The study protocol was registered in the PROSPERO (International Prospective Register of Systematic Reviews) database (CRD42021240578) [16]. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 statement was adapted to report this systematic review of reviews [17] and is presented in Multimedia Appendix 1 [17].

A systematic literature search was performed on 7 databases for all publications up to March 2021: Ovid MEDLINE, Ovid Embase, CINAHL, PsycINFO, Web of Science, the Cochrane Database of Systematic Reviews, and the Database of Abstracts of Reviews of Effects. The Peer Review of Electronic Search Strategies checklist [18] was used to guide the development of the search strategy. The electronic search strategy was initially developed in MEDLINE by a reviewer (THD) and was then peer-reviewed by a group of experts in relevant fields (KB, MA, NW, PPJ, and PS) and a librarian to ensure its comprehensiveness. The search strategy combined controlled vocabulary and keywords, including synonyms, antonyms, and acronyms related to adherence, intervention, and cancer, and was adapted for each database. We did not limit the publication date but limited the search to the English language, human studies, and reviews only (refer to Multimedia Appendix 2 for full search strategies).

In addition to the database search, bibliographies of selected studies were also hand-searched to identify relevant studies not detected by the electronic search.

The studies had to meet all the following criteria to be eligible for inclusion:

One reviewer (THD) conducted the searching, deduplication, and initial screening of titles and abstracts of all studies found. A second reviewer (ARMF) conducted a random independent assessment of the identified papers and reviewed the screening results of the first reviewer. Two reviewers independently screened all full texts of potentially eligible papers. When necessary, any differences between the 2 reviewers were discussed until consensus was reached or resolved by a third reviewer. Covidence site, operated by Veritas Health Innovation Ltd [19], was used for data screening, selection, and management.

The methodological quality of the included systematic reviews was independently assessed by 2 reviewers (THD and ARMF), adapting the A Measurement Tool to Assess Systematic Reviews (AMSTAR) 2, which has demonstrated satisfactory reliability and construct validity [20]. AMSTAR 2 is a tool used to evaluate the methodological quality of systematic reviews, which includes randomized and nonrandomized studies of interventions, including 10 domains and 16 items or questions. The answering options were yes, partial yes, or no/no information (corresponding to low or high risk of bias). We used the findings from the AMSTAR 2 critical appraisal to understand the certainty of the evidence base of the systematic reviews. Disagreements were resolved by discussion.

As AMSTAR 2 does not combine individual item ratings to create an overall score, the scheme for interpreting weaknesses detected in critical (7) and noncritical (9) items, proposed by Shea et al [20], was applied. The overall confidence in the results of the review was classified as high, moderate, low, or critically low, according to the number of critical and noncritical weaknesses identified in the systematic review under appraisal.

Data were independently extracted by 2 reviewers (THD and ARMF) in a standardized table (Multimedia Appendix 3), which was pilot-tested, for 7 random eligible studies, and then, results were compared and agreed upon. THD extracted data for the remaining eligible studies, which were then reviewed by ARMF, with discrepancies resolved through consensus. The corresponding authors of the included studies were contacted for further information or clarification, if necessary. The extracted data included the type of review, research questions, type of interventions, search strategies, search period limits, characteristics of included studies, quality assessments, methods of analyses, and findings. The included reviews were expected to have high heterogeneity in terms of interventions, comparators, outcome measures, study populations, and methodologies. Therefore, statistical pooling through meta-analysis was not appropriate.

The results of this review are presented in the following order: search results, characteristics of included reviews, quality of systematic reviews, description of interventions, and outcomes of included reviews by group—scoping, systematic, and literature reviews. Owing to the heterogeneity of the included reviews, the findings are presented in a narrative format and refer to meta-analyses performed by the authors of the included reviews whenever available.

The search strategy identified 2098 unique results from 7 databases, 1 from the reference lists of the included studies, and 1 from the automatic alerts of the databases. Title and abstract screening identified 51 studies for full-text screening, of which 29 (57%) met the inclusion criteria. Details of the excluded studies and reasons for exclusion are provided in Multimedia Appendix 4 [2,7-9,12,21-46]. A high level of concordance was achieved between the 2 reviewers in the screening process, with disagreement in only 10% (5/51) of cases. These 5 papers were discussed by the 2 reviewers, and consensus was achieved. The selection process is illustrated in the flowchart (Figure 1).

Among the 29 reviews, 12 (41%) were systematic reviews, of which only 1 (3%) conducted meta-analyses (Multimedia Appendix 5). Methodological quality was low or critically low overall, with at least 2 out of 16 AMSTAR 2 appraisal items [20] not met in all systematic reviews. The quality assessment of the 7 critical AMSTAR 2 domains is presented in Multimedia Appendix 6. Only 8% (1/12) systematic reviews from the study by Arthurs et al [21] received moderate overall confidence ratings in the reported results, which meant that this systematic review may provide an accurate summary of the results of the included studies to address the questions of interest. Moreover, 33% (4/12) and 59% (7/12) of systematic reviews received low and critically low overall confidence ratings, respectively, meaning that the summarized results of these studies may be inaccurate and that the conclusions need to be interpreted carefully. The best adherence was found for using the components of the PICO (population, intervention, comparison, and outcome) framework when describing the search question and inclusion criteria (item 1) and describing the included studies in adequate detail (item 8). The item that most reviews (9/12, 75%) failed to meet was providing a justification for excluding individual studies (item 7). For the critical domains, 42% (5/12) reviews referred to a review protocol (item 2), and 25% (3/12) reviews provided a list of excluded studies and justified their exclusion (item 7). Nearly all reviews (11/12, 92%) accounted for risk of bias when interpreting the results (item 13). Most of the reviews (10/12, 83%) used a satisfactory technique for assessing the risk of bias in individual studies (item 9), and 75% (9/12) of reviews conducted a comprehensive literature search (item 4). The only review with meta-analyses adhered to the item of using appropriate methods for statistical combination of the results (item 11) and investigated publication bias (item 15). More details on the bias assessments of all 16 AMSTAR 2 items are provided in Multimedia Appendix 7 [2,7-9,12,21-27].

This overview of reviews aimed to synthesize evidence from available reviews on interventions to improve MA to OACMs in adults with cancer. To the best of our knowledge, this is the first study to achieve this goal. Among the 29 included reviews, only 1 (3%) conducted meta-analyses and 17 (59%) did not follow systematic methodologies in identifying, analyzing, and reporting literature. Consequently, it was impossible to perform quantitative analyses. Nevertheless, including literature reviews in the narrative synthesis is useful for understanding the breadth of the study field. The only systematic reviews of moderate quality focused on therapeutic patient education interventions in ambulatory care settings [21]. The other 11 systematic reviews on the topic of interest had low or critically low confidence rating. Therefore, the results of the included reviews should be interpreted with caution.

The comparability of the study results is limited because of the high heterogeneity of the included reviews (Multimedia Appendix 5) and studies within each review [9,13]. The content of adherence-enhancing interventions is varied [29]. In addition, there are differences in the characteristics of patients whose adherence has been influenced [9]. Furthermore, comparability is constrained owing to different adherence measurements [2,7,8,52]. Accordingly, this review summarizes the main themes of the included reviews rather than comparing them.

This review suggests that single strategies to promote adherence (eg, education, reminders, or monitoring) are not sufficient to improve adherence. Multidimensional interventions that used collective strategies to promote adherence (education, reminder, cognitive, behavioral, and affective) were potentially more effective. Our findings are in line with earlier reviews of interventions to improve adherence in various chronic conditions [45,46] and those focusing on cancer [9,50,54,58]. These findings also resonate with the report of the World Health Organization that MA is a multidimensional phenomenon determined by 5 dimensions (social and economic, health system, condition-related, therapy-related, and patient-related) [70]. Thus, multicomponent interventions applying different strategies are needed to address the multifaceted adherence phenomenon [70].

The described theoretical frameworks were neither clear nor validated. One-third of reviews reported on the scattered use of cognitive and behavioral theories in only a few studies [12]. Although the authors [12,48] emphasized the importance of using theoretical grounding in planning, designing, and evaluating outcomes of multilevel interventions, a few [13,23] argued that the effect of this was quite modest. This uncertainty is in line with a meta-analysis of 683 studies that quantified the impact of theory-driven interventions on adherence [71]. The limited use of theory to design interventions means that no conclusions can be drawn regarding the importance and effectiveness of theoretically derived interventions. Furthermore, the complex and multifaceted factors contributing to nonadherence represent another challenge in the selection of an appropriate conceptual model to design interventions. Perhaps, a combination of theories may better explain the diverse barriers and facilitators of MA and provide a stronger direction to formulate interventions. Future research should pay more attention to this aspect of adherence interventions.

The use of digital solutions to enhance adherence to cancer treatment has been increasing in the past decade [47,55]. The literature has emphasized the potential of digital platforms to facilitate oral antineoplastic adherence among people with cancer [11]. Medication nonadherence can be intentional or unintentional. Intentional nonadherence is a patient’s conscious decision not to take a drug, for example, because of unpleasant side effects [72]. Unintentional nonadherence is unplanned by a patient, for example, because of forgetfulness [73]. Therefore, the interventions require different modes of action. A variety of measures, such as patient education and good patient-provider communication, can enable patients to better report and manage therapeutic side effects [55]. Technologies can enhance these measures by providing patients with rapid, continuous, and easy access to both educational resources and symptom self-management strategies, also facilitating communication between patients and their care teams [11,55]. Personal lifestyle and electronic triggers (eg, SMS text messages) remind and motivate patients to take their medication, so that it becomes an integral part of their daily activities [11,52]. In both cases, digital platforms (eg, mobile apps) can enable real-time monitoring of patient self-management [11]. However, this is an emerging field, and most studies have focused only on evaluating the acceptability, usability, and feasibility of interventions. The effectiveness of digital MA interventions in clinical oncology practice is poorly supported [47,48]. Future research should not only focus on determining the effect of digital interventions on adherence but also on identifying barriers to delivering high-quality personalized care to end users [11].

Using frameworks and engaging stakeholders in the design and development of digital interventions is crucial. Design frameworks help in planning the resources needed for each stage of the design and to mobilize them effectively and efficiently [48]. The involvement of stakeholders is central in ensuring that the intervention meets the needs of the target audience and in increasing its sustainability [49]. Nevertheless, strategies involving stakeholders (eg, patients, caregivers, oncology clinicians, nurses, pharmacists, and the community) have rarely been reported [48]. The involvement of professionals in the intervention development processes was very limited [49,55]; only 2 studies [74,75] mentioned patients’ and clinicians’ participation. Most interventions did not use or, at least, did not report the use of design frameworks in the development processes [48]. Given the rapid increase of technology applications in MA and the importance of this aspect in intervention development, it is worthy of future research into the involvement of stakeholders and the design framework used in the development of adherence interventions.

The findings from this review show that the use of digital solutions alone may be insufficient and may require cultural adaptive change [57]. Health care professionals’ interaction with patients is pivotal to augmenting the effect of these interventions [51]. Nurses and pharmacists are uniquely positioned to promote adherence to oral cancer therapies [10,51,53]. Findings suggest that clinical support (eg, tailoring education to meet patients’ needs) and symptom assessment and management provided by nurses empowered patients’ ability to adhere to treatments [8,28,55]. Future interventions in cancer should maximize the advantage that health professionals can contribute to patients’ MA with the support of digital technology.

Finally, this review suggests that to consolidate evidence on the effects of MA interventions in cancer, further work is needed using rigorous methods, such as prospective randomized designs in large samples of patients. Study outcomes should not only be limited to adherence rates but also the long-term effects of interventions and meaningful clinical outcomes, such as decreased symptoms and adverse effects of therapy, inhibited disease progression, and increased patient survival and quality of life. These suggestions are consistent with the results of some other systematic reviews of interventions to promote adherence to OACMs that have been published to date [2,9].

This review has inevitable limitations owing to the limited existing high-quality quantitative analytic evidence, which also demonstrates a high risk of bias. Similarly, the significant heterogeneity across and within reviews and studies did not allow statistical analyses beyond reporting of results from the only meta-analysis and narrative analyses performed by the authors of the included reviews. Throughout the process, we relied on published evidence rather than aggregated data from individual studies. Therefore, a definitive assessment of the overall strength of evidence and the effectiveness of current interventions to enhance adherence to anticancer medicines among adults with cancer is not possible. Finally, only the reviews published in English were included. Thus, there is a risk of missing the relevant literature published in other languages. However, comprehensive searches were conducted using different databases to minimize this limitation as much as possible.

Despite these challenges, this review suggests the potential effectiveness of multicomponent interventions to promote adherence to OACMs in adults. This review highlights the role of digital health in enabling and enhancing multicomponent adherence interventions. Nurses and pharmacists are in unique positions and play an important role in facilitating and motivating patient adherence behavior in oncology treatments. These processes can be facilitated without creating a burden if they are integrated into the current routine practices with the support of technology. The findings from this review support the need for future research in developing evidence-based digital multicomponent interventions to assist people with cancer in adhering to their oral therapies. This review also underscores the importance of stakeholders’ involvement and the use of a design framework in the development of interventions to increase translatability and sustainability in real oncology practices. Given the rapidly increasing use of oral antineoplastic medicines and the dramatic availability of digital tools worldwide, research in this field is expected to increase rapidly.