Systematic review of strategies to justify validity of instrumental variables in observational studies of dementia and neurodegenerative disease

Abstract

AbstractBackgroundInstrumental variable (IV) analysis is a methodology used to study causal effects of treatments or exposures. If properly implemented, causal effects can be estimated even in the presence of unmeasured confounding. . Central to IV analysis is the identification of a valid instrument that satisfies three assumptions – one is empirically verifiable (Restriction) but the remainder require substantive knowledge and convincing argumentation (Exclusion restriction and Exchangeability). The objective of this study is to systematically review the strategies used to validate an IV in observational studies of dementia and neurodegenerative disease. This systematic review is pertinent given the increasing use of real‐world evidence to guide public health and regulatory decision‐making in lieu of randomised controlled trials.MethodsWe conducted our search using PubMed, PsycINFO, and Web of Science. We included published observational studies investigating the causal effect of some risk factor, exposure, or treatment using IV analysis and using either a sample of: i) adults without diagnosis of dementia or neurodegenerative disease and an outcome of incident diagnosis of dementia or neurodegenerative disease; or ii) adult patients diagnosed with dementia, major neurocognitive disorder, or neurodegenerative disease and a clinical outcome related to incident dementia, mortality, hospitalisation, institutionalisation, serious adverse outcomes, medication adherence, cognition, neuropsychiatric symptoms, and biomarkers of neurodegeneration. We will extract and categorise text information directly arguing for the assumptions of a valid IV and the type(s) of falsification tests used.Results927 unique articles were retrieved. Screening for eligibility and data extraction by two study authors are currently underway. Results will be presented at the conference.ConclusionsReviewing how variables have been reasoned to be valid IVs may help improve our understanding of the causal effects of treatments or exposures on dementia and neurodegenerative disease which may have relevance in shaping future therapeutic and management approaches.