A synthetic control arm from observational data to estimate the background incidence rate of an adverse event in patients with Alzheimer’s disease matched to a clinical trial population

Abstract

AbstractBackgroundIdentifying a potential safety issue in Alzheimer’s disease (AD) clinical trials is challenging due to the blinded nature of the studies and the number of adverse events (AEs) that occur in this patient population independent of the investigational treatment. An accurate estimate of the incidence rate of an AE in nontreated individuals with AD may help to identify unexpected safety findings.MethodThe objective of this retrospective case‐control study was to generate a synthetic control arm from observational data to estimate the background incidence rate of one AE (pneumonia) in individuals with AD. We used high‐dimensional propensity scoring to match individuals aged 50 to 89 years with AD with to controls from 01/01/2014 to 31/12/2018 in PharMetrics, a US claims database. The inclusion and exclusion criteria were matched to a Phase III clinical trial in AD evaluating crenezumab. Additional inclusion criteria included at least two diagnoses of AD using ICD‐9/ICD‐10 in 90 days and ≥90 days of continuous enrollment prior to the index date. The main outcome was the 5‐year incidence rate of pneumonia in patients with AD. A zero‐inflated Poisson model was designed to predict the expected count of pneumonia as a function of AD diagnosis, propensity score, patient age at index date, sex, disease history, and comorbidity using the Charlson Comorbidity Index.ResultsIn our preliminary analyses, 31,107 individuals with AD were identified from the database; 19,457 (63%) were female; mean age was 75.7 (SD, 8.3) years. Within 5 years of the reported date of AD diagnosis, 4,692 patients were diagnosed with pneumonia. The incidence rate (95% CI) of pneumonia in individuals with AD was 113.07 (109.84‐116.31) per 1,000 patient‐years. Increasing age (incident rate ratio [IRR], 1.06), being male (IRR, 1.5), and patient comorbidities (IRR, 1.25) were associated with an increased risk of pneumonia in individuals with AD.ConclusionA synthetic control arm from observational data may help to estimate the background incidence rate of one AE, pneumonia, in individuals with AD and can help to identify a potential safety issue in a blinded clinical trial.