Scolaris Content Display Scolaris Content Display

Cochrane Database of Systematic Reviews Protocol - Intervention

Electronic symptom monitoring for patients with advanced cancer

Authors' declarations of interest

Version published: 11 July 2023 Version history

https://doi.org/10.1002/14651858.CD015732
Collapse all Expand all

Abstract

Objectives

This is a protocol for a Cochrane Review (intervention). The objectives are as follows:

To assess the effects of electronic symptom monitoring (ESM) on patient survival, hospital admission, emergency visit, changes in treatment protocol, adverse events, patient‐reported outcomes (PROs), and cost‐effectiveness for patients with advanced cancer.

Background

As a reflection of disease progression and treatment efficacy, symptoms can often offer vital and first‐line information about advanced cancer. These symptoms may link to the progression of the disease and are often first perceived by patients. Besides traditional methods of monitoring the above symptoms via printed questionnaires, telephone, or routine clinical reviews, the wide use of electronic connected devices and systems during the last decade has paved the way for electronic symptom monitoring (ESM) in oncology (Basch 2020; Basch 2022; Nipp 2019).

ESM refers to symptom monitoring relying on electronic systems via the World Wide Web, a smartphone app, or an automated telephone interface (Basch 2020). ESM systems are evaluated symptom reporting systems that collect the symptom data remotely and are used to communicate symptoms between patients and healthcare providers in a clinical consultation, thereby potentially influencing clinical decision‐making (Gandrup 2020). The innovation in ESM overcomes several challenges of former symptom monitoring methods to provide real‐time rather than retrospective feedback from any internet‐accessible device (Denis 2019). However, several randomised controlled trials (RCTs) have presented mixed evidence about the efficacy of ESM in treating patients with advanced cancer, which may be a source of confusion for clinicians and patients.

Description of the condition

Definition of advanced cancer

The American Cancer Society states that advanced cancers are those that "cannot be cured", and can be locally advanced or metastatic (American Cancer Society 2013). The definition of 'metastatic' is consistent across cancers, while the definition of 'locally advanced' is not. Although cancers with higher stages (e.g. T3/4N0M0 prostate adenocarcinoma) or recurrent cancers may be hard to cure, they should not arbitrarily be considered as locally advanced cancer as they remain potentially curable.

The criteria for locally or locoregionally advanced cancer is that the cancer is not resectable or curable by any treatment modalities after investigators' assessment, for example the tumour has invaded essential anatomical structures, and surgical ablation of the tumour is predicted to be lethal. The criteria were widely used in a variety of solid cancers, including adrenocortical carcinoma (Fassnacht 2015), breast cancer (Robertson 2016), cholangiocarcinoma (Malka 2014), cervical cancer (Tewari 2017), urothelial carcinoma (Petrylak 2017), oesophageal cancer (Sun 2021), head and neck squamous cell carcinoma (Lee 2021), renal cell carcinoma (Motzer 2021), hepatocellular carcinoma (Abou‐Alfa 2018), non‐small‐cell lung cancer (Nestle 2020), pancreatic cancer (Hammel 2016), prostate cancer (Denham 2019), thyroid cancer (Brose 2014), gastric cancer (Kang 2017), melanoma (Hodi 2018), and sarcoma (Schöffski 2016). However, in several types of solid cancer, including colon (Foxtrot Collaborative Group 2012), rectal (Bahadoer 2021), and ovarian cancer (Kehoe 2015), patients with 'advanced cancer' could still receive tumour ablation surgery with the intent to cure. As tumour ablation is predicted to be inadequate to treat these cancers, it always co‐ordinates with chemotherapy or other therapies. Hence, in these cases, advanced cancer remains incurable.

In addition, the literature rarely mentions 'advanced' brain cancers (including glioma and glioblastoma), Herrlinger 2019, and haematologic cancers (including acute leukaemia or chronic myeloid leukaemia), Othus 2016, because all of these have a poor prognosis and are considered almost incurable, and thus could be considered as advanced cancer. However, there are exceptions for haematologic cancer patients who have received bone marrow transplantation, as they may be completely cured (Li 2023).

The prognosis of advanced cancers differs extensively based on cancer type. For example, the median overall survival in patients with advanced melanoma who received ipilimumab is less than 24 months (Hodi 2018; Schachter 2017), while the 3‐year disease‐free survival rate of advanced rectal cancer could be more than 60% to 70% (Bahadoer 2021; Hong 2014). A cutoff of prognosis should therefore not be incorporated into the definition of advanced cancer.

Overall, advanced cancer is metastatic or unresectable/inoperable/incurable cancer.

Symptoms in patients with advanced cancer

Common symptoms experienced during advanced cancer include pain, confusion, delirium, anorexia, fatigue, anxiety, depression, breathlessness, insomnia, nausea, and constipation (Carlson 2016; Gilbertson 2011; Haun 2017; Henson 2020; Teo 2021; Vanderwerker 2005), with more than one symptom typically experienced at any specific time (Lage 2020). Accurate symptom assessment is the prerequisite for the evaluation and treatment planning of advanced cancer, and effective symptom management is deeply linked to patients’ quality of life (QoL), treatment compliance, and survival (Cleary 2022; Subbiah 2021).

Common treatment modalities for advanced cancer can be used alone or in combination, including chemotherapy, radiation therapy, surgery, hormone therapy, targeted therapy, immunotherapy, palliative care, and hospice care. Moreover, the individual treatment modalities also lead to some specific symptoms, most of which are related to side effects. For example, symptoms of radiotherapy mainly feature erythema, desquamation, telangiectasia, taste disturbance, hair loss, and sore skin (Ahmad 2012). Patients who receive chemotherapy commonly experience symptoms such as febrile neutropenia, diarrhoea, and mucositis (Wei 2021). Immunotherapy may initiate autoimmune responses that could further lead to severe hypotension, fever, and renal dysfunction (Kroschinsky 2017). Opioids used in palliative care may lead to increased cardiovascular adverse events, potential addiction, and symptoms including constipation, depression, increased pain, hormonal dysfunction, and weak bone (Baldini 2012). Apart from the symptoms mentioned above, patients often expect treatment goals that are higher than goals developed by their oncologists, and goal inconsistency is associated with patient anxiety (El‐Jawahri 2020). All these symptoms identified the potential need for continuous monitoring during advanced cancer treatment.

Description of the intervention

According to the definition, ESM systems do not automatically collect vital signs by wearable or implantable sensors; are not a tablet or computer collecting data in the waiting room before a clinic visit; and are not self–management‐only systems.

Before monitoring, patients may be required to complete a sociodemographic questionnaire listing their age, sex, race, relationship status, employment, and education (Basch 2022; Nipp 2019). Additional medical information could be obtained by integrating ESM with the electronic health record system.

For electronic systems online, a specific website or software with dashboards and corresponding questionnaires is distributed to patients via desktop/laptop computers or mobile phones (Leahy 2021; Nipp 2019). In some studies, if patients were unable to complete the computerised symptom assessment, they could use paper versions instead (Nipp 2019). Voice communication functions feature an automated telephone interface, which may lead to automatic unidirectional or bidirectional communications between patients and navigators. Using the above technologies, navigators could administer surveys, provide feedback to patients via electronic devices, and populate databases with patient‐reported outcomes (Leahy 2021; Mir 2022). Navigators could be clinicians, medical assistants, or clinical nurses (Basch 2022; Nipp 2019; Posadzki 2016; van den Hurk 2022).

Variations exist in symptom monitoring and interactive features of each ESM system. The most popular feature of ESM is navigating symptom data with alerts to health professionals for severe symptoms (Basch 2016; Nipp 2020). However, some ESM systems were explicitly designed to focus on patient self‐management with tailored advice and ways to contact their medical navigators if needed (Ruland 2013a; Ruland 2013b). Hybrid approaches also exist so that ESM systems can complete both functions (Absolom 2017; Mir 2022). Information collected by ESM may be categorised into self‐reported symptom items, QoL items, adverse events items, and direct information interaction between patients and navigators. It should be noted that the information collected from different classes may overlap. Most ESM systems adopted validated questionnaires or scales to facilitate information collection (Basch 2022; Bryant 2020; Nipp 2019; Strasser 2016).

Moreover, the design of ESM systems may endow them with more functions beyond symptom reporting and monitoring. ESM systems may allow patients to review their symptom reports over time; provide tailored automated patient advice on managing symptoms; or include a forum for patients to communicate with one another (Warrington 2019). These additional features of ESM systems may improve user experience.

The frequency of information feedback also varies among ESM systems. ESM systems may utilise one of three types of frequency for patient feedback:

  • time points‐based feedback: patients are asked to complete surveys at single or multiple time points at a certain timeframe (daily or weekly, for years or until they have discontinued all cancer treatment) (Basch 2022). This type of feedback is previously scheduled and is timed regularly;

  • real‐time information feedback: patients can report their symptoms or contact navigators in real time by leaving messages, completed forms, or voice feedback (Posadzki 2016; Pusic 2021; Strasser 2016). This type of feedback is based on unexpected situations and is non‐anticipatory, which is an important difference from time points‐based feedback;

  • hybrid feedback: patients experience hybrid approaches of the above‐mentioned two types of feedback, such as real‐time feedback during navigator working hours, and time points‐based feedback or seeking emergency service ESM systems provided during their break (Basch 2022; Schenker 2021).

How the intervention might work

Symptoms can be common and debilitating for patients with advanced cancer and anticancer treatment (Cleeland 2013; Reilly 2013). These symptoms often go unmeasured and unrecognised between routine visits. In a report focusing on symptomatic toxicities during anticancer treatment, a low agreement between patients and physicians was observed, with physician‐reported toxicity rates lower than those reported by patients (Di Maio 2015). Up to 30% of symptoms were reported by patients but not physicians, including anorexia, constipation, and hair loss, demonstrating that traditional symptom monitoring system urgently needs improvement (Haynes 2002).

Hence, in an attempt to resolve this disconnect, in recent years clinicians have been advised to consider using patient‐reported outcomes (PROs) to augment clinical judgement, where patient‐reported outcome measures (PROMs) are widely used (Posadzki 2016; Yang 2018). PROMs with self‐rated scales are developed to improve the detection of the patient's subjective experience (Appleyard 2021; Bevans 2014; Casey 2022).

There are two main types of PROMs: those designed to evaluate the burden of a specific disease, and others that focus on the general assessment of perceptions (Calvert 2019; Greenhalgh 2009; Greenhalgh 2018). Using various questionnaires, PROMs are introduced for evaluation of severity, symptoms burden, impacts on daily life, and emotional stability. PROMs can also evaluate the personal experience of care and subjective perception of the treatment (Ciani 2020; Fava 2019).

With the use of PROMs, patient engagement has been greatly improved, and patients are increasingly becoming involved in the decision‐making progress. Some studies have found the introduction of PROMs to be associated with increased survival in patients with advanced cancer (Denis 2019). Upon acknowledgement of early symptoms, timely responsiveness, including symptom management counselling, supportive medications, and medicine dose modifications, could prevent adverse downstream consequences (Denis 2019). In a study that found PROMs could prolong the overall survival of patients during routine cancer treatment compared to usual care, the authors also noted that patients with PROMs were able to tolerate a prolonged continuation of therapy (Denis 2019; Robins 1991). PROMs have also prompted patient‐clinician discussions and communications, given that some patients may have previously felt hindered from contacting their clinicians due to fears of being bothersome (Snyder 2010; Weaver 2007; Yang 2018). PROMs assist in streamlining consultations by guiding communication that focuses on patient concerns (Hopwood 1998; Wilkie 2003; Yang 2018). As a result, a decrease in symptom severity and improved patient health outcomes are found following PROs use.

Before the ubiquity of electronic devices, the detection of patient symptoms relied more on paper‐based materials, the use of which poses a great challenge to organise and analyse the data of each patient. Disadvantages of paper‐based materials have been proposed for clinicians in recognising each patient's condition, as systematic analysis showed clinicians preferred visual representations of PROs data via histograms, bar charts, and line graphs with comparison data (Bainbridge 2011). From the patient's perspective, the lack of electronic reminders to complete questionnaires also becomes a barrier to improving communication (Basch 2007). Patients also preferred to complete the questionnaire on an electronic device rather than face‐to‐face (Mullen 2004).

In summary, electronic systems overcome the flaws of traditional paper‐based ways, have revolutionised the healthcare sector, and promote real‐time personalised monitoring and therapeutic care, especially in the condition of chronic diseases (Mir 2022). ESM allows the timely performance of medical care not only in routine traditional clinical settings but also in homes, workplaces, and even travel locations (Mir 2022).

Why it is important to do this review

During the last decade, the wide application of smart devices and systems have paved the way for ESM in oncology, which overcomes several flaws of previous PROMs. ESM has been introduced for patients with advanced cancer or those receiving anticancer treatment to efficiently manage individualised physical and psychological symptoms (Basch 2016; Basch 2020; Basch 2022; Bryant 2020; Mir 2022; Nipp 2019; Nipp 2020).

For patients, it is essential to clarify whether the benefits of ESM systems offset the costs. The introduction of digital devices requires tangible and intangible costs of learning and operating the system, especially for patients lacking smart device experience. Several RCTs have presented mixed evidence regarding the potential benefits of ESM systems for patients with advanced cancer. Basch and colleagues claimed that the benefits appeared greater for computer‐inexperienced patients, as they may also have less‐developed health communication skills and thereby benefit more from a structured programme of ESM (Basch 2016). They also reported that almost 40% of patients chose to use an automated telephone interface rather than a web interface, particularly older patients, those living in rural areas, and those with lower educational levels (Basch 2020). However, Nipp and colleagues found that older patients with advanced cancer did not comprehensively benefit from the ESM intervention in terms of emergency room visits and survival due to their cognitive function and the availability of social support (Nipp 2020; Pamoukdjian 2017).

From the clinician's perspective, ESM may improve doctor‐patient communication efficiency, facilitate disease management, and set up alert and reminder mechanisms (Mir 2022; Wujcik 2022). To date, many clinicians consider that ESM could improve the quality of communication with patients and that self‐reported information is helpful for patient care (Basch 2020; Basch 2022; Mir 2022). However, a few clinicians felt confused when confronted by the complex information gathered by ESM systems (Basch 2020). Consensus on a variety of details (including preferred feedback frequency, patient indications, contraindications, etc.) has not been reached. Clinicians may need more information to consider whether to use ESM systems in an independent population of patients.

For researchers, the designing, assessing, and reporting of ESM systems significantly varies based on several different metrics, such as patient engagement, survival status, QoL, physical symptoms, and psychosocial symptoms (Basch 2022; Judson 2013; Steel 2016; van den Hurk 2022). Different designations of ESM systems and related studies have reported different outcomes (Warrington 2019), without clear priorities. For instance, the CaSSy (Cancer Support System) and Symptom Tracking and Reporting system both showed improvements in global QoL, where CaSSy included a care co‐ordinator in cognitive behavioural therapy and psycho‐oncology (Basch 2016; Steel 2016). However, the WebChoice system did not significantly impact patient QoL (Ruland 2013a). More instructions may be needed to address what outcomes should be reported to evaluate the efficacy of ESM.

In 2019, a systematic review focusing on electronic systems for patients with anticancer treatment was published (Warrington 2019). The ratio of RCTs in the included studies was < 25%, which decreased the credibility of conclusions. Furthermore, the final search date was 2017, thereby precluding the timely provision of recommendations.

Lastly, our review may have an additional global significance in this post‐COVID‐19 era. Many patients with advanced cancer cannot physically access healthcare services owing to the impact of the COVID‐19 pandemic in low‐, middle‐, and high‐income countries (Guven 2020; Ranganathan 2021). Remote review and symptom monitoring are suggested when the risk of COVID‐19 exposure outweighs the benefits of offline visiting a hospital (NICE 2021).

Overall, this review will be helpful for patients with advanced cancer, clinicians, and researchers, in the post‐COVID‐19 era with advanced telemedicine techniques.

Objectives

To assess the effects of electronic symptom monitoring (ESM) on patient survival, hospital admission, emergency visit, changes in treatment protocol, adverse events, patient‐reported outcomes (PROs), and cost‐effectiveness for patients with advanced cancer.

Methods

Criteria for considering studies for this review

Types of studies

We will include parallel or cluster‐randomised controlled trials (RCTs) evaluating the effects of electronic symptom monitoring (ESM) on patients with advanced cancer.

We will also include online clinical trial results and summaries of unpublished RCTs. We will attempt to contact the authors of unpublished studies if sufficient data cannot be achieved publicly (i.e. for risk of bias assessment). If we are unable to obtain complete study data, we will categorise the study as awaiting classification.

We will exclude cross‐over trials, quasi‐RCTs, non‐randomised clinical trials, cohort studies, case‐control studies, case series, case reports, letters, reviews, and non‐clinical studies.

Types of participants

We will include participants diagnosed with advanced cancer, regardless of the anticancer treatment received.

Inclusion criteria

  • Participants with locally advanced cancers, metastatic cancers, or haematological malignancies.

Exclusion criteria

  • Participants with physical or cognitive deficits that preclude using ESM.

  • Participants inaccessible to the devices for ESM.

  • Participants with emergent medical conditions who are unable to report their symptoms.

  • Participants with haematological cancer who receive or plan to receive bone marrow transplantation.

We will include studies involving a subset of eligible participants if results are reported separately for the eligible subset. If separate data are not available, we will only include such studies if ≥ 80% of participants adhere to the inclusion criteria. We will exclude data from any such studies in sensitivity analyses to test the robustness of the primary meta‐analyses.

Types of interventions

We will include studies comparing the effects of ESM versus usual care or another ESM system.

  • Intervention group: participants will have received ESM (as defined in Description of the intervention) via the internet, smartphone applications, or an automated telephone interface. We will exclude wearable devices that automatically record vital signs without symptom reporting and monitoring functions, and self‐monitoring applications without monitoring from the healthcare team. The information ESM collected should include at least one of the following items:

    • any self‐reported symptoms items;

    • any quality of life (QoL) items;

    • any adverse events items;

    • any direct information interaction between patients and navigators.

  • Control group: participants will have been followed up by any usual care or another ESM method. Typically, usual care means any healthcare service without ESM. We will also consider the use of self‐operated electronic applications on smart devices without monitoring from professional personnel as usual care.

Types of outcome measures

Primary and secondary outcomes are defined below. However, whether or not a study reported on our prespecified primary and secondary outcomes will not be an inclusion criterion for this review.

There will be no restrictions on follow‐up durations.

Primary outcomes

  • Overall survival (OS, the time from the beginning of the intervention to death due to any cause).

Secondary outcomes

  • Hospital admission (proportion of participants admitted to hospital for any reason).

  • Emergency visit (proportion of participants visiting emergency department at least once).

  • Changes in the treatment protocol (proportion of participants failing the original oncological treatment protocol or changing to a novel oncological treatment protocol).

  • Patient engagement (proportion of symptom report completions or the number of times participants access the system).

  • Adverse events related to the ESM application.

  • Progression‐free survival (PFS, the time during and after the intervention that participant lives with the disease but it does not worsen)

  • Disease‐specific survival (DSS, the time from the beginning of the intervention to the death of participant due to a specific disease)

  • Patient‐reported outcomes (PROs) (any report of the status of a participant’s health condition that comes directly from the participant without interpretation of the participant’s response by a clinician or anyone else) measured by validated tools:

    • QoL;

    • physical functioning;

    • emotional functioning/well‐being;

    • symptom scores.

  • Cost‐effectiveness.

We anticipate that different studies may report PROs by different validated tools. If data on more than one of these scales are provided for a given trial, we will rank the outcome measures. According to studies in core outcome sets, we consider QoL assessed by EQ‐5D, EORTC Core Quality of Life Questionnaire (EORTC QLQ‐C30), Expanded Prostate Cancer Index Composite (EPIC), and 36‐item Short Form Health Survey (SF‐36) as the first hierarchy for the comprehensive assessment of PROs in patients with advanced cancer (Ciani 2021). However, we will include all PROs, self‐reported or reported with the assistance of interviewers, at any time point during follow‐up.

Search methods for identification of studies

We will conduct systematic searches for RCTs. We will place no restrictions on language or date of publication when searching the electronic databases and other resources. Translations will be carried out if necessary.

Electronic searches

We will search the following electronic databases from inception:

  • the Cochrane Central Register of Controlled Trials (CENTRAL, via the Cochrane Library);

  • Embase (OvidSP) (from 1974 onwards);

  • MEDLINE (OvidSP) (from 1946 onwards).

The search strategy for MEDLINE (OvidSP) is shown in Appendix 1. Where appropriate, we will combine search strategies with the Cochrane Highly Sensitive Search Strategies for identifying RCTs and controlled clinical trials (Lefebvre 2022).

Searching other resources

We will search the following trial registries from inception:

  • ClinicalTrials.gov;

  • World Health Organization International Clinical Trials Registry Platform (WHO ICTRP).

We will also screen the reference lists of published reviews and eligible studies for additional publications.

Data collection and analysis

Selection of studies

Two review authors (LL, TW) will independently screen the titles/abstracts of records identified by the search for potentially relevant studies. We will obtain the full‐text articles of those studies deemed potentially relevant, and two review authors (LL, TW) will independently decide on inclusion or exclusion. In case of disagreement between review authors, a third review author (Y Cao) will be consulted to make a final decision. We will collate multiple reports of the same study so that each study, rather than each report, is the unit of interest. We will record the study selection process in a PRISMA flow diagram (Moher 2009). We will describe reasons for exclusion of the excluded studies in 'Characteristics of excluded studies' tables.

Data extraction and management

Two review authors (LL, TW) will independently pilot the form to extract data and check for agreement before entering the data into Review Manager Web (Review Manager Web 2022). We will note in the 'Characteristics of included studies' table if outcome data were reported in an unusable way.

We will extract the following study characteristics.

  • Methods: study design, number of study centres and location, study setting, withdrawals, date of the study, follow‐up.

  • Participants: number, mean age, age range, gender, type of treatments, inclusion criteria, exclusion criteria, and other relevant characteristics.

  • Interventions:

    • Symptom monitoring system

      • whether a commercial or in‐house electronic system is used

      • which devices are used for ESM

      • whether patients and clinicians use the same interface

      • whether basic characteristics of patients are displayed

      • whether it is integrated with electronic health record system

    • Symptom reporting by patients

      • which symptom measures are incorporated into the questionnaire

      • how to grade or score the symptoms

      • when to report the symptoms (time‐based, real‐time, or both)

    • Symptom reporting reminders

      • when to remind the patients of symptom reporting

      • how to remind the patients of symptom reporting

    • Over‐time symptom review

      • who could review the symptom data over time and when the data could be reviewed

      • the form in which the over‐time symptom data are presented

    • Alert of worsening symptoms

      • the criteria or threshold score of alert

      • how to alert the patients and healthcare team

      • how to manage the alerted worsening symptoms

    • Education on symptom management

      • the contents of educational materials

      • how to deliver educational materials to patients

    • Communication about patients' symptoms

      • whether a function for patient‐healthcare‐team communication is provided

      • whether a forum for patients' communication is provided

  • Comparisons: if the control group uses another type of ESM, the details of ESM will be reported in detail; if the control is a non‐ESM intervention, the details of the control intervention and differences between comparisons will be provided.

  • Outcomes: main and other outcomes specified and collected, and time points reported.

  • Notes: funding for the trial, notable conflicts of interest of trial authors, ethical approval.

We will first contact the authors of relevant studies to request important missing data. If we receive no response, we will use statistical methods for different types of missing data, where this is possible (Dealing with missing data).

Assessment of risk of bias in included studies

Two review authors (LL, Y Chen) will independently assess risk of bias for each study using version 2 of the Cochrane risk of bias tool (RoB 2) (Higgins 2022; Sterne 2019).

We will assess risk of bias for all primary and secondary outcomes. We will consider the risk of bias for the studies that contribute to a given outcome with the longest follow‐up.

Any disagreements will be resolved by discussion or by involving a third review author (Y Cao). We will assess risk of bias in the included studies according to the following domains:

  • bias arising from the randomisation process;

  • bias due to deviations from intended interventions;

  • bias due to missing outcome data;

  • bias in measurement of the outcome;

  • bias in selection of the reported result.

For cluster‐RCTs, we will additionally assess risk of bias arising from the timing and recruitment of participants according to the RoB 2 variant for cluster trials.

We will use the signalling questions in the RoB 2 tool and rate each domain as 'low risk of bias', 'some concerns', or 'high risk of bias'. The effect of interest will be the effect of assignment to intervention, for which an intention‐to‐treat analysis will be preferable. We will summarise the risk of bias judgements across different studies for each of the domains listed for each outcome. The overall risk of bias for the result will be the least‐favourable assessment across the domains of bias. We will use the RoB 2 Excel tool to implement RoB 2 (www.riskofbias.info/welcome/rob-2-0-tool/current-version-of-rob-2).

Measures of treatment effect

We will report the following effect measures for included studies and subsequent analyses. For dichotomous data, we will calculate risk ratios (RR) with 95% confidence intervals (CIs). If studies use continuous scales of measurement to assess the effects of the intervention, we will use mean differences (MD) with 95% CIs. If studies use different scales or measurements, we will use standardised mean difference (SMD). Moreover, if studies use time‐to‐event outcome metrics, we will seek appropriate overall measures of effect (e.g. hazard ratio with 95% CIs). If this is not possible, we will use median and interquartile ranges to present data.

Unit of analysis issues

The participant will be the analytical unit. If a study allocates participants into multiple groups which will not be distinguished in this review (e.g. ESM versus usual care 1 versus usual care 2), we will attempt to obtain participant‐level data from the study authors and analyse data as 'ESM versus usual care'; if the participant‐level data are not accessible, we will synthesise the data of 'ESM versus usual care 1' and 'ESM versus usual care 2' and halve the control group to avoid double‐counting.

Dealing with missing data

For studies published in the past 20 years, we will contact the corresponding authors of the study to obtain the missing data. The maximum waiting period allowed will be one month. If the data remain unavailable (e.g. the authors don't reply or the study was published before 2002), we will analyse only the available data (Deeks 2022).

For survival outcomes, we will adopt an intention‐to‐treat analysis, which means the effect of assignment to the interventions at baseline, regardless of whether the intended interventions are received. The reason for this choice is that we are interested in finding evidence of the intervention at a health system level, and evaluating the efficiency of the intervention in a real‐world setting. For other outcomes, we will use an as‐treated analysis to reflect the risk from the procedure and individual advanced cancer patient‐level results.

Assessment of heterogeneity

We expect that the type of advanced cancer will add to the heterogeneity of the results, as will differences in participants, interventions, and reported outcomes between included studies. We will use the I2 and Chi2 statistics to estimate the percentage of statistical heterogeneity between trials. We will utilise the following rules for interpreting the I2 statistic (Deeks 2022):

  • 0% to 40% may indicate slight heterogeneity;

  • 30% to 60% may indicate moderate heterogeneity;

  • 50% to 90% may indicate substantial heterogeneity;

  • 75% to 100% may indicate very substantial heterogeneity.

Assessment of reporting biases

We will construct a funnel plot to assess asymmetry if ≥ 10 studies are included in a meta‐analysis, which reflects whether results are influenced by publication bias. We will use Egger's methods for continuous data, Egger 1997, and Begg's methods for dichotomous and time‐to‐event data, Begg 1994, to further test funnel plot asymmetry.

Data synthesis

We will only perform meta‐analysis when PICO elements are adequately similar among studies, with sufficient data in an acceptable format (McKenzie 2022). We will use a random‐effects model for meta‐analyses as we anticipate substantial heterogeneity between included trials due to the types of advanced cancer and ESM system. With this approach, the CI for the pooled average intervention effect is wider than would be obtained with a fixed‐effect approach, leading to a more conservative interpretation.

When meta‐analysis is not applicable due to incompletely reported outcome or effect estimate, different effect measures, or bias in the evidence, we will synthesise and present evidence by tabulation qualitatively. The data available will determine the synthesis methods we apply:

  • summarising effect estimates: effect estimates available without measures of precision;

  • combining P values: P values available;

  • vote counting based on the direction of effect: directions of effect available.

When the available data vary among included studies, we will consider synthesising and presenting the evidence by all three methods at the same time.

Subgroup analysis and investigation of heterogeneity

We will only perform subgroup analysis for primary outcomes in the case of sufficient participant information and individual data.

We plan to analyse the following subgroups where possible:

  • types of ESM;

  • types of participants (in curative treatments or hospice treatments);

  • types of advanced cancer;

  • study design (parallel or cluster‐RCTs).

We note the limitation of subgroup analyses and their observational nature, which require consideration when interpreting the results. We have chosen the above subgroups as they may elucidate some heterogeneity encountered in the included studies.

Sensitivity analysis

If uncertainty arises owing to the inclusion of a study with missing data/imputed data in the meta‐analysis, we will perform sensitivity analysis by excluding this study.

We will conduct a sensitivity analysis by excluding 'low risk of bias' and 'some concerns' studies to test the stability of the results irrespective of study quality.

We also plan to perform a sensitivity analysis on the model of meta‐analysis to test if a fixed‐effect model may alter the findings.

Summary of findings and assessment of the certainty of the evidence

We will create a summary of findings table for the following outcomes with the longest follow‐up duration (Schünemann 2022). We will note outcomes with shorter follow‐up duration in the summary of findings table.

  • Overall survival

  • Hospital admission

  • Emergency visit

  • Changes in the treatment protocol

  • Patient engagement

  • Adverse events related to the ESM application

We will use the GRADE approach to rank the certainty of the evidence by employing GRADEpro GDT software (GRADEpro GDT). Domains that can decrease the certainty level of a body of evidence include risk of bias or limitations in the detailed design and implementation (assessed by RoB 2), unexplained heterogeneity or inconsistency of results, imprecision of results, indirectness of evidence, and high probability of publication bias. We will use footnotes to support all decisions to downgrade the certainty of the evidence, and will make comments to aid the reader's understanding where necessary. Three review authors (LL, TW, and Y Chen) will assess the certainty of the evidence, with any disagreements to be resolved by consulting another review author (Y Cao) where needed.