Service capacity and amyloid-related imaging abnormality burden of antiamyloid therapy for early Alzheimer’s disease: a parameterised modelling study of a one-million-person
TL;DR
Treated patient numbers are modest relative to the assessment volume required to identify them, and the binding constraints are diagnostic throughput and MRI capacity rather than the treatment decision itself, so investment in referral pathways and imaging capacity is likely to determine delivery more than further refinement of triage assays.
Abstract
Background Anti-amyloid monoclonal antibodies are the first therapies to modify the biology of Alzheimer’s disease, but their delivery requires biomarker confirmation, repeated intravenous administration and scheduled magnetic resonance imaging (MRI) surveillance. The service capacity implied by adoption at population scale has not been widely quantified for a defined catchment. Objective To estimate, for a catchment of one million people, the annual number of patients initiating anti-amyloid therapy, the associated MRI and infusion workload, and the expected number of amyloid-related imaging abnormality (ARIA) events, and to identify which parameters dominate that estimate. Methods A deterministic cascade model was constructed representing referral, specialist assessment, plasma biomarker triage, confirmatory testing, clinical eligibility screening and treatment initiation, with resource use and genotype-stratified ARIA risk applied to the treated population. Input parameters were drawn from published cohort and trial data where available and declared as assumptions where not. Two scenarios were evaluated: all clinically eligible patients treated, and treatment withheld from APOE e4 homozygotes. Uncertainty was quantified by Monte Carlo simulation with 20 000 draws from triangular distributions, and parameter influence by one-way sensitivity analysis. Face validity was assessed against a published memory-clinic eligibility cohort. Results In the base case, 3 200 referrals produced 1 760 candidate assessments per year, of which 299 were amyloid-positive and 114 (6.5 percent of those assessed) met clinical eligibility; 97 were treated when APOE e4 homozygotes were excluded. Approximately 15 patients required assessment for each patient treated. Plasma triage generated 445 positive results with a positive predictive value of 0.61 and 352 confirmatory tests. Treatment generated 568 MRI examinations and 2 956 infusion visits in the first year, with 3.1 symptomatic ARIA-E and 19.3 ARIA-H events expected annually, falling to 1.5 and 12.7 respectively under APOE exclusion. Monte Carlo intervals were wide (treated patients 62 to 228). Referral rate, amyloid positivity among those assessed, and clinical eligibility dominated the variance; plasma assay accuracy did not affect the number treated. The modelled eligibility fraction of 6.5 percent was close to the 6.3 percent reported in an independent memory-clinic cohort. Conclusion Treated patient numbers are modest relative to the assessment volume required to identify them, and the binding constraints are diagnostic throughput and MRI capacity rather than the treatment decision itself. Because plasma assay accuracy influenced confirmatory testing volume but not the number treated, investment in referral pathways and imaging capacity is likely to determine delivery more than further refinement of triage assays.