Effectiveness of Personalized Surveillance and Aftercare for Breast Cancer

1. Treatment of subjects: personalized follow-up Personalized Surveillance Plan (PSP): Around the first surveillance mammogram (i.e. one year after end of treatment), the personalized surveillance plan (PSP) is generated by means of the PSP decision aid, called the 'Breast Cancer Surveillance Decision Aid' (16). This Aid incorporates the INFLUENCE tool 3.0, which is a sequel of the INFLUENCE tool 2.0 (15) and will be developed during this project. Compared to INFLUENCE 2.0, INFLUENCE 3.0 will additionally include patients treated with neoadjuvant therapy a broader population than INFLUENCE 2.0 (including patients treated with neoadjuvant therapy) and will include a more recent population in order to provide more contemporary risk estimates that are applicable in a broader population. During an outpatient clinic visit, around the first surveillance mammogram (i.e. approximately one year after the diagnosis), the INFLUENCE 3.0 prediction tool, as part of the PSP decision aid, is completed by the HCP (i.e. surgical oncologist or nurse specialist) and patient together by filling in data on patient, tumour and treatment characteristics. The estimated personal risk will be explained to the patient and summarized on a leaflet that also outlines the options possible for the patient (e.g. annual mammogram or less frequent, duration of follow-up, how to deliver the result from the mammogram). The patient receives a personal account with which she can, at home, complete the PSP decision aid that provides information about different surveillance options and a value clarification exercise to help the patient to get insight in her personal needs and preferences. This helps the patient to consider the pros and cons of the different options. A summary sheet based on patients' answers, combined with the result of the first surveillance mammogram, is used in the next consultation to make shared decisions on a personalized surveillance plan (PSP). Personalized Aftercare Plan (PAP): To support creation of this PAP, a patient decision aid (PtDA) will be used, which assesses patients' needs, offers information and provides a summary of patients' needs and preferences regarding the aftercare trajectory. The content of this PtDA will be developed in five cocreation sessions with a multidisciplinary team of researchers, patient representatives and care providers. First, needs assessment studies among patients and care providers will be conducted, which results serve as input for the content of the PtDA. This content will be critically revised by the team and rewritten to B1 language level (Common European Framework of Reference for Languages). Usability will be tested, consisting of think-aloud sessions with patients and interviews by telephone among health care professionals. During aftercare consultation(s) in the first year after the end of patients' treatment, the HCP (i.e. nurse or nurse specialist) will introduce the PtDA to the patient. Next, patients access the online PtDA to complete a needs assessment and receive information about possible effects of breast cancer, available options and choices that she has concerning her aftercare trajectory and available resources for help and support. Patients can weigh options and fill in preferences and considerations. Once patients have completed the PtDA, a summary sheet will automatically be created, containing an overview of patient-reported needs, preferences and considerations, which can be used as a base for final decision-making on the PAP in a consultation with their care provider. These decisions will compose the PAP, which will most likely include decisions on organisation of aftercare (e.g. further support or referrals, mode of contact, involved care providers) and signals to seek care for and contact details. Since patients' needs may vary over time, the care provider can introduce the PtDA multiple times during the aftercare trajectory. Also the PAP might be re-evaluated and adapted during the aftercare trajectory, depending on patients' needs and arranged contact frequencies. 2. Sample size calculation: The sample size was estimated using a user-written script for mITS designs in the Statistical Analysis System (SAS) software program. Since there are two primary parameters, a Bonferroni correction will be used to correct for multiple testing and therefore set the statistical significance level at alpha=0.025 (two sided). The effects the investigators wish to measure are a difference of 1.52 on the Cancer Worry Scale (CWS; range 6-24) and a difference of 4.8 on the EQ-VAS score (range 0-100), which is part of the EQ-5D. This decision is based on the aim to detect a small to moderate difference of 0.4 times the standard deviation, which was found to be around 3.8 for the CWS and 12 for the EQ-VAS score in previous studies. From a clinical point of view, a difference of 1.52 on the CWS is relevant for the purpose to estimate the effectiveness of personalized follow-up: even a small decrease in cancer worry can lead to improved quality of life. A difference of 4.8 on the VAS score is considered to be clinically relevant as well. A correlation is expected between the first two measurements within one hospital of 80%, and this correlation is expected to drop to 50% when comparing the first with the last measurement. In addition, an intraclass correlation coefficient of 0.15 is assumed. Taken 25% loss-to-follow up into consideration, each hospital will have to include four patients per period of three weeks in order to detect a difference of 1.52 on the CWS and 4.8 on the VAS score with 84%. The total inclusion time per hospital is 26 periods (78 weeks, or approximately 18 months), which amounts to a number of 104 patients per hospital and a study population of N=1,040. 3. Statistical analyses: An overview of the demographic and clinical characteristics will be provided using descriptive statistics. Continuous data will be expressed as a mean with the standard deviation (SD), or the Interquartile range (IQR) where appropriate. Categorical data will be expressed as frequencies (%). All questionnaires will be analysed in accordance with their corresponding manual. Self-composed questions (i.e. perceived risk of recurrence, adjusted questions from the CQ-Breast Index, demographics) will be analysed per item. To assess the effectiveness of personalized surveillance and aftercare, all outcome parameters will be compared between the current-care and personalized-care groups. Time series patterns will be visualized pre- and post-personalization to assess possible change in pattern after implementation of the personalization, for each hospital separately and combined. In this way the investigators can identify any underlying trends, seasonal patterns and outliers. To test the change in level and slope associated with the personalization and to control for other (confounding and overall trend) effects, segmented regression analyses will be used in which piecewise regression lines are fitted to each segment of time series, allowing each segment to exhibit different trends. To correct for correlation between repeated measurements residual plots against time will be visually examined, which can additionally be statistically tested using the Durbin-Watson statistic. Autocorrelation will consequently be adjusted for by including the autocorrelation parameter in the segmented regression model. Intention-to-treat analyses are done to estimate the effectiveness of personalized follow-up on the outcomes of the questionnaires. The Bonferroni correction will be used to adjust for multiple testing. Patients included during the transition phase will be analysed as receiving current care or personalized care dependent on whether PSP and PAP was applied. Sensitivity analyses will be performed with and without the patients included during the transition period, since there may be differences between care provided during and after the transition phase. In case of missing data, the investigators will record the percentage of drop-out and missing at each follow-up timepoint. If necessary, multiple imputation (if the assumptions for this technique are met) will be performed to ensure accurate analysis. Thereafter, meta-analyses will be performed to evaluate the average intervention effect per patient group across all hospitals and the overall effect across all patient groups and hospitals. A cost-effectiveness and cost-utility analysis will be performed comparing costs and effects of "personalized follow-up care" versus "usual care", using a two-year (study-based; based on data from the current study) and lifetime (model-based; based on the study and extrapolations by means of data from literature) time horizon. Information will be derived on the impact on personalized follow-up care on cancer worries, QoL, healthcare costs and (shift in) resources. Direct healthcare costs based on activities extracted from the EHRs (e.g. number of mammograms, consultations) will be multiplied by costs described in 'Benchmark costs from the Netherlands' or from the 'Nederlandse Zorgautoriteit' (NZa). For the costs of the interventions (decision support tools), an activity-based costing method will be performed for development, use and maintenance of the decision support platforms. Indirect costs that will be taken into consideration are healthcare consumption outside the hospital and health-related productivity losses. Productivity losses will be calculated by means of the Friction cost method, according to the International Society of Pharmaco-economical Organization and Research (ISPOR) guidelines. The cost-effectiveness will be expressed in incremental costs per patient with a clinically relevant improvement on the CWS as primary outcome of the NABOR study. The cost-utility will be expressed in incremental costs per quality adjusted life years (QALYs) gained, obtained from the EQ-5D-5L (including the VAS). Long term consequences will be based on both the study results and literature to extrapolate the patient outcomes for a lifelong time horizon. The cost-effectiveness analysis will be performed according to the guidelines for economic evaluations of the Dutch Zorginstituut (ZIN). 4. Handling of data Data sources for our study: * NCR (Netherlands Cancer Registry): data (e.g. patient-, tumor- and treatment-related characteristics) are already gathered and part of the general data collection of the NCR * Additional patient data from the EHR: collected during the project * PROFILES: a data registry collecting answers on questionnaires during the project Handling of data: For obtaining the data form the NCR, additional patient data from the EHR and answers on questionnaires through PROFILES, the study participants will provide Informed Consent. Registration will be done by the PROFILES registry. In order to send questionnaires to participants, participants will be asked permission for registration of their name and email-address or postal address in the PROFILES registry. In order to combine answers on questionnaires with data from NCR and the EHR later on, the PROFILES registry also needs their patient number and date of birth. Participants will also be asked permission for registration of their patient number and date of birth in the PROFILES registry. After the patient signs Informed Consent, the HCP will register the patient in PROFILES. HCPs can obtain the personal data and sequence number in the PROFILES registry, but have no access to the answers on questionnaires. Accessibility of data: All the collected data is patient data, this cannot be made publicly available. Data from the NCR and additional EHR data can be requested, a local privacy committee evaluates requests. PROFILES data (i.e. answers on questionnaires) can be requested as well, and linked to NCR data if needed.