BACKGROUND: Several observational studies have reported an association between drugs with strong anticholinergic (AC) properties and a higher risk of dementia, especially for antidepressants (AD). However, these studies did not generate measures of absolute risk under complete adherence to these treatment strategies.
OBJECTIVES: To estimate the per-protocol effect of adhering to the following treatment strategies on 12-year risk of dementia: (1) initiation of AD with AC properties (“AC AD”) and continuation for ≥4 months, until the development of a contraindication, or (2) initiation of AD without AC properties (“non-AC AD”) and continuation for ≥4 months, until the development of an indication, or (3) no initiation of AD over follow-up.
METHODS: We used CPRD GOLD electronic health records of UK adults aged ≥50 years without a diagnosis for dementia or cognitive impairment and with no use of antidepressants in the last 12 months to emulate a sequence of 204 nested target trials, beginning in each of the months between 2001-2017. In the baseline month of each emulated trial, we classified individuals into one of three strategies: “AC AD” initiators, “non-AC AD” initiators, and non-initiators. Individuals were censored if and when they deviated from their assigned treatment strategy. We used pooled logistic regression to estimate the per-protocol effect of these strategies on incident dementia via 12-year risk differences (RD) and risk ratios (RR), adjusting for baseline and post-baseline variables using inverse-probability weighting.
RESULTS: Eligible individuals included 29,408 initiators of AC AD (80,329 person-years of follow-up), 19,753 initiators of non-AC AD (58,728 person-years of follow-up), and a selected random sample of 1,114,975 non-initiators of AD (5,608,493 person-years of follow-up). The estimated 12-year risk (95% CI) of dementia was 5.7% (4.4%; 7.5%) under the “AC AD” strategy, 6.2% (4.3%; 7.5%) under the “non-AC AD” strategy, and 5.5% (5.0%; 6.0%) under the “no AD” strategy. The estimated RD (“AC AD” vs “non-AC AD”) was -0.5% (-2.2%;1.6%) and the corresponding RR was 0.92 (0.69; 1.35). The estimated RD (“AC AD” vs “no AD”) was 0.3% (-0.9%; 1.3%) and the corresponding RR was 1.05 (0.85; 1.23). The estimated RD (non-AC AD vs. “no AD”) was 0.8% (-1.0%; 2.1%) and the corresponding RR was 1.14 (0.82; 1.37).
CONCLUSIONS: Our results suggest that the AC properties of ADs do not increase the risk of dementia, with 95% CI that are compatible with both a decrease and an increase in the 12-year risk of about 2%. Similarly, the use of AD, regardless of their AC properties yields estimates compatible with both a 1% decrease and a 2% increase in the 12-year risk of dementia.
