(C29) Network Analysis of Anticholinergic Medication Use in a Cohort of People Aging with HIV: Identification of Anticholinergic Medication Drivers of Physical Frailty States

Publication Number: 973

Background: The concurrent use of medications with anticholinergic properties has been shown to exacerbate frailty. While traditional statistical methods have been employed to investigate the relationship between anticholinergic burden and frailty, the application of network science has been limited.

Objectives: This study aims to identify potential anticholinergic medication drivers of frailty states in people living with HIV (PLWH) in Canada by using network analysis.

Methods: Data were obtained from the baseline visit of 824 participants enrolled in the Positive Brain Health Now (+BHN) study, a prospective study of PLWH aged ≥ 35 years recruited between 2014 and 2016 across Canada. Medications with anticholinergic activity were identified using the Anticholinergic Cognitive Burden (ACB) scale. Physical frailty was determined using a modified Fried frailty phenotype (FP). Undirected anticholinergic co-medication networks were generated, with each node representing an anticholinergic and an edge linking two nodes representing the concomitant use of the corresponding anticholinergics by the same participant. The network was visualized, and network parameters were analyzed using NodeXL. Network comparison was conducted by comparing graph properties and community patterns between frailty states and any significant increase in the importance of specific anticholinergics in the frail state estimated by the Neighbor Shift Score and ∆Betweenness using the NetShift web app.

Results: The prevalence of frailty states was: frail (15%), prefrail (45%), and robust (40%). The global anticholinergic network showed a sparse network of 30 nodes and 85 unique edges. Quetiapine-bupropion combination was the most frequent co-medication (6%). Based on centrality metrics, the top five most “influential” medications were amitriptyline, trazodone, bupropion, quetiapine and atenolol. The network comparisons between frailty states identified the critical anticholinergic drivers as Frail vs Robust - trazodone, loperamide, amitriptyline, and codeine; Prefrail vs Robust - morphine, amitriptyline, loperamide, risperidone, trazodone and bupropion; Frail vs Prefrail - codeine, trazodone, loperamide, quetiapine, atropine, morphine.

Conclusions: Our results show that the critical node drivers between frailty states in the anticholinergic network were mostly antidepressants and opioids. However, network analysis is more suited for hypothesis generation. It would need to be complemented with traditional pharmacoepidemiologic methods to control for confounding and selection bias to confirm the co-medication patterns observed. Future network analysis will need to investigate if these findings are driven by comorbidities or medications using bipartite networks.