Effectiveness of symptom monitoring on electronic patient-reported outcomes (ePROs) among patients with lung cancer: a systematic review and meta-analysis

Abstract

Symptom monitoring using electronic patient-reported outcomes (ePROs) has demonstrated benefits for patients with cancer, yet the systematic effects for lung cancer remains unknown. This study performed a literature search in Medline, Embase, Cochrane library, CINAHL and APA PsyInfo before April 23rd, 2025, and identified 5755 papers. 18 (0.31%) papers from 11 studies conducting symptom monitoring on ePROs and sending alerts of severe symptoms were included. The meta-analysis showed significant improvement in health-related quality of life (SMD = 2.44, P < 0.001) among patients with lung cancer, with an intervention duration of less than 6 months, 6 months and more than 6 months. When excluding studies that sent alerts to patients themselves, overall survival for lung cancer patients was significantly prolonged (HR = 0.54, 95% CI [0.22, 1.31], P = 0.031). Symptom burden, physical functioning, and healthcare service utilization was also advantaged, but implementation process and cost-effectiveness data was insufficient (Trial registration: CRD420251000397).

Introduction

Lung cancer is the second most common malignancy and the leading cause of cancer-related death worldwide¹. In 2020, there were over 2.2 million new cases and nearly 1.8 million deaths from lung cancer, which accounted for 11.4% and 18.0% of the global new cases and deaths of cancer, respectively¹. The incidence and mortality rates are highest in economically developed countries². The initial stage of lung cancer is usually a symptom-free course with nonspecific symptoms, so therefore most lung cancers are diagnosed at an advanced stage^3,4. In addition, a large proportion of patients with lung cancer (48%) are diagnosed with distant metastases⁵. Consequently, adverse symptoms like cough, pain, shortness of breath, and hemoptysis arise from advanced lung cancer and greatly impact their health-related quality of health (HRQoL) and survival⁶. Although innovations in lung cancer treatments have improved the survival rate noticeably, these treatments cause severe symptoms during the treatment period and thereafter, which can impose a worsened HRQoL, distress, functional disability, more healthcare utilization, and financial burden among the patients^7–9. Nevertheless, clinicians fail to detect and address adverse symptoms up to half the time. Although clinician assessments are currently the main method for symptom monitoring in cancer, they often miss a significant number of the symptoms that patients experience, owing to short visit times and low efficiency in patient-clinician communications^8,10,11.

Patient-reported outcomes (PROs) refer to health outcomes reported directly from patients or their proxies, without any interpretation by clinicians or anyone else, and it is considered to show more accurate evaluations of patients’ subjective symptoms than objective assessments by clinicians^12,13. Electronic patient-reported outcomes (ePROs) (PROs reported via tablets, mobile phones, and other electronic devices) may optimize symptom monitoring for patients with cancer¹⁴. Empirical evidence suggests that the routine collection of ePROs either between visits or at visits, followed by alerts or advice activated based on individual conditions sent to patients and/or healthcare providers, can lead to better HRQoL, physical function, symptom control and prolonged survival for patients with cancer^15–18. This is achieved by promoting timely responses to concerning symptoms by clinicians and enhancing awareness and capabilities of self-management among the patients^19,20. Existing literature employed ePROs-based symptom monitoring as an intervention in randomized controlled trials (RCTs) to evaluate the effects on health outcomes. These interventions varied in setting, duration, participants, controlled scenarios and targeted outcomes, and the effects may differ from different study designs²¹. Previous meta-analyses have quantified the significantly positive effects of ePRO-based symptom monitoring interventions on a series of health outcomes among patients with various types of cancer, such as breast or prostate cancer, under multiple types of treatments, including chemotherapy, immunotherapy, and radiotherapy^19,22–24. Given these advantages, it is recommended to incorporate ePROs-based symptom management in routine oncology care during systemic cancer treatment^25–27.

Lung cancer is always accompanied by a high risk of deterioration and complex side-effects related to treatments, and several studies also demonstrated the superiority of symptom monitoring on ePROs also in facilitating care for patients with lung cancer^28–34. However, there is a lack of synthesis analysis on the effectiveness of ePROs among patients with lung cancer. As a result, although ePROs are significant indicators of the disease and treatment on health, the routine use of ePROs is still not widely recognized among lung cancer³⁴. Moreover, despite a great importance to identify the optimal ePROs-based interventions, there is limited evidence to validate heterogeneity in effectiveness derived from different intervention durations, ePROs measurements, and treatments^21,35,36. Additionally, a practice and scalable approach is substantial for integrating such practice into routine care, but only few studies have discussed the implementation fidelity and cost-effectiveness of ePROs-based interventions for lung cancer^37,38.

Therefore, this systematic review and meta-analysis aims to synthesize the results of existing RCTs that included patients with lung cancer to explore the effectiveness of ePRO-based symptom management interventions on the HRQoL and overall survival. This study also targets to elicit the effectiveness of different types of interventions, and to clarify the adherence, burden, satisfaction, implementation, and cost-effectiveness of these interventions among patients with lung cancer.

Results

Study selection

Three rounds of literature search were conducted in Medline, Embase, Cochrane library, CINAHL and APA PsyInfo before April 23rd, 2025. The initial literature search yielded a total of 5755 articles from five databases. We eliminated 1683 duplicates and 438 book chapters, conference abstracts, commentaries, etc, automatically and manually before screening. Next, after screening titles and abstracts of the remaining 3634 articles, 160 articles were retrieved for full-text review. We excluded 145 articles that did not meet eligibility criteria, the information and reasons for excluding these articles were documented in Supplementary Table 1. We included 2 articles when screening for references of the full-text reviewed articles, and 1 article identified on March 1st, 2025. No additional paper was retrieved on April 23rd, 2025. Overall, 18 articles from 11 studies were ultimately deemed eligible in this review. The selection procedure is presented in Fig. 1.

Fig. 1.

Study selection flowchart.

Description of the included studies

Study characteristics

Table 1 shows the characteristics of the included studies. Seven articles derived from four RCTs originated in the United States of American (USA)^39–45, three articles from two RCTs conducted in China^30,31,46, three articles from a RCT in France^47–49, two articles reported results from a RCT in the Netherlands^28,29, one in the United Kingdom (UK)⁵⁰, one from Canada⁵¹ and one from Switzerland⁵². All trials were conducted between 2006 and 2024 and all articles were published from 2009 to 2025. Two studies^41,42,52 were cluster RCTs, Billingy et al.^28,29 utilized a stepped wedge design in RCT, only two studies were single-center trials^39,40,45, and the others were multicentre RCTs. Each study compared symptom monitoring with ePROs to a control group with two arms, except for two three-armed RCTs from USA^39,40,45 and one from the Netherlands^28,29. Six study protocols were retrieved (two were published^53,54 and four were in supplementary file) to help extract study methodology.

Table 1.

The characteristics of the studies included in this review

Study (year), country	Study design	Participants	Sample size, attrition rate	Intervention	Primary outcomes a	Secondary outcomes a
Basch et al.39,40 USA	RCT; single center; 3 groups; random permutated blocks; 12 months; ITT.	Adults receiving outpatient chemotherapy; with metastatic breast, genitourinary, gynecologic, or lung cancers.	766; I = 441, C = 325 (lung cancer: I = 43, C = 26); b 70%.	Website for computer-experienced subgroup, and wireless touchscreen tablet computers or freestanding computer kiosks for computer-inexperienced subgroup; ePROs reported at each clinic visit, and those with home access were encouraged to complete it remotely within 72 h of the visit; alerts were triggered based on predefined conditions and sent to nurses.	HRQL (EQ-5D) at 6 months (P < 0.001) Overall survival at 1 year (P = 0.05) Overall survival at 7 years (P = 0.03)	ED visit (P = 0.02) Hospitalizations (P = 0.08) Duration of Chemotherapy (P = 0.002)
Basch et al.41,42 USA	Cluster RCT; 52 centers; 2 groups; random permutated blocks stratified by rural and urban location; 12 months; ITT.	Practices were asked to consecutively approach and enroll up to 50 adults; with metastatic cancer of any type; receiving treatment with chemotherapy, targeted oral therapy, and/or immunotherapy.	1191; I = 593, C = 598 (no detailed information on the number of lung cancer patients, but reported the number of thoracic cancer); b 57%.	Website or telephone system; ePROs reported weekly and patients received symptom advice booklets.; automated reminders via email, text, or automated call; alerts were sent to nurses via automated emails.	Overall survival at 12 months (P = 0.86)	HRQoL (QLQ-C30) at 1 (P = 0.003), 3 (P = 0.002), 6 (P = 0.006), 9 (P = 0.03), 12 months (P = 0.24) Physical functioning (QLQ-C30) at 1 (P = 0.21), 3 (P = 0.02), 6 (P = 0.003), 9 (P = 0.02), 12 months (P = 0.68) Symptom control (QLQ-C30) at 1 (P = 0.003), 3 (P = 0.002), 6 (P = 0.02), 9 (P = 0.045), 12 months (P = 0.32)
Berry et al.43,44 USA	RCT; 2 centers; 2 groups; block design; during T2 treatment; no information for ITT or per-protocol.	Adult ambulatory patients; with any cancer diagnosis; starting a new medical or radiation treatment regimen.	660; I = 327, C = 333 (lung cancer I = 18, C = 23); b 90%.	Touch-screen, notebook computers; ePROs reported at two visits (before starting treatment and 4–6 weeks later) in clinic waiting rooms; a two-page, color graphical summary of report was automatically generated and provided to the clinical team before the targeted T2 clinic visit; alerts in the reports were flagged by color and height of a bar graph.	Likelihood of SQLIs being discussed (P = 0.032) Depression (SDS-15) (P = 0.02)	Clinic Visit Duration (P = 0.352) Clinician Evaluation of the Intervention: highly usable for identifying appropriate areas of SQLIs, guiding the interview, promoting communication, and for identifying appropriate areas for referral. Nurses were more likely than clinicians to rate the intervention as highly useful in identifying appropriate areas of SQLIs (P = 0.004) and for referral (P = 0.015)
Billingy et al.28,29 Netherlands	Stepped wedge RCT; 13 centers; 3 groups; involving sequential transition of the participating hospitals in a randomized order; 12 months; ITT.	Adult patients with cytologically/histologically proven or radiological suspect small or non-small cell lung cancer; starting treatment with radiotherapy, surgery, chemotherapy, immunotherapy, targeted therapy; ECOG Performance Status of 0, 1, or 2; c internet access.	515; I1 = 160 d, I2 = 89 d, C = 266; b 81%.	Web application; ePROs reported weekly; reminders were sent through push notifications or emails; in the active approach group, alerts were sent to healthcare providers via email, instructing them to contact the patient within 24 h; in the reactive approach group, alerts were sent to patients via pop-up notifications and email, advising them to contact the hospital within 24 h.	HRQoL (QLQ-C30) at 15 weeks (P < 0.001), 6 months (P < 0.001) and 12 months (P = 0.006) Overall survival at 12 months: active approach group (P = 0.16), reactive approach group (P = 0.23)	Progression-Free Survival at 12 months (P = 0.09). Physical functioning (QLQ-C30) at 15 weeks (P < 0.001), 6 months (P < 0.001) and 12 months (P = 0.009) Fatigue (QLQ-C30) (P = 0.009), dyspnea (P = 0.001), constipation (P = 0.001) at 15 weeks
Dai et al.30,31 China	RCT; 3 centers; 2 groups; online central randomization module; during postoperative hospitalization, and 4 weeks post discharge or when adjuvant therapy was commenced; per protocol.	Age 18–75 years; clinically diagnosed with lung cancer (stage I-IIIA); scheduled to undergo surgery; able and willing to fill out electronic questionnaires on smartphones or tablets.	166; I = 65, C = 69; b 81%.	Website; ePROs reported daily post-surgery and twice weekly post discharge for up to 4 weeks; patients received automatic short message reminders to complete the survey; alert triggered at redefined threshold and sent to the treating surgeon; surgeons responded to alerts within 24 h, providing consultation, education, medication guidance, and clinic visit suggestions.	Number of symptom threshold events (MDASI-LC) at discharge (P = 0.007) and at 12 months (P = 0.04)	Composite symptom score post discharge (MDASI-LC) (P = 0.005) and at 12 months (P = 0.31) Composite physical interference score post discharge (MDASI-LC) (P = 0.001) and at 12 months (P < 0.001) Composite affective interference score post discharge (MDASI-LC) (P = 0.014) and at 12 months (P = 0.002) HRQoL post discharge (MDASI-LC) (P > 0.05)
Denis et al.47,48 French	RCT; 5 centers; 2 groups; method of randomization not described; 24 months; ITT.	Adult patients with nonprogressive small cell or non-small cell lung cancer; staged as at least cTxN1/pTxpN1 to TxNxM+ cancer; last treatment (surgery, adjuvant chemotherapy, combined chemotherapy, conventional or stereotactic radiotherapy, first- or second-line chemotherapy) less than three months before random assignment; patients with metastatic lung cancer not progressing on tyrosine kinase inhibitor treatment or maintenance chemotherapy; performance status between 0 and 2; initial symptom score of less than 7 (sum of five self-assessed symptoms: appetite loss, fatigue, pain, cough, and breathlessness).	121; I = 60, C = 61; b 71%.	Application; ePROs reported weekly; an alert email was sent to the oncologist when symptoms met predefined criteria; the application provided an individualized schedule for imaging based on patient symptoms.	Overall survival at 12 months (P = 0.002) and 24 months (P = 0.03)	Quality of Life (FACT-L) at 6 months (P = 0.04) Performance status at first detected relapse (P < 0.001). Optimal treatment initiation (P < 0.001)
Lizée et al.49 French					The incremental ICER of the e-PRO-based strategy compared to conventional follow-up was €12,127 per LYG and €20,912 per QALYS gained The probabilities that the e-PRO-based strategy is very cost-effective (WTP of €30,000/QALYS) and cost-effective (WTP of €90,000/QALYS) were 97% and 100%, respectively	The total cost of care was €802,995 in the experimental group and €707,109 in the control group The total incremental LYG and QALYS in the experimental arm compared to the control arm during the study period were 7.9 years and 4.6 QALYS, respectively
Kearney et al.50 UK	RCT; 7 centers; 2 groups; automated interactive voice response telephone randomization system; 4 chemotherapy cycle; ITT.	Adult patients with a diagnosis of breast, lung, or colorectal cancer; commencing a new course of chemotherapy treatment; receiving outpatient chemotherapy.	112; I = 56, C = 56 (lung cancer: I = 26, C = 13); b 52%.	Mobile phone-based system; ePRO survey daily in the morning, evening, and at any time they felt unwell; patients received written feedback on the mobile phone interface, comprising tailored self-care advice; reports sent to clinicians with an ‘amber alert’ indicated toxicities that were not severe or life-threatening but required early intervention; a ‘red alert’ indicated severe toxicities; clinicians were advised to contact patients within 1 h of receipt of a red alert.	Chemotherapy-related morbidity of six common chemotherapy-related symptoms (P = 0.04) Reports of hand–foot syndrome (P = 0.031)	Severity and distress of the six symptoms (severity: P = 0.033, distress: P = 0.028). Fatigue (severity P = 0.18, distress: P = 0.13)
Kuo et al.51 Canada	RCT; 2 centers; 2 groups; randomization stratified by treating oncologist, planned chemotherapy treatment; ECOG status (0 or1 vs 2 or greater); 6 months; no information for ITT or per-protocol.	Histologic or cytologic diagnosis of incurable stage IIIB or IV non-small cell lung cancer; no prior chemotherapy; plans to commence first-line chemotherapy; ECOG performance status 0–3. c	102; I = 44, C = 51; b 94%.	Handheld pocket personal computer with a touch screen and stylus; ePROs survey before each chemotherapy cycle (every 3 weeks), and at each follow-up visit until disease progression; report was presented to the treating oncologist before the patient’s assessment, with trends highlighted in a 2-page graphical report; Patients in the control group completed the same PRO survey at the same time without the data being presented to the oncologist.	PC referral rate (P = 0.54) Time to PC referral (P = 0.98)	Number of first-line chemotherapy cycles administered (P = 0.58) Referral to homecare nursing (P = 0.25), pain clinic (P = 0.29), palliative radiation (P = 0.80), home oxygen (P = 0.19), bisphosphonate (P = 1.00), transfusions (P = 0.50). Appetite stimulants (P = 0.89) HRQoL (eLCSS-QL) (P > 0.05)
Parikh et al.45 USA	RCT; single center; 3 groups; electronically randomization stratified by cancer type; 6 months; ITT.	Adult patients diagnosed of incurable or Stage IV lung or gastrointestinal cancer; receiving primary oncology care; currently receiving or planned to receive IV chemotherapy within 2 weeks; have a smartphone that can receive SMS text messages and has Bluetooth capability.	108; I1 = 24 e, I2 = 25 e, C = 33 (lung cancer: I1 = 8, I2 = 10, C = 7); b 76%.	Text message for weekly ePROs survey and wearable Fitbit device for step data; patients in ‘PROStep + nudge’ group received a text encouraging them to contact their oncology team if they reported a severe symptom; ‘PROStep’ and ‘PROStep + nudge’ group both reported ePROs, ePROs data summarized in a dashboard for both clinicians and patients’ PGHD was summarized and provided in paper or electronic form to the patient’s oncology clinician on the morning of the clinic visit.	Patient-perceived understanding of symptoms (P = 0.85) and functional status (P = 0.59)	Adherence to ePROs surveys was 63.5%. Hospitalization (P = 0.78) PC consults (P = 0.61)
Strasser et al.52 Switzerland	Cluster RCT; 8 centers; 2 groups; block randomization method via fax; 6 weeks; ITT.	Adult patients with incurable, symptomatic solid tumors; receiving new outpatient chemotherapy with palliative intent; expected tumor response rate ≤ 20%; clinically not cognitively impaired; symptomatic defined as at least one ESAS symptom score ≥ 3.	264; I = 145, C = 119 (lung cancer: I = 25, C = 25); b 39%.	Palm device; ePROs and estimated nutritional intake, body weight change immediately reported before the weekly oncologists’ visit; a cumulative, longitudinal monitoring sheet was printed immediately and provided to the oncologists.	HRQoL (QLQ-C30) at 6 weeks (P = 0.06)	Symptom distress score (QLQ-C30) (P = 0.003) Symptom management performance (P = 0.06) Coping, and treatment burden, at differences of 0.54 (95% CI [0.41, 0.67]) and 0.52 (95% CI [0.39, 0.66]) between groups
Zhang et al.46 China	RCT; 28 centers; 2 groups; computer-generated sequential number; 6 months or the end of therapy; no information for ITT or per-protocol.	Adult patients receiving cancer immunotherapy; ECOG performance status of 0 or 1; c life expectancy of at least 6 months; able to use smartphones or computers to include information in the app with or without the help of their caregivers.	278; I = 141, C = 137 (lung cancer: I = 13, C = 19); b 92%.	Application; ePROs reported weekly and pictures uploaded of examination results; reminder messages to patients who did not reply for two days, if no reply was obtained, follow-up by telephone was conducted; standardized advice was sent to patients for grade 1/2 irAEs; text message, email, and application alerts were sent to the healthcare team simultaneously for grade 3/4 irAEs.	Incidence of severe (3/4 irAEs) (P = 0.01) ED visits (P = 0.01) Treatment discontinuation (P = 0.02) Death (P = 0.28)	HRQoL (QLQ-C30) at 6 months (P = 0.001) Time spent implementing the ePRO model (P < 0.001)

RT, randomized controlled trial, eROs electronic patient-reported outcomes, HRQoL health related quality of life, ED emergency department, SQLIs symptoms and quality-of-life issues, ICER incremental cost-effectiveness ratio, LYG life year gained, QALYSs quality-adjusted life years, PC palliative care, irAEs immune-related adverse events, EQ-5D EuroQol Five Dimensions Questionnaire, QLQ-C30 the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire core 30 items, SDS-15 Self-Rating Depression Scale 15 items, MDASI-LC MD Anderson Symptom Inventory-Lung Cancer, FACT-L Functional Assessment of Cancer Therapy-Lung, eLCSS-QL the electronic version of Lung Cancer Symptom Scale.

^a The P value < 0.05 referred to better effects on HRQoL, overall survival, progress free survival, physical functioning, symptom control and other health outcomes, less healthcare services utilization, and better adherence in the intervention group.

^b I referred to intervention group, C referred to control group.

^c EGOG performance status referred to Eastern Cooperative Oncology Group performance status, which is a scale used to assess how a patient’s disease is progressing and how it affects their daily living abilities. It measures a patient’s level of functioning, including their ability to care for themselves and perform daily activities. This status is particularly important for determining a patient’s ability to tolerate therapies, especially in serious illnesses like cancer.

^d I1 and I2 referred to two intervention groups, the active approach group and reactive approach group, which sent alert signals to healthcare providers and patients themselves, respectively.

^e I1 and I2 referred to two intervention groups, the ‘PROStep’ and ‘PROStep + nudge’ group, and patients in the two groups reported ePROs at the same time and their clinicians received the same alerts if sever symptom occurred. In addition, patients in the ‘PROStep + nudge’ received a message that encouraged them to seek for medical service after the alert.

The randomization was primarily performed by computers, two studies allocated participants through permuted blocks via computer^39,40, two studies used computer-generated sequential numbers^30,31,46, Billingy et al. ^28,29 involved sequential transition of the participating hospitals in a randomized order, three studies randomized participant stratified by rural and urban location^41,42, treating oncologist, treatment and ECOG (Eastern Cooperative Oncology Group)⁵⁵ status⁵¹, and cancer type⁴⁵, respectively. On the other hand, Strasser et al.⁵² and Kearney et al.⁵⁰ performed randomization via fax and interactive voice response telephone system, respectively. The study in France did not reveal the methodology for randomization^47–49.

Characteristics of patients

Only three studies solely enrolled patients with lung cancer, among which, studies from the Netherlands^28,29 and France^47–49 involved patients with small or non-small cell lung cancer during any type of treatments, while Dai et al.^30,31 targeted patients with stage I-IIIA lung cancer post-surgery. The other eight studies recruited patients with multiple types of cancer, including lung cancer, undergoing various treatments such as chemotherapy (six studies), radiotherapy (one study), or immunotherapy (one study). A total of 4283 patients were involved in this review, at least 1160 patients with lung cancer were involved, since one study only did not mention the exact number of lung cancer patients involved^41,42. The sample size ranged from 102⁵¹ to 1191^41,42. Attrition rate fluctuated from 39%⁵² to 90%^43,44, and if not reported, retention rate was calculated by dividing the number of patients successfully followed up by the number of participants in the study.

Characteristics of ePROs-based symptom monitoring interventions

In terms of the intervention duration, most were no more than 6 months (eight studies), only interventions in two studies lasted for 12 months, and one for 24 months. Categorizing by the methods patients reported, ePROs reporting tools could be classified into websites (four studies), applications (three studies), and handheld systems or touch-screen notebook (four studies). Parikh et al.⁴⁵ combined wearable devices “Fitbit” for step data and applications for ePROs reporting.

Among the eleven studies included, seven required patients in the intervention group to report ePROs weekly, Berry et al.^43,44 and Kuo et al.⁵¹ collected ePROs data from patients in the intervention group at every visit. Dai et al.^30,31 prompted patients in the intervention group to report ePROs daily post-surgery and twice weekly post discharge for up to 4 weeks, and Kearney et al.⁵⁰ demanded patients in the intervention group to report ePROs on a daily basis or at any time they felt unwell.

Several approaches were taken to remind patients in the intervention group of reporting ePROs routinely. Reminders were sent primarily, solely, and most frequently via notifications on the reporting application (two studies), automated text messages (two studies) or emails (one study). Two studies reminded patients in the intervention group of weekly ePROs reporting via combined ways, Billingy et al.^28,29 used both push notifications and emails whereas Basch et al.^41,42 utilized texts and automated calls. The other four studies did not provide information on reminders. The reminders could be taken as a way to strengthen adherence and compliance.

Ten out of eleven studies generated alerts according to a predefined algorithm. The majority of the ePROs-based interventions sent alerts by email (four studies) or through web-based or application systems (three studies), two studies sent alerts through text messages and one study alerted patients through multiple ways, including pop-up notifications and emails. However, Kuo et al.⁵¹ did not generate alerts, but presented all ePROs data with trends highlighted in a 2-page graphical report to the oncologists.

Only two studies sent alerts to both healthcare providers and patients, and the others sent alerts only to healthcare providers, either designated clinicians (four studies), or nurses (two studies), or the healthcare team (two studies), reminding healthcare providers of responding patients’ symptoms timely. Specifically, Billingy et al.^28,29 allocated consented participants into either one control group or two intervention groups differing by alert management, ‘active’ alerts to healthcare providers or ‘reactive’ alerts advising patients to contact the hospital within 24 h. Parikh et al.⁴⁵ allocated participants into arm A (control), or arm B or C (intervention groups), whereby arms B and C sent texts encouraging patients who reported a severe symptom to contact their oncology team, as well as an electronic health message sent to their oncologists, moreover, arm C received additional brief text message feedback including an ‘active nudge’ question asking “Do you plan on discussing these symptoms with your oncologist at your upcoming visit? ”.

Some studies further displayed different types of alerts based on severity of ePROs assessed symptoms, and provided additional services. Berry et al.^43,44 presented alerts flagged by color and height of a bar graph. Dai et al.^30,31 encouraged surgeons responded to alerts within 24 h, and to provide consultation, education, medication guidance, and clinic visit suggestions. Denis et al.^47,48 provided an individualized schedule along with the alerts. Kearney et al.⁵⁰ offered patients with written feedback on the mobile phone interface, comprising tailored self-care advice after reporting ePROs, and alerts were classified into an ‘amber alert’ that indicated toxicities that were not severe or life-threatening but required early intervention, and a ‘red alert’ that indicated severe toxicities, and clinicians were advised to contact patients within one hour of receipt of a ‘red alert’. Zhang et al.⁴⁶ sent standardized advice to patients who reported mild symptoms and sent alerts to the healthcare team simultaneously if severe symptoms occurred.

Characteristics of controls

Participants from the control groups of the studies by Berry et al.^43,44 and Kuo et al.⁵¹ received monitoring alone, and all participants from the control groups in the other studies received usual care.

Characteristics of outcome measures

To collect and assess ePROs data during routine follow-up, studies either used existing symptom measurements, or developed new ePROs questionnaires. The majority of studies selected common cancer symptoms that were tested in pilot trials or recognized with clinical consent, from the National Cancer Institute-Common Terminology Criteria for Adverse Events (CTCAE) to form the ePROs survey (seven studies)⁵⁶. Berry et al.^43,44 utilized a self-developed quality of life (QoL) questionnaire after validating it in several pilots^57–59, the other three studies used MD Anderson Symptom Inventory-Lung Cancer module (MDASI-LC)⁶⁰, the electronic version of Lung Cancer Symptom Scale (eLCSS-QL)⁶¹, and Edmonton symptom assessment score (ESAS)⁶², respectively.

Five measures were applied to measure HRQoL among participants in nine studies, and two included studies did not measure HRQoL. Six studies applied the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire C30 (EORTC QLQ-C30), one used single-item QoL scale (SIQoL)⁶³, the other studies made use of Functional Assessment of Cancer Therapy-Lung (FACT-L)^64,65, eLCSS-QL and EuroQoL EQ-5D Index⁶⁶, respectively.

Risk of bias

Risk of bias assessments according to the revised Cochrane risk-of-bias version 2 tool are detailed in Fig. 2. We combined papers from the same study such as Denis et al.^47,48 and Lizée et al.⁴⁹. Among eleven studies included in this review, eight articles performed intention-to-treat analysis, with three and five studies graded as low risk of bias^28,29,45,52 and some concerns^{39–44,47–50}, respectively. One study with per-protocol analysis showed some concerns^30,31. Kuo et al.⁵¹ and Zhang et al.⁴⁶ did not provide information on principals for data analysis, and the two articles of theirs were ranked as some concerns. Concerns of risk of bias were raised due to the deviations from the intended interventions (six articles), measurements of the outcome (twelve articles), and selection of the reported result (four articles).

Fig. 2.

Risk of bias summary of the included studies.

Meta-analysis

HRQoL

For RCTs with more than one paper, or one paper reported HRQoL outcomes at several time points, only one paper and results from one time point was included in a meta-analysis. Six studies were included in the meta-analysis due to changes in “global HRQoL”, a composite score of all aspects of HRQoL, of baseline, mean, and standard deviation (SD). Among the six studies, Billingy et al.^28,29 reported results at 15 weeks, 6 months and 12 months in 2 papers. Zhang et al.⁴⁶ reported results at 3 months and 6 months. Three studies included patients with various types of cancer and reported the number of patients with lung cancer, but one study failed to report the number of patients with lung cancer and only reported the number of thoracic cancer patients. We provided the standard mean differences (SMDs) and 95% confidence intervals (CIs) to illustrate the pooled effects of HRQoL.

Data from 6 studies that included 3135 participants were used to evaluate the effectiveness of ePROs-based symptom monitoring on global HRQoL. At least 390 and 397 patients with lung cancer were involved in the intervention group and control group, respectively. The results of the random effects model were demonstrated in Fig. 3. First, we included the two studies that only involved patients with lung cancer and both reported HRQoL outcomes at 6 months^29,47. Fig. 3a illustrated that ePROs-based intervention could significantly enhance HRQoL (SMD = 2.44, P < 0.001) but with a wide 95% CI (−1.35, 6.22). Second, we involved five studies, consisting of two papers that exclusively enrolled lung cancer patients and three that involved multiple cancer types (with unspecified numbers of lung cancer patients, precluding assessment of heterogeneity across cancers)^{29,39,46,47,52}. As presented in Fig. 3b, ePRO-based interventions also promoted cancer patients’ HRQoL (SMD = 0.68, 95% CI [−0.33, 1.68], P < 0.001). Third, we added the paper that did not report the number of patients with lung cancer⁴¹, and assumed that the number of thoracic cancer patients could represent the weight of lung cancer (Fig. 3c), finding that the effects of ePROs was still significantly positive (SMD = 0.58, 95% CI [−0.30, 1.47], P < 0.001).

Fig. 3. Results of meta-analysis for the effects of ePROs-based intervention on HRQoL.

a Presents the pooled effects for patients with lung cancer; b presents the results after including studies that involved lung cancer and clarified the number of lung cancer participants; c the results after including all studies that involved lung cancer and reported means and standard deviations, where the number of lung cancer participants were not reported, the number of thoracic cancer patients were used. ePROs electronic patient-reported outcomes, SD standard deviation, SMD standardized mean differences, HRQoL health related quality of life.

For subgroup analysis, we analysed the heterogeneity in the effects on HRQoL among different intervention durations and lengths of treatment regimens. In Fig. 4a, the pooled results of the subgroup analysis for duration of intervention presented a significant improvement in HRQoL (SMD = 1.09, 95% CI [0.18, 2.01], P < 0.001). Different duration of ePROs-based symptom monitoring all showed significant effects on HRQoL (P < 0.001), the effect of a duration of 6 months was the highest (SMD = 1.41, 95% CI [−0.89, 3.71]), followed by a duration for more than 6 months (SMD = 1.14, 95% CI [−0.83, 3.10]) and less than 6 months (SMD = 0.64, 95% CI [−048, 1.76]). Furthermore, as implemented in Fig. 4b, the effect from a synthesis treatment for cancers (SMD = 1.58, 95% CI [0.50, 2.65], P < 0.001) was higher than that from chemotherapy (SMD = 0.14, 95% CI [−0.07, 0.35], P < 0.001).

Fig. 4. Subgroup analysis for effects of ePROs-based intervention on HRQoL.

a Shows the effects grouped by different duration of interventions, categorized as ‘less than 6 months’, ‘6 months’, ‘more than 6 months’; b shows the effects grouped by patients undergoing different treatments, classified as ‘chemotherapy’ and ‘all types of treatment’. ePROs electronic patient-reported outcomes, SD standard deviation, SMD standardized mean differences, HRQoL health related quality of life.

We conducted sensitivity analyses by excluding one study at a time, and found ePROs-based symptom monitoring and management still significantly improved HRQoL (Supplementary Fig. 1).

Overall survival

Five papers from four studies assessed the impact of ePROs-based intervention on the overall survival with hazard ratios (HRs) (Fig. 5). Basch et al.^40,42 reported overall survival for multiple cancers from two studies followed up at 12 months and 7 years, respectively. Billingy et al.²⁹ reported HRs for two intervention groups with either active (T1) or reactive alert (T2), compared with the control group, Denis et al.^47,48 elicited the interim analysis and final results of HR. In Fig. 5a, we pooled results from two studies that only included patients with lung cancer and reported HRs at 12 months, the random effects model illustrated no significant difference between intervention and control groups (HR = 0.62, 95% CI [0.51, 0.88], P = 0.098)^28,47. As is shown in Fig. 5b, when excluding the HR from reactive alert group, ePROs-based intervention significantly prolonged survival (HR = 0.54, 95% CI [0.22, 1.31], P = 0.031)^28,47. After including additional results from three studies that covered multiple cancers, and keeping results measured at 24 months by Denis et al. instead of the ones collected at 12 months, the effect of ePROs-based intervention for survival among patients with lung cancer increased to an HR of 0.81 (95% CI [0.67, 0.97], P = 0.074) (Fig. 5c)^28,40,42,47. We used leave-one-out approach to ensure robustness (Supplementary Fig. 2).

Fig. 5. Results of meta-analysis the effects of ePROs-based intervention on overall survival.

a Presents the pooled effects for patients with lung cancer; b showed the results when including studies that used active alert approach; c showed the results when including all studies that involved lung cancer and reported hazard ratios. ePROs electronic patient-reported outcomes, HR hazard ratio.

Symptom burden

Out of eighteen papers, seven papers reported symptom burden. Two studies calculated a composite score of symptom burden, one resulted in 16.1% more patients in the intervention group gaining symptom control benefits compared with the control group (OR, 1.50 [95% CI, 1.15–1.95]; P = 0.003) (EORTC QLQ-C30)⁴¹, the other reported the symptom score was lower by 0.63 between the intervention group and control group (MDASI-LC)³⁰. The two studies also provided scores for each symptom. Basch et al.^41,42 reported the changes in pain, nausea, vomiting, diarrhea, constipation, fatigue, insomnia, appetite loss, and dyspnea, in which fatigue (mean change from baseline, −4.90 [−8.12, −1.68]; P = 0.003) was significantly improved at the endline of the trial. Dai et al.^30,31 stated fluctuations in fatigue, distress, appetite loss and mood scores during intervention, and indicated lower burden for all symptoms. In some studies, only outcomes of certain symptoms were reported, for instance, Billingy et al.^28,29 clarified the scores of fatigue, dyspnea, constipation and all symptoms from the EORTC QLQ-C30 symptom scales. Kearney et al.⁵⁰ provided severity and distress from nausea, vomiting, mucositis, hand-foot syndrome, diarrhea, and fatigue. The remaining two studies focused solely on depression- or distress-related symptoms. Berry et al.⁴⁴ assessed depressive symptoms using the SDS-15⁶⁷ and reported a 1.21-point reduction in SDS-15 scores within the intervention group. Strasser et al.⁵² evaluated symptom distress using the sum of nine ESAS items. Other key symptoms such as anxiety and sleep problems was measured in four studies, but the scores were not provided^{30,31,41–45}.

Physical functioning

Six papers from three studies reported physical functioning outcomes. Assessed with the EORTC QLQ-C30, Basch et al.⁴¹ reported that 13.8% more patients in the intervention group experienced physical function benefits than in the control group. Billingy et al.^28,29 demonstrated an increase in physical functioning in the intervention group increased by 1.73, 1.78, and 1.83 at 15 weeks, 6 months and 12 months, respectively. Strasser et al.⁵² compared the difference of 5.16 in physical functioning between the intervention and control arms. With MDASI-LC, Dai et al.^30,31 found a decrease composite physical interference score by 0.72 and 0.86 in the intervention group after the intervention and at 12-month follow-up.

Healthcare service utilization

Five papers reported the impact of ePROs-based symptom management on healthcare service utilization. For emergency room (ER) visits, Basch et al. (P = 0.02)³⁹ and Zhang et al. (P = 0.01)⁴⁶ both found significantly more patients visited ER in the control group than in the intervention group. For hospitalization, Basch et al.³⁹ reported 45% and 49% of patients were hospitalized in the intervention group and in the control group, respectively (P = 0.08). Parikh et al.⁴⁵ found no significant differences in hospitalization between control and intervention arms (P = 0.78). For palliative care (PC) visits, Kuo et al.⁵¹ concluded that PC referral rate (P = 0.54) and time to PC referral did not differ between groups (P = 0.98). Parikh et al.⁴⁵ also drew the same conclusions that there was no difference of PC consults between groups (P = 0.61). For treatment discontinuation, Kuo et al.⁵¹ found no difference in the number of first-line chemotherapy cycles administered between arms (P = 0.58). While Zhang et al.⁴⁶ reported significantly less treatment discontinuation in the intervention group (P = 0.02). Moreover, Berry et al.⁴³ illustrated that there was no significant difference of duration of clinical visits between groups (P = 0.352), but patient-clinician communication was significantly enhanced in the intervention group (P = 0.032).

Study adherence, burden and implementation

Four studies provided potentially useful data on study adherence, burden and implementation. As for patient adherence, Parikh et al.⁴⁵ reported that the adherence to ePROs surveys was of 63.5% and to Fitbit step count collection was 52.7%, mean composite adherence to both ePRO and Fitbit was 47.8%. As for burden and implementation, Strasser et al.⁵² evaluated the study burden from the perspectives of patients by asking patients to rank their coping ability from ‘no effort at all’ to ‘a great deal of effort’, and to describe their burden from ‘not at all’ to ’severely’, which resulted in differences of 0.54 (95% CI [0.41, 0.67]) and 0.52 (95% CI [0.39, 0.66]) for coping and burden between groups. From the perspectives of researchers, Zhang et al.⁴⁶ measured the burden of implementing ePROs-based intervention by comparing the time spent implementing the ePRO survey, and on average 8.2 min and 36.1 min were spent on the intervention group and control group, respectively (P < 0.001). From the perspectives of healthcare providers, Berry et al.^43,44 involved clinicians and nurses, which acknowledged that the ePROs-based intervention was highly usable for identifying concerning conditions and promoting patient-clinician communication⁴³.

Cost-effectiveness

Only one study calculated the cost-effectiveness of the effectiveness and cost of conducting a RCT that imposed ePROs-based symptom monitoring on patients with lung cancer⁴⁹. In the cost-effectiveness analysis, the total cost of care was €802,995 in the intervention group and €707,109 in the control group, and the total incremental life year gained (LYG) and quality-adjusted life years (QALYs) in the experimental arm compared to the control arm during the study period were 7.9 years and 4.6 QALYs, respectively, underscoring the surveillance of lung cancer patients using ePROs reduced the follow-up costs and was cost-effective.

We assessed quality of evidence according to Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) framework, and the quality of evidence for HRQoL and overall survival was moderate (Supplementary Table 2).

Discussion

In this review, we retrieved data from 18 papers from 11 RCTs to examine the effectiveness of ePROs-based symptom monitoring for patients with lung cancer. Among the 11 included studies, only 3 (27%) exclusively enrolled lung cancer patients, while 8 (72%) recruited mixed-cancer populations. A total of 4283 patients were involved in this review, including at least 1160 patients with lung cancer. Sample sizes ranged from 102 to 1191 participants across studies, with attrition rates varied substantially, from 39% to 90%. The interventions primarily collected ePROs via websites (4/11, 36%), mobile applications (3/11, 28%), or handheld/touch-screen devices (4/11, 36%). Most interventions (72%, 8/11) lasted for less than 6 months.

The intervention was mainly comprised of a reminder, ePROs reporting via digital devices, alerts for concerning issues sent to patients themselves and/or healthcare providers, and additional services. As a way to enhance patients’ adherence and compliance to study protocol, the majority of studies automatically reminded patients in the intervention group of reporting symptoms weekly (7/11, 64%). Given that addressing alerting symptoms timely was the primary mechanism to improve health outcomes, most studies (9/11, 82%) sent alerts to healthcare providers only to prompt clinicians to cope with patients’ symptoms (active alert). On the other hand, two studies also sent alerts to patients themselves (reactive alert), trying to eliminate clinicians’ burden of handling alert signals and to compare the effectiveness of reactive and active alerts. Nudge strategies, such as asking patients whether they would take action after receiving the alerts, or highlighted or flagged alerts in the symptom reports, were utilized in three studies included^43,45,51 and one protocol retrieved⁶⁸ (the final results have not been published yet) to bridge the core process of ePROs-based intervention, which was to facilitate patients or healthcare providers to notice the concerning symptoms before they deteriorated. There were substantial evidence supporting the notion that ePROs-based symptom monitoring could strengthen HRQoL and prolong overall survival among patients with lung cancer. We also explored the heterogeneity among different duration of interventions and patient treatments, and described the implementation fidelity and cost-effectiveness of ePROs-based intervention.

Examined with the revised Cochrane risk-of-bias version 2 tool, the risk of bias among included studies were of low risk or some concerns. Overall, 8 (73%) and 1 (9%) study out of 11 studies used intention-to-treat and per protocol analysis, respectively. 27% (3/11) of the studies indicated a low risk of bias and 73% (8/11) had some concerns. In light of the GRADE framework, the overall quality of evidence and methodology was moderate.

Six studies (6/11, 55%) reported the mean and SD in HRQoL, two of them exclusively targeted patients with lung cancer, and four of them enrolled multiple cancers and did not reveal heterogeneity tests for different cancers. The meta-analysis showed a significant improvement in HRQoL at 6 months in the two studies that only involved lung cancer (SMD = 2.44, 95% CI [−1.35, 6.22], P < 0.001)^29,47. The results were robustly tested with leave-one-out approach. Although only two studies were included, this is the first pooled result to declare an increase of 2.44 in HRQoL of ePROs-based symptom monitoring among lung cancer. The effectiveness in enhancing lung cancer patients’ QoL may be derived from the ePROs’ capability of capturing and managing cancer related symptoms before they deteriorated, and promoting patient-clinician communications^25,69,70. Then, we examined the effects of all six studies that involved lung cancer to integrate the effects of ePROs-based interventions in oncology, and found the results were significantly positive but again with a wide 95% CI^{29,39,41,46,47,52}, which were consistent with previous conclusions about other cancers, underscoring the benefits of ePROs-based intervention usage in oncology care^19,38. Subgroup analyses for effects on HRQoL among different duration of interventions illustrated that the effect size diminished over time. After separating the effects for different treatments, ePROs-based interventions demonstrated better effects when synthesizing the impacts on patients under various treatments than under chemotherapy. The difference of HRQoL measured between groups were larger among patients during immunotherapy than during chemotherapy, suggesting ePROs-based intervention leveraged in other treatments, especially immunotherapy, may lead to optimal effects^46,71. However, the pooled effects of six studies should be interpreted with caution, since they could not represent the exact effect size for ePROs-based interventions in lung cancer, and could only indicate that positive impacts existed in studies that covered lung cancer participants. Further trials and meta-analysis may consider exploring heterogeneity related to study participants’ cancer type and treatments.

In terms of overall survival, HRs from two studies (2/11, 18%) were employed and the pooled results of these studies revealed positive but not significant improvement among lung cancer patients at 12 months^29,47, and the results were reliable with robustness tests. Although Billingy et al.^28,29 proved there were significant advances in HRQoL, symptom control and physical functioning with reactive alerts sent to patients themselves, the effect sizes were relatively smaller than with active alerts. After excluding the results from the reactive alert group, ePROs significantly improved overall survival (HR = 0.54, 95% CI [0.22, 1.31], P = 0.031), which was identical to results of similar meta-analyses^21,72,73. Evidence suggested the effect may diminish over time, but we failed to elicit the trend for survival among patients with lung cancer with limited records. Few researchers have included overall survival in their outcomes in protocols, but the results have not been reported yet, and more analyses may be performed with these data^54,68. The prognostic values and long-term effects of ePROs for survival may be explored in future studies.

In addition, we reviewed evidence for effectiveness of ePROs-based symptom monitoring for patients with lung cancer on alleviating symptom burden^{28–30,41,44,50,52}, strengthening physical functioning^{28–31,41,52}, reducing healthcare service utilization^{39,43,45,46,51}, better patient adherence, and lowering implementation burden for researchers^43,45,46,52. Currently, we acknowledge that extrapolating these findings to different populations may be risky, because the sustainability, generalizability and scalability of ePROs-based interventions for lung cancers, and burden and satisfaction from patients’ perspectives were not thoroughly explored. In the included studies, patient experiences were assessed minimally via two simple questions in ref. ⁵². To address this limitation, mixed-methods approaches and implementation science frameworks could be utilized to help incorporate such interventions into routine care and explain intervention mechanisms^22,35,74. Trials with ePROs-based intervention was highly cost-effective, however, only one study reported related results for patients with lung cancer⁴⁹. Nixon et al.⁷⁵ demonstrated the advantages of implementing ePROs-based symptom monitoring in reducing costs and improving QALYs with Markov model, there is still a lack for trial-based evidence that could quantify the cost-effectiveness of ePROs-based interventions in oncology. On the one hand, symptom burden, physical functioning, and mental well-being are aspects of HRQoL and greatly associate with overall survival. Moreover, symptoms, physical and mental outcomes, and healthcare service utilization were important indicators and mediator to show the effectiveness of ePROs-based intervention. On the other hand, adherence, burden, and cost-effectiveness related outcomes were key indicators to test implementation fidelity of ePROs-based interventions. Consequently, concluding the results of ePROs-based interventions on these outcomes may contribute to thoroughly understanding the effectiveness of ePROs-based interventions, explaining the mechanisms between ePROs-based intervention and HRQoL and survival, and providing empirical evidence to utilize ePROs in clinical practice for patients with lung cancer^13,14,76,77.

There are several limitations in this review and should be considered carefully when drawing a conclusion. First, this review only included a small number of studies and limited evidence were found to explain the effects of ePROs-based symptom monitoring on each outcome. With limited evidence, the directionality and significance of ePROs-based intervention on lung cancer was more worthy of reference, rather than the absolute pooled effect sizes of HRQoL and overall survival. Although the studies acknowledged the importance of ePROs, most of them used ePROs as outcome measures or as intervention for feasibility tests, only few RCTs conducted ePROs-based interventions and included patients with lung cancer^21,78. Second, only two studies were included to calculate the pooled effects for HRQoL and overall survival among patients with lung cancer, potentially hindering the reliability of the effect sizes. The majority of included studies targeted patients with not only lung cancer, but also other cancers, and heterogeneous effects for different cancers were not reported. Despite the small number of studies included and small sample sizes, this is still the first meta-analysis to reveal the effect sizes of ePROs-based interventions on lung cancer. We also provided the pooled effects for HRQoL and survival from all included studies where data were available to help support our results and future research, and performed sensitivity analyses to ensure robustness. Similarly, the implications for intervention implementation were minimal due to insufficient data. Third, the included studies employed ordinal scaling properties with different measures of HRQoL, and thus, SMDs may not provide precise effect sizes. Future studies could consider linking PRO measurement scores, e.g., using item response theory methods, to explain the meanings of HRQoL derived from different scales. Fourth, we could not explore every possible heterogeneity with limited data, such as the differences in effects among states or countries and outcome measurements, though, essential examinations of effects among different intervention duration and treatments were investigated. Last, publication bias may not be eliminated, even with comprehensive data searching in five databases.

Symptom monitoring on ePROs and subsequent alerts sent to patients or healthcare providers have been recognized as a package of intervention which could benefit in HRQoL, overall survival, and other outcomes for cancers in RCTs. Previous reviews have demonstrated the effectiveness of these interventions in various cancers, and pooled the effect sizes for cancers, or an isolate cancer, such as breast cancer or prostate cancer^{19,21,22,38,73,79–81}. To our knowledge, this is the first meta-analysis to examine the effectiveness of ePROs-based symptom monitoring among patients with lung cancer. Furthermore, we were novel to explore the heterogeneity for effects on HRQoL and overall survival, and we demonstrated the impacts on other health outcomes, implementation process evaluation, and cost-effectiveness. The findings provide robust evidence to explain the mechanisms of ePROs-based interventions, and to support the potential of integrating ePROs-based symptom monitoring into routine clinical care for patients with lung cancer. Researchers, clinicians, and healthcare systems may prioritize the implementation of tailored ePROs interventions, particularly those incorporating real-time alerts. Moreover, the feasibility and effectiveness of reactive or active alerting approaches may still need exploration, and possible measures, such as nudging, could be taken to optimize adherence and effectiveness. Furthermore, the observed heterogeneity in intervention effectiveness highlights the need for adaptive frameworks that optimize duration, measurement tools, and alert mechanisms to enhance patient-clinician communication and self-management capabilities. Future research efforts may address barriers such as digital literacy, patients’ burden and experiences, and cost-effectiveness.

This systematic review synthesizes evidence in 18 papers from 11 RCTs to demonstrate the substantial role of ePROs-based symptom monitoring in lung cancer management. This is the first study to illustrate the effect size of such interventions on HRQoL and overall survival for lung cancer patients. However, caution should be taken in interpreting the results of this review due to the small number of studies included, heterogeneity, as well as the wide 95% CIs. Future research may prioritize pragmatic trials evaluating long-term effects for survival and other outcomes, the implementation process, and health economics indicators for intervention combining ePROs. This study will help advocate for ePROs as a cornerstone of remote symptom monitoring among lung cancer, enhancing patient self-management and patient-clinician communications.

Methods

Search strategy

The review was registered in The International Prospective Register of Systematic Reviews (PROSPERO) (registration number: CRD420251000397) and reported following the standards of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 statement (PRISMA 2020 Checklist)⁸². Medline, Embase, Cochrane library, CINAHL and APA PsyInfo were searched on September 19th, 2024, March 1st, 2025 and April 23rd, 2025, using broad search terms and MeSH terms. The search strategy was guided by the Cochrane methodology using the PICOS tool (Population, Intervention, Comparison, Outcome and Study design), with keywords used across the databases, adapting Boolean operators and MeSH vocabulary⁸³. The search terms were optimized to capture studies related to “lung cancer”, “digital health intervention”, “ePROs”, “symptom monitoring” and “randomized controlled trials (RCTs)”.

Study eligibility criteria

Studies were included if they met the inclusion criteria in Table 2. Briefly, RCTs or cluster RCTs of ePROs-based symptom management in adults with lung cancer, implementing digital health interventions and published in English before April 23rd, 2025 (date of last search) were considered for inclusion. Studies including mixed disease groups were eligible, if lung cancer related data could be extracted. Any type of digital health intervention was eligible for inclusion, except for interventions only via telephone calls or message texts because of their limited reflection of contemporary digital health tools. In this review, primary outcomes encompassed HRQoL, defined as multidimensional functioning and well-being across physical, mental, and social domains, which aligned with one of the first and most widely acknowledged definitions of HRQoL^84,85, and survival outcomes. Secondary outcomes involved aspects of HRQoL outcomes, including symptom burden (physical symptoms such as pain, fatigue and sleep problems, and mental symptoms such as distress, anxiety and depression) and physical functioning (limitations in daily activities and physical capabilities)^19,21,37,38, and implementation-related outcomes such as healthcare service utilization, patient-clinician communication, study adherence, burden, implementation, and cost-effectiveness^{13,14,76,77,86}. However, studies reporting any primary outcome listed in Table 2 were included due to pragmatic consideration. Studies were excluded if they were book chapters, conference abstracts, commentaries, opinion articles, reviews, meta-analyses, unpublished data, and so on, or if full texts or complete data were unavailable. The full search strategies can be found in Supplementary Note.

Table 2.

Inclusion criteria defined by PICOS

Inclusion criteria
Participant	Patients with any type, stage or treatment of lung cancer; aged ≥18 years
Intervention	Participants in the intervention group report ePROs electronically via websites, computers, mobile phones, tablets, etc, with responses to patients themselves and/or clinical teams
Comparator	Participants in the control group received usual care (standard care, traditional care or routine care in researcher’s settings)
Outcomes	Outcomes included one of the following indicators: HRQoL, overall survival, symptom burden, physical functioning, psychological well-being, healthcare service utilization, patient-clinician communication, study adherence, burden, implementation, and cost-effectiveness
Study design	RCTs (single-center, multicentre, cluster, stepped wedge, or etc)

ePROs electronic patient-reported outcomes. HRQoL health related quality of life.

Data extraction

Two researchers (Y.X. and X.G.) independently screened each of the identified publications based on their titles and abstracts, and then full texts based on eligibility criteria using Endnote X9. The identified papers were collated and duplicates were removed. Backward and forward reference searching was used to identify additional papers. Authors were contacted if full texts or data were unavailable. Subsequently, the two researchers worked independently to extract data from the included studies in order to complete a predesigned electronic form based on the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions⁸⁷. The form included: publication information (title, journal, and year); author names; study details (country, duration of interventions, design, blinding and randomization, method, retention rate); participants (eligibility, sample size, demographics, type of cancer and treatment); intervention (frequency, electronic platform used, ePRO questionnaire); results (primary and secondary outcomes, time points and measurements, with SDs, means, HRs, 95% CIs, and statistical significance). A third reviewer (WZ) was consulted in case of any disagreements.

Assessment of risk of bias

The revised Cochrane risk-of-bias version 2 tool was used to assess the quality of the studies on five aspects: bias arising from the randomization process, bias due to deviations from intended interventions, bias due to missing outcome data, bias in measurement of the outcome and bias in selection of the reported result⁸⁸. Each domain of bias was determined by answering several questions. Three researchers (Y.X., X.G., and W.Z.) independently categorized included articles as “low risk of bias”, “some concerns” to “high risk of bias” using the algorithms within the tool⁸⁸. A study was classified as “low risk of bias” if all five domains demonstrated a low risk of bias; “some concerns” if at least one domain raised some concerns, provided no domain was deemed at high risk; and “high risk of bias” if at least one domain any domain was rated as high risk or if multiple domains presented concerns⁸⁸.

Meta-analysis

Analyses were prepared in Excel forms and performed using STATA (version 18.0, StataCorp LLC). Within-group difference in mean for the composite score of HRQoL and their SDs, and HRs and 95% CIs for intervention and control groups were employed to determine the intervention effects. We calculated the SMD in each study to combine data measured with different scales, by dividing the differences in mean between groups by the SD⁸⁹. Studies were excluded from meta-analysis if they did not report SD or 95% CI. The effect sizes of the studies were pooled using the random-effects model because the true heterogeneity was suggested to be high after examining I² statistic (97.2%) and P value (<0.001)⁹⁰. The robustness was assessed with sensitivity analyses utilizing the leave-one-out approach, and by including results of different populations, measurements, or time points. We conducted a subgroup analysis to explore potential heterogeneity of the effectiveness of different treatment types and intervention duration.

Quality of evidence

The quality of evidence was assessed using GRADE criteria and was graded from “high”, “moderate”, “low” to “very low”⁹¹. The framework evaluated evidence quality through five key domains: risk of bias, imprecision, inconsistency, indirectness, and publication bias. Given that all analysed studies were RCTs (which initially qualify as high-quality evidence under GRADE criteria), the evidence rating began at the “high” level and was then progressively downgraded whenever any of the five domains demonstrated as “serious,” “very serious,” “likely,” or “very likely”^21,92.

Author contributions

Y.X.: Conceptualization, Data curation, Methodology, Writing-original draft, Writing-review & editing. X.G., W.Z., Y.W. and Z.S.: Data curation, Methodology, Writing-review & editing. P.H.: Conceptualization, Supervision, Writing-review & editing. All authors have read and approved the manuscript.

Data availability

Data is provided within the manuscript or supplementary information files.

Code availability

The code that supports the findings of this study is available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41746-025-01812-x.