The intervention's efficacy is being evaluated using a randomized, controlled trial among patients undergoing a course of chemotherapy at two Department of Veterans Affairs (VA) oncology clinics located in Ann Arbor, MI and Fargo, ND. Three hundred patients with solid tumors undergoing chemotherapy and reporting at least one core symptom at a moderate level are being assigned to either 10 weeks of weekly ATSA contacts or 10 weeks of weekly ATSA contacts plus caregiver alerts and access to the Cancer CarePartners website. The duration and intensity of the intervention is premised on past work suggesting that ATSA's benefits peak at 10 weeks [11
The goal of the study is to determine if, when compared to a comparison group receiving ATSA alone, patients who receive ATSA and whose caregivers receive access to Cancer CarePartners will report: significantly less total symptom severity; and secondarily improved physical functioning, adherence to chemotherapy, and more appropriate utilization of health services at 10 weeks. The study also will determine whether caregivers provide more social support and report less caregiver burden and distress.
Patients and caregivers will be surveyed at intake, 10, and 14 weeks. Further, we will track patient participation with ATSA and (in the experimental arm) caregiver use of the Website and social support provided, allowing for indicators of intervention dose to be incorporated into our analyses. Finally, all patients will receive routine oncology, hospice, and/or palliative care.
Study patients must be 21 years or older, cognitively intact, English speaking, able to hear, and able to use the phone. They must have home access to a touch-tone telephone. They must have a solid tumor; for example, prostate, lung, colorectal, breast, non-Hodgkin lymphoma, or head/neck cancer. Patients may have new or recurrent disease in early or late stage. All patients must be within 10 weeks of initiating chemotherapy and, if they are being treated for recurrent disease, must have experienced a treatment free interval of 10 weeks between the new and past treatment. Patients may live alone or with others, with or without an in-home caregiver. Patients are excluded if they have been diagnosed with a hematologic malignancy or their treatment regimen involves bone marrow transplantation because such patients' symptom profiles differ from those of other cancer patients. Also, patients will be excluded if they (a) have untreated mental illness or cognitive impairment limiting their ability to participate with ATSA, (b) cannot identify at least one potential consenting caregiver, (c) are institutionalized or enrolled in hospice prior to enrollment, or (d) plan to discontinue receiving the majority of care from one of the study sites during the follow-up period.
For the purposes of this study, a potential caregiver is defined as a person living inside or outside of the patient's household who has at least monthly contact with the patient, either by phone or in person, and is willing to help the patient manage his/her symptoms. S/he must be 21 years or older, cognitively intact, English speaking, and able to hear and speak English for telephone interviews. S/he must have access to a telephone and computer with a high-speed Internet connection.
Caregivers need not be providing direct care prior to the study; indeed, a major motivation for this intervention is that many well-intentioned friends and family members would be willing to provide more cancer support, but currently do not do so because there is no structure in place to make that possible. Caregivers will be excluded if they, themselves, are on active treatment for cancer. They will also be excluded if they have poorly controlled mental illness, moderate cognitive impairment, or are unable to be interviewed by telephone.
Recruitment & Randomization
Study staff review the schedules of participating chemotherapy infusion clinics weekly to identify potentially eligible patients. The list of potentially eligible patients is provided to staff oncologists for their approval prior to attempting recruitment.
Eligible patients are approached by research assistants after their second week of chemotherapy. Research staff explain the study to them, screen for eligibility, offer enrollment, and obtain informed consent. Patients agreeing to participate are asked to identify at least four potential caregivers using an algorithm based on the Norbeck Social Support Scale to determine the ranking of choices according to their involvement in the patient's life [32
]. Research assistants then obtain contact information for the potential caregivers. Potential caregivers are recruited by research assistants by phone or in person if they accompany the patient to clinic. The research assistants explain the study, screen for eligibility, and obtain written informed consent. When a potential caregiver declines participation in the study, the research assistants contact the remaining potential caregivers on the list following the ranking provided by the patient. If no potential caregiver consents to participate, the patient is excluded from the study.
Patient participants are asked to identify preferred calling times for their ATSA calls. In addition, they are instructed on the use ATSA and rehearse an ATSA symptom assessment with the research assistant while in clinic. All patient participants are told to contact the project manager if they have any questions during their participation in the study and are given the telephone number for their respective oncology clinic for clinical issues not related to the study or when advised to do so by ATSA.
Basic sociodemographic and clinical data are kept on patients and caregivers who are ineligible for or refuse enrollment in order to understand the reach of the intervention in the broader population of eligible patients. In particular, we will describe the extent to which the requirement of caregiver Internet access serves as a barrier to recruitment among economically disadvantaged or rural patients.
After completing the baseline assessment, participating patients are randomized to either the control or intervention arm by a computer minimization program that balances the arms for type of cancer (lung vs. other) and the type of caregiver (in-home vs. out of home). The minimization procedure, known as 'adaptive randomization' balances trial arms according to the history of previous allocations. Arms are being balanced for lung cancer since patients with lung cancer have a symptom pattern distinct from patients with other solid tumors [34
]. Arms are being balanced for type of caregiver because in-home caregivers have more access to the patient than do caregivers living at a distance and, hence, more opportunities for providing support.
Patients randomized to the control arm of the study receive the Symptom Management Toolkit followed by weekly ATSA calls for 10 weeks. Caregivers in the control arm receive an email with a PDF copy of "What you need to know about cancer" from the National Cancer Institute (available from http://www.cancer.gov/cancertopics/wyntk/colon-and-rectal/page1
). This booklet on cancer discusses possible causes, symptoms, diagnosis, treatment, emotional issues, and questions to ask the doctor.
Patients randomized to the intervention arm receive the Symptom Management Toolkit and weekly ATSA calls for 10 weeks, as received by control-group participants. In addition, however, their caregiver receives weekly emails prompting them to log into the Cancer CarePartners website when their patient reports any core symptom severity of four or higher. The website informs caregivers of their patients' symptom assessment scores and provides them with advice for how to help, as described above. For the first 4 weeks in the study, staff monitor each caregiver's use of the website. When caregivers do not log into Cancer CarePartners within 48 hours of an assessment, staff telephone them to troubleshoot as needed.
Data are being gathered from patient and caregiver surveys, ATSA, the Cancer CarePartners website, and medical record review.
Patient & caregiver surveys
Research assistants interview patients and caregivers by phone at baseline, 10 weeks post baseline, and 14 weeks post baseline. Patient surveys assess the study primary outcome (i.e. symptom severity) and secondary outcomes (i.e. physical and mental functioning, adherence to chemotherapy, and utilization of health services). Surveys also gather data on potential mediators of intervention effects (i.e. self-efficacy and actualized social support). Goals of the caregiver surveys are to measure caregiver outcomes (i.e. distress and burden). Patients and caregivers receive $15 for each completed survey.
ATSA tracks call attempts, call outcomes (completed vs. incomplete), and reported symptom scores. The number of completed calls relative to call attempts will be used to describe patient uptake of the intervention as well as the dose received.
Cancer CarePartners website
The website tracks the number of email alerts sent to caregivers, numbers of log-in attempts, the web pages visited, the time spent per page, the tasks chosen, and caregivers' responses to the weekly task list survey. These data will be used to describe the usability of the website, as well as caregivers' choices of caregiving tasks from suggested options.
A registered nurse reviews the patient's electronic medical record at week 14 or following withdrawal or death to assess covariates (e.g. comorbidities, supportive care received) and secondary outcomes (utilization, adherence with chemotherapy, and adverse events). The nurse reviews all inpatient and outpatient oncology notes written from four weeks prior to 14 weeks after informed consent. All modalities of cancer treatment and adjunct supportive care received during the observation period are noted (including consultations from palliative care services or referrals to hospice or home health nursing). Surgical procedures, the dose and duration of radiation treatment, and the dose and duration of chemotherapy (both recommended and actualized) will be noted, as will any medication changes (and reasons thereof).
Description of Patient Measures
Summed symptom severity score
(source: patient baseline, 10 week, and 14 week surveys). A summed symptom score will be generated based on the sum of symptom severity scores (scored 0-10) across the 8 core symptoms.
Adherence to chemotherapy
(source: medical record audit). Adherence to chemotherapy will be determined by the total number of dose reductions, infusion delays, and agent discontinuations.
(source: patient baseline, 10 week, and 14 week surveys) Physical functioning will be measured using the SF-36 summary measure for physical function which has an alpha reliability exceeding 0.80 [36
Utilization of services
(source: medical record audit). The utilization of inpatient and outpatient services will be measured in terms of the number of cancer-related: hospitalizations, visits to emergency or urgent care, scheduled and unscheduled contacts with oncology (in person or by phone), and referrals to home health or hospice.
(source: patient baseline and 10 week surveys). The Cancer Behavior Inventory
(CBI) Version 2 will be employed to measure patient self-efficacy. The CBI has an internal validity of .94 and covers 7 aspects of self-management, including: maintenance of activity and independence, seeking and understanding medical information, stress management, coping with side effects of treatment, accepting cancer/maintaining positive attitude, affective regulation, and seeking support [37
(source: patient baseline and 10 week surveys). Two dimensions of social support will be examined: actualized and perceived. Actualized social support will be measured using the Inventory of Socially Supportive Behaviors (ISSB) [38
]. The ISSB is a 40 item measure that assesses the receipt of directive guidance (e.g. caregiver suggested the patient take some action), nondirective support (e.g. caregiver expressed concern), positive social exchange (e.g. the caregiver talked with patient about non-cancer related interests), and tangible assistance (e.g. caregiver drove patient to appointment). The ISSB has excellent internal consistency and good test-retest reliability (> 0.90) [38
]. Perceived social support will be measured using the Social Provisions Scale (SPS). The internal consistency scores of the six subscales and total scale of the SPS are above .70 and .90 respectively.
Description of Caregiver Measures
(source: caregiver baseline, 10 week, and 14 week surveys). The Caregiver Reaction Assessment (CRA) measures 5 dimensions of caregiver burden, including: caregiver esteem, lack of family support, impact on finances, impact on schedule, and impact on health. The CRA has good internal consistency (alpha 0.80-0.90) and a high degree of reliability [39
Caregiver distress (source: caregiver baseline, 10 week, and 14 week surveys)
We will employ the Center for Epidemiological Studies-Depression Scale (CES-D) to measure depression in CarePartners. The CES-D is a scale widely used to measure depression in caregivers because of its ease of use (it has 20 items) and high reliability (alpha = 0.90) [40
Mastery of caregiving
(source: caregiver baseline and 10 week surveys). Caregiver mastery will be measured using the problem solving skill subscale of the caregiver self-efficacy instrument by Zeiss; this subscale has a reliability of .83 [41
Research assistants rehearse an ATSA call with every patient participant at the time of enrollment. They review call data from ATSA regularly to identify patients who do not respond to scheduled calls. When patients do not respond to multiple ATSA attempts over 48 hours (at any period throughout the 10 weeks of the trial), research assistants contact them to determine the reason; reasons for non-compliance are tracked. Support is offered to patients whose reason for noncompliance relates to technical issues.
Similarly, for the first 4 weeks on the study, caregivers' Web logs are reviewed to identify caregivers who either have not been able to sign-in or create a task list. Research assistants contact caregivers whose Web logs indicate noncompliance and call caregivers to identify the cause. Depending on the reason for the difficulty, research assistants offer to guide the caregiver by telephone during their next online session. Research assistants track all reasons for noncompliance with the website.
Patients and caregivers are not permitted to use ATSA or the website until the baseline survey is completed. Data is entered into a clinical data tracking system that alerts staff of survey due dates and incomplete data. The date the baseline survey is completed determines the dates for the 10 and 14 week surveys. Up to 10 attempts are made to contact each subject or caregiver for scheduling an interview. Abbreviated versions of all interviews are available for patients who have difficulty completing the entire survey. Patients and caregivers receive a $15 gift card for each survey they complete. Monthly updates with comments on completeness of data are maintained and reviewed by the Project Manager.
The success of randomization is monitored quarterly by comparing the demographic and clinical characteristics of patients across study arms. Clinical staff is blinded to the assignment arm, as are the analysts examining the data.
Monitoring and Reporting of Adverse Events
The study includes several mechanisms for monitoring and identifying adverse events. First, symptom data from ATSA are reviewed weekly by the Principal Investigator (PI) who is a physician. When a patient reports multiple symptoms at severe levels (7 or above), the PI reviews the medical record to insure that the symptoms have been recognized by oncology staff. In the event that there are no progress notes indicating that the patient has been recently seen or contacted, the PI calls the oncology nurse case manager to alert them to the patient's reports.
When patients fail to respond to ATSA calls on two consecutive days, research staff calls the patient to ensure their safety. When patients cannot be reached, study staff informs the caregiver and oncology nurse case manager. Patients who are hospitalized are given the option to suspend their ATSA calls. Patients enrolling in hospice are given the option to come off study.
Subjects and clinicians have access to a toll-free number they can call with study-related problems. Caregivers have access to the research team's contact information through the website. Adverse clinical events are reported immediately to the clinician of record as well as the IRB. All deaths are immediately investigated by a physician not affiliated with the study in order to determine the cause.
Sample Size Calculation
Testing the primary outcome using two sided t-testing will require at least 224 dyads (patients and caregivers) to complete the study (112 per arm). This sample size affords 80% power to detect a difference as small as 0.33 standard deviations in mean summed symptom severity (Hypothesis 1) and social support (Hypothesis 2) between arms using 0.05 as the level of significance. This consistent with standards for clinical relevance and those seen in similar interventions [26
]. Moreover, this sample size is sufficient to accommodate longitudinal analyses with repeated measures. To account for 25% attrition (consistent with prior trials of ATSA) and ensure that 112 dyads in each group are available for analysis, the goal is to enroll 300 dyads in the trial [26
Graphical analyses will include: (1) examining distributions of the various numerical outcome variables using box plots and histograms to investigate skewness, gaps and outliers; (2) side-by-side box plots and back-to-back histograms to graphically rule out baseline differences in outcomes between the two study groups; (3) plots of cross-sectional outcome means over time to assess the longitudinal trends within and between the two arms; and (4) scatter plots of change in outcome variables versus change in potential mediator variables. Variables with highly skewed distributions will be transformed. When such transformations are not successful, we will employ generalized linear models in the primary analyses. In addition, we will consider dichotomizing the variables using clinically-relevant cut points.
We will test for possible baseline differences in the outcomes and characteristics of the patients and caregivers in the two study groups including but not limited to: patient and caregiver socio-demographic and clinical characteristics, symptom scores, physical and mental functioning, social support, caregiver relationship type (spouse vs. other), and caregiver location (in-home vs. out-of-home). If any significant baseline differences are identified, intervention effects will be estimated using multivariable analysis adjusted for baseline variables as well as stratified analysis using propensity scores.
To ensure the validity of findings, baseline characteristics of patients and caregivers who drop out of the study will be compared by study group. We will also summarize and compare reasons for attrition. If any differences are found, the appropriate covariance adjustments or propensity score stratifications will be employed in the primary analysis. To ensure the generalizability of our findings, we will also compare the baseline characteristics of those completing the study against those of subjects who drop out.
Primary Analytic Strategy
The primary effect of the intervention will be determined using data obtained at 10 weeks, and the sustained effect will be determined using data obtained at 14 weeks. Cross group comparisons will be made based on an intention-to-treat basis. We will also test the hypotheses with a longitudinal design by fitting a linear mixed-effects modelor a generalized linear mixed effects model to estimate effect averaged over 10 and 14 weeks, adjusting for potential within-person correlation and baseline values of the outcome variables [15
When missing values occur, we will determine whether missingness is random or associated with patient or caregiver characteristics using standard regression models. As long as we do not observe any bias in patterns of dropout or missing data (see attrition analysis above), we will use regression techniques allowing for missing at random. Models will include 10 and 14 week assessment data as longitudinal outcome values, an intervention effect indicator, and baseline values of the outcome variable and other covariates as fixed independent variables. If the analyses indicate that missingness depends on unobserved outcome values(, we will account for missing data using a pattern mixture model [43
]. This technique allows the use of data from all participants (not just completers) and provides an unbiased intervention effect estimate.
We will compare summed symptom scores at 10 weeks between the control and intervention groups using a multiple regression model. The model will include a dummy variable for study group assignment as the main independent variable and will be adjusted for baseline symptom scores. While we do not expect the groups to be unbalanced with respect to other potential predictors of the outcome, we will adjust for these if randomization is not fully successful. The parameter estimate of the group assignment dummy variable and its statistical significance will estimate and test for the outcome difference between the two study groups at week 10, adjusting for baseline values of the outcome. A similar approach will be used to test the intervention effect on secondary outcomes (namely physical functioning, service utilization, and chemotherapy adherence). To determine if there is maintenance of the intervention effects on each outcome, we will use a similar regression model using the corresponding measurements at week 14, and a longitudinal model for the outcomes at 10 and 14 weeks.
Our analytic approach for testing Hypothesis 2 will be similar to that described above. We will use multiple regression models to compare group differences in patient-reported social support scores as measured by the Inventory of Socially Supportive Behaviors at 10 weeks, adjusting for baseline values of the ISSB [38
]. The model will include a dummy variable for study group assignment as well as the baseline value of the ISSB scores, and the parameter estimate of the dummy variable will estimate the intervention effect on caregivers at 10 weeks. A similar approach will be used to test the intervention effect on caregiver burden and distress.
To determine whether the improvements in patient outcomes are mediated by patient self-efficacy and social support, we will employ the approach originally described by Baron and Kenny and more recently further developed by others [44
]. In addition to the linear models relating outcomes to study group, we will fit two additional models relating mediator to the study group, and the model relating the outcome to the mediator and study group. To establish mediation, the study group effect has to be significant in all but the last model. If in the last model the study group effect is not significant, then we will conclude that complete mediation has occurred. If in the last model the study group effect is reduced, but is still significant, then we will conclude that partial mediation has occurred. To test for mediation, Sobel's test will be carried out, and percent of variation in outcomes mediated by patient or caregiver mediators will be determined [49