• Our primary objective was to systematically analyze existing literature on the interventions to increase BC screening (participation in BC screening programs, mammography uptake) and, where possible, to perform meta-analyses of the effectiveness of these interventions, by type of intervention implemented, underserved status of the target population, and other relevant contextual variables.

The protocol of this systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO registration number: CRD42023393352). The study adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines for transparent and comprehensive reporting [19].

In this review, the Population, Intervention, Comparator, Outcome, and Study design (PICOS) system was used to develop the literature search strategy [20].

• Population: Apparently healthy or asymptomatic females in the target age groups (which may differ in different countries) for BC screening. Both the general population and underserved groups are eligible for inclusion. Underserved groups are identified by characteristics such as low-SES, underinsurance, immigrant background, ethnic minority status, or cultural/religious barriers, present in the majority (>50%) of the study population.

Exclusion criteria included non-English language studies, studies that did not investigate outcomes related to mammography uptake, and studies focused on breast cancer survivors or other individuals not considered healthy. Research papers that concentrate on treatment conditions, rehabilitation, non-intervention studies, biomedical or treatment research, pharmaceutical research, descriptive research, and those lacking valid outcome measures were also excluded.

The following databases were searched: Cochrane Central Register of Controlled Trials, Medline (via Ovid), Global Health (via Ovid), Biological Abstracts (via Ovid), Scopus, Web of Science citation indexing, Google Scholar.

This review aimed to fill the temporal gap left after Agide et al.’s groundwork that included 22 studies published between 2004 and 2016 [21]. These reports were assessed for eligibility according to our selected inclusion criteria. Ultimately, 20 of these studies were included in our analysis. In order to address the temporal gap left, our investigation was extended to include all eligible papers published between 2017 and 2023. The full search string is available in Supplementary Table S1.

Articles retrieved from the electronic databases were exported directly as MS Excel files and imported into Rayyan software for de-duplication and screening [22]. At least two independent reviewers screened the abstract (AF, ARA) and full-text articles (AF, ARA, SO, NY, DJH). The review process is presented in the PRISMA Flow chart (Figure 1).

Prior to the actual data extraction two independent reviewers (AF, SO) developed the data extraction form. This was adapted from the Cochrane Effective Practice and Organisation of Care (EPOC) guidelines [23]. The form underwent pilot testing, following which all relevant data were extracted in duplicate. The completion of the data extraction forms involved at least two independent review authors (AF, ARA, SO, NY, DJH). Any conflicts that arose during the review process were resolved through consensus. A third author (GVH) was consulted for arbitration when consensus was not reached.

To appraise the quality of the randomized studies included, the Risk of Bias tool (RoB-2) for randomized controlled trials (RoB-2), developed by the Cochrane Collaboration, was used [24]. To appraise non-randomized studies, the Risk of Bias in Non-randomized Studies of Interventions (ROBINS-I) tool was used [25]. The risk of bias for each eligible study was assessed by at least two independent review authors (AF, DJH, WD). Judgments were categorized as “low risk,” “some concerns,” “high risk” for each study appraised using RoB-2 tool, and “low risk,” “moderate risk,” “serious risk,” “critical risk” for each study appraised with the ROBINS-I tool. Judgments were categorized as “No information” when details about the methods were insufficient, lacking, or not applicable. All conflicts were resolved by consensus, and a fourth author (GVH) was invited to arbitrate where necessary.

For studies examining categorical outcomes, we extracted proportions, percentages, and unadjusted odds ratios (ORs) along with 95% confidence intervals (CIs). When available, adjusted odds ratios (aORs) and their 95% CIs from multivariable analyses were also extracted.

A meta-analysis of randomized studies was conducted using MetaXL Version 5.3 (EpiGear International 2016) [26]. These findings were visualized using forest plots.

To account for study-level variability in the outcomes, random-effects meta-analysis was used [27]. All estimates were pooled and presented as OR (95%CI) and p-values.

To assess how much confidence to place in the findings on the effectiveness of the interventions, the ‘Grading of Recommendations, Assessment, Development, and Evaluation’ (GRADE) approach was used by the review authors [30].

Of the 53 studies, 44 employed randomized designs [31–72, 82, 83] while 9 used non-randomized designs [73–81].

Several studies were conducted in the United States (n = 28) [31, 38, 40, 41, 43–46, 48, 52, 55–59, 62, 64, 67–69, 72, 73, 75, 78–80, 82], Canada [42, 54], and Australia [36], as well as in Europe, including Italy [66], Spain [61, 65], France [37, 49], Belgium [50], Latvia [81], UK [34, 35, 39], Ireland [76]. Furthermore, studies were conducted in the Middle East including Iran [33, 51, 60, 70], Türkiye [71, 77], Saudi Arabia [32], and Israel [74], and East Asia, including South Korea [53] and Taiwan [83].

Of the 54 included studies, 26 focused on underserved groups, including ethnic minorities (n = 15) [43, 44, 48, 52, 56, 59, 64, 67, 68, 72–75, 78, 79], immigrant communities (n = 6) [36, 44, 53, 56, 71, 77], with low-literacy or other cultural or religious barriers (n = 5) [44, 47, 72, 74, 77], with low-SES (n = 4) [31, 39, 44, 82], uninsured or underinsured (n = 4) [69, 73, 79, 80].

While most studies focused on populations falling in the target age groups commonly identified for BC screening of 45–74 years [84], 24 studies also included younger individuals (from 18 years on) [32, 33, 38, 43, 45–48, 51–53, 56, 57, 60, 64, 67, 68, 71, 72, 74, 75, 77, 79, 80] and 3 studies included older individuals (up to 84 years) [56, 57, 69].

Most studies followed-up participants for a period comprised between 1 week and 3 months (n =19) [32, 33, 36, 39, 40, 47, 48, 50, 51, 56, 58, 60, 62, 65–68, 79, 83] or between 4 and 6 months [34, 35, 42, 53, 54, 61, 62, 71–75, 77, 81, 82] (n =15). However, there were also studies with a follow-up period comprised between 7 and 12 months [37, 38, 41, 43, 45, 48, 49, 66, 67, 69, 76] or of over a year [44, 52, 57, 59, 64].

The complete list of study characteristics is shown in Table 1.

In total, 14 types of interventions aimed at increasing BC screening uptake were identified.

In particular, 25 studies employed navigation services [36, 39, 41, 46, 48, 53, 56–59, 62–64, 69, 72, 73, 75, 77, 78, 80, 83], 16 studies employed educational interventions [32, 43, 44, 51–53, 60, 64, 68, 70, 72, 75, 77–80], 13 employed telephonic interventions [31, 36, 39, 40, 46, 47, 54, 57, 63, 74, 81–83], 12 employed reminders [34, 36, 39, 40, 54, 57, 62, 64, 71, 73, 80, 83], 10 employed invitation letters [34–36, 42, 49, 50, 57, 71, 76, 83], 8 employed decision aids [37, 38, 40, 45, 61, 65–67], 7 employed linguistically adapted strategies (e. g., using materials translated into different languages, using culturally appropriate language or employing communication strategies effective for speakers of a particular language or dialect) [36, 44, 48, 55, 68, 72, 73], 7 employed culturally sensitive approaches (e. g., tailored to align with the values and religious beliefs and practices of a specific community or group) [43, 44, 47, 52, 72, 74, 78], 5 employed digital-based interventions (e. g., web, DVD, other platforms) [41, 48, 55, 71, 82], 4 employed smartphone based-interventions (apps/SMS/social media campaigns) [56, 63, 80, 81], 3 employed printed materials (card, brochure, leaflet, flyer) [37, 57, 69], 2 employed health consultations with physicians [58, 73], 2 gave economic support or vouchers for testing [69, 73] and 2 asked for the mediation of a family member [56, 79].

Of the 23 studies describing interventions based on a specific conceptual framework or model, 9 employed the Health Belief Model (HBM) [32, 40, 51, 56, 60, 70, 71, 74, 82]. The following most common were the Transtheoretical Model (TTM) (n = 5) [33, 40, 69, 74, 82], the Theory of Planned Behavior (TBP) (n = 3) [40, 70, 79] and the Social Cognitive Theory (SCT) (n = 3) [46, 48, 78]. Interventions were also based on the Nudge approach [66, 81], Theoretical frameworks of decisions aids [66, 67], Volunteerism [78], the Cognitive Dissonance Theory [78], Culture-Based Health Beliefs [47], the Social marketing theory [72], the GATHER Consultancy Technique [60], Elaboration Likelihood Model [79], the Exemplification theory [67], a Cultural Competence approach [74], the Health Promotion Model [71], the Social norms theory [49], the Self-regulation model of Health Behavior [38], Motivational interviewing [32] and a Loss-framed approach [31].

Among randomized studies appraised with the ROB-2 tool (Figure 2), 12 were judged to be at low risk of bias [34–36, 40, 45, 46, 52, 62, 63, 65, 82, 83], 27 yielded some concerns [32, 33, 37–39, 41, 43, 44, 47–49, 51, 53–55, 57–60, 64, 66–72] and four had a high risk of bias [31, 56, 61, 71]. In particular, one study [71] had a high risk of bias due to deviations from intended interventions, two studies [56, 61] had a high risk of bias due to bias in measurement of the outcome, and one study [31] had a high risk of bias due to both deviations from intended interventions and bias in measurement of the outcome.

Among non-randomized studies appraised with the ROBINS-I tool (Figure 3), no studies had a low risk of bias, two yielded some concerns [74, 80], five had a serious risk of bias [73, 75, 76, 78, 79] and two had a critical risk of bias [77, 81]. Because of the frequent serious and critical risk of bias detected among non-randomized studies, as well as the high variability in terms of study design, settings and populations, a meta-analysis of non-randomized studies was not performed.

The full risk of bias assessment is available at Supplementary Figures S14, S15.

The pooled effects of interventions to increase the uptake of BC screening overall, and among underserved groups (as a subgroup), are presented in Figures 4, 5, respectively. Randomized studies investigating the effect of interventions to increase mammography uptake [31–35, 37, 39, 40, 42–44, 46, 48–54, 56–62, 65–67, 71–79, 81–83], repeated mammography [70], and combined indicators for either scheduled or obtained mammography [36, 47], were included in the meta-analysis.

The GRADE assessment revealed moderate certainty in the effectiveness of interventions for increasing BC screening uptake. The certainty was downgraded due to unexplained heterogeneity. The summary table is available at Supplementary Table S2.

This review highlights the significant impact of interventions for increasing mammography uptake. Overall, these could increase uptake by 55%, with an even greater effect of 80% among underserved groups. This indicates that while underserved groups participate less in cancer screening, these groups have great potentials for improved participation.

Educational interventions emerged as the most effective strategy for increasing BC screening uptake, nearly tripling the odds of participation overall. These interventions were primarily delivered in person through lectures, workshops, and events. For example, a study featured a “Pink Party” (a community event in a festive setting) where participants learned about the importance of BC screening from BC survivors and medical professionals. Encounters with trained community members, such as promotoras, also offer personalized education and support, addressing individual concerns and encouraging health behaviors based on social norms. Additionally, digital approaches, such as videos shown in hospital waiting areas, help explain the BC screening process and improve comfort levels in clinical settings.

Navigation services have also proven highly effective, doubling the odds of BC screening uptake both overall and among underserved groups. These services offer comprehensive support to guide individuals through the healthcare system. For instance, navigators build collaborative relationships with participants, identify their needs, and work to overcome barriers. If transportation is an issue, the navigator can arrange for transportation support. Navigators might also accompany participants to their screenings when necessary. In another example, navigators provided participants with educational materials and coached them on questions to ask their healthcare providers, enhancing their confidence and engagement in the screening process. Indeed, other types of communication strategies about the BC screening program directed towards the target group proved highly effective.

Especially among underserved groups, reminders and telephonic interventions were highly effective, increasing the odds of participation by up to three times. For example, a follow-up reminder letter with a set date can be sent a few weeks after the initial invitation if no appointment has been scheduled. Reminders can be personalized by being sent in the participant’s preferred language, including a quote or signature from their own GP or other referral healthcare providers, or by providing simplified information about the BC screening process. Telephonic interventions also offer a personalized approach to encourage mammography uptake. In one study, participants received calls where trained healthcare workers employed a loss-framed strategy, highlighting the risks of undetected malignancies and emphasizing the high efficacy of mammograms in early detection. For instance, a caller might say, “By getting a mammogram now, you can catch any potential issues early, significantly increasing the chances of successful treatment.” Telephonic interventions can also lead to more comprehensive navigation services, where callers assist with appointment scheduling, address concerns, and coordinate other necessary services.

Finally, community-sensitive approaches tailored to the values, religious beliefs, and practices of specific groups have also proven highly effective. For example, religious institutions can play a role through community health advisors who provide culturally sensitive educational materials on BC screening. In a study involving Arab women, phone calls were used to address cultural barriers. To counter the belief that cancer is an immutable fate, callers referenced religious teachings that emphasize personal health responsibility, stating, “Both Muslim and Christian teachings highlight the importance of taking care of your health. Early detection through mammography can save lives.” These culturally sensitive conversations also tackled practical concerns, such as fear of pain, by explaining that discomfort lasts only a few seconds and suggesting scheduling the mammogram a week after menstruation to reduce sensitivity. Additionally, callers addressed concerns about the gender of the healthcare provider, reassuring participants that many facilities offer the option to choose a female doctor if preferred, thus respecting personal comfort and cultural preferences.

Evidence suggests that interventions are often more effective when implemented together [85]. By integrating the strategies highlighted in this article, multiple barriers could be addressed simultaneously, ensuring participants receive comprehensive support throughout the breast cancer screening process. For example, a program might begin with an invitation letter, followed by a reminder letter in the participant’s preferred language. If no response is received, a navigator could then follow up by phone, offering to schedule the appointment and providing logistical support as needed. However, further research is warranted to assess the combined effectiveness of these strategies and to determine their impact on BC screening rates.

While the majority of studies included in this meta-analysis did not explicitly assess the cost-effectiveness of the interventions, this aspect was occasionally discussed in a speculative or theoretical manner. For example, interventions requiring minimal personnel involvement, such as paper-based or digital letters and reminders, were often considered more likely to be cost-effective due to their low operational costs. Future research should explicitly evaluate the cost-effectiveness of these strategies to better inform their practical application and potential for widespread adoption.

This study comprehensively reviews a broad range of interventions, providing a holistic view of strategies to increase BC screening uptake. By specifically examining the impact of interventions on underserved groups, the study addresses a critical gap in the literature, providing actionable insights for healthcare providers and policymakers.

The substantial sample size across the studies included in this review (161,141 individuals, with 14,720 from underserved groups) adds robustness to the findings and enhances the generalizability of the results.

According to the GRADE assessment, the findings are attributed with moderate certainty, indicating that the results are generally reliable. However, the study findings also exhibit unexplained heterogeneity, which is defined as the presence of variation in true effect sizes across different studies. The inclusion of studies conducted in diverse populations and settings, along with the use of comparators other than usual care, might contribute to this heterogeneity, necessitating cautious interpretation of the results.

To investigate the impact of specific studies, leave-one-out analyses were conducted. Three main outliers were identified. The studies by Alizadeh-Sabeg et al. [33] and Kim et al. [53] had minimal impact due to their low relative weight. In contrast, excluding the third outlier, Beauchamp et al. [36] (B), slightly lowered the overall effect size, suggesting some sensitivity to individual studies, without reducing heterogeneity. Sensitivity analyses excluding studies with a high risk of bias largely confirmed the reliability of the study results. A random-effects model was applied to account for unexplained heterogeneity.

Our review aimed to fill the temporal gap left after Agide et al.’s groundwork that included 22 studies published between 2004 and 2016 [21]. These reports were assessed for eligibility according to our selected inclusion criteria. Ultimately, 20 of these studies were included in our analysis, along with an additional 34 studies published between 2017 and 2023. While this study focused on the type of study by setting (individual-based interventions, community-based interventions, and interventions via religious, cultural promoters and lay workers), our study investigated the effectiveness of interventions by format and content (educational interventions, telephonic interventions, navigation services, linguistically adapted interventions, culturally sensitive interventions, digital-based educational intervention, smartphone-based interventions, decision aids, invitation letters, reminders, health consultations with physicians, printed materials, economic support, family member mediation), providing a more granular understanding of which specific strategies and formats are most effective in increasing participation, especially among underserved communities.

To our knowledge, this is the first recent meta-analysis on interventions to improve mammographic screening with emphasis on a diverse spectrum of underserved groups (eg. individuals with low-SES, underinsured or uninsured, with an immigrant background, belonging to an ethnic minority, or facing possible cultural or religious barriers). In fact, while other reviews on the topic have been published, these would focus exclusively on specific subgroups such as ethnic minority women [86], or women living in low-income and middle-income countries [87]. Our findings, in conjunction with insights from these existing resources, can guide the development of comprehensive strategies to enhance mammographic screening participation among underserved groups.

This review is part of the project “ENTER: Equity in breast cancer screening in Flanders” [88]. In line with its overarching goal of addressing disparities in BC screening participation among women with low socioeconomic status in Flanders (Belgium), a practical objective of this review was to pinpoint interventions - and their features - that could be seamlessly integrated into the existing Flemish BC screening program, considering its logistical, contextual and social context.

Considering costs, practicality and impact, a reminder letter with culturally and linguistically adapted content emerged as the most promising option for tryout in the ENTER project and is in the meantime being implemented in a pilot RCT.