IntroductionBackground

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” []. PROs can be more sensitive and reliable than outcomes reported by clinicians, particularly when reporting adverse events [,]. 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 [] and improve patient management and outcomes.

Many different types of ePRO systems have been developed to monitor and manage diseases, treatments, and symptoms [] 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 [,]. ePROs are being increasingly integrated as part of routine oncology clinical care in the United Kingdom, the United States, and Australia [-]. 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 [].

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 [,]). Reviews have also been conducted to examine the impact of ePROs on quality of patient care [].

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 [,]. 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 [,]. 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.

MethodsSearch 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 [] 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 []. 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. Inclusion criteria.

Inclusion criteria and determinants

Population: those with any type of cancer diagnosis or cancer stage, aged ≥16 yearsIntervention: electronic patient-reported outcome interventions where participants report outcomes electronically (web-based, computer, mobile phone, tablet, etc), and responses are subsequently made available to clinical teamsComparison: usual care or other control conditionsOutcomes: validated questionnaires measuring health-related quality of life (primary outcome), symptoms, psychological well-being, satisfaction with care, health care use, survival, and progression-free survival (secondary outcomes)

Note: The PICO (Population, Intervention, Comparison, Outcomes) framework to identify determinants [] was used for the inclusion criteria.

Exclusion Criteria

We excluded studies based on the criteria presented in .

Textbox 2. The criteria used to exclude studies.

Exclusion criteria

Book chapters, conference abstracts, commentaries, opinion articles, reviews, meta-analyses, unpublished data, and so onNot an electronic patient-reported outcome (ePRO) interventionNot focusing on patients with cancer and patients aged ≥16 yearsNot a randomized controlled trial (eg, nonrandomized trial, correlational study, or case study)Conditions differ except for the ePRO intervention (eg, the ePRO group received another intervention that was not received by the control group)Data published elsewhereNo control group or an inappropriate control group (eg, where the control condition received an ePRO intervention or another intervention not received by the experimental group)Not a health-related outcome measured using validated questionnaires: no health-related quality of life, physical symptoms (eg, nausea, vomiting, pain, breathlessness, and fatigue), psychological symptoms (eg, anxiety and depression), satisfaction with care, health care use outcomes (hospital admission, emergency department visit, and chemotherapy completion), survival, or progression-free survivalePRO results were not fed back to cliniciansFull text not availableScreening

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 []), 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 [] 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 []. 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 [].

ResultsResults 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 (). 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.

‎Figure 1. Illustration of the flow of studies through the systematic review process. ePRO: electronic patient-reported outcome. 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 (). The characteristics of the 19 included papers are presented in , and a summary of the studies is presented in . 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 []. 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%) [,] did not provide a definition of standard or usual care.

Table 1. The included papers categorized by study (year and country), methods, sample size, and participant characteristics.Study, year; countryMethodsSample sizeParticipant characteristicsAbsolom et al [], 2021; United KingdomRCTa; 2 groups; random permuted blocks via an automated 24-hour systemEligible=782; randomized=508; Ib=256, Cc=252; DO: I=42, C=19Age (y) mean: I=55.9 (SD 12.2), C=56 (SD 11.3)
Sex: Md=102, Fe=406
Diagnosis and stage of disease: colorectal, breast, or gynecologic cancers; cancer of any stage
Basch et al [], 2016; United StatesRCT; 2 groups each, with 2 subgroups based on level of prior computer use; computer system using randomly permuted blocks Eligible=1107; randomized=766; CEf participants: I=286, C=253; CIg participants: I=155, C=72; DO: I=69, C=89 (CE participants); I=46, C=26 (CI participants)Age (y), median: I=61 (IQR 30-91), C=62 (26-88)
Sex: M=322, F=444
Diagnosis and stage of disease: breast, genitourinary, gynecologic, or lung cancers; metastatic
Basch et al [], 2017; United StatesRCT; 2 groups each, with 2 subgroups based on level of prior computer use; computer system using randomly permuted blocksEligible: not stated; randomized=766; I=539, C=227; DO (overall): 249Age (y; overall), median: 61 (IQR 26-91)
Sex: M=322, F=444
Diagnosis and stage of disease: breast, genitourinary, gynecologic, or lung cancers; metastatic
Basch et al [], 2022; United StatesRCT; 2 groups; cluster designEligible=1444; randomized=1191; I=593, C=598; DO: I=278, C=230Age (y), median: I=64 (IQR 29-89), C=62 (IQR 28-93)
Sex: M=496, F=694; 1197 randomized and 1191 analyzed; however, data on sex are only listed for 1190 patients
Diagnosis and stage of disease: cancer of any type; metastatic
Denis et al [], 2017; FranceRCT; 2 groups; minimization programEligible: not stated; randomized=133; I=67, C=66; DO: I=7, C=5Age (y), median: I=65.2 (IQR 35.7-86.9), C=64.3 (IQR 42.7-88.1)
Sex: M=81, F=40; 133 randomized and 121 analyzed
Diagnosis and stage of disease: nonprogressive small cell lung cancer or non–small cell lung cancer; advanced (at least cTxN1 or pTxpN1 to TxNxM+ at least stage II cancer)
Greer et al [], 2020; United StatesRCT; 2 groups; computer-generated randomization scheme stratified by cancer typeEligible=500; randomized=181; I=91, C=90; DO: I=11, C=1Age (y), mean: I=52.85 (SD 13.74), C=53.76 (SD 12.08)
Sex: M=84, F=97
Diagnosis and stage of disease: diverse malignancies; not stated
Kearney et al [], 2009; United KingdomRCT; 2 groups; automated interactive voice response telephone randomization systemEligible: not available due to incomplete data; randomized=112; I=56, C=56; DO: I=27, C=27Age (y), mean: I=55.1 (SD 10.6), C=56.9 (SD 10.5)
Sex: M=26, F=86
Diagnosis and stage of disease: breast, lung, or colorectal cancer; irrespective of stage
Maguire et al [], 2021; Austria, Greece, Ireland, Norway, and the United KingdomRCT; 2 groups; repeated measures parallel group evaluator-masked stratified trialEligible=1222; randomized=840; I=422, C=418; DO: I=134, C=122Age (y), mean: I=51.9 (SD 12.4), C=52.9 (SD 12.1)
Sex: M=151, F=678; 840 randomized and 829 analyzable
Diagnosis and stage of disease: breast cancer, colorectal cancer, Hodgkin disease, or non-Hodgkin lymphoma; stages 0 to IV (not metastatic breast or colorectal cancer)
Pappot et al [], 2021; DenmarkRCT; 2 groups; cluster randomizationEligible=682; randomized=682; I=347, C=335; DO: I=11, C=22Age (y), median: I=53 and C=53 (range 21-82)
Sex: M=0, F=682
Diagnosis and stage of disease: breast cancer; not stated
Post et al [], 2013; United StatesRCT; 2 groups; method of randomization not describedEligible=93; randomized=60; I=31, C=29; DO: I=4, C=6Age (y), mean: I=49.5 (SD 10.7), C=52.1 (SD 8.5)
Sex: M=0, F=60
Diagnosis and stage of disease: breast cancer; primary (stages I-III)
Riis et al [], 2020; DenmarkRCT; 2 groups; computer-generated sequenceEligible=177; randomized=134; I=65, C=69; DO: I=5, C=7Age (y), mean: I=64.4, C=64.2
Sex: M=0, F=129
Diagnosis and stage of disease: breast cancer; primary (early breast cancer, stages I-III)
Riis et al [], 2021; DenmarkRCT; 2 groups; computer-generated sequenceEligible=177; randomized=134; I=65, C=69; DO: I=5, C=7Age (y; overall), mean: 64.3
Sex: M=0, F=129
Diagnosis and stage of disease: breast cancer; primary (early breast cancer, stages I-III)
Tolstrup et al [,], 2020, 2022; DenmarkRCT; 2 groups; open-label, computer-randomized trialEligible=200; randomized=146; I=73, C=73; DO: I=6, C=2Age (y), median: I=66, C=66
Sex: M=78, F=68
Diagnosis and stage of disease: melanoma; metastatic (stages III-IV)
Velikova et al [], 2004; United KingdomRCT; 3 groups; random permuted blocks by telephoneEligible=439; randomized: 286; I=144, attention-C=70, C=72; DO: I=60, attention-C=35, C=27Age (y), mean: I=55.1 SD (13.02), attention-C=54.8 (SD 12.4), C=54.7 (SD 11.67)
Sex: M=76, F=210
Diagnosis and stage of disease: cancer of any type; mixed stage
Velikova et al [], 2010; United KingdomRCT; 3 groups; telephone by the research officeEligible=439; randomized=286; I=144; attention-C=70, C=72; DO: I=59, attention-C=36, C=25Age (y), mean: I=54.8 (SD 12.9), attention-C=55.2 (SD 11.79), C=54.9 (SD 11.76)
Sex: M=64, F=194
Diagnosis and stage of disease: cancer of any type; mixed stage
Wheelock et al [], 2015; United StatesRCT; 2 groups; block designEligible=102; randomized=100; I=59, C=41; DO: I=9, C=6Age (y), mean: I=54.78 (SD 8.66), C=3.3 (SD 10.79)
Sex: M=0, F=100
Diagnosis and stage of disease: breast cancer; primary (stages I-III)
Yang et al [], 2019; ChinaRCT; 2 groups; scheme generated by independent statistical personnel using a computerEligible=58; randomized=58; I=31, C=27; DO: 0Age (y), mean: I=51.1 (SD 8.98), C=53.96 (SD 8.58)
Sex: M=38, F=20
Diagnosis and stage of disease: cancer of any type; not stated
Zhang et al [], 2022; ChinaRCT; 2 groups; open-label trialEligible=364; randomized=300; I=150, C=150; DO: I=9, C=13Age (y), mean: I=57.6 (SD 12.6), C=60.1 (SD 12.7)
Sex: M=206, F=72
Diagnosis and stage of disease: cancer of any type; life expectancy was at least 6 months; not stated

aRCT: randomized controlled trial.

bI: intervention.

cC: control.

dM: male.

eF: female.

fCE: computer-experienced.

gCI: computer-inexperienced.

Textbox 3. Summary of included studies.

Studies summarized

Countries: United States (5/15, 33%), United Kingdom (3/15, 20%), Denmark (3/15, 20%), China (2/15, 13%), France (1/15, 7%), and multiple European countries (1/15, 7%)Randomized controlled trials: 2 groups (14/15, 93%) and 3 groups (1/15, 7%)Study size: ranging from 58 to 1191 patients, with a total of 5446 patients; 5497 patients randomizedSex: 83.6% (4553/5446) were femaleStage of disease: primary cancer (3/15, 20%), metastatic cancer (4/15, 27%), and any stage or not specified (8/15, 53%)Study Quality

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

Table 2. Results from the risk-of-bias assessment performed using the revised Cochrane risk-of-bias tool for randomized trials.Study, yearRandomization and allocation (selection bias)Blinding of participants (performance bias)Missing outcome data (attrition bias)Blinding of outcome assessment (detection bias)Selective reporting (reporting bias)Absolom et al [], 2021Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasBasch et al [,], 2016 and 2017Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasBasch et al [], 2022Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasDenis et al [], 2017Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasGreer et al [], 2020Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasKearney et al [], 2009Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasMaguire et al [], 2021Low risk of biasLow risk of biasLow risk of biasLow risk of biasLow risk of biasPappot et al [], 2021Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasPost et al [], 2013Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Some concerns (did not include a prespecified analysis plan)Riis et al [,], 2020 and 2021Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasTolstrup et al [,], 2020 and 2022Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasVelikova et al [,], 2004 and 2010Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasWheelock et al [], 2015Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Some concerns (provided minimal details of a prespecified analysis plan)Yang et al [], 2019Low risk of biasLow risk of biasLow risk of biasLow risk of biasLow risk of biasZhang et al [], 2022Low risk of biasLow risk of biasLow risk of biasSome concerns (outcome assessors were aware of the intervention received by study participants)Low risk of biasePRO Intervention ComponentsOverview

describes the components of the included interventions.

Table 3. Intervention components and the effect on review primary outcome and secondary outcomes.Study, yearIntervention componentsReview primary outcome, secondary outcomes, and effect
Symptom monitoringSymptom managementCommunicationAlert managementTiming of alertsQuality of life (primary outcome)Patient survivalSymptomsHospital admissionsEmergency department visitsChemotherapy completionSatisfaction with careAbsolom et al [], 2021Patients and clinicians (linked to electronic patient records)Tailored advice for patients; reports sent to cliniciansNRaClinical team shared email address; monitored by nursesReal time+b—c+–d—–—Basch et al [,], 2016 and 2017Patients and cliniciansReports sent to cliniciansNREmail sent to nurses; not monitored 24 hoursPrinted at each clinic visit+++–++—Basch et al [], 2022Patients and cliniciansTailored advice for patients; reports sent to cliniciansNREmail sent to designated admin staff who forwarded it to relevant nurseReal time; reports at clinic visits+—+————Denis et al [], 2017Patients and cliniciansReports sent to cliniciansNREmail sent to oncologistReal time++———— —Greer et al [], 2020Patients and cliniciansGeneric advice for patients; reports sent to cliniciansNREmail sent to clinicianNot specified–—–––—–Kearney et al [], 2009Patients and cliniciansTailored advice for patients; reports sent to cliniciansNRDedicated 24-hour pager system; clinicians should contact patients within 1 hour for severe symptomsReal time——+————Maguire et al [], 2021Patients and cliniciansTailored advice for patients; reports sent to cliniciansNRAlerts sent to clinicians on dedicated handsetsReal time+—+————Pappot et al [], 2021Patients and cliniciansNRNRNRShown to clinicians after completion at each visit–—––—–—Post et al [], 2013Patients and cliniciansTailored advice for patientsNRNRPrinted at each clinic visit–—–————Riis et al [,], 2020 and 2021Patients and cliniciansReports sent to cliniciansThe patient could request a consultation through the ePROMe systemPrincipal investigator monitored questionnaire and emailsNot specified–—–——––Tolstrup et al [,], 2020 and 2022Patients and cliniciansPatients advised to contact clinical team for severe symptomsNRNRLog in to system to view before consultation+—–————Velikova et al [,], 2004 and 2010Patients and cliniciansNRNRNRPrinted at each clinic visit+—+———–Wheelock et al [], 2015Patients and cliniciansTailored advice for patients; reports sent to cliniciansFree text to report concerns and ask questionsMonitored by designated nurse practitionerReal time——–————Yang et al [], 2019Patients and cliniciansPatients advised to follow medication plan in case of severe painReal-time consultationNRViewed when patients request a consultation+—+————Zhang et al [], 2022Patients and cliniciansTailored advice for patients; reports sent to clinical teamConsult team via app at any timeAn oncology specialist and 2 nurses from each centerViewed before consultation+–+—+– —

aNR: not reported.

bStatistically significant effect (P<.05).

cDid not measure this outcome.

dNo statistically significant effect (P>.05).

eePROM: electronic patient-reported outcome measure.

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 [].

Symptom Management

Although all ePROs gave the clinical team access to patient reports, 7 (37%) of the 19 studies [-,-,] 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 [-,-,,]. Greer et al [] 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 [,,,,,,]. In some instances (2/15, 13%), the advice was to contact the clinical team [,,], whereas others (5/15, 33%) provided links to self-management techniques and advice [,,,,]. 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 [,].

Communication

Of the 15 ePROs, 4 (27%) facilitated patient communication with the clinical team [,,-]. 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 [,]. Only 1 (7%) of the 15 studies specified 24-hour alert monitoring []. 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 []. ePRO questionnaire responses were only integrated into electronic patient record systems in 1 (7%) of the 15 studies []. 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 [,,,,,]; in 6 (40%) of the 15 studies, reports were reviewed before consultations [-,,,-,]; and in other studies, reports were reviewed weekly (1/15, 7%) [], only if patients requested a consultation (1/15, 7%) [], or if no timing was specified (2/15, 13%) [,,].

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%) [], weekly (7/15, 47%) [,,-,,,], or every third month (2/15, 13%) [,,]. Other studies based ePRO completion around clinical time points: before each clinic visit (2/15, 13%) [,,,], before each cycle of chemotherapy (2/15, 13%) [,], or on days 1 to 14 of each chemotherapy cycle (2/15, 13%) []. 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 [,].

Patient Adherence to Allocated Intervention

Data on patient adherence to the trial interventions were available in 10 (67%) of the 15 studies [-,-]. Patient adherence was not standardized across the studies; rather, it was assessed and reported in various ways. Of the 19 papers, 4 (21%) [,,,] reported the percentage of intervention completions versus expected intervention completions across the whole study; 4 (21%) [,,,] reported completion rates by time point, either for individual patients or as an average; 4 (21%) [,,,] reported the percentage of patients across the whole study who completed the intervention as per protocol; and 1 (5%) [] 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% [,]. 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 [] and Absolom et al [] reported a decrease over time (from 100% to 73% and from 72% to 58%, respectively), whereas Riis et al [,] reported no significant change over time (P=.37).

Primary Outcome: Quality of Life

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

Table 4. Intervention types along with the study primary outcome and secondary outcomes.Study, yearIntervention typeStudy primary outcomeStudy secondary outcomesAbsolom et al [], 2021eRAPIDa, an online eHealth system for patients to self-report symptomsSymptom controlImpacts on hospital services (process of care measures) and cost-effectivenessBasch et al [,], 2016 and 2017 Reporting of 12 common symptoms via STARb, a web-based interfaceHRQOLcEDd visits, hospitalizations, overall survival, and survival at 1 yearBasch et al [], 2022Electronic symptom monitoring with PROe surveysOverall survivalPhysical function, symptom control, and HRQOLDenis et al [], 2017Web-mediated follow-up of symptomsOverall survivalPerformance status at first relapse, progression-free survival, and HRQOLGreer et al [], 2020Smartphone mobile appAdherence, symptom burden, and quality of lifePatient satisfaction with treatment and health care useKearney et al [], 2009 Mobile phone–based remote monitoring ASyMSfSymptom scores and the occurrence of 6 symptoms that are components of the total symptom score—gMaguire et al [], 2021Remote monitoring via the ASyMSSymptom burdenHRQOL, supportive care needs, anxiety, self-efficacy, and work limitationsPappot et al [], 2021ePROh questionnaire of symptom toxicitiesNumber of patients with ≥1 treatment adjustments Number of patients with ≥1 hospitalizations, ≥1 events of febrile neutropenia, number of patients with treatment postponed >7 days, as well as completion of the scheduled 6 cycles of chemotherapy was registered, and compliance to ePROPost et al [], 2013PDA-delivered symptom communicationEffects on pain, depression, and fatigue symptomsStudy feasibility, patient and clinician responses to study participation, and intervention effects on HRQOL and communication self-efficacyRiis et al [,], 2020 and 2021A patient-initiated follow-up program customized to the needs of the individualSatisfaction with care and unmet needs Use of consultations, adherence to treatment, and quality of life; number of in-person, telephone, and email consultations; and patient satisfactionTolstrup et al [,], 2020 and 2022Web-based symptom reporting using AmbuFlexNumber of severe adverse events (grades 3-4)Service use (eg, number of telephone consultations as well as outpatient and inpatient visits), length of time toxicities experienced, and length of time steroids required; HRQOL; and associations between toxicity severity and HRQOLVelikova et al [,], 2004 and 2010Touch screen HRQOL questionnairesHRQOL, physician-patient communication, and clinical managementProcess measures (tests, drugs, and medical records), continuity of care, and patient satisfactionWheelock et al [], 2015 An online health questionnaire with a component of remote follow-up called SIS.NETiQuantify the time between symptom reporting and remote evaluation of symptomsCompare use of health care resources (breast cancer–related visits, total number of medical appointments, and laboratory and imaging studies)Yang et al [], 2019 A mobile phone app (Pain Guard)Remission rate of painMedication adherence, improvements in HRQOL, frequency of breakthrough cancer pain, incidence of adverse reactions, and patient satisfactionZhang et al [], 2022ePRO follow-up mobile appIncidence of serious (grades 3-4) immune-related adverse events, ED visits, HRQOL, time spent implementing the ePRO model, rate of treatment discontinuation, and death—

aeRAPID: electronic patient self-reporting of adverse events: patient information and advice.

bSTAR: Symptom Tracking and Reporting.

cHRQOL: health-related quality of life.

dED: emergency department.

ePRO: patient-reported outcome.

fASyMS: advanced symptom management system.

gNot applicable.

hePRO: electronic patient-reported outcome.

iSIS.NET: system for individualized survivorship care, based on patient self-reported data, with review by nurse practitioners, targeted education, and triage.

Of the 15 studies, 13 (87%) measured HRQOL, of which 9 (69%) found a statistically significant effect. Only 5 (38%) of these 13 studies [,,,,] 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 [] 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 [] found that global HRQOL scores for the ePRO group were significantly higher than those for the control group (P<.001). Basch et al [] 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 [] (SMD 4.06, 95% CI 2.65-5.46; P<.001) and Velikova et al [,] (SE 2.84, 95% CI 13.64-2.37; P=.006). Velikova et al [] 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 [] 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 [] 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 [] 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 [] 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%) [,,,,] found no statistically significant differences in relation to HRQOL between groups, and 2 (13%) did not examine HRQOL [,,].

Meta-Analysis

For RCTs with >1 paper (eg, Basch et al [,]), 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 ( [,,,,,,,]). 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 ( [,,,,]). 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.

‎Figure 2. Forest plot of the meta-analysis exploring the effect of any health-related quality of life measure at 6 months. ‎Figure 3. Forest plot of the meta-analysis exploring the effect of treatment on Functional Assessment of Cancer Therapy–General scores at 3 months. ES: effect size. Secondary OutcomesPatients’ Survival

Of the 19 papers, 4 (21%) reported survival data [,,], with Basch et al [] performing a follow-up survival analysis. Basch et al [] 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 [] 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 [] 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 [] 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 [], the Kaplan-Meier method with log rank tests [,], and chi-square tests []. 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 . Of these 14 papers, 7 (50%) reported a positive effect on symptoms at 5 to 6 months or after 6 cycles of chemotherapy [,-,,,]. Velikova et al [] 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 [] 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 [] 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 [] 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 [] 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 [] reported a positive effect on symptoms at 2 weeks and 4 weeks. Yang et al [] 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 int