Examining the Effectiveness of Electronic Patient-Reported Outcomes in People With Cancer: Systematic Review and Meta-Analysis

Introduction

Background

A patient-reported outcome (PRO) has been defined as “a measurement of any aspect of a patient’s health status that comes directly from the patient” [1]. PROs can be more sensitive and reliable than outcomes reported by clinicians, particularly when reporting adverse events [2,3]. Electronic PROs (ePROs) refer to an electronic administration of the PROs. ePRO systems can present summary reports of patients’ health-related quality of life (HRQOL) and symptoms to the clinician in real time to enhance communication [4] and improve patient management and outcomes.

Many different types of ePRO systems have been developed to monitor and manage diseases, treatments, and symptoms [5] in a variety of patient groups. Guidelines published by the European Society for Medical Oncology recommend using ePROs for symptom monitoring in routine oncology care during systemic cancer treatment due to evidence of benefits for communication, satisfaction of care, treatment adherence, symptom control, HRQOL, hospital admissions and visits, and survival [6,7]. ePROs are being increasingly integrated as part of routine oncology clinical care in the United Kingdom, the United States, and Australia [8-11]. One study exploring the use of ePROs in clinical practice has shown that symptom and quality-of-life scores reported by patients are clinically meaningful and relevant: cough and mobility scores were lower for people with poorer performance status, and patients undergoing chemotherapy and radiotherapy showed improvements in some symptoms [12].

Previous reviews have examined the features of ePRO systems (eg, exploring the integration of ePROs into clinical practice and identifying features that may be associated with patient engagement and patient-centered outcomes [13,14]). Reviews have also been conducted to examine the impact of ePROs on quality of patient care [15].

Objectives

To our knowledge, no existing reviews have grouped randomized controlled trials (RCTs) to estimate an effect size through a meta-analysis to establish clinical benefit. Some meta-analyses evaluated the capacity of telehealth or eHealth interventions to enhance HRQOL in patients with cancer [16,17]. However, many of these studies included complex interventions with various components, not just the ePRO alone. Similarly, other reviews have not specifically explored interventions that provide results to clinicians, and they have not explored the specific components of the ePRO interventions [18,19]. The primary objective of our review was to examine the effect of administering ePROs to patients with cancer on HRQOL compared to usual care. The secondary objectives included the comparison of survival, symptoms, psychological well-being, health care use, and satisfaction with care between participants receiving the ePRO intervention and those receiving usual care.

Methods

Search Strategy

The review protocol was registered with PROSPERO (CRD42020175007). Subsequent protocol changes included the requirement that ePRO results be fed back to clinicians for review. The reporting of this review was guided by the standards of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement. We used the PRISMA guidelines [20] to ensure that all relevant information detailing the processes we followed as well as the findings of this review were included within the manuscript. The completed PRISMA checklist can be found in Multimedia Appendix 1 [20]. Seven databases (PubMed, MEDLINE, PsycInfo, Cochrane Central Register of Controlled Trials, Web of Science, Embase, and CINAHL) were searched systematically on July 16, 2020 (search 1), and June 20, 2022 (search 2). Similar keywords were used across the databases, adapting Boolean operators and Medical Subject Headings vocabulary. The search terms related to the electronic dimension of the “ePRO” system, “patient reported outcomes,” “cancer,” and “randomized controlled trials.”

Inclusion Criteria

Full-text research papers in English published in peer-reviewed journals from January 2000 to June 2022 were included in the review. Given the technological advances in this field over recent years, studies published before 2000 were excluded because the ePRO systems they described would not be comparable to currently available ePRO systems. The inclusion criteria are listed in Textbox 1.

Exclusion Criteria

We excluded studies based on the criteria presented in Textbox 2.

Screening

The identified papers were collated and duplicates removed. The screening of titles and abstracts was conducted independently by 2 reviewers (BK and MBP). Full texts were located for any papers meeting the inclusion criteria and again reviewed by 2 reviewers (BK and MBP). Authors were contacted if full texts could not be obtained. A third reviewer (ST) was consulted in case of any disagreements. Backward and forward reference searching was used to identify additional papers.

Data Extraction

Data were extracted and recorded by 2 researchers (BK and MBP) and included sociodemographic and clinical information, type of ePRO system, feature included in the ePRO (according to the taxonomy of system features [14]), study design and characteristics, type and validity of outcome measure, dropouts, sample sizes, and data used to compute the effect size.

Risk-of-Bias Assessment

The risk of bias was assessed by 3 researchers independently (BK, MBP, and AV) according to the revised Cochrane risk-of-bias tool for randomized trials [22] before making a final collaborative decision. Studies were categorized as low risk of bias, some concern, or high risk of bias.

Data Analysis

The outcomes used and the time points of assessments across the studies were assessed for consistency. If a sufficient number of studies reported common outcomes but measured them in different ways, data were combined as standardized mean differences (SMDs). This is a commonly used summary statistic in meta-analysis that expresses the magnitude of the effect in each study compared to the variability observed. It is calculated by taking the difference in mean outcomes between the groups and dividing it by the SD of the outcome among participants [23]. Where outcomes used the same measurement scale, we combined data as the mean difference. We completed 2 separate meta-analyses: one looking at any HRQOL measure closest to 6 months and the other examining Functional Assessment of Cancer Therapy–General (FACT-G) scores at 3 months; we looked at this particular measure and time point specifically because it was the most commonly used. For all analyses, a maximum likelihood random effects meta-analysis was performed.

The I² statistic was used to assess the presence of heterogeneity. There was no observed heterogeneity (I²=0%), but the 95% CIs were wide (95% CI 0%-64.8% and 95% CI 0%-79.2%), suggesting that the true heterogeneity could plausibly be high; therefore, random effects were chosen. The causes of heterogeneity were not explored, although, as sensitivity analyses, we repeated the meta-analyses with fixed effects. Forest plots were used to visually present the results. Stata (version 14; StataCorp LLC) was used for all analyses, specifically the metaan command [18].

Results

Results of the Search

The search was conducted in 2 stages. A total of 10,965 papers were identified across the 2 searches (2020 and 2022; search 1: n=7281, 66.4%; search 2: n=3684, 33.6%). The full texts of 37 papers were reviewed (search 1: n=19, 51%; search 2: n=18, 49%). Of these 37 papers, 13 (35%) met the inclusion criteria and were included in the review (Figure 1). References from the included papers were reviewed, and this led to 6 additional articles being identified (search 1: n=5, 83%; search 2: n=1, 17%). Thus, overall, 19 papers were deemed eligible after full-text review.

‎

Included Studies

Nineteen papers from 15 RCTs were identified. The second paper of a study was only used if it provided some additional data not presented in the first paper. Reasons for exclusion are summarized in the flow diagram (Figure 1). The characteristics of the 19 included papers are presented in Table 1, and a summary of the studies is presented in Textbox 3. All patients received systemic anticancer therapy. Each study included in the review compared patient reporting of symptoms to the clinical team via an ePRO to a control group. The majority of the studies (14/15, 93%) compared just 2 groups; however, 1 (7%) of the 15 studies included a third attention-control group where participants completed the intervention, but the results were not fed back to the clinical team [24]. The control comparison group was defined in the majority of papers as standard or usual care. Standard or usual care generally consisted of regular appointments with oncology specialists to assess symptoms, followed by appropriate management. Patients were also encouraged to contact the clinical team by telephone if they experienced any problems between appointments. Of the 15 studies, 2 (13%) [25,26] did not provide a definition of standard or usual care.

Study Quality

The risk-of-bias assessment for each study is summarized in Table 2.

ePRO Intervention Components

Overview

Table 3 describes the components of the included interventions.

Symptom Monitoring

All identified ePROs required patients to monitor and report symptoms and included the facility for a clinician to view results. The majority of the ePROs (14/15, 93%) collected data on a wide range of symptoms, whereas 7% (1/15) focused specifically on pain [41].

Symptom Management

Although all ePROs gave the clinical team access to patient reports, 7 (37%) of the 19 studies [24-26,37-39,41] did not actively send reports to the clinical team. Of the 14 systems, 9 (64%) incorporated a facility to alert clinicians if patients reported severe symptoms or a change in symptoms over time [27-31,33-36,40,42]. Greer et al [32] sent all reports (not just those reporting severe symptoms) to the clinical team. Of the 15 interventions, 9 (60%) provided tailored self-management advice for patients based on the problems reported and their severity [26,27,30,33,34,40,42]. In some instances (2/15, 13%), the advice was to contact the clinical team [28,29,40], whereas others (5/15, 33%) provided links to self-management techniques and advice [26,30,33,34,41]. Some (2/15, 13%) used algorithms based on symptom severity to indicate whether patients should receive self-management advice or be advised to contact the clinical team [27,32].

Communication

Of the 15 ePROs, 4 (27%) facilitated patient communication with the clinical team [35,36,40-42]. Communication facilities included the ability to contact the team at any time through the app, initiate or request the need for a consultation, and use free text to report concerns and ask questions.

Alert Management

The majority of the ePROs (6/15, 40%) sent alerts by email; however, in 2 (13%) of the 15 studies, dedicated handsets or pager systems were used [33,34]. Only 1 (7%) of the 15 studies specified 24-hour alert monitoring [33]. Reports were usually sent to designated clinicians; however, in 1 (7%) of the 15 studies, reports were sent to an administrative team who then directed them to an appropriate member of the nursing team [30]. ePRO questionnaire responses were only integrated into electronic patient record systems in 1 (7%) of the 15 studies [27]. The remaining studies (2/15, 13%) used stand-alone web-based systems that required the clinical team to log in to view ePRO responses.

Timing of Alerts

The timing of delivery of ePRO reports to clinicians varied: in 6 (40%) of the 15 studies, electronic reports or alerts were provided in real time [27,30,31,33,34,40]; in 6 (40%) of the 15 studies, reports were reviewed before consultations [24-26,28,29,37-39,42]; and in other studies, reports were reviewed weekly (1/15, 7%) [32], only if patients requested a consultation (1/15, 7%) [41], or if no timing was specified (2/15, 13%) [32,35,36].

Delivery of the Intervention

The frequency of expected ePRO completion varied across the studies. Some of the studies requested reports at specified time intervals: daily (1/15, 7%) [41], weekly (7/15, 47%) [26,27,30-32,37,38,42], or every third month (2/15, 13%) [35,36,40]. Other studies based ePRO completion around clinical time points: before each clinic visit (2/15, 13%) [24,28,29,39], before each cycle of chemotherapy (2/15, 13%) [25,34], or on days 1 to 14 of each chemotherapy cycle (2/15, 13%) [33]. None of the included studies provided data detailing the fidelity of intervention delivery. Of the 15 studies, 2 (13%) mentioned administrative errors where patient data were not collected due to questionnaires not being given, but this referred to outcome data only and not to the ePRO intervention [32,38].

Patient Adherence to Allocated Intervention

Data on patient adherence to the trial interventions were available in 10 (67%) of the 15 studies [24-30,33-40]. Patient adherence was not standardized across the studies; rather, it was assessed and reported in various ways. Of the 19 papers, 4 (21%) [25,30,34,40] reported the percentage of intervention completions versus expected intervention completions across the whole study; 4 (21%) [27,33,35,36] reported completion rates by time point, either for individual patients or as an average; 4 (21%) [28,29,37,38] reported the percentage of patients across the whole study who completed the intervention as per protocol; and 1 (5%) [26] reported the percentage of participants who completed reports. Each paper reported an individual adherence rate, and the figures reported were between 50% and 100%, with only 11% (2/19) reporting figures <70% [27,40]. None of the studies reported adherence in terms of whether the interventions were fully or partially completed. Of the 4 papers that presented adherence by time point, Kearney et al [33] and Absolom et al [27] reported a decrease over time (from 100% to 73% and from 72% to 58%, respectively), whereas Riis et al [35,36] reported no significant change over time (P=.37).

Primary Outcome: Quality of Life

Table 4 highlights the intervention focus for the included papers along with the primary outcome and secondary outcomes.

Of the 15 studies, 13 (87%) measured HRQOL, of which 9 (69%) found a statistically significant effect. Only 5 (38%) of these 13 studies [24,26,32,34,40] used a specific tool to measure psychological well-being; in most cases (n=8, 62%), this was measured with a general HRQOL tool. Tolstrup et al [38] found that, at 48 weeks, the intervention group patients had higher mean scores than the control group patients (mean difference 0.06, 95% CI −0.00 to 0.13; P=.05). Yang et al [41] found that global HRQOL scores for the ePRO group were significantly higher than those for the control group (P<.001). Basch et al [28] found statistically significant improvements for the intervention arm at 6 months for 3 EQ-5D subdomains (mobility: P=.02, self-care: P=.01, and anxiety and depression: P=.01). Improvements in FACT-G scores were reported by Maguire et al [34] (SMD 4.06, 95% CI 2.65-5.46; P<.001) and Velikova et al [24,39] (SE 2.84, 95% CI 13.64-2.37; P=.006). Velikova et al [24] found statistically significant changes in FACT-G physical well-being and FACT-G functional well-being subscale scores in particular (P=.03). Zhang et al [42] found higher total mean scores for HRQOL in the intervention group at 6 months (mean 74.2, SD 15.1, 95% CI 71.7-76.9 vs mean 64.7, SD 28.5, 95% CI 61.0-68.4; P=.01), particularly physical function (mean 84.9, SD 10.5, 95% CI 82.9-88.5 vs mean 68.8, SD 20.7, 95% CI 65.8-72.5; P=.001). Basch et al [30] found that patients in the ePRO group had significantly greater HRQOL than those in the usual care group (odds ratio 1.41, 95% CI 1.10-1.81; P=.006). Absolom et al [27] found that participants in the intervention group reported better overall health on the EQ-5D visual analog scale at 18 weeks (mean 75.6, SD 18.0 vs mean 68.7, SD 20.4; mean 4.48, 95% CI 1.11-7.86; P=.009) and 12 weeks (mean 74.0, SD 16.6 vs mean 71.4, SD 19.5; mean 3.50, 95% CI 0.35-6.66; P=.03), but there was no difference at 6 weeks (mean 74.0, SD 17.3 vs mean 71.4, SD 19.5; mean 1.36, 95% CI 21.66-4.39; P=.38). Denis et al [31] reported that HRQOL at 6 months was stable or that it improved more in the experimental arm (81% vs 59%; P=.04). Of the 15 studies, 4 (27%) [25,26,32,35,36] found no statistically significant differences in relation to HRQOL between groups, and 2 (13%) did not examine HRQOL [28,29,40].

Meta-Analysis

For RCTs with >1 paper (eg, Basch et al [28,29]), only 1 paper was included in the meta-analysis. Of the 15 studies, 8 (53%) were included in the meta-analysis exploring the effect of any HRQOL measure closest to 6 months (Figure 2 [24,27,28,31,32,34,35,38]). Overall, treatment at 6 months demonstrated an average small improvement (SMD 0.07, 95% CI –1.24 to 1.39), although with a wide 95% CI, suggesting that an effect in either direction is possible. There was relatively little variability in reported effect sizes, which ranged from –0.22 to 0.56, although the 95% CIs surrounding these values were often wide. Of the 15 studies, 5 (33%) were included in a meta-analysis of FACT-G scores at 3 months (Figure 3 [24,27,31,32,34]). Here too, the effect of treatment on FACT-G scores at 3 months showed a small average improvement with a wide 95% CI (SMD 0.28, 95% CI –1.22 to 1.78), again suggesting that the true effect of treatment could be positive, negative, or null. Repeating these analyses using fixed effects did not change the results.

‎

‎

Secondary Outcomes

Patients’ Survival

Of the 19 papers, 4 (21%) reported survival data [28,31,42], with Basch et al [29] performing a follow-up survival analysis. Basch et al [28] found a statistically significant difference in patients’ overall survival at 1 year (75% vs 69%; P=.05) and in quality-adjusted survival (mean 8.7 vs 8.0 mo; P=.004). Basch et al [29] also explored longer-term survival and reported a significant difference between groups (mean 31.2 months for the ePRO group compared to mean 26.0 months for the usual care group; P=.03). Denis et al [31] reported an improvement of 7 months in overall survival for the intervention group (hazard ratio 0.32, 95% CI 0.15 to 0.67; P=.002). Zhang et al [42] found no significant differences in survival between the 2 groups (hazard ratio 0.38, 95% CI 0.07-1.99; P=.28). The methodologies used by the studies to calculate survival rates were different and included logistic regression [28], the Kaplan-Meier method with log rank tests [29,31], and chi-square tests [42]. Due to the inconsistent methods of comparison, we compared these descriptively rather than in a meta-analysis.

Patient Symptoms

Of the 19 papers, 14 (74%) reported the effect of the intervention on patients’ symptoms, as shown in Table 3. Of these 14 papers, 7 (50%) reported a positive effect on symptoms at 5 to 6 months or after 6 cycles of chemotherapy [24,28-30,34,39,42]. Velikova et al [24] found statistically significant changes in FACT-G physical well-being subscale scores in the intervention group compared to the control group (P=.006). Maguire et al [34] found that symptoms were better controlled in the intervention group, remaining at pretreatment levels, whereas the symptoms of control group participants increased initially (least squares absolute mean difference –0.15, 95% CI –0.19 to –0.12; P<.001).

Other studies reported the impact of the ePRO on symptoms at 3 months. Absolom et al [27] found that FACT-G physical well-being subscale scores showed that there was better symptom control for participants in the electronic patient self-reporting of adverse events: patient information and advice arm at 6 weeks (difference of least squares means 1.08, SE 0.49, 95% CI 0.12 to 2.05; P=.03) and 12 weeks (difference of least squares means 1.01, SE 0.49, 95% CI 0.05-1.98; P=.04), but there was no difference at 18 weeks (difference of least squares means 0.2, SE 0.51, 95% CI 20.81-1.20; P=.70). Basch et al [30] found a significant improvement in symptom control in the intervention group compared to the control group (odds ratio 1.50, 95% CI 1.15-1.95; P=.003). Kearney et al [33] found that patients who had received the intervention reported lower levels of fatigue than those in the control group (67% vs 81%; P=.04).

Yang et al [41] reported a positive effect on symptoms at 2 weeks and 4 weeks. Yang et al [41] found significant improvements in various symptoms (nausea and vomiting: W=272; P=.01; constipation: W=261; P=.008; fatigue: W=211.5; P=.001; and pain: W=177; P<.001) in the intervention group compared to the control group. Of the 19 papers, 6 (32%) reported that the ePRO interventions had no statistically significant effect on symptom control [25,26,32,35,37,40], whereas 1 (5%) paper did not report this outcome [31].

Hospital Admissions

Of the 19 papers, 4 (21%) reported the effect of the intervention on hospital admissions. Basch et al [28] reported that the patients in the intervention group were less frequently hospitalized at 1 year (incidence rate: 45% vs 49%; P=.08). The remaining papers (3/4, 75%) reported no significant difference in hospital admissions [25,27,32]. All studies assessed the statistical significance differently and at different time points; for example, Absolom et al [27] used incidence rate ratio over 18 weeks, whereas Pappot et al [25] used odds ratio over 6 cycles of treatment.

Emergency Department Visits

Of the 19 papers, 3 (16%) reported a positive effect of the ePRO intervention on emergency department (ED) visits. Basch et al [28] found that patients receiving the intervention were less frequently admitted to the ED (34% vs 41%; P=.02), as did Zhang et al [42] (23% vs 41%; hazard ratio 0.46, 95% CI 0.26-0.81; P=.01). Greer et al [32] reported fewer ED visits resulting in hospitalization (P=.05). All studies assessed the statistical significance differently and at varying time points, using incidence at 1 year [28], hazard ratio after the intervention (6 mo) [42], and mean (SE) during the 12-week study period [32].

Chemotherapy Completion

Of the 19 papers, 5 (26%) reported the effect of the intervention on chemotherapy completion. Basch et al [28] found that the intervention group remained on chemotherapy longer (mean 8.2 vs 6.3 mo; P=.002). Pappot et al [25] found that the 6 scheduled cycles of chemotherapy were completed with treatment adjustments in 34% (ePRO arm) and 40.6% (usual care arm) of the participants (P=.10), but this result was not significant. Absolom et al [27], Riis et al [35], and Zhang et al [42] found no significant difference in treatment adherence at the end of the study.

Patient Satisfaction With Care

Of the 19 papers, 4 (21%) reported the effect of the intervention on patient satisfaction with care [24,32,35,39]. None of the papers reported a statistically significant difference between the groups. Velikova et al [39] found that between 79% and 89% of the patients, regardless of the study arm, rated their quality of care as very good/excellent. Similarly, Riis et al [35] found that participants in both groups were highly satisfied with their follow-up care.

Discussion

Principal Findings

To our knowledge, this is the first meta-analysis to explore the effects of ePROs on health-related outcomes in adults with cancer. All interventions involved patients reporting their symptoms, which clinicians could view and monitor. The presentation of patients’ ePRO responses to clinicians varied across the studies. Of the 14 systems, only 1 (7%) embedded reports within the electronic patient record system [27]. Some of the studies (6/15, 40%) provided results to clinicians in real time [27,30,31,33,34,40], whereas others (6/15, 40%) distributed results before consultations [24-26,28,29,37-39,42]. Of the 8 systems that included clinician alerts and reported HRQOL, 6 (75%) had a positive effect [27-31,34,42]. Of the 9 systems that alerted clinicians and reported symptoms, 6 (67%) had a positive effect [27-30,33,34,42]. This suggests that alerting clinicians is a key component of ePROs and likely to lead to positive effects on symptoms and HRQOL. None of the studies explored whether the intervention was more effective for particular patient groups, such as a particular sex or cancer type.

The meta-analysis showed an improvement in HRQOL at 6 months, although caution should be taken when interpreting the results due to the wide 95% CI. FACT-G scores at 3 months showed a small average improvement but again with a wide 95% CI. Due to the heterogeneity of the studies, specifically the different outcome measures and the different data collection time points, not all studies were included in the meta-analysis. Only 5 (33%) of the 15 studies were included in the FACT-G 3-month meta-analysis, with the weighting predominantly spread across 3 (60%) studies [27,32,34]. These 3 studies all provided advice for patients and sent reports to clinicians. Of the 15 studies, 8 (53%) were included in the HRQOL 6-month meta-analysis, with the majority of the weighting spread across 4 (50%) studies [27-29,32,35,36]. Of the 15 studies, 9 (60%) reported a statistically significant effect on HRQOL. The measures used to assess HRQOL, the specific HRQOL elements that improved, and the time points when significant differences were recorded varied across the studies.

Symptoms were the most commonly reported secondary outcome, and a positive effect was reported in 8 (57%) of the 14 studies where this was measured. The symptoms tracked by ePROs varied across each of the studies however all studies covered common treatment side effects such as fatigue, pain, and nausea. A variety of different tools were used to measure symptoms; therefore, it is difficult for comparisons to be made. Some symptom assessment tools covered a wide range of symptoms, whereas others focused on single symptoms. A narrative review has highlighted the potential benefits of ePRO use in outpatient care in terms of symptom monitoring and facilitating more timely interventions [6]. However, the review does not explore individual components of the interventions, which may significantly impact effectiveness.

Only 3 (20%) of the 15 studies measured survival [28,31,42], of which 2 (67%) reported improved survival in the intervention group [28,31], whereas 1 (33%) found no difference [42]. Many of the other studies only included participants with early disease treated with curative intent; therefore, including survival is not appropriate. Studies exploring the use of ePRO interventions in patients with advanced disease should consider including survival in their outcome measures. A meta-analysis exploring the prognostic values of PROs for survival in cancer suggests that there is evidence of a relationship between PRO data, particularly physical functioning as measured by the European Organisation for Research and Treatment of Cancer Quality of Life Core Questionnaire-30 [43]. Another meta-analysis of 6 studies tentatively suggested that patient monitoring via PROs could have a positive effect on survival [44], but only two-thirds of the studies included ePROs, and only one-third were RCTs. In terms of the impact on care and hospital use, only 2 (13%) of the 15 studies [28,42] reported a significant effect on ED visits. The study by Basch et al [28] also reported a significant impact on hospital admissions and chemotherapy completion. In Basch et al [28,29], patients in the intervention arm remained on chemotherapy for significantly longer. The intervention is similar to other interventions in that symptom reports are available to clinicians, alerts are sent to the nursing team, and results are printed at each clinic visit. The study by Basch et al [28] discusses the direct actions taken after patients report problems. These actions included symptom management counseling, supportive medication changes, referral to the ED or hospital, chemotherapy dose modification, and imaging or test orders. All these actions may have had an impact on chemotherapy tolerance and continued adherence.

Of the 15 studies, 14 (93%) had some methodological concerns, which could indicate a potential risk of bias. However, no study was categorized as high risk of bias in any of the categories. Blinding was not possible, given the nature of the intervention.

Limitations

There are several limitations to this review, which make it difficult to draw firm conclusions regarding the effectiveness of ePRO interventions in clinical practice. In some of the studies (7/15, 47%), the data presented were insufficient for use in the meta-analysis. Authors were contacted where possible, but the required data were not always available. The ability to combine and compare results was further weakened by the varied outcome measures collected at different time points. Many of the studies (12/15, 80%) included a higher proportion of female participants. Male participants may not be as engaged with ePRO interventions or may not openly discuss health problems; therefore, ePRO interventions may not be as effective for male participants.

Comparison With Prior Work

ePROs have been successfully implemented in clinical practice in other diagnostic groups. In rheumatology, ePROs were found to be feasible for use and provided a unique insight into patient experience [45]. Similarly, in diabetes, the completion of preconsultation ePROs was reported to be feasible and acceptable [46]. The AmbuFlex system was successfully used in follow-up across 9 diagnostic groups, including heart disease, epilepsy, asthma, and some cancers. ePROs were completed and the results used to indicate whether a follow-up appointment was necessary. In line with the studies identified in this review, all these systems included clinician review of symptoms to direct patient management and clinical decisions.

New and more effective oncology treatments are being introduced, meaning that the patient population and clinical workload are increasing. The European Society for Medical Oncology has published guidelines on ePRO use in oncology, providing further evidence on the effectiveness of this approach [7]. Symptom reporting could be useful in helping clinicians to manage a larger number of patients for a longer period of time. The findings from this review can highlight which interventions (and components) may be most useful to support clinical practice and improve patient care. Including a facility to send alerts to clinicians is an important feature that should be included in ePROs aiming to improve HRQOL. Regardless of the features included, any ePRO system will only be a success if patients and clinicians fully engage with it [47]. Patients need to regularly complete PROs, and clinicians need to review results and respond.

Conclusions

The aim of this systematic review and meta-analysis was to synthesize all current evidence on the effectiveness of ePROs in RCTs. In total, 19 papers of 15 RCTs were identified. Nearly two-thirds (9/15, 67%) of the interventions showed positive effects on HRQOL and symptoms in adults with cancer. However, caution should be taken in interpreting the results of this review due to the heterogeneity in the interventions, outcome measures, and data collection time points, as well as the wide 95% CIs. This systematic review should act as a driver for further RCTs to be conducted to explore the effectiveness of ePROs using validated outcome measures. This will help to orientate health care professionals toward the most suitable setting and ePRO features to develop and propose optimal care to patients with cancer.