The review protocol, including the search strategy, was prospectively registered in the International Prospective Register of Systematic Reviews (PROSPERO, CRD420251047109). This systematic review and meta-analysis were conducted in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 [28] and PRISMA-S (Preferred Reporting Items for Systematic Reviews and Meta-Analyses literature search extension) [29]. The completed PRISMA 2020 checklist and PRISMA-S checklist are provided in Checklists 1 and 2.

We conducted a comprehensive literature search across multiple electronic databases, including Web of Science, PubMed, Embase, Cochrane Library, CINAHL, PsycINFO, CNKI, and Wanfang databases from inception to April 30, 2026. CINAHL and PsycINFO were searched simultaneously via the EBSCOhost platform, whereas the other databases were searched separately through their respective platforms.

The initial search strategies were independently developed by our research team based on a thorough review of relevant literature [13,17,19], and were further refined through intensive group discussions and consultations with information specialists. The search strategy used a combination of MeSH terms and free-text keywords related to (1) EMA (eg, “experience sampling,” “momentary assessment,” and “daily diary”) and (2) oncology (eg, “neoplasm,” “cancer,” and “tumor”). Additionally, the reference lists of relevant systematic reviews and meta-analyses were manually screened to identify further eligible studies. Furthermore, during the manuscript revision stage, we reran the exact search strategies across all databases to update the literature and ensure the inclusion of the most recent publications. The detailed search strategies for each database are provided in Multimedia Appendix 1.

No study registries, search filters, or additional sources (eg, gray literature, conference proceedings, and organizational websites) were searched or applied. The search strategy did not undergo formal peer review. Apart from database searching, manual screening of reference lists, and author contact for missing or unclear compliance data, no additional search methods were used.

Studies were considered eligible for inclusion if they met the following criteria: (1) Population: Patients had a pathologically confirmed diagnosis of cancer. (2) Methodology: The study used EMA or the experience sampling method, including electronic daily diaries. (3) Outcome measures: clear quantitative documentation of EMA compliance metrics. (4) Study design: observational studies. (5) Sampling frequency: data were collected at least once per day. (6) Publication requirement: peer-reviewed articles in English or Chinese with full-text available.

Studies were excluded based on the following predefined criteria: (1) nonempirical studies: commentaries, editorials, conference abstracts, and studies lacking original data. (2) Intervention-embedded protocols: real-time assessments conducted as integral components of experimental interventions (eg, EMA used to deliver a treatment), as these may confound compliance attributable to assessment procedures alone. (3) Duplicate datasets: Secondary publications derived from identical EMA datasets (retained only the earliest publication per dataset). (4) Dyadic or multiparticipant studies: Studies that simultaneously monitored patients and their caregivers (eg, spouses or family members), as the primary focus of this meta-analysis was the independent compliance of oncology patients.

Identified references were uploaded into Rayyan, a web-based review management tool for literature screening. All duplicates were removed in Rayyan. Two authors (MGY and DC) screened the titles and abstracts independently to identify potentially eligible studies. Then all full-text articles were assessed according to the inclusion and exclusion criteria by both reviewers. Disagreements were resolved through discussion and, when necessary, consultation with a senior reviewer (GC). Full-text studies that did not meet the inclusion criteria were excluded with reasons, and the study selection process was described using the PRISMA flow diagram.

Data from included studies were independently extracted using the collection form developed based on the adapted STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) Checklist for Reporting EMA Studies (CREMAS) [30]: (1) Basic information included title, first author, year of publication, country, study purpose, study duration, and sample size. (2) Demographic characteristics included age, gender, cancer type, cancer stage, treatment, clinical status, etc (3) EMA-related features included EMA device type, application name, operating system, EMA data sampling scheme (time-based or event-based), duration of individual assessments, number of items (ie, number of items to be answered for each measurement), daily measurement frequency, assessment window (ie, the length of time that each measurement had a chance to be answered before it was considered unanswered), training for the EMA, incentive, participation rate, completed prompts (the number of prompts answered), delivered prompts (the number of prompts sent), attrition, and strategy to deal with unanswered prompts. The compliance rate was calculated as completed prompts divided by delivered prompts. For studies that reported only compliance rates without explicitly providing the number of answered prompts, we derived the answered prompts by multiplying the reported compliance rate by the total number of prompts (ie, scheduled study d and the planned daily assessments). For studies that did not report detailed data related to compliance, we attempted to contact their corresponding authors by email.

The database search identified 5049 records. After duplicate removal and title and abstract screening, full-text articles were assessed for eligibility, and 23 studies [18,21,34-54] were included in the final synthesis, with a total sample of 13,565 participants. Figure 1 presents the detailed process of study selection, Table 1 presents the characteristics of the included studies, and Table 2 presents the detailed information of the EMA design. We assessed potential overlap in study populations and confirmed that all included studies were based on independent samples, with no evidence of duplicate data.

Participants were predominantly female (64.7%, 8760/13,543 among studies reporting sex). The included studies were conducted across 7 countries, with the United States contributing the majority (n=15, 65.2%) [18,21,35,37,41,43-51,53]. Sample sizes varied substantially, ranging from 6 participants to 12,433 participants. Most studies were conducted in outpatient settings (n=22, 95.7%) [18,34-54]. Additional characteristics of the study populations are summarized in Table 1. (More comprehensive information can be found in Multimedia Appendix 2)

To our knowledge, this is the first meta-analysis to quantitatively synthesize EMA compliance among oncology populations, directly addressing our three primary research objectives. In alignment with our first aim, we systematically characterized EMA deployment strategies across 23 studies [18,21,34-54]. Regarding our second objective, the pooled analysis revealed a generally favorable overall compliance within the moderate to high range [19]. However, contrary to our expectations, subgroup and meta-regression analyses indicated that no robust study-level design characteristics or procedural factors consistently explained compliance variability. Consequently, the wide prediction interval suggests that compliance remains highly context-dependent [55]. Finally, reflecting our third objective, while EMA is a feasible methodology for real-world symptom tracking in cancer care, this interpretation must remain cautious in light of substantial heterogeneity, risk of bias, and the low certainty of the current evidence.

The EMA compliance observed in this review was broadly comparable to existing EMA benchmarks and meta-analyses. Stone and Shiffman recommended an 80% compliance threshold as a useful methodological benchmark for EMA studies [56]. Our findings align closely with this threshold, as well as with previous meta-analyses reporting pooled compliance rates of approximately 75% in individuals with substance use disorders [25], 86% in older adults [26], and 78% in children and adolescents [27]. Specifically, the compliance among patients with cancer was slightly lower than the rates reported for older adults, but exceeded those found in individuals with substance use disorders and was comparable to the levels seen in pediatric and adolescent populations. The disparity may stem from the relatively regular daily routines of older adults and the lower EMA prompting frequency typically used in studies involving this demographic [57], which together facilitate sustained participation. In contrast, patients with cancer often contend with severe symptom burdens and treatment-induced discomfort [42,43,48], which may hinder their ability to complete all assessments promptly, resulting in slightly lower compliance relative to older adults. However, compared with individuals with substance use disorders or youth cohorts, patients with cancer tend to exhibit stronger health management motivation and well-established compliance behaviors shaped through rigorous medical follow-ups [58,59]. Furthermore, one possible explanation is that participants who enroll in EMA studies may be more motivated or more comfortable with health-related monitoring, although this cannot be confirmed using aggregate study-level data.

The notable gap between the high EMA compliance observed in this study and the typically lower rates of clinical treatment compliance (20%‐60%) [60] highlights unique behavioral dynamics. Crucially, the pooled compliance estimate should not be equated with clinical treatment compliance. EMA completion is a short-term research behavior involving repeated self-reported prompts [9,10], whereas treatment compliance is shaped by broader clinical, behavioral, and health-system factors [60]. This divergence likely stem from the low-burden nature of EMA tasks [12-14], the shorter durations [12-14], and a potential “Hawthorne effect” where frequent digital prompts foster sustained engagement [61]. While these findings suggest patients with cancer are capable of consistent real-time monitoring, selection bias must be acknowledged, as study participants may represent a more motivated subgroup with higher baseline compliance [62]. Nonetheless, EMA remains a high-fidelity tool for capturing the symptomatic triggers, such as acute pain or fatigue, that often precipitate clinical noncompliance [12-14].

Focusing solely on comparative averages may overlook the profound variability observed across oncology studies, as evidenced by the wide 95% prediction interval of 48.59%‐93.41%. This wide prediction interval reflects substantial between-study variation and highlights the context-dependent nature of EMA compliance in oncology populations. While such variability presents statistical challenges, the pooled estimate remains valuable as a benchmark when interpreted alongside the prediction interval [33], signaling that researchers should set context-specific targets rather than expecting uniform rates. Furthermore, the absence of significant predictors in our meta-regression suggests that compliance is likely governed by complex, nonlinear interactions among disease severity, treatment toxicity, and digital literacy, rather than simple design metrics like prompt frequency alone [19].

Subgroup analyses revealed that EMA compliance remained remarkably consistent across diverse study characteristics. Specifically, no significant differences were identified by country region, cancer type, or technical configurations such as device type and assessment windows. However, these null findings should be interpreted cautiously, as many subgroup comparisons were underpowered due to limited study counts within specific strata [63]. Similarly, factors such as study length, sampling frequency, and items per assessment did not significantly alter compliance rates. These findings may imply that oncology patients’ engagement with EMA may be driven more by intrinsic factors, such as health motivation and perceived utility, than by extrinsic protocol adjustments [64,65]. Consequently, researchers may have considerable flexibility in designing EMA protocols without necessarily compromising participant compliance, provided the clinical burden remains manageable.

Overall, these findings highlight that EMA is a feasible, scalable methodology for oncology populations, supporting its integration into personalized health management [66]. Future studies should focus on developing standardized reporting and compliance metrics to enhance cross-study comparability and reproducibility. Additionally, given that traditional design features did not significantly drive compliance in our analysis, future research should investigate more complex, nonlinear predictors of engagement, such as digital literacy and real-time symptomatic burden, through mixed-methods approaches. Finally, exploring the long-term sustainability of EMA in clinical practice is essential to maintain high-fidelity data collection throughout the cancer survivorship continuum without imposing excessive patient burden.

This study has several limitations. First, there was substantial heterogeneity among the included studies, and the available study-level variables did not fully explain this heterogeneity. Some subgroup analyses were constrained by small sample sizes, warranting cautious interpretation of these results. Second, many studies provided incomplete reporting of implementation details, such as participant training, incentive strategies, assessment windows, response latency, missing data handling, or the exact denominator used to calculate compliance. This lack of methodological transparency directly increases the potential risk of bias, particularly regarding selection and reporting biases. Third, this review included only published studies in English and Chinese, potentially introducing reporting biases and small-study effects. Finally, the exclusion of studies due to missing or nonextractable prompt-level compliance data may introduce a degree of reporting bias, as studies with lower compliance might be less likely to report these metrics. Consequently, these compounded factors, including severe inconsistency, potential reporting bias, and the observational nature of the primary literature, resulted in a very low to low certainty of evidence within the GRADE framework. Therefore, our pooled compliance estimate should be interpreted as an approximate, context-dependent benchmark rather than a universal standard.

This study suggests that patients with cancer can achieve moderate-to-high compliance with EMA protocols. However, the wide prediction interval, substantial heterogeneity, possible reporting bias, and very low certainty of evidence indicate that compliance is highly context-dependent. Consequently, the pooled estimate should therefore be interpreted as an approximate benchmark for planning oncology EMA studies rather than as a universal expected rate. For future research, maximizing feasibility requires careful, transparent, and patient-centered implementation. Researchers should consistently report compliance denominators, distinguish delivered from prompts, and clarify response windows alongside missing data handling. Furthermore, comprehensive reporting on training protocols, incentives, and attrition patterns is essential. Prospective studies must evaluate how varying combinations of design features impact compliance across distinct clinical cohorts and cancer stages. In terms of translation, EMA holds substantial value for capturing real-time symptoms, behaviors, and psychological states in patients’ daily lives. However, before widespread adoption in routine oncology care, identifying sustainable designs that preserve data quality without exacerbating patient burden remains a critical priority. Standardized reporting, enhanced methodological transparency, and a deeper understanding of patient-level compliance determinants are pivotal to transitioning EMA from a promising research tool to a reliable clinical monitoring strategy.