The CAREST-study design and intervention will be reported in accordance with the CONSORT statements for eHealth interventions [39] and parallel group randomised trials [40], the SPIRIT 2013 statement [41, 42], and in accordance with the recommendations and guidelines for internet intervention research in psycho-oncology [23].
Study design
The CAREST-study is a multi-centre, randomised controlled trial, comparing online self-help training with care as usual in breast cancer survivors. A sample of 454 women with curatively treated breast cancer will be recruited from 8 hospitals scattered over the Netherlands. The participating hospitals are the Maasstad hospital in Rotterdam, St Antonius hospital in Utrecht, Admiraal de Ruyter hospital in Vlissingen, Reinier de Graaf hospital in Delft, Antonius hospital in Sneek, St. Elisabeth hospital in Tilburg, Catharina hospital in Eindhoven, and UMCG in Groningen (all situated in the Netherlands). After completion of the baseline measure, women will be randomised to either the online self-help or control group. Follow-up assessments are at 3 months (post-intervention), 9 months, and 24 months after baseline. Additionally, every 3 months participants will be asked to fill out a short measure about their healthcare use. Two reminders will be sent by e-mail one and two weeks after the first invitational e-mails. Eventually, participants who do not complete the questionnaires will receive a phone call from the researcher to remind them.
Participant eligibility
Women are eligible to participate if they had a diagnosis of breast cancer 1–5 years ago; have no signs of local or regional recurrence or metastatic disease; are capable of filling out questionnaires in Dutch; if their age at disease onset was 18 years or older; and if they have access to a computer with an internet connection. There are no exclusion criteria.
Recruitment settings and procedure
Patients will be recruited in two ways. First, four hospital sites are expected to recruit participants through oncology nurses, nurse practitioners, or oncologists, who will ask eligible patients to participate in the study. These patients will be informed about the study during their regular check-up at the outpatient clinic. When women show interest in participating in the study, they receive a comprehensive information letter. In three hospitals patients will be phoned by the researcher two weeks after receiving the information letter. They will be asked whether they have any questions and if they are interested to participate in the study. Second, the remaining four hospital sites are expected to recruit participants by a comprehensive information letter sent to them by mail. These patients will also be phoned by the researcher about two weeks after receiving the information letter to ask whether they are interested to participate in the study. When patients decide to participate, they are asked to return the included informed consent form with the reply-paid envelope within a week. Moreover, all participants will be informed about the possibility for psychological counseling nearby, so they know where to turn to in case they need (more) help.
Randomisation
After completing the baseline measure, every patient will be randomly assigned to either the online self-help training or care as usual with an allocation ratio of 1:1. Block randomisation (block size 10) will be carried out through a sealed envelope system, for each hospital separately. Both the participants and the researchers are blinded for the randomisation process, but not for the randomisation outcome. Statistical analysis will be done by a statistician blind for randomisation outcome.
Intervention
After conducting a survey in patients from the Dutch association of cancer patients’ organizations, the online self-help training “Less fear after cancer” was developed by the Helen Dowling Instituut, an institute for psycho-oncology in Bilthoven, the Netherlands.
“Less fear after cancer” is a tailored online self-help training based on cognitive behavioural therapy. Participants start the training by filling out the FCRI [3], after which they get (automated) feedback about their scores and a suggestion about which modules to follow. The FCRI and all modules are visible on worksheet of the intervention (see Fig. 1). First, participants follow two basic modules: 1) Psycho-education about FCR, its symptoms and learning to recognize symptoms of fear; and 2) The basic principles of cognitive behavioural therapy (this module is divided in two parts). After these basic modules women can choose from the following four the modules that are relevant to their situation: 1) How to stop rumination, behavioural techniques to stop ruminating; 2) Action, making an action plan about what one can do when fear of recurrence pops up; 3) Relax, audio files with relaxation practices; and 4) Reassurance, how and when to seek reassurance. Each module consists of an informative part and a practical part in which participants are motivated to do exercises or assignments in daily life. Participants are advised to take a week for each module they choose, so most participants will need four to six weeks depending on how many modules they do. It is explained that the more time they invest, the more effect they can expect from the training, but participants eventually choose themselves how much time is actually spent on the training. For every patient, the intervention will be available for three months.
The most important functionality of the online self-help training “Less fear after cancer” is the worksheet, because it gives an overview of the modules and access to the intervention. By clicking on a module, participants can access the information (texts, videos, audio files) and exercises of the self-help training. Other functionalities include a library with the information and forms in pdf format, videos, audio files, and a mailbox for technical assistance.
“Less fear after cancer” is fully automated and primarily non-guided and is delivered without professional support from a therapist. In this study, personal online support by an e-mail coach (the researcher) is available for the women that indicate a need for this. The e-mail coach can give technical assistance and eventually refer participants to their general practitioner or medical specialist when they indicate a need for professional help.
Usual care
The control group of this RCT has access to usual care. Care as usual may differ somewhat between hospitals and may include psychosocial care from within the hospital or elsewhere. In the Medical Consumption Questionnaire, use of psychosocial care will be assessed. Care as usual will be available in both conditions, in the intervention condition the online self-help is extra.
Outcomes
All participant outcomes will be gathered using online self-report questionnaires hosted by SurveyMonkey.com. Participants will receive an invitational e-mail with a link to complete the questionnaires online. The questionnaires at baseline, 3 months, 9 months, and 24 months include questions on socio-demographic and medical variables.
Primary outcomes
Fear of cancer recurrence will be assessed using the 43-item Dutch version of the FCRI [3]. The FCRI consists of statements rated on a 5-point Likert scale ranging from 0 (not at all or never) to 4 (a great deal or all the time). The FCRI includes seven subscales: triggers, severity, psychological distress, coping strategies, functioning impairments, insight, and reassurance. The triggers-subscale evaluates the presence of potential stimuli activating FCR. Psychological consequences of FCR are evaluated by the subscales psychological distress and functioning impairments. The insight scale measures the level of self-criticism towards FCR intensity. The reassurance- and coping strategies-scales measure a variety of coping strategies than can be used to cope with FCR including denial, wishful thinking, cognitive avoidance, and reassurance. The severity subscale assesses the presence and severity of intrusive thoughts or images associated with FCR and this scale can be used separately as a brief screening instrument of FCR and as an outcome measure [3]. The severity subscale is the primary outcome measure. The coping strategies- and functioning impairments-scale scores at baseline will be used in the predictor analysis. The original 42-item French-Canadian version of the FCRI had a good internal consistency (Cronbach’s α = 0.95 for the total score and α = 0.89 for the severity subscale) and stable over a 1-month interval (r = 0.89, p < 0.001) [3]. The scale has a robust factor structure and the results support construct validity with other self-report scales assessing FCR (r’s 0.68 to 0.78) or related constructs (r’s 0.43 to 0.66) and quality of life (r’s −0.20 to −0.36) [3]. The Dutch version of the FCRI (FCRI-NL) is currently being validated.
Fear of cancer recurrence will also be assessed with the Dutch version of the Cancer Worry Scale (CWS) [43]. The CWS assesses concerns about developing cancer or developing cancer again and the impact of these concerns on daily functioning. The Dutch version of the CWS consists of 8 items that are rated on a 4-point Likert scale ranging from 1 (never) to 4 (always). Higher scores indicate more frequent worries about cancer. A cut-off score of 13 (low ≤13, high ≥14) turned out to be optimal for detecting severe levels of FCR [44]. Moreover, the CWS is a reliable questionnaire (Cronbach’s α = 0.87) and evidence has been found to support the construct validity [44].
Secondary outcomes
Healthcare costs will be assessed with the Medical Consumption Questionnaire (MCQ), a questionnaire to assess non-disease specific healthcare costs [45]. More precisely, the volume of used healthcare will be assessed with the MCQ. Afterwards, the Dutch Manual on Cost Investigations will be used to calculate the healthcare costs [46].
Furthermore, the Dutch translation of the EuroQol-5D (EQ-5D), a generic measure of health status, will be used for the economic evaluation [47, 48]. The EQ-5D comprises five domains: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each domain consists of one question with three answer categories: (1) no problems, (2) some problems, and (3) extreme problems [47]. A health state can be derived by combining the scores from each dimension [47]. This results in a 5-digit number, for example state 12233. This health state indicates no problems with mobility, some problems with self-care and usual activities, and extreme problems with pain/discomfort and anxiety/depression [47, 48]. EQ-5D health states may be converted into a EQ-5D index by applying predetermined weights to the five domains [49]. The Dutch EQ-5D tariff will be used to value this EQ-5D index [47]. The EQ-5D index gives a societal-based global quantification of the patient’s health status on a scale ranging from 0 (death) to 1 (perfect health) [50]. For economic evaluation, the EQ-5D index scores will be used to determine quality-adjusted life years (QALYs) [50]. Patients will also be asked to rate their overall health status on a visual analogue scale (EQ-5D VAS), a quantitative self-rating of health status in which patients are asked to rate their current health state on a 0 (worst imaginable health status) to 100 (best imaginable health status) scale [48]. The EQ-5D is a ‘user-friendly’ questionnaire, with acceptable reliability and validity for various populations [51–53].
Psychological distress will be assessed with the corresponding subscale of the FCRI-NL [3].
Process outcomes
In addition to the self-report questionnaires, technical data on the use of the intervention will be gathered in the intervention group. For example, frequency of logins, duration of logins, and website activity will be evaluated.
Other outcomes
The use of extra (psychological) help will be assessed with the Medical Consumption Questionnaire (MCQ) [46]. Furthermore, help or referral by the e-mail coach will be registered and added to the MCQ score.
Psychosocial problems and risk factors will be assessed with the Psychosocial Distress Questionnaire-Breast Cancer (PDQ-BC) [54]. The PDQ-BC is a multi-dimensional screening instrument specific for breast cancer patients. It consists of nine subscales using 35 items assessing psychological risk factors (i.e. trait anxiety and (lack of) social support) and state anxiety, depressive symptoms, social problems, physical problems, body image, financial problems, and sexual problems. All items are answered on a 4-point Likert scale, ranging from 1 (not at all) to 4 (very much) [54]. For most subscales, high scores indicate more psychosocial problems, except for body image and social support for which higher scores indicate fewer problems [55]. The PDQ-BC appears to have a sufficient internal consistency, and good construct validity, test–retest reliability, and sensitivity to change. Furthermore, the PDQ-BC subscales state anxiety and depressive symptoms have a satisfactory sensitivity and specificity [54–56].
Self-efficacy for online self-help will be assessed with a questionnaire which was especially assembled for the current study. Bandura [57] argued that all-purpose measures of perceived self-efficacy usually have limited explanatory and predictive value because most of the items in an all-purpose test may have little or no relevance to the domain of functioning. Scales of perceived self-efficacy should be tailored to the particular domain of functioning that is the object of interest. Therefore, we collected many potentially useful items from various self-efficacy questionnaires [58–62]. In consultation with both professionals and patients, we improved and reduced the items to a 15 item questionnaire tailored to assess self-efficacy for our online self-help training. The items are rated on a 5-point Likert scale, ranging from 1 (not like me) to 5 (totally like me). The items are divided in three domains: 1) general internet use (3 items); 2) health related coping strategies (7 items); and 3) patients’ expectations on online self-help training for fear of cancer recurrence (5 items).
A subsample of patients (n = 16) will be asked about their experience with the online self-help in a semi-structured interview, to evaluate the online self-help training and to detect possible ways to further improve the training. There are different profiles of FCR, which vary according to its severity and the type of coping strategies used. Patients will be selected based on their baseline score on the FCRI and will represent different FCR-profiles: mild FCR-severity and low coping, mild FCR-severity and high coping, moderate FCR-severity and high coping, moderate FCR-severity and low coping [63]. From the first 50 participants who finished the online self-help training, four participants from each FCR-profile group will be randomly picked. If women refuse to participate, another participant will be randomly picked from that group.
Sample size calculation
The sample size calculation is based on a clinically relevant improvement on the FCRI severity subscale at 3 months. With an effect size of d = 0.3, a minimum number of 2.28 points on the FCRI severity subscale could be detected. Based on our experience with the FCRI, this seems to be a clinically relevant difference. In total 454 patients need to be included (227 in each group) to statistically detect the minimum effect size of d = 0.3 between mean FCRI severity subscale scores of both groups with a power of 0.8 and a two-sided alpha of 0.05. The power analysis program G*Power 3.1.7 was used to calculate the effect sizes [64]. Since in previous online intervention studies amongst breast cancer patients about half of all invited patients expressed an interest in participating and another 35 % was lost after randomization [65, 66], the aim of this study is to ask a minimum of 900 patients to participate in the study.
Statistical analysis
Primary analyses
Baseline characteristics in both groups will be compared to check if randomisation has resulted in an equal distribution of the baseline variables. Data will be analysed according to the intention-to-treat (ITT) principle. In the primary analyses, post-training scores will be compared between the two groups, controlling for baseline symptom levels. Analyses will be performed using t-tests and linear mixed models including exploratory predictor analyses. An advantage of linear mixed models is the optimal use of available data. The primary analysis is aimed at comparing the online self-help with care as usual at 3 months on the FCRI severity subscale. Analysis of co-variance will be performed to test whether the outcome variables differed between the online self-help and care as usual, using baseline level as covariate. The stability of the results will be analysed using the data from 9 months and 24 months after baseline, again using linear mixed models to control for the dependency caused by the repeated measurements. Time will be analysed as a categorical predictor with four levels (baseline - 3 months - 9 months - 24 months). Linear mixed models with a specified covariance pattern model will be used to examine the course of FCR [67]. The fixed-effects parameters of the models will be estimated with maximum likelihood. Inspection of the Log likelihood ratio test, Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC) with restricted maximum likelihood (REML) will be used to find most suitable covariance pattern model (e.g., compound symmetry, autoregressive, unstructured).
For the secondary analyses, significant predictors will be selected using exploratory predictor analyses. Then multivariate hierarchical regression analysis will be performed to assess predictors of effect. The predictors that are found statistically significant will be added as interaction terms with condition (online self-help versus care as usual). Drop outs (attrition) will be closely investigated and predictors of drop outs will also be analysed by multivariate regression analysis.
Cost-effectiveness analyses
The economic evaluation will consist of a cost-effectiveness analysis and a cost-utility analysis, both done from a societal perspective [50]. The cost-effectiveness ratio will represent the costs per significantly improved participant, while the cost-utility analysis will represent the costs per additional quality-adjusted life year (QALY). The time horizon will be life time. As the study period is limited in time, cost and effect will be modelled in time, using the assumption that the spontaneous recovery is 2 years. The effects of this assumption on the cost-effectiveness ratio will be tested by testing the scenarios of spontaneous recovery after 6 months, 1 year and 5 years.
Costs will be estimated from a societal perspective and will thus include the costs related to the intervention, all other healthcare costs and non-healthcare costs during the time horizon of the study. Healthcare consumption will be measured with the MCQ. Healthcare consumption includes all non-disease specific healthcare used in the previous 3 months, such as visiting the general practitioner or other healthcare providers, emergency room visits, hospitalisation, and medication use. Then, the guidelines as descripted by the Dutch Manual on Cost Investigations will be used to calculate the healthcare costs [46]. For healthcare where no guideline or standard prices are available, real cost prices will be determined or, when available, derived from the health care provider administration.
QALYs will be calculated from EQ-5D health states using the Dutch EQ-5D tariff [47]. Non-parametric bootstrap simulations will be used to estimate uncertainty intervals around the ICERs, in order to deal with the most likely skewed distributions of costs. Cost-effectiveness acceptability curves will be calculated to show the probability that the intervention is cost-effective in comparison with the control group, given varying thresholds for the willingness-to-pay for gaining one unit of effect, i.e. a QALY or a significantly improved participant. The robustness of the results will be explored using one-way sensitivity analyses in which the input variables for assessing both cost and effectiveness are varied.