Lay Coach-Led Early Palliative Care for Underserved Advanced Cancer Caregivers: The Project ENABLE Cornerstone RCT

Multi-Year Funded Merit Award - Years 6-7: For this 2-year merit award funding, we propose to continue enrollment beyond 24-weeks and extend follow-up with participants long-term, through the patient's end of life and for 24 weeks post-patient death to test the effectiveness of Cornerstone on caregiver and patient outcomes compared to usual care.

Aim 1: Test the long-term effects of Cornerstone on caregiver outcomes at the patient's end of life.

Hypothesis Primary: Compared to usual care, Cornerstone caregiver participants will report lower distress as measured by the Hospital Anxiety and Depression Scale in the 24 weeks prior to death. We considered: A) a terminal decline longitudinal design with baseline and at least two time points prior to a patient's end of life (weeks 24 and 12 prior to death) comparing CGs randomized to ENABLE Cornerstone vs. usual care; B) 80% power, (1-β=.8) to detect a between-group mean difference, if it exists, in the outcome scales included in Specific Aim 1 (CG distress, QOL, burden); C) a significance level adjusted with a False Discovery Rate approach (5% FDR) assuming intervention effects on at least half of the effective outcome scales included in Specific Aim 1, i.e., α=.05×.5=.025; D) intra-subject correlation of ρ=.5 between repeated measurements; and E) data from 150 CGs with deceased patients (approximately equally split between study groups), composed of 67 deaths that occurred during the initial 104 study weeks plus additional 83 patient deaths during 132 additional weeks of recruitment and follow-up. Under these assumptions, for a continuous outcome such as the primary outcome scales (HADS-Anxiety and HADS-Depression), the variance of the within-subject average of three assessments prior to death is V_avg=(σ^2⁄3)(1+2ρ), where σ^2 is the variance of the outcome and ρ is the intra-subject correlation. Standardizing the outcome (σ^2=1) to compute a detectable difference as an effect size (Cohen's d), results in the variance of the within-subject average estimated at V_avg=(1⁄3)∙(1+2∙.5)=.67. Thus, the detectable time-averaged mean difference between the study groups over the three timepoints, given a sample size n for each of the two study groups, can therefore be estimated as d=(Z_(α⁄2)-Z_(1-β))∙(√((2V_avg)⁄n)), where Zk is the kth percentile of the standard normal distribution. So, with a sample size of n=75 per study group the detectable difference is d≈.34; and in terms of precision of the estimate, with a standard error of ├ √((2V_avg)⁄n)┤= .13 the sample size allows a 97.5% confidence interval for d of half-width .30.

Hypotheses Secondary: Compared to usual care, Cornerstone participants will report better a) QOL (PROMIS Global Health 10)40,54 and b) lower burden (Montgomery-Borgatta Burden Scale)39 in the 24 weeks prior to death.

Aim 2: Test the long-term effects of Cornerstone on patient outcomes at end of life.

Hypotheses: Compared to usual care, Cornerstone participants' care recipients will report a) lower distress as measured by the Hospital Anxiety and Depression Scale (HADS), b) better QOL, c) better quality of death and dying, and d) lower healthcare utilization, including EOL healthcare utilization, in the 24 weeks prior to death. A longitudinal model retrospective from patient death (i.e., terminal decline) fitted with linear mixed methods will be used for each outcome. The focus of inference will be the between-group difference in mean outcome at weeks 12 and 24 before death. A random effect for subject will be fitted to account for covariance among repeated measures on the same individuals. The fixed-effect part of the model will include coefficients for group, baseline values, and coefficients for time retrospective from death by group (i.e., multiple degree-of- freedom time-by-group interaction) that accommodate at least 2 time-points. Model estimated means by group at weeks 12 and 24 prior to death will be estimated from the model, and linear contrasts will be used to estimate a between-group time-averaged difference over the 24 weeks before death. If necessary, we will conduct covariate adjustment for baseline factors unbalanced between the groups or predictive of dropout.

Aim 3: Test the long-term effects of Cornerstone on caregiver bereavement outcomes.

Hypotheses: Compared to usual care, Cornerstone caregiver participants will report a) lower distress (HADS), b) higher QOL, c) lower burden, d) higher resilience, and e) lower grief over 24 weeks post-patient death. For bereavement scales, similarly to the above analyses, we will fit longitudinal models with time coefficient indicators for 12 and 24 weeks post-death and estimate intervention effects via linear contrasts.

***Parent Study - Years 1 - 5: This hybrid type I randomized effectiveness-implementation trial is designed to answer the question "Is ENABLE Cornerstone efficacious in enhancing caregiver and patient outcomes?" The primary aims are to assess the effectiveness of ENABLE Cornerstone over 24 weeks (Primary and Secondary Aim). The investigators will also explore implementation cost and cost effectiveness, including healthcare utilization (Aim 3) and other potential mediator/moderator effects of the intervention on coping, social support, and preparedness (Exploratory Aim). The study will recruit 206 family caregivers over the age of 21 reporting that they are the primary unpaid provider of support to a close family member or friend who has been newly-diagnosed with an advanced stage cancer within the past 60 days. Half of the participants will be randomized to ENABLE Cornerstone (n=103) which consists of 6 brief, weekly in-person/telephone sessions followed by monthly follow-up every 4 weeks as needed by participants. The other half will be assigned to a usual care condition (n=103). The randomization scheme, stratified by site (UAB and MCI) in block lengths of 8, will be executed in REDCap, a clinical trials management software program using a computer-generated algorithm. Assessments will be administered via telephone, by a research coordinator blind to group assignment. Participants will complete outcome assessments upon enrollment (T1) and at 12 (T2), 24 (T3) and every 12 weeks thereafter. The T2 assessment captures the short-term outcomes of the most intensive part of the intervention. The T3 assessment will capture the long-term outcomes of the intervention. Subsequent assessments explore the sustained effect of ENABLE Cornerstone at the patient's end-of-life and post death. The primary outcome is whether the intervention affects the "slope" of FCG distress at 24 weeks.

An intention to treat (ITT) approach will be used for all analyses. That is, all caregivers and patients will be included in their respective assigned conditions, regardless of their degree of participation in the study. Primary data analysis will begin with descriptive statistics for baseline caregiver and patient characteristics and outcomes. The investigators will examine balance between study groups with respect to baseline characteristics using effect sizes such as the standardized mean difference for numerical variables and Cramer's V for categorical variables. Conceptually relevant baseline factors showing non-trivial imbalances between groups will be then used as adjusting covariates in the longitudinal group comparisons. Distributional assumptions will be examined and, when appropriate, we will employ inferential and modeling procedures robust to distributional assumptions such as normality. Patterns of missing data will be examined and whether baseline characteristics are associated with dropout.

Conceptually relevant baseline factors predictive of dropout will then be used as adjusting covariates in the longitudinal group comparisons. Mixed-effect modeling techniques and covariate adjustment will reduce the impact of missing data, as the missingness is not assumed completely at random (MCAR) but conditionally (on the covariates) at random (i.e., MAR, a milder assumption). We will use the latest versions of SAS and R for all analyses and reports.

For specific aims 1 and 2, a longitudinal model fitted with linear mixed methods will be used for each outcome. The focus of inference will be the between-group difference in outcome change from baseline over 12 and 24 weeks, modeled by a time by group interaction. A random effect for subject will be fitted to account for covariance among repeated measures on the same individuals. The fixed-effect part of the model has the form: Y=b_0+b_1?Cornerstone+b_2?Week12+b_3?Week24+b_4?Cornerstone�Week12+b_5?Cornerstone�Week24 , where Y ? is the expected value (i.e., the mean) of the outcome according to the predictor combinations, Cornerstone is a binary indicator of group assignment (1=ENABLE Cornerstone, 0=usual care), and Week24 and Week 12 are binary indicators of the follow-up time points. For the usual care group, the mean outcome at baseline is estimated by the b_0 coefficient, the mean outcome at Week 12 is estimated by b_0+b_2, and the mean outcome at Week 24 is estimated by b_0+b_3. For the ENABLE Cornerstone group, the mean outcome at baseline is estimated by b_0+b_1, the mean outcome at Week 12 is estimated by b_0+b_1+b_2+b_4, and the mean outcome at Week 24 is estimated by b_0+b_1+b_3+b_5. The overall treatment effect, ?, will be computed as the average between-group difference in change from baseline (averaged over Week 12 and Week 24) estimated with a linear contrast, ? =(1\⁄2)?(b_4+b_5). From the perspective of significance testing, the test for the interaction effect, a 2-degree of freedom test, is the test of difference in change from baseline between the study groups (since the groups are randomized and the mean outcome at baseline should be similar). If necessary, the team will conduct covariate adjustment for baseline factors unbalanced between the groups or predictive of dropout.

For specific aim 3, this within-trial analysis will be conducted separately from the perspectives of healthcare payers and FCGs and patients following current recommendations. Because the ENABLE Cornerstone intervention potentially affects FCG and patient distress and QOL and, by consequence, potentially affects healthcare utilization and costs, upfront implementation costs will be weighed against potential savings in healthcare costs and effectiveness measured by improvements in quality of life adjusted years (QALY) over the follow-up period.