Respiratory diseases like asthma and chronic obstructive pulmonary disease (COPD) lowers quality of life and increases mortality. Much of this burden comes from exacerbations, acute episodes where patients experience worsening symptoms and reduced lung function. Preventing these events is a key goal of treatment, which is why exacerbations are often the primary outcome in clinical trials.
However, these exacerbations are relatively rare, making it expensive to run studies to determine treatment effect. Traditionally, clinical trials often last a year, with large numbers of participants. Researchers are now looking for ways to make better use of high-resolution lung function data to shorten trials while still keeping a strong connection to the clinically relevant exacerbation risk.
One major shift has been the increased use of home spirometry. These portable devices allow patients to track their lung function multiple times a day. This creates rich datasets that can capture day-to-day fluctuations in airflow and lung function. Past studies have found that certain patterns in these fluctuations, such as long-term variability in peak expiratory flow (PEF), are linked to a higher risk of exacerbations [1].
Another promising strategy is using surrogate endpoints: events that are strongly correlated with exacerbations but occur more frequently. A notable example is the CompEx endpoint [2], which combines changes in PEF, use of alleviating inhalers, and symptoms. Clinical trials using CompEx have shown that it can dramatically reduce the trial duration and number of participants needed to evaluate new treatments.
A new approach builds on this idea but takes a more individualized approach. Ludvig Jakobsson is an industrial PhD student at the Fraunhofer-Chalmers Research Centre for Industrial Mathematics in Gothenburg, in a collaboration with AstraZeneca and the Department of Mathematical Sciences at Chalmers University of Technology and University of Gothenburg. Instead of relying on global thresholds for all patients, his approach uses statistical modelling to define events at the patient level. Specifically, PEF is modelled using a discrete-time mixed-effects hidden Markov model (MHMM). This method captures both population-level trends and individual variability, while estimating each subject’s underlying disease state over time.
Figure 1: Schematic view of a hidden Markov model.
Figure 2: Estimated latent states for a simulated lung function time series.
The MHMM approach builds on earlier work in statistical modelling of time series in medicine [3] and uses inference techniques like the SAEM algorithm to estimate both individual and population parameters. By modelling lung function this way at the individual level, the hope is to create trial designs that are shorter, more efficient, but still tightly linked to exacerbation risk and thus patient lung health.
References
[1] Jacob Leander, Mats Jirstrand, Ulf G. Eriksson, and Robert Palmér. A stochastic mixed effects model to assess treatment effects and fluctuations in home-measured peak expiratory flow and the association with exacerbation risk in asthma. CPT: Pharmacometrics & Systems Pharmacology, 11(2):212–224, December 2021
[2] Anne L Fuhlbrigge, Thomas Bengtsson, Stefan Peterson, Alexandra Jauhiainen, Göran Eriksson, Carla A Da Silva, Anthony Johnson, Tariq Sethi, Nicholas Locantore, Ruth Tal-Singer, and Malin Fagerås. A novel endpoint for exacerbations in asthma to accelerate clinical development: a post-hoc analysis of randomised controlled trials. The Lancet Respiratory Medicine, 5(7):577–590, July 2017.
[3] Maud Delattre and Marc Lavielle. Maximum likelihood estimation in discrete mixed hidden Markov models using the SAEM algorithm. Computational Statistics & Data Analysis, 56(6):2073–2085, June 2012.