MODELING SCREENING AND PREVENTION OF ALZHEIMER'S DISEASE

Purpose: To build a model identifying candidates for screening and subsequent treatment with a therapy that delays the onset of Alzheimer's disease (AD). Background: AD is treated with symptomatic drugs, but no cure or effective preventive treatment currently exists. The invasiveness, cost, and reported specificity/sensitivity of screening for AD vary widely. Specificity and sensitivity depend on the time before onset (thus the screening age). Consensus is lacking on the definition and incidence of mild cognitive impairment (MCI) which may precede the onset of AD. Methods: A time-to-event simulation model was built using literature-based distributions for time to: onset of MCI, onset of AD, discontinuation of treatment, and death. Modeled persons for whom death would precede AD onset do not get AD. Persons with negative screens do not get treatment. Additional literature-based inputs include specificity and sensitivity of screening for AD, and the efficacy of a hypothetical therapy measured in percentage increase in time to disease onset. It’s assumed that the absolute delay of age of AD onset is a function of length of treatment, regardless of the age when treatment occurs. Output tables indexed by effectiveness and mean treatment length were generated for various starting ages. Outcomes for this analysis are Alzheimer's-free years (AFYs) and MCI-free years (MCI-FYs). The model also produces QALYs, person-years of treatment, adverse events, and false-positive screens. Probabilistic sensitivity analyses were performed to assess the variability of outcomes and are expressed as 95% confidence intervals. Results: At ages 60 and 70, 221 and 264 per 1000 unselected persons screened would be recommended for treatment. For a treatment resulting in a 50% increase in time to progression and initiated at age 60, an average 5 years of treatment adds 42.9 MCI-FYs (20.9-72.0) and 20.8 AFYs (5.0-41.8). Screening at age 70 with a mean 5 years of treatment adds 65.5 MCI-FYs (37.2-99.6) and 62.1 AFYs (32.7-98.3), while averaging 3 years of treatment reduces the benefits to 49.7 MCI-FYs (27.4-76.9) and 47.5 AFYs (23.3-77.6). To prevent one AD case at age 70, the average number needed to screen is about 135 and the NNT 35. Conclusions: Time-to-event modeling offers flexibility in examining uncertainty when there are many choices of clinical strategies as well as targeting screening and treatment to particular ages and subpopulations.