Identifying and triangulating all available evidence on the effect of blood lipids and statins on dementia outcomes

Background Much published literature exists on the effect of blood lipid levels, and lipid‐regulating agents (LRA) such as statins, on the risk of dementia and related outcomes. Several approaches, including observational studies, randomised controlled trials (RCTs) and Mendelian randomisation (MR) studies (which uses genetic variants as proxies for the risk factor of interest) have been used to try and address this question, each with distinct assumptions and limitations. However, no previous review has sought to systematically identify and integrate these distinct study designs as part of a single synthesis exercise. This comprehensive systematic review and qualitative triangulation of the evidence aims to address this. Method Following standard systematic review methods, 16109 potentially eligible records were identified by a systematic search and screened as part of the review. Of these, 127 records reporting 105 primary studies were included, the majority of which (n=92; ∼88%) were observational studies. Data were extracted using a piloted form, and the risk of bias in included studies was assessed using the appropriate domain‐based tool (e.g. ROBINS‐I for non‐randomised studies of interventions). Result Distinct study designs were only available for some exposure (lipids/statins) and outcome (all‐cause dementia, Alzheimer’s disease, vascular dementia) pairs. A common theme across study designs was the absence of vascular dementia outcomes. A key example of the triangulation approach if the results for the effect of statins on all‐cause dementia. Evidence from the meta‐analysis of observational studies found a protective effect (HR: 0.78, 95%CI: 0.69‐0.89), in contrast to both the RCT (OR:1, 95%CI: 0.60‐1.65) and MR (RR:0.9, 95%CI: 0.29‐2.83) approaches. This discrepancy could potentially be explained by immortal time bias, which typically increases the observed protective effect of the intervention, as several observational studies were found to suffer from this bias. Conclusion Evidence on the effect of blood lipid levels, and lipid modifying agents, is available from study with distinct methodological designs. By identifying and integrating this evidence as part of a single synthesis exercise, the strength of the conclusions can be increased. Future research should specifically focus on vascular dementia as an outcome, as there is a lack of evidence on this condition.