This systematic review and meta-analysis followed the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement guidelines (Multimedia Appendix 1) [34]. The protocol has been registered at PROSPERO (no. CRD42024550203). There were no deviations from the registered protocol.

The inclusion and exclusion criteria of this study are summarized as follows (Table 1).

Randomized controlled trials (RCTs) and clinical trials (nonrandomized, comparative interventional trials) were included. Review papers, systematic reviews, meta-analyses, cross-sectional studies, case-control studies, pilot studies, feasibility studies, editorials, commentaries, letters, opinion pieces, conference abstracts, gray literature, and non–peer-reviewed sources were excluded.

This study included participants who met the following criteria: (1) patients aged 18 years and older, (2) patients diagnosed with cancer, and (3) patients taking oral SACTs. Patients younger than 18 years, studies that included mixed-age groups of participants, patients with cancer taking nonoral SACTs, and patients with cancer exclusively receiving injectable SACTs were all excluded.

The digital interventions were categorized according to the existing literature and the Cochrane Effective Practice and Organisation of Care (EPOC) taxonomy of health system interventions. EPOC outlined 4 categories of information and communication technology that health care organizations use for managing and delivering health care: health information systems, the application of information and communication technology, smart home technologies, and telemedicine [35].

To improve their adherence to oral SACTs, patients with cancer who used digital interventions, such as mobile apps, web-based platforms, wearable devices, telemedicine interventions, reminder systems (eg, text message reminders), virtual support groups, or web-based communities, were included. Studies using nondigital interventions to enhance adherence were excluded.

As there is no gold standard for measuring adherence and its associated outcomes, studies that reported adherence to oral SACTs, measured by various methods including self-reported adherence measures (such as the Morisky Medication Adherence Scale Score [36]), pharmacy refill data, medication event monitoring systems (including smart pill bottles and electronic pill caps), and biological markers, and presented as continuous or dichotomous data, such as the medication possession ratio [37], the proportion of days covered [37], or the proportion of adherence or nonadherence, were included in this review. Any studies that did not contain outcome measures related to adherence and studies that used adherence measures based solely on subjective reporting (unless validated self-reported measures were used) were excluded.

A comprehensive electronic database search was conducted on MEDLINE, Embase, APA PsycINFO, and CINAHL Plus from their inception to May 31, 2024, as this review began in June 2024. MEDLINE and Embase are widely recommended for studying health care interventions [38], while APA PsycINFO and CINAHL Plus, although narrower in scope, are also well suited for this field. These databases focus on subject-specific rather than population-based information. Although there is no established guideline for the number of databases to include in a search, the combination of 2 broad and 2 focused databases is considered appropriate for the subject area of this review. Various structured search strategies were used, using controlled vocabulary and keywords based on the study’s inclusion and exclusion criteria (Table 1) (Multimedia Appendix 2).

The title and abstract of papers retrieved from the electronic databases search were first screened by 2 reviewers (FA and WCL) independently according to the selection criteria (Table 1) using the predesigned electronic screening form. Each paper was rated as “included,” “further check,” or “excluded.” The intraclass correlation coefficient (2-way mixed-effects model with absolute agreement [39]) and 95% CI were calculated for the consistency between 2 reviewers (FA and WCL) in record screening. Any discrepancy was resolved by discussing between reviewers and, if necessary, with a third reviewer (LCC) to reach a consensus. The full texts of potentially eligible papers were further reviewed independently by 2 reviewers (FA and WCL) to conclude the selection of studies.

The data for each study were independently extracted by 2 reviewers (FA and WCL) using the standardized and piloted electronic data extraction sheet. Disagreements were adjudicated by a third reviewer (LCC). Study information (study title, lead author, country, and year of publication), study design, setting, targeted population (cancer and oral SACT), intervention (digital apps), comparison, outcome measures, and follow-up period were extracted. Study results, including continuous data (such as mean adherence scale score and SD) and dichotomous data (such as the proportion of adherent or nonadherent patients), were retrieved. If raw data are unavailable, risk ratio, hazard ratio, mean (SD), median (range) of adherence duration, or any other results that can be converted into raw data were extracted. Duplicates were identified using EndNote 20 (Clarivate Analytics) through its default 1-step auto-deduplication process, which applies the matching criteria of “author,” “year,” and “title.” This process was used to aid in screening the studies.

Controlling the risk of bias in a systematic review is crucial, as bias can distort the true effect of interventions [40]. Quality assessment of all included studies was conducted using version 2 of the Cochrane Risk of Bias Assessment Tool (RoB 2) for RCTs as recommended by the Cochrane Collaboration [41]. By assessing bias across 5 critical methodological aspects of each RCT, namely, the randomization process, deviations from the intended intervention, missing outcome data, outcome measurement, and selection of reported results [41], the included studies were categorized into “’low risk of bias,” “some concerns,” or “high risk of bias” using the RoB 2 tool. The results were subsequently tabulated. Risk of bias assessment was conducted independently and in duplicate by the 2 reviewers (FA and WCL).

All outcomes were compared between the exposed group (digital intervention users) and the nonexposed group (those receiving standard care). The proportions of poor adherence were synthesized using a random-effects model (Der-Simonian and Laird method [42]). The pooled results were reported as odds ratio and 95% CI. The heterogeneity was assessed with the I² test (%). If appropriate, the mean difference and 95% CI of the adherence scale scores between the exposed and nonexposed groups were calculated and synthesized. The meta-analysis was conducted in STATA (Release 14; StataCorp LLC).

Of the 844 records identified from the electronic databases search, 181 duplicates were deleted. After screening titles and abstracts, 614 records were removed due to the irrelevance to digital interventions in patients with cancer receiving oral SACTs (n=426), being not fully published original interventional papers (n=159), not assessing medication adherence (n=16), involving patients younger than 18 years (n=10), not being in English (n=2), and both arms using digital interventions (n=1). After the full-text screening of the remaining 49 studies, 36 were excluded, leaving 13 studies (2611 participants) for inclusion in this review (Figure 1). The intraclass correlation coefficient between the 2 reviewers (WCL and FA) is 0.886 (95% CI 0.868-0.902), indicating good consistency. Since both authors demonstrated consistency and agreement at the full-text screening stage, the intraclass correlation coefficient was calculated solely for the abstract screening.

The 13 included RCTs, published from 2016 to 2024, were conducted in various countries: the United States (n=3) [21,33,43], South Korea (n=2) [22,24], France (n=2) [44,45], Egypt (n=1) [23], Finland (n=1) [46], Australia (n=1) [47], Colombia (n=1) [14], Singapore (n=1) [48], and Turkey (n=1) [25]. The studies involved patients with breast cancer (n=5) [21,22,24,47,48], various types of cancer (n=5) [25,33,43-45], chronic myeloid leukemia (n=1) [46], colorectal or gastric cancer (n=1) [23], and multiple myeloma (n=1) [14]. Digital interventions included mobile apps (n=4) [22,24,25,43], reminder systems (n=4) [21,33,47,48], telephone follow-ups (n=3) [23,44,45] and interactive multimedia platforms (n=2) [14,46]. According to the EPOC taxonomy [35], 7 RCTs used smart-home technologies [22,24,25,33,43,47,48], 4 used telemedicine [23,33,44,45], and 2 used information and communication technology [14,46] (Table 2). There were 1305 patients in the digital intervention group and 1306 patients in the control group.

The 13 included RCTs raised concerns primarily related to the randomization process, missing outcome data, and outcome measurement; there were no high risks identified in any of the 5 areas of bias. The randomization was conducted by the principal investigator (KM) in one study [46] and lacked blinding in another [23]. In several studies, adherence outcomes were derived subjectively from patient-reported data via self-completed questionnaires [14,22,25,43-48]. In addition, challenges in differentiating missed appointments from actual medication nonadherence [21] and the possibility of smart pill bottles being opened without medication intake [24] further compounded measurement bias (Multimedia Appendix 3).

The challenges in recording outcome measures were found in 2 studies [21,24]. The authors of these RCTs made assumptions about the absence of urine samples as an indicator of nonadherence and the correlation between opening smart bottles and actual medication intake. While both studies used a sampling check or additional survey to support their assumptions, these diverse approaches contributed to increased heterogeneity and potential biases in this meta-analysis.

Four studies used mobile apps to integrate educational materials into their platforms [22,24,25,43]. Although the app (ILOVEBREAST) by Kim et al [22] functioned as a game, it still served as an educational tool for patients. Standard features of these mobile apps include side effects and symptom management [22,25,43], lifestyle guidance [43], and addressing adherence concerns [24,43]. Two of these studies incorporated additional digital technologies into their mobile apps, such as smart pill bottle reminders [24] and integrated Fitbit for monitoring physical activity [43] (Multimedia Appendix 4).

Moreover, standard features across mobile apps and other digital technologies included disease management and patient education about specific cancer types. Three studies directly targeted adherence through their digital technologies, either by questioning patients about their adherence [23,44] or by measuring it [24]. The remaining studies indirectly addressed adherence by focusing on related features. Some text messages covered a variety of content related to not only medication adherence but also physical activity, healthy diet, well-being, side effects management, physician recommendations, and providing support [21,47]. In addition, 3 studies used digital interventions to identify problems, particularly symptoms and toxicities [25,44,45]. In 1 study, health care professionals were able to access patient data and communicate with nurse navigators via a web portal [45] (Multimedia Appendix 4).

The delivery mode of digital technologies in the 13 RCTs varied. Mobile apps involve self-administration by patients, constituting a passive delivery method, although 2 studies personalized the app experience with features such as customized medication dosing timetables and symptom recording [25,43]. Reminder systems, either via text message or phone call, were passively delivered through telecommunication companies [48] or an interactive voice response system [33], with reminders predominantly generic. Telephone follow-ups were tailored to individual patients and proactively delivered by trained nurses [44,45] or a single principal investigator [23]. Interactive multimedia platforms, although passively delivered, provided bespoke content. One study combined multimedia interactive platforms with face-to-face counseling sessions delivered by trained nurses [46] (Multimedia Appendix 4).

The duration of digital interventions in the 13 RCTs ranged from 3 weeks [22] to 3 years [21], with 1 study comprising 44 months in 11 undefined-length cycles [23]. Reminder systems were predominantly weekly, except for some studies conducted daily [33] or biweekly reminders [21]. Several studies used reminder systems to enhance adherence to oral SACTs. These systems varied, with some studies using smartphone messages [25,46,48], smart pill boxes [24], or telephone calls [33] to remind patients about their medication. Mobile apps were recommended for daily [22,24] or weekly use [25], except 1 study with unspecified frequency [33]. Telephone follow-ups varied from weekly [23] to less regular pattern [44,45]. One study combined follow-up phone calls with a web portal for web-based communication and patient information sharing [45]. Multimedia interactive platform engagement varied from monthly [14] to unspecified frequencies [46], with text messages being sent daily in 1 study [46] (Table 2).

Adherence was the primary outcome in 11 RCTs, while 2 studies assessed it as a secondary outcome [45,47]. Various subjective measures, including surveys [14,22,25,43-46,48], relative dose intensity (RDI) [33,45], pill count [23], self-reported missed doses [47], and a smart pill bottle [24], were used across the 13 RCTs. One study used a more objective measure of adherence using time-to-adherence failure, defined by urine aromatase inhibitor metabolite assay results [21] (Table 2).

The pooled result from 11 studies [14,21,23-25,33,43,45-48] showed that users of digital technology had a significantly lower risk of poor adherence to oral SACTs than nonusers (odds ratio 0.60, 95% CI 0.47-0.77; I²=73.1%) (Table 3). A trend was observed where smaller studies favored the digital intervention group [14,25,46], while larger studies favored the control group or showed no significant difference [21,33,43,45,48]. However, definitive conclusions cannot be drawn due to substantial heterogeneity (I²=73.1%) [40]. In 1 study, only the proportion of medium adherence was reported, with no significant difference observed between the intervention (92/183, 77.2%) and control (91/183, 81.3%) groups [44].

Two studies reported adherence scale scores [22,44]. Although the results were not pooled, the mean difference was calculated (Table 4). These 2 studies generated positive mean differences, indicating that digital technology users experienced an increase or improvement in oral SACT adherence compared with nonusers. The mean (SD) of the RDI for the intervention group and the control group were 0.89 (0.03) (n=122) and 0.92 (0.03) (n=117) in one study [33], and 0.84 (0.26) (n=255) and 0.80 (0.21) (n=265) in another study [45]. A value of RDI<0.8 indicated underadherence, as reported in 1 study [33].

This study investigated the efficacy of digital interventions in improving adherence to oral SACTs and found that digital intervention users had a significantly lower risk of poor adherence to oral SACTs than nonusers. In addition, digital technology users demonstrated improved or increased adherence scores compared with nonusers.

Interactive and patient-focused digital supports have revolutionized the possibilities for improving medication adherence [16]. An overview of reviews indicates that incorporating digital technologies with direct clinician contact is likely to increase adherence [31]. A systematic review confirmed the efficacy of digital interventions in improving short-term treatment adherence among patients with cancer receiving oral chemotherapy [32]. Our pooled meta-analysis results also support this, as they showed a significantly reduced risk of poor adherence to oral SACTs among users of digital tools.

The efficacy of digital tools in achieving success can be attributed to various factors, for example, providing instructional resources, dosage aids, engagement with health care providers, digital medicine, self-monitoring, and quickly implementable technical methods [16]. Patient awareness of their drug regimen and the goals, benefits, and potential adverse events is critical for optimal adherence [49]. Digital can offer medication information and instructional help as educational resources [22,24,25,43]. Digital-based interventions such as personalized dosing schedules help patients organize and improve drug adherence [43]. Face-to-face counseling, proposed as a single consultation experience, was also included in our review for its potential to enhance patient adherence [46,50].

Medication adherence is crucial in oncology therapy, yet low adherence rates, as low as 14% for some cancer regimens, significantly impact patient health outcomes and strain health care systems and budgets [51]. This indicates that personalized interventions may improve adherence [51,52]. With more than 4.57 billion web users globally, 91% are accessing it via mobile devices, and smartphone usage—projected to increase by 8% annually [53], as well as digital health tools including phones and wearable devices—offer promising avenues for enhancing health care outcomes, cost-effectiveness, and patient acceptance [27].

Telemedicine offers greater flexibility than in-person interventions, allowing for addressing nonadherence wherever and whenever it occurs, such as between appointments or outside of clinic settings [54]. Telemedicine for reminder and follow-up phone calls was also a method of implementation used in several studies examined [23,33,44]. Digital medicine involves tools such as electronic pill bottles and wearable electronic devices. These devices enhance adherence and can track when containers are opened, although this does not verify intake [55]. Moreover, digital treatments may have drawbacks, including the cost and time needed for transferring or connecting with electronic equipment [16].

One study investigated whether using 1 or 2 digital tools improved adherence [56]. Both groups received weekly automated voice responses over 8 weeks, with the intervention group receiving additional daily text messages for 21‐28 days. Results suggested that the extra text messages improved adherence and symptom management in patients taking oral anticancer agents. Another similar study showed that additional text messages could positively impact patients by promoting behavior change and improving self-care [28]. This highlights the potential for diverse clinical outcomes with varying types and quantities of digital tools.

Furthermore, social inequality is often correlated with the reduced use of digital technology in health care, contributing to a digital health divide [57]. For instance, older adults are less likely to use the web [58] or smartphones [59], and individuals with lower incomes face greater barriers to web access [60]. This inequality results in disparities in access to digital tools and hampers the implementation of digital interventions in health care [61]. To enhance accessibility, patients and health care professionals need to be involved in the development of these interventions, ensuring that they meet the needs of diverse patient populations. In addition, educational campaigns should aim to raise awareness and provide training on digital tools while also challenging stereotypes about older adults’ technological capabilities and reinforcing patients’ confidence in maintaining their privacy when using such interventions [61].

This review focuses on managing medication adherence at home for patients with cancer who are prescribed oral SACTs. All studies included are RCTs, considered the gold standard for measuring intervention efficacy [62]. We excluded single-group pre-post test designs to ensure randomization and aimed to cover various contemporary digital tools to assess their efficacy on medication adherence. One study had a 3-year follow-up, offering valuable insights into long-term impact [21]. The pooled meta-analysis results provide an integrated understanding of digital tools’ efficacy in supporting medication adherence among patients with cancer.

While digital interventions hold promise, we acknowledge several limitations in this study, including various cancer types and oral SACT classes introducing disease uniqueness and drug response variability, potentially impacting medication adherence and intervention efficacy.

Despite including only RCTs, these studies exhibited considerable variability in research design, data collection methods, outcome measures, and the digital interventions used, as well as diversity in the cancer types investigated. The inability to conduct a patient-blinded experiment due to patient expectations of additional digital support is recognized [23,25]. Follow-up phone calls by different health care professionals may introduce bias [44,45]. Furthermore, reliance on subjective self-monitoring or self-reporting for medication adherence evaluation poses potential errors [24,33,46]. Small sample sizes in some trials may limit statistical power and significance between intervention and control groups.

This heterogeneity is inherent to the subject matter [63]. Methodological heterogeneity was notable (I²=73.1%), but it was accounted for by using a random-effects model in the meta-analysis, which assumes a normal distribution of underlying effects [40]. Also, due to the significant heterogeneity, the publication bias assessment test was not conducted to avoid presenting potentially misleading results. Acknowledging these limitations is crucial for interpreting the research results and allows readers to evaluate the significance and scope of the study more comprehensively. Another limitation of the study was that subgroup analyses were not conducted due to lack of data. This could have been used to investigate heterogeneous results or ask specific questions about a cancer type or intervention type.

This review included a variety of adherence and outcome measures due to the lack of consensus on these metrics. While self-reported adherence may be less robust due to recall bias and social desirability effects [64], only those studies using validated tools widely accepted in adherence research were included. Although these tools facilitate low-burden data collection, self-reported adherence may not always accurately reflect actual behavior, necessitating cautious interpretation of results. This diversity in outcome measures provides a comprehensive view of adherence-related consequences, which is crucial for understanding the broader context of digital interventions but may also complicate the ability to draw definitive conclusions.

Cancer populations encompass low-, middle-, and high-income regions globally, each with varying access to digital technologies and health care systems. Most studies have been conducted in high-income regions, which limits the generalizability of the results to low- and middle-income areas. In addition, the limited and diverse regional patient inclusion across these studies may further restrict the applicability of the findings to broader conditions [23-25,43].

Future interventions should be developed that focus on patient-centered, motivation-driven, and culturally adapted digital tools and be tailored for individuals with different types of cancer or oral SACTs. Efforts should focus on minimizing the threshold and difficulties associated with using digital tools and ensuring accessibility and ease of implementation for patients of all ages. Investigating patients’ preferences for digital interventions could also increase usage rates. Monitoring health care professionals’ responses and perspectives on digital interventions, alongside tracking patients’ medication adherence, would provide valuable insights. To prevent alert fatigue [21], future research could explore optimal timing and frequency for implementing digital interventions. Qualitative studies could be conducted to delve deeper into the experiences of digital intervention users in real-world therapeutic settings, complementing quantitative findings.

Considering the growing use of oral SACTs and their higher patient acceptance over intravenous therapy, addressing medication adherence is vital in clinical oncology. Digital interventions offer effective support, enhancing adherence to oral SACTs and improving treatment outcomes while providing convenience for patients. This study highlights the significant benefits of digital technology in promoting adherence. Future research should focus on refining and personalizing digital tools to better meet individual patients’ needs.