Novel causal inference methods to inform clinical decision on when to discontinue symptomatic treatment for patients with dementia

PROJECT SUMMARY/ABSTRACT Appropriate use of acetylcholinesterase (AChEIs) and memantine can meaningfully improve the health outcomes and quality of life among people with Alzheimer’s disease-related dementia (ADRD). Deprescribing of these symptomatic medications can help mitigate medication burden and associated adverse events in this

population, particularly given the high level of multimorbidity and pill burden. However, no current US guideline exists on deprescribing of these medications in ADRD. Existing non-US guideline recommendations are largely consensus-based and should be strengthened through higher levels of evidence. Two pivotal questions need

to be answered first: 1) what is the long-term effect of symptomatic dementia medications? and 2) when is suitable to discontinue these medications? Ideally, answers to these questions would come from randomized controlled trials, but conducting trials evaluating multiple treatment duration or discontinuation strategies

simultaneously with large enough sample sizes in each arm would be cost-prohibitive. Observational data from dementia medication use in the real-world setting provides a unique opportunity. However, treatment duration or discontinuation strategies necessarily involve interventions on time-varying treatment decisions. Evaluating

the time-varying medication use on health and patient-centered outcomes must appropriately control for complex time-varying confounding that renders conventional regression invalid. Novel causal inference methods, including Robins’ g-formula and a three-step weighting approach (cloning, censoring, weighting) can

appropriately account for such time-varying confounding and generate estimates of absolute risks while preventing immortal time bias. By emulating the valid analyses of trials, causal analyses of observational data are also cost-efficient and have greater generalizability. Using data collected in a large survey linked with

electronic health databases, we will characterize the utilization pattern of symptomatic dementia medications and examine factors that influenced treatment discontinuation (Aim 1). We will then use novel causal inference methods to estimate the long-term effect of continuous treatment (Aim 2), and to evaluate different treatment

discontinuation strategies (Aim 3) with regard to incidence of clinical and patient-centered outcomes and health service utilization. We will use data from the Health and Retirement Study (HRS)-Medicare linked dataset. The nationally representative, longitudinal, NIA-funded HRS survey provides validated measures on cognitive

impairment and dementia. The linkage to Medicare provides extensive information on medication, clinical characteristics, and health care utilization. The expected outcome of this study is an understanding of the effects of long-term use of dementia medications and the impact of different treatment discontinuation

strategies on outcomes. The findings of this study will provide a scientific basis for the development of evidence-based guidelines and the planning of clinical trials in the deprescribing of symptomatic dementia medications in people with ADRD to improve their care.