UC Irvine

When studying drug effects using observational data, time-related biases such as immortal time bias, time-window bias, and time lag bias may exist and result in spurious associations. Many of the recent studies investigating the effects of metformin on dementia risk were based on large health care administrative databases and might be subject to time-related biases. This systematic review aims to assess if time-related biases exist in previous studies investigating the association between metformin use and dementia risk among diabetes patients. The electronic databases of PubMed, Web of Science, and ProQuest were searched for the terms “Metformin” AND “dementia” OR “Alzheimer’s Disease” OR “cognitive decline” OR “cognitive impairment.” These databases were searched from inception through 03/10/2021. Only English language articles and human subjects research were eligible. In total, twelve retrospective cohort studies, two case-control studies, and two nested case-control studies were identified. Twelve studies reported a reduced risk of dementia associated with metformin use, two articles reported increased risk, and two articles indicated no significant association between metformin use and dementia risk. In these sixteen studies, immortal time bias existed in eleven articles, time lag bias was found in seven articles, time-window bias was identified in three articles, while only one article was not identified any time-related biases. Most previous studies investigating the association between metformin and dementia risk did not carefully considered time-related biases. Future observational studies may need to take these time- related biases into consideration.