Breast Cancer Drug Analysis Flags Dementia Signals for Anastrozole in Women Over 50

A sweeping analysis of more than two decades of spontaneous adverse event reports submitted to the US Food and Drug Administration has found no overall excess of dementia reports among women with breast cancer taking endocrine and targeted therapy drugs. But the study, published in Frontiers in Pharmacology, identified a more complicated picture beneath that neutral headline: within specific subgroups of patients, the analysis detected positive safety signals for narrow-spectrum dementia events, most notably for the aromatase inhibitor anastrozole used as monotherapy in women over 50 years of age. The authors, a team from Shanghai University of Traditional Chinese Medicine and Shanghai Jiao Tong University School of Medicine led by Hongfeng Chen, emphasize that these findings are hypothesis-generating rather than proof of causation, and that they require confirmation in prospective studies.

Hormone receptor-positive breast cancer accounts for roughly two-thirds of breast cancer cases, and systemic treatment for this subtype is anchored by endocrine and targeted therapy strategies. In early-stage disease, adjuvant endocrine therapy reduces the ten-year recurrence risk by an estimated 30 to 50 percent and improves overall survival by 15 to 20 percent, while in metastatic disease endocrine agents are frequently combined with targeted drugs such as CDK4/6 inhibitors to extend progression-free survival. Because these regimens are typically administered for five to ten years, any long-term neurological consequence of treatment would carry substantial public health weight. Prior research on cognition during endocrine therapy has focused mainly on older drugs, suggesting that aromatase inhibitors may impair memory and concentration by suppressing estrogen synthesis, while the selective estrogen receptor modulator tamoxifen might exert partial protective effects. The cognitive safety of newer targeted agents and of combination strategies has remained largely undefined.

To address that gap, the researchers mined the FDA Adverse Event Reporting System, or FAERS, a public database of safety reports submitted by patients, healthcare professionals and pharmaceutical companies. They retrieved all records from January 2010 through June 2025, removed duplicates, and restricted their analytical population to women aged 18 or older with a confirmed breast cancer diagnosis. From roughly 21 million available cases, filtering identified 190,579 cases in adult female breast cancer patients, of which 32,968 involved endocrine and targeted therapy drugs. After excluding cases with potential confounders of cognitive function, such as brain metastases, central nervous system disease progression, or concomitant use of benzodiazepines, anticholinergics or chemotherapy agents, 32,746 cases remained for analysis. Within that cohort, 247 cases involved narrow-spectrum dementia events and 8,084 involved broad-spectrum events, with the study drugs listed as the suspected cause in 56 and 1,430 cases respectively.

The analysis used a case/non-case disproportionality design, the standard framework of modern pharmacovigilance. Rather than measuring incidence, it compares the frequency of a target adverse event among reports mentioning a given drug against its frequency among reports mentioning all other drugs. The team applied two complementary metrics: the proportional reporting ratio with a chi-square statistic, and the Empirical Bayes Geometric Mean calculated by the Multi-Item Gamma Poisson Shrinker, which applies Bayesian shrinkage to pull extreme estimates toward the null when case counts are small, thereby reducing false positives. A positive signal required meeting both thresholds simultaneously: a lower 95 percent confidence bound of the EBGM of at least 2.0 and a PRR above 2 with chi-square above 4. Dementia events were systematically identified using standardized MedDRA queries, which group related preferred terms into narrow and broad categories to limit coding-related misclassification.

Across the entire population, the results were reassuring. For narrow-spectrum dementia, the EBGM05 was 1.27 with a PRR of 1.32, and for broad-spectrum dementia it was 1.02 with a PRR of 1.03, both well below the signal thresholds. In other words, women with breast cancer receiving endocrine and targeted therapies as a whole were not over-represented in dementia reports. However, the picture changed when the data were stratified by treatment strategy. Among monotherapies, anastrozole monotherapy produced a positive narrow-dementia signal with 20 reported cases, an EBGM05 of 4.00 and a PRR of 4.68, while exemestane monotherapy, another aromatase inhibitor, also crossed both thresholds with nine cases, an EBGM05 of 2.10 and a PRR of 2.64. Both signals were concentrated in female adult breast cancer cases.

Age stratification sharpened the pattern further. In the overall age-group analysis, none of the subgroups reached the PRR threshold, although the over-65 group for narrow dementia approached it with a PRR of 1.93. But in a targeted analysis of anastrozole monotherapy among reporters over 50 years old, the narrow-dementia signal remained positive with 20 cases, an EBGM05 of 2.52 and a PRR of 3.67, meeting both pre-specified criteria. Two exploratory signals, meeting the PRR criterion but not the EBGM05 threshold, emerged for combination strategies: aromatase inhibitor plus CDK4/6 inhibitor for narrow dementia, and the triple combination of aromatase inhibitor, gonadotropin-releasing hormone agonist and CDK4/6 inhibitor for broad dementia in patients aged 18 to 50. Notably, sensitivity analyses that pooled all three clinically available CDK4/6 inhibitors, including palbociclib and ribociclib, abolished the combination signals entirely, indicating that the disproportionality was specific to abemaciclib-containing regimens rather than a class-wide effect. Abemaciclib is the member of the class with documented higher blood-brain barrier penetrance.

A second sensitivity analysis tested the robustness of the broad-spectrum findings by excluding 29 non-specific preferred terms related to sleep, psychotic and mood symptoms and retaining 31 cognition-specific terms. This refinement cut total broad-dementia cases by 42.6 percent, from 8,084 to 4,640, and completely eliminated the exploratory signal for the triple combination, which fell to an EBGM05 of 0.63. The authors interpret this as evidence that much of the broad-category case burden reflects multifactorial neuropsychiatric symptoms, including psychological distress and polypharmacy in patients with higher disease burden, rather than dementia itself. One novel exploratory signal did appear after refinement, for a quadruple combination of aromatase inhibitor, CDK4/6 inhibitor, GnRHa and SERM, but with only seven cases the authors treat it as preliminary and in need of validation.

The temporal profile of the reported events offers a clue to monitoring. The median time to onset was 120 days for narrow-spectrum cases and 77 days for broad-spectrum cases, and most narrow events occurred within 179 days of therapy initiation, while most broad events occurred within 539 days. Demographically, patients with narrow-spectrum cases were overwhelmingly older, with 83.9 percent aged over 65, and the narrow cases were more likely to be reported by healthcare professionals and to have non-fatal outcomes. The authors suggest that for women over 50 on anastrozole monotherapy, cognitive screening with tools such as the Montreal Cognitive Assessment within one to six months of starting treatment could be considered, while follow-up for younger patients on GnRHa-containing combinations might extend to 18 months, though they stress that clinical decisions should be individualized and that the observed signals are not contraindications.

Mechanistically, the team points to estrogen depletion as the most plausible explanation for the anastrozole signal. Aromatase inhibitors block peripheral estrogen synthesis, and in postmenopausal women, whose brains already depend on low but critical levels of estrogen for synaptic plasticity and neuroprotection, this suppression may compound an existing hormonal deficit. In younger patients, drug-induced peripheral estrogen suppression combined with GnRHa-mediated artificial menopause could create a compounded deficit, offering a framework for the exploratory signal in that group. For abemaciclib, preclinical evidence of cellular senescence induction in neuroblastoma models and the drug’s brain penetrance suggest a possible impact on neural homeostasis. The authors caution, however, that these are interpretive frameworks, not demonstrated mechanisms.

The study carries the inherent limitations of spontaneous reporting systems. FAERS lacks denominator data, so absolute incidence rates cannot be calculated, and its retrospective, voluntary nature precludes causal inference. Underreporting of mild cognitive events, notoriety bias, and reporting variations by geography and source all distort the data, and abemaciclib’s 2017 approval may have inflated its reporting in later years through the Weber effect. Residual confounding from unmeasured comorbidities and missing details such as dosage and disease stage further constrain the analysis. The authors note that an umbrella review has linked endocrine therapy to cognitive impairment affecting memory and executive function while finding overall dementia risk inconsistent across studies, which they say helps explain the neutral overall signal in their data. Their conclusion is measured: endocrine and targeted therapy for breast cancer shows no overall dementia safety signal in real-world reporting, but subgroup-specific findings for aromatase inhibitor monotherapy, particularly anastrozole in women over 50, deserve prospective investigation to determine their clinical relevance.

Subject of Research: Pharmacovigilance analysis of dementia-related risk associated with endocrine and targeted therapy in breast cancer

Article Title: Dementia-related risk associated with endocrine and targeted therapy strategies in breast cancer: a real-world pharmacovigilance analysis

Article References: Cui, R., Yin, Y., & Chen, H. (2026). Dementia-related risk associated with endocrine and targeted therapy strategies in breast cancer: a real-world pharmacovigilance analysis. Frontiers in Pharmacology, 17, Article 1855112. https://doi.org/10.3389/fphar.2026.1855112

Image Credits: AI Generated

DOI: 10.3389/fphar.2026.1855112

Keywords: breast cancer, dementia, pharmacovigilance, FAERS, anastrozole, aromatase inhibitors, endocrine therapy, CDK4/6 inhibitors, abemaciclib, adverse events, disproportionality analysis, cognitive decline