June 2022
Volume 63, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2022
A Hierarchical Bayesian Change Points Method to Study the Long-Term Natural History of Diseases
Author Affiliations & Notes
  • Liangbo Linus Shen
    Ophthalmology, University of California San Francisco, San Francisco, California, United States
  • Lucian V Del Priore
    Ophthalmology, Yale School of Medicine, New Haven, Connecticut, United States
  • Joshua Warren
    School of Public Health, Yale University, New Haven, Connecticut, United States
  • Footnotes
    Commercial Relationships   Liangbo Shen Boehringer Ingelheim, Code C (Consultant/Contractor); Lucian Del Priore Boehringer Ingelheim, Code C (Consultant/Contractor); Joshua Warren None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2022, Vol.63, 315 – F0146. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Liangbo Linus Shen, Lucian V Del Priore, Joshua Warren; A Hierarchical Bayesian Change Points Method to Study the Long-Term Natural History of Diseases. Invest. Ophthalmol. Vis. Sci. 2022;63(7):315 – F0146.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : It is often impractical to follow patients over decades to study the long-term natural course of a chronic disease. Here, we developed a statistical method for investigating the long-term natural history of diseases using data from patients followed over short durations.

Methods : We manually delineated geographic atrophy (GA) lesions on 1607 fundus photographs of 318 eyes in the age-related eye disease study (AREDS). We developed Bayesian entry time realignment (BETR), a hierarchical Bayesian change points method (Fig. 1), which estimated a patient’s disease progression rate and disease duration at enrollment based on a hypothesized progression model. The best model was selected based on the lowest deviance information criterion (DIC). We tested BETR in 100 rounds of simulated trial data generated from GA data in the AREDS. We then applied BETR in the actual GA data from the AREDS.

Results : BETR identified the correct model in 100 out of 100 simulations when the true disease progression model was the first order, second order, or exponential model. The intraclass correlation coefficient (ICC) between the estimated and true disease duration and progression rate ranged from 0.73 to 0.93. Applying BETR in patients with GA in the AREDS showed that the second order model (DIC = -358) was better than the first order model (DIC = -6) and exponential model (DIC = 365). The results remained the same among 100 rounds of random subsamplings of 159 eyes. BETR reconstructed an approximately 30-year natural history of GA progression among individual eyes (Fig. 2).

Conclusions : BETR could identify the correct long-term disease progression model and estimate patient-level disease progression parameters with high accuracy, demonstrating BETR as a promising method to study long-term natural history of diseases. The square root of GA area enlarged linearly and at different rates among different eyes over approximately 30 years.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.

 

Overview of BETR. BETR can estimate the disease progression rate (β1i), baseline disease duration (δi), and measurement error (εij) for each patient based on a hypothesized progression model (f(t)). Then we can realign the entry times of the patients by a factor of δi to reconstruct a long-term natural history of a disease based on datasets with short follow-up durations (from the left to right figure).

Overview of BETR. BETR can estimate the disease progression rate (β1i), baseline disease duration (δi), and measurement error (εij) for each patient based on a hypothesized progression model (f(t)). Then we can realign the entry times of the patients by a factor of δi to reconstruct a long-term natural history of a disease based on datasets with short follow-up durations (from the left to right figure).

 

Application of BETR in patients with GA from the AREDS.

Application of BETR in patients with GA from the AREDS.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×