-driven initial programme theories (IPTs), and then test (confirm, refute, or refine) them with additional data to produce the refined programme theories [29].

A programme theory is an idea about how the programme works, i.e. if we do X, then Y will happen because of Z [27]. Through this ideation process, we observe patterns in published literature and develop plausible theories that inform the programme design and implementation in different settings [30]. In this realist review, we operationalise mechanisms as resources and reasoning. Resources refer to strategies or components introduced by the interventions in a specific context to elicit change. Reasoning refers to the behavioural response of participants triggered by these resources [31]. Such an operationalisation helps differentiate if data contributes contextually or mechanistically. We will utilise CMO configurations organised as if…then…because statements to answer the following research question: What are the underlying contexts and mechanisms that lead to the success of gender transformative approaches in reducing adolescent pregnancy in low- and middle-income countries? We will follow the iterative process described by Pawson et al. (2005) to conduct our realist synthesis review [28]. While we follow the five key steps outlined in the article, we also adapt the sub-steps to fit our purpose (Fig. 1). The protocol for this realist review is registered in PROSPERO (CRD42023398293), and we will follow the RAMESES publication standards for this realist synthesis [32].

Fig. 1Stages 1a and 1b — Clarify the scope of the review and development of initial programme theories

The research question for this review was developed as a part of the first author’s (S. S.) doctoral study proposal. As a next step, to refine the purpose of the review, we conducted a rapid assessment of seven systematic reviews on gender transformative approaches targeting adolescent health [20, 22, 24, 33,34,35,36]. This search helped build the first list of interventions for developing the IPTs. We identified 18 studies from the seven systematic reviews and used them for IPTs formation. A codebook based on the CMO components was developed to guide data extraction to identify relevant factors contributing to the IPTs [37]. One author (S. S.) designed the codebook, and two other authors revised it in an online discussion (S. M., I. O. A.). The codebook is provided in Additional Table 1. A list of experts in the field of sexual and reproductive health (SRH), programme design, realist evaluation, and adolescent health was prepared based on identified literature. We approached 12 experts via mail for their involvement in the feedback round, 7 of whom agreed.

Using the codebook, we extracted data on the study context, intervention details, strategies, implementation techniques, outcomes, and possible mechanisms that explained how different intervention modalities from the studies retrieved in stage 1 may be linked to reducing adolescent pregnancy. We then mapped the extracted information onto an initial intervention framework, outlying the aggregated information with the outcome of interest (see Fig. 2). The context was divided into five categories based on the ecological framework [38], and the intervention was separated into two parts, intervention strategies, and implementation components (hereon called mechanism resources). The implementation components were organised using the TIDieR checklist [39]. The mechanisms and outcomes represent a composite of data presented in the intervention documents. The framework is only a representation of the granular overview of the components and will aid the development of IPTs [37, 40]. Based on the extracted CMO data and the framework, eight IPTs were proposed (Fig. 3). These IPTs were further revised by the review team and experts in two online meetings (questions for the experts can be found in the Additional file).

Fig. 2

Initial intervention framework

Fig. 3

Initial programme theories

We observed that different resources were often used in combination with each other and elicited different mechanisms to impact adolescent pregnancy. For example, financial and life skill training in disadvantaged communities often used hands-on techniques in a safe space with peer interactions to elicit different mechanisms. While learning new life skills like negotiation or management increased the participant’s capability and motivated them to apply these newly acquired skills in daily life, learning financial skills triggered the accumulation of human capital and decreased dependence on men. These mechanisms may increase participants’ aspiration to look for alternatives to marriage and transactional sex and thereby decrease the risk of adolescent pregnancy. Another example is the economic strengthening applied in schools and communities in resource poor settings with high levels of poverty. These interventions integrated resources like interactive games and provided cash transfers to the family. For participants from socioeconomically low-income families with low literacy, cash transfers increased available funds, increasing their family’s flexibility to invest in education. Prolonged education triggered the increase in decision-making capacity and agency. Both these mechanisms can lead to a feeling of self-reliance and an increase in bargaining power and thereby lead to changes in risky sexual behaviour, delay marriage, and subsequently lower the risk of pregnancy. Figure 3 illustrates several pathways in which interventions may influence our outcome of interest. However, these are only the preliminary findings and will undergo further changes as the review progresses.

Stages 2 and 3 — Search for evidence and appraisal of evidence

In the second stage, we developed a search strategy based on the critical components of the research question: adolescent population, adolescent pregnancy, gender transformative programmes, and LMICs (Additional Table 2). In this review, the primary outcome of interest — adolescent pregnancy — was defined per the threshold set by the World Health Organization as ‘pregnancy in a female adolescent under the age of 20’ [41]. This includes intended, unintended, and unwanted pregnancies. Further, we define gender transformative approaches as ‘programmes and interventions that foster critical examination of gender norms, create opportunities for individuals to actively challenge gender norms, promote gender equality, and address power inequities between persons of different genders’ [42]. All studies included in this review will be assessed based on this definition to be qualified as ‘gender transformative’ (see Fig. 1 of Additional file).

We will examine GTAs targeting the adolescent population (10–19 years) at any individual, group, or community level. In cases where the target population includes older participants, documents will be selected if results are disaggregated by age or if the majority of the sample comprises adolescents. All types of studies and documents describing interventions (i.e. non-experimental, experimental, quantitative, qualitative, mixed methods, implementation reports, project reports, policy briefs, blog posts) except systematic reviews, which examine a gender transformative intervention with adolescent pregnancy as one of the outcomes and that was conducted in an LMIC context, will be included. Interventions targeting both adolescent girls and boys as study participants will be considered. Only documents that provide a link between an intervention and the outcome of interest will be included. Documents published in English only will be included. The authors note that while translation software may support screening of non-English languages, data extraction in a non-English language is best conducted by a fluent speaker of that language, and the benefit of including them may be outweighed by the significant resources required [43].

The main aim of the systematic search is to identify additional evidence that will be eligible to refine, refute, or confirm the initial programme theories. The databases in our scope include (but are not limited to) the following: Embase, MEDLINE, ERIC, PsycINFO, CINAHL, Gender Studies Database, Reproductive Health Library, Studies in Family Planning, Reproductive Health Matters, International Family Planning Perspectives, and Population and Development Review. Additionally, grey literature will be included by locating resources on websites including OpenGrey, Advocates for Youth, Family Health International, Guttmacher Institute, Interagency Youth Working Group, International Center for Research on Women, Pathfinder International, Population Council, United Nations Population Fund, United Nations Children’s Fund, World Health Organization, and USAID. Based on realist methodology, the literature search process will be iterative and will embrace forward and backward citation tracking and contacting authors in cases where we need more information.

Using the above search strategy, we will pilot a standardised title and abstract screening form and full-text screening form to conduct a pilot exercise on a sample of 10 abstracts and 5 full-text articles to calibrate and test the review forms. Conflict will be mediated by another review author not involved in the screening process. We will use Rayyan, a free web tool designed to help researchers working on knowledge synthesis projects for deleting duplicates and conducting title and abstract screening [44]. Data on CMO configurations will be extracted using the codebook mentioned in stage 1, piloted by three authors using five articles to ensure codebook consistency. We will compare the results of the three authors and modify the codebook accordingly.

For the appraisal process, each article will be assessed based on relevance (whether it fits the inclusion criteria), richness (whether it can contribute to IPT building), and rigour (whether the information produced is credible and uses trustworthy methods) [32]. No article will be excluded or included based on just one criterion but on the overall value added to the research question (Fig. 4). The documents rated as ‘high’ according to our appraisal will be extracted first followed by those rated medium and low [45]. The authors will meet regularly to discuss their findings and modify the IPT.

Fig. 4

Appraisal tool — relevance, richness, rigour [46]

Stages 4 and 5 — Synthesise evidence to develop refined programme theory and disseminate recommendations

In the fourth stage, the extracted data on context, mechanism and outcome underlying the interventions of each document will be analysed. We will delineate CMOCs as ‘if…then…because’ statements. Similar CMOCs will be combined to construct richer CMOC statements, differentiating various contexts and mechanisms [47]. Finally, we will iteratively refine the earlier proposed IPTs into programme theories (PTs) by testing them against the comprehensive CMOCs. Each step of this process will be discussed by members of the review team. The PTs will be sent to the experts for their feedback, following which another cycle of refinement may be required. In the final stage, the refined programme theories will be shared with the broader audience via academic papers in open-access journals, conference and institutional presentations, and policy recommendations.