Patient preference attributes in eHealth interventions for cancer‐related fatigue: A scoping review

Abstract Introduction Cancer‐related fatigue (CRF) is one of the most reported long‐term effects breast cancer patients experience after diagnosis. Many interventions for CRF are effective, however, not for every individual. Therefore, intervention advice should be adjusted to patients' preferences and characteristics. Our aim was to develop an overview of eHealth interventions and their (preference sensitive) attributes. Methods eHealth interventions were identified using a scoping review approach. Eligible studies included breast cancer patients and assessed CRF as outcome. Interventions were categorised as physical activity, mind–body, psychological, ‘other’ or ‘combination’. Information was extracted on various (preference sensitive) attributes, like duration, intensity, peer support and costs. Results Thirty‐five interventions were included and divided over the intervention categories. (Preference sensitive) attributes varied both within and between these categories. Duration varied from 4 weeks to 6 months, intensity from daily to own pace. Peer support was present in seven interventions and costs were known for six. Conclusion eHealth interventions exist in various categories, additionally, there is much variation in (preference sensitive) attributes. This provides opportunities to implement our overview for personalised treatment recommendations for breast cancer patients struggling with CRF. Taking into account patients' preferences and characteristics suits the complexity of CRF and heterogeneity of patients.


| INTRODUCTION
In the Netherlands, one in seven women is diagnosed with breast cancer at some point in their lives (Eijkelboom et al., 2020). Early diagnosis, for example by the national screening programme, and improved treatment have increased the survival rates over the years (Netherlands Cancer Registry, 2021). As the number of breast cancer survivors increases, there are more survivors suffering from the longterm effects of having had cancer and its treatment. One of the most prevalent, but still underreported, long-term effects is cancer-related fatigue (CRF) (Bower et al., 2006;de Ligt et al., 2019;Minton & Stone, 2008;Ruiz-Casado et al., 2021). The National Comprehensive Cancer Network (NCCN) defines CRF as 'a distressing, persistent, subjective sense of physical, emotional, and/or cognitive tiredness or exhaustion related to cancer or cancer treatment that is not proportional to recent activity and interferes with usual functioning' (Berger et al., 2020).
Many different interventions exist to help patients in their struggles to prevent or cope with CRF. The NCCN guidelines (Berger et al., 2020) describe two broad types of interventions, pharmacological and non-pharmacological. The latter is divided into different categories of which physical activity, mind-body therapy and psychological interventions are the most prevalent (Berger et al., 2020;Pearson et al., 2018). In physical activity, patients are motivated to exercise to increase energy expenditure (Conn et al., 2011;Pearson et al., 2018), examples of mind-body interventions are mindfulness, meditation or yoga (Carlson et al., 2017;Pearson et al., 2018), and psychological interventions consists of, for example, psychoeducation, cognitive behavioural therapy (CBT) or supportive-expressive therapy (Fors et al., 2010;Pearson et al., 2018).
Instead of focussing on one specific category, it is also possible to combine categories within one intervention (Pearson et al., 2018).
Non-pharmacological interventions are often delivered in a faceto-face setting. However, patients experience several barriers to follow face-to-face interventions, amongst others travel time, costs and lack of transportation (Stubblefield, 2017). To overcome these barriers, interventions can also be delivered online, as eHealth interventions. There is not one clear definition for eHealth, but health and technology are two common terms (Oh et al., 2005). Therefore, in this paper, we define eHealth interventions as health interventions that have an online, or technological, component that is necessary and relevant to support patients throughout the intervention, with or without healthcare professional. These can be mobile phone applications, websites with assignments or blended care with additional face-toface consults, but not solely a conferencing system to facilitate teleconsults without additional intervention components.
Effectiveness of an intervention on an outcome measure, for example CRF, is important to decide what intervention an individual patient should follow. Hilfiker et al. (2018) ranked categories of interventions by effectiveness on CRF to help patients and healthcare professionals decide what intervention to follow. However, even though an intervention is found to be effective in general, it might not help an individual patient. Randomised controlled trials (RCTs) reporting an overall significant effect on CRF also revealed that not all patients showed clinically relevant change; some patients even worsened (Abrahams et al., 2017;Bruggeman-Everts et al., 2017;Yun et al., 2012). Furthermore, there is not one gold-standard intervention that works best for all patients with CRF (Bower, 2014). So, as Hilfiker et al. (2018) also suggested, it is relevant to look into preferences of patients when suggesting an intervention.
Factors like personal characteristics and preferences influence whether a patient follows an intervention as intended. For example, participation in RCTs is lower compared to participation in randomised patient preference trials, where patients are divided to their preferred arm (Wasmann et al., 2019). Another example relates to the categories of interventions, a patient who exercised before diagnosis is more likely to successfully follow an exercise intervention (Pickett et al., 2002). In contrast, if an intervention does not fit the personal characteristics and preferences of an individual patient, this might cause a drop in motivation, less time investment and thus fewer to no impact on the outcome measure (Cillessen et al., 2020). Therefore, to do justice to the individual, when advising an intervention, this advice should be personalised to the individual patient, combining the type and severity of CRF, personal characteristics and preferences regarding interventions.
In case preferences of patients can be related to attributes of interventions, these attributes can help patients and healthcare professionals by selecting an intervention that matches patients most, or at least have as many overlap with preferences as possible. For example, flexibility regarding duration and intensity can be an important attribute since patients have different time investment possibilities. Additional preference sensitive attributes in eHealth are having an introductory training, contact with a healthcare professional (HCP), peer support, mode of content delivery, costs and effectiveness (Phillips et al., 2021).
To be able to link patient preferences to attributes of existing interventions, an overview is needed in which these two aspects are combined. Within the best of our knowledge, attributes of interventions related to patient preferences have not been reviewed in combination with existing eHealth interventions. This study therefore aims to answer two research questions: (1) What eHealth interventions exist to help breast cancer patients with CRF? and (2) What (preference sensitive) attributes make up these interventions?

| METHODS
To create an overview of eHealth interventions for CRF in breast cancer patients, a scoping review approach was used, as this fits the broad scope of this study (Arksey & O'Malley, 2005;Tricco et al., 2018). For reporting, the Preferred Reporting Items for Systematics Reviews-Scoping Review (PRISMA-ScR) checklist (Tricco et al., 2018)

| Eligibility criteria
We aimed to find eHealth interventions in which (1) the study population included or completely consisted of breast cancer patients and (2) CRF was assessed as (one of the) outcome measure(s). Interventions could be studied with any design and were excluded if they only described the development. The specific eligibility criteria are outlined in full texts and discussed with S. S., M. V., K. W. and A. W. to reach consensus on which to include. If interventions were described in several papers, information was combined.
For all interventions, information was extracted and charted into an overview in Excel. The information was based on several data items (see Table 2), which were related to general information of the intervention, (preference sensitive) attributes and patient T A B L E 1 Inclusion and exclusion criteria to select the eligible studies in this scoping review To compare the eHealth interventions, interventions were divided into the three non-pharmacological categories described in the introduction (physical activity, mind-body and psychological). In case neither or more than one of these three categories fit, interventions were placed in the 'other' or 'combination' category. The (preference sensitive) attributes and patient characteristics related to successful, adherent and dropped-out patients were analysed one by one to compare them, if possible, between the intervention categories.

| Consultation
As an addition to the PRISMA-ScR checklist, the methodological framework of Arksey and O'Malley (2005) proposes to add consultation. Therefore, we held two meetings for consultation: one with experts on CRF and one with experts on the user perspective (November 2021). In these meetings, preliminary results were presented, and experts were asked to give input on these results. The input was processed to get to the results presented below.

| RESULTS
We identified 344 unique articles of which 43 matched the eligibility criteria. These articles described a total of 35 interventions. With additional searches for more information on the interventions, an extra 18 articles were identified and included as well. Figure 1 shows the flow diagram of the selection of studies. Table S1 shows the full overview of all interventions with all information related to the data items.

| eHealth interventions
To answer the first research question, 'what eHealth interventions exist to help breast cancer patients with CRF?', the 35 interventions are described below.

| (Preference sensitive) attributes of interventions
For the second research question, 'what (preference sensitive) attributes make up these interventions?', the (preference sensitive) attributes as listed in Table 2 are compared below. First, the attributes that are preference sensitive are described, after which additional attributes are described that might be relevant to patients as well.

| HCP contact
For all categories, there were interventions with HCP contact. For the physical activity interventions, 4/5 had HCP contact (Delrieu et al., 2018;Falz et al., 2021;Galiano-Castillo et al., 2013;, of which three had synchronous contact. In 2/7 mindbody interventions, professionals were involved synchronously (Carlson et al., 2019;Zernicke et al., 2013). Six of the 13 psychological interventions had professional involvement. Two had synchronous contact (Kelleher et al., 2021;Owen et al., 2017), two asynchronous contact (Dozeman et al., 2017;Mendes-Santos et al., 2019), one intervention started with two face-to-face synchronous sessions, after which contact continued asynchronously via email (Abrahams et al., 2015) and one had 'health professional monitoring' without description on whether this is (a)synchronous (Yun et al., 2012). Both interventions in the 'other' category had professional involvement (Cairo et al., 2020;Kapoor & Nambisan, 2020), one synchronous and the other asynchronous, and in the combined interventions, 5/8 had professional involvement (Grossert et al., 2016;Nápoles et al., 2019;Smith et al., 2019;Zhou et al., 2020), of which three had asynchronous contact by giving daily or weekly written information.
The remaining 16 interventions without HCP contact can be followed anonymously. For some, it is necessary to have an email address or mobile phone number to create an account. Of note is that in a study setting, for some interventions, the research team was available to help with technical issues or called to stimulate adherence (Corbett et al., 2016;Henry et al., 2018;Lengacher et al., 2018;Puszkiewicz et al., 2016;Subnis et al., 2020;van den Berg et al., 2012;Zachariae et al., 2018).

| Peer support
Peer support was included in only seven interventions. One of the 'combined' interventions had an introductory group meeting (Smith et al., 2019), whereas in other interventions (psychological, Owen et al., 2017;Willems et al., 2015, and'other', Kapoor &Nambisan, 2020) peer support was facilitated through an online discussion forum or chat sessions. Two mind-body interventions had weekly interactive sessions with peers (Carlson et al., 2019;Zernicke et al., 2016) and in one 'combined' intervention face-to-face peer contact was arranged (Zhou et al., 2020).
Information was presented to users in various ways, namely, via video (n = 21), audio (n = 16), text (n = 26) or with images/graphics and visualisation (n = 11). Most psychological interventions used text and audio, whereas all mind-body interventions had audio tracks.
There were 21 interventions that had assignments, assessments or exercises, that is, in physical activity, psychological or combined interventions. Some interventions had other options, like vignettes with information (n = 2, psychological interventions), quizzes (n = 3, physical activity and psychological interventions) or an activity tracker to support in reaching activity goals (n = 2, physical activity and combined interventions). Table S1 describes the modes of content delivery per intervention.

| Costs
Only six studies described the costs of their interventions. For two interventions, parts of the sessions or exercises in the app were free of charge with the possibility to buy more content (Puszkiewicz et al., 2016;Subnis et al., 2020). Two interventions described that participants of the study had the possibility to use the intervention free of charge, indicating that otherwise costs were involved (Smith et al., 2019;Spahrkäs et al., 2020a). The e-mindfulness-based cognitive therapy (eMBCT) intervention with weekly HCP feedback was covered by insurance  and the health and wellness coaching app with daily HCP contact had a subscription fee of $65 per month (Cairo et al., 2020).

| Study results (effectiveness)
In Tables 3-7, the results of the interventions on the various fatigue outcomes are shown, showing that 24 (69%) interventions had significant results. Four (11%) interventions did not find significant results and for eight (23%) interventions, a study was still ongoing. One intervention is counted twice as the feasibility trial had nonsignificant results, but the RCT is still ongoing (Delrieu, Anota, et al., 2020;.
In all categories, there was at least one intervention that had a significant effect on CRF. In physical activity interventions, 2/5 inter-

| Usage
Of the preference sensitive attributes described above, duration and intensity are both related to expected usage. However, it is also relevant to know whether patients used the intervention as proposed.

| Experiences with intervention
For 21 interventions, patients were asked for their experience with the intervention. Experiences were asked both qualitative and quantitative. Overall, patients were positive about the interventions, although sometimes technological issues were reported. In Table S1, experiences are described more extensively.

| Cancer treatment
The timing of the intervention in relation to the primary treatment of cancer differs. Some interventions were delivered during primary treatment, whereas others were delivered after primary cancer treatment. For interventions after treatment, it differed whether these were in the first 5 years after treatment or in a later phase (Fosså et al., 2008;Mols et al., 2005;Thong et al., 2009). Division into these three periods is specified in Table S1.
Almost all interventions were for patients after primary treatment. Five interventions were delivered only during primary treatment; these were one physical activity intervention (Delrieu, Anota, et al., 2020), two mind-body interventions (Carlson et al., 2019;Kubo et al., 2018) and two interventions of the combined category (Grossert et al., 2016;Zhou et al., 2020). Four interventions were delivered to patients both during and after treatment; these were one mind-body intervention (Mikolasek et al., 2021), one psychological intervention (Owen et al., 2017) and two in the combined category (Smith et al., 2019;Spahrkäs et al., 2020a).
Reasons for drop-out (Table 8) were related to the expectation patients had of the intervention/study, technical and health issues, other reasons, or patients were lost to follow-up. The characteristics of drop-outs (Table 9), (non-)adherent patients (Table 10) and other characteristics/relations (Table 11) were only described in 15 interventions (Bray et al., 2017;Bruggeman-Everts et al., 2017;Dozeman et al., 2017;Henry et al., 2018;Kubo et al., 2018;Mikolasek et al., 2018;Owen et al., 2017;Smith et al., 2019;Spahrkäs et al., 2020b;van den Berg et al., 2013;Yun et al., 2012;Zachariae et al., 2018;Zernicke et al., 2014). Characteristics were related to demographics and health status at baseline, and relations were found between CRF and adherence or CRF and other outcome variables. Not all interventions found specific characteristics for drop-outs or adherent patients, some studies reported on a comparison in which no differences were found (Bruggeman-Everts et al., 2017;Henry et al., 2018;Owen et al., 2017;Smith et al., 2019;van den Berg et al., 2013;Zachariae et al., 2018), and results differed per study. For example, in one study on a psychological intervention (Insight, Henry et al., 2018), there was no difference in age when comparing drop-outs to the participants, whereas another study, also on a psychological intervention (PROSPECT, Bray et al., 2017), revealed that drop-outs were younger.

| DISCUSSION
For the first aim of our scoping review, we found 35 eHealth interventions for breast cancer patients with CRF that could be divided into the categories physical activity, mind-body, psychological, 'other' or a combination of categories. The second aim was to develop an overview of (preference sensitive) attributes that make up these interventions, showing there is much variation in attributes, both between and within the different categories.
This variation in attributes presents opportunity to personalise treatment advice towards preferences of an individual patient. Nevertheless, in preference studies, results are usually summarised for the participant group as a whole. For example, participants in the discrete T A B L E 9 Characteristics reported of the drop-outs in the various studies Demographics • No difference in age (Henry et al., 2018) • Being younger (Bray et al., 2017) • Men  • Lower educational level Kubo et al., 2018) • Lower income (Kubo et al., 2018) • Higher social functioning  • Less occupied with household activities  • Shorter time since diagnosis (Kubo et al., 2018) Health • Higher rates of antidepressant use (Bray et al., 2017 Zernicke et al., 2014) T A B L E 1 0 Characteristics and relations reported of (non-) adherent participants in the various studies Intervention usage • Higher baseline anxiety/depression leads to less training time (Bray et al., 2017) • More module use was predicted to the indicatory red/orange 'traffic lights'  • More module use was predicted by higher perceived relevance of modules  • Less module use was related with having a partner  Baseline/personality characteristics One interesting finding in relation to the review of Mustian et al. (2017) is that even though we did not aim to find effect sizes per category or attribute, we found that in the combination category, all interventions had a significant effect on CRF. Additionally, in psychological interventions, this was the case for 9/13 interventions, with the other four still ongoing. This is in line with their results after primary treatment, where psychological interventions with or without exercise have a larger effect size than exercise alone (Mustian et al., 2017).
Previous reviews on (eHealth) interventions also reported their results in Tables 3-7 with information on a number of intervention aspects, like content, mode of content delivery, duration, intensity and treatment status of patients (Corbett et al., 2019;Seiler et al., 2017;Vannorsdall et al., 2020;Xie et al., 2020). These reviews mostly continue with a meta-analysis, whereas our goal was to use these tables to develop an overview of (preference sensitive) attributes. Like us, these reviews found variation, or heterogeneity, in the included interventions. Reasonably, this was seen as an disadvantage, as heterogeneity makes it more difficult or even impossible to pool results to find an overall effectiveness (Corbett et al., 2019;Seiler et al., 2017;Vannorsdall et al., 2020;Xie et al., 2020). In contrast, we see this variation as an opportunity for patients to select an intervention that matches their preferences and characteristics best. Additionally, next to using the variation to determine 'what fits whom best', again, it is also interesting to investigate 'what works best for whom.'

| (Preference sensitive) attributes of interventions
As previous studies have not extensively discussed the variation in attributes between interventions, below, we discuss some of the distinctions found in this scoping review.
The intensity of an intervention seemed to be related to the category of intervention: mind-body interventions had daily expectations, psychological interventions weekly usage and exercise interventions were in between with three sessions per week. This can be extended to the interventions that combine categories, as these had varying intensities. It might be that the intensity of these interventions is related to either of the three options, depending on the category that prevails within the intervention.
Interventions without HCP contact were said to be anonymous.
In some, as described in the results, there is still contact with the research team possible. However, this contact is non-existent in a non-study setting, and therefore, these interventions were noted down as anonymous for patients.
Costs were only reported in six interventions. During a study, the intervention is still in a developing/testing phase, so the costs might still be unknown. However, costs are an important attribute to patients (Phillips et al., 2021) and, in line with inquiring experiences of users, costs might influence these results.
T A B L E 1 1 Other relations reported between patient characteristics and outcome variables Baseline characteristics/comorbidities • No difference in improvement of outcome by age (Henry et al., 2018) • Younger (<56) participants had larger effect on fatigue (Spahrkäs et al., 2020b; • Baseline levels of fatigue (Owen et al., 2017), demographic/clinical variables (Smith et al., 2019;Zernicke et al., 2016) and education and cancer status (Spahrkäs et al., 2020b) were no predictors of outcome on fatigue Outcome variables • Fatigue was correlated with depression  and sitting time  • Change in self-efficacy score was associated with fatigue symptoms (Smith et al., 2019) • Clinically significant improvement is predicted by (Yun et al., 2012): • Being moderately to severely fatigued • Having sleep problems at baseline • Having comorbidities at baseline • More evident effect of intervention is seen with (Yun et al., 2012): • Lower Brief Pain Inventory (BPI) severity score • Higher sleep quality index I and II score Adherence • Effect on fatigue was found when users used the module fatigue  • Number of cores completed was not associated with improvement in fatigue (Zachariae et al., 2018) • High/medium users had more reduction in fatigue than low/non users (Spahrkäs et al., 2020b) The mode of content delivery varied between the interventions included in this review. In the DCE of Phillips et al. (2021), only the main mode of content delivery was included, but in our overview, all options are included. As patients might like combinations as well, all modes of content delivery are relevant to report, instead of only the dominant one.
Even though usage is not reported in all studies, it is relevant to know if patients practiced with the intervention as supposed. On the one hand, this information specifically shows how well the expected usage of an intervention fitted patients and thus how well the intervention fitted the patients. On the other hand, if the intervention was effective while used less, it can also indicate that adjustments to the expected usage can be made. This then might result in a broader range of duration and intensity for this intervention and thus more patients with whom the preferences align.
The timing of the intervention relative to cancer treatment is based on the inclusion criteria of the interventions. However, this does not mean that in future prescription of these interventions, this is the only period in which patients can follow the intervention. It is the period for which the intervention is evaluated and thus the results are valid, but it is still possible to suggest the intervention to patients outside this range in case the intervention further fits the patients' preferences and characteristics.

| Patient characteristics
The importance of asking patients' preferences and linking these to a suitable intervention is also shown by the reasons for drop-out (Table 8). Preferences towards duration, intensity and HCP contact are relevant as well as technical issues patients might face.
Next to preferences, patient characteristics might be of importance to determine what intervention works best for which individual patient; however, this is not investigated sufficiently. In 15 interventions, characteristics were analysed and only one of those explicitly stated in their research question they were going to do this to determine the beneficial individual effect (Dozeman et al., 2017). Some studies focussed on just one characteristic, for example age, instead of subgroup of patients in which differences can occur. Therefore, in future intervention studies, it is important to include an analysis on intervention effectiveness for subgroups of patients to help and identify patients for whom the intervention worked best.

| Strength and limitations
The two consultation sessions organised with experts are a strength of this study. Using preliminary results as input, the discussion during these sessions gave new insights in important attributes by including the perspective of experts on CRF and the user perspective.
The quality of the individual studies was not assessed. This step is not necessary to reach the goal of developing an overview of interventions and their attributes, and in the PRISMA-ScR checklist, the critical appraisal of individual sources of evidence is optional (Tricco et al., 2018).
In this review, we included 35 eHealth interventions for breast cancer patients experiencing fatigue. This patient group was selected because the number of survivors is rising (Netherlands Cancer Registry, 2021; World Health Organization, 2021) and patients experience barriers for face-to-face interventions (Stubblefield, 2017). It could be that we missed studies from the way our search string was set up, for example if fatigue were a secondary outcome and not mentioned in the title/abstract/keywords. Also, interventions could be missed because of the databases searched. Some interventions are used without being tested in research setting; however, these cannot be identified in a systematic way as we did leading up to the same level of proof. Next to that, there are more interventions for patients with (cancer-related) fatigue that might not have been studied with breast cancer patients but can still be useful for them. As an example, the CBT for insomnia intervention included in this review, SHUTi (Ritterband et al., 2012;Zachariae et al., 2018), was tested (although not always on fatigue) in different participant groups (Christensen et al., 2016;Hagatun et al., 2018).
Even though it is possible we missed some eHealth interventions for breast cancer patients with CRF, and more interventions are being developed over time, our overview shows there is a lot of variation within attributes. For clinical implementation, when the overview is used to help patients and healthcare professionals selecting a fitting intervention, adding extra interventions will extend the intervention options, but most likely not the attribute ranges.

| Future research
Future research could focus on the implementation of our overview in a healthcare setting to support each individual patient with an intervention for CRF that fits them best. For this, a number of next steps need to be taken. First, more research needs to be done into preferences breast cancer patients have in relation to the attributes found in this review. Next, a method has to be developed to link the preferences of an individual to interventions included in the overview of this scoping review. This leads to personalised treatment recommendations for the patient. With this, it is important that the recommendation is transparent in terms of why a certain intervention fits a patient.
This will help both the healthcare professional as well as the patient.

| CONCLUSION
In this scoping review, we created an overview of existing eHealth interventions for breast cancer patients with CRF. Interventions were divided into the categories physical activity, mind-body, psychological, 'other' or a combination of categories. This overview outlines that there is variation in preference sensitive attributes related to time investment, introduction training, HCP contact, peer support, mode of content delivery, costs and effectiveness of interventions. Future research could show how this overview creates possibilities for patients to follow an intervention that better aligns with their preferences. This will, ideally, increase adherence, effectiveness and satisfaction, decrease CRF and, with that, improve quality of life of patients after breast cancer.