Identifying adverse event clusters and treatment effects in randomised trials using latent class analysis: an application in the COV-BOOST trial
Abstract
Background: Randomised controlled trials (RCTs) are essential for evaluating treatment efficacy and safety, particularly in vaccine trials, where adverse events are systematically collected and assessed. The high volume of reported adverse events often presents challenges in discerning meaningful associations between treatments and adverse events. Traditional analytical approaches rely on multiple pairwise treatment comparisons of individual adverse events, which can be difficult to interpret. This study explores the use of latent class analysis (LCA) to identify adverse event profiles and estimates the average treatment effects (ATE) for these in an RCT setting.
Methods: Data from the COV-BOOST trial (ISRCTN registration 73765130), which assessed the safety and efficacy of seven COVID-19 booster vaccines, was used. This trial monitored a predefined set of 17 adverse events, graded by severity, over a 7-day period following immunisation. LCA was applied to classify participants into adverse event subgroups based on shared event profiles. After identifying these classes, the ATE was estimated for each identified adverse event profile to quantify the causal impact of different vaccine boosters. This was achieved by regressing treatment on the latent adverse event class membership outcome.
Results: Fatigue, pain, and headache were commonly reported adverse events across all vaccine arms. LCA identified three distinct adverse event profiles among participants: (class 1) minimal AEs, (class 2) mild AEs, and (class 3) severe AEs reports. Four vaccines, ChAd after BNT/BNT primer, Ad26, CVn, and MOD after both primer vaccines, were associated with higher reactogenicity, with most recipients falling into class 2 or 3.
Conclusion: This study demonstrates the utility of LCA in identifying adverse event profiles within an RCT and uniquely integrates the latent adverse event class membership of participants as an outcome for ATE estimation. By examining the varying effects of booster vaccines in the COV-BOOST trial, this method advances the understanding of these vaccines’ safety. The successful application of this innovative approach in this study could pave the way for its adoption by researchers in future clinical trials across various medical domains. More broadly, integrating LCA with ATE estimation provides a novel framework for enhancing vaccine pharmacovigilance and improving safety assessments for medical interventions.
Related
- Identifying Lyapunov functionals with a PINN : application to a hyperbolic transport equation
- Mutual Information as a Tool for Optimal Classification: Application to Identifying Rapid-Responding Behaviour
- Transporting summary measures of relative effects from randomised trials to the treated patient population: an application to breast cancer endocrine therapy
- Estimating heterogeneous treatment effects from randomised trials: a comparison of the risk modelling and treatment effect modelling approaches
- Identifying and Measuring Treatment-Effect Heterogeneity across Sites through Within-Site Randomization
BibTeX
@article{khoueiry2026identifying,
title = {{Identifying adverse event clusters and treatment effects in randomised trials using latent class analysis: an application in the COV-BOOST trial}},
author = {Lea Maria Khoueiry and Eleanor Van Vogt and Leila Janani and Victoria Cornelius and Saul N. Faust and Alasdair P. S. Munro and Xinxue Liu and Suzie Cro and the COV-BOOST study group and Katrina Cathie and David Baxter and Marcin Bula and Krishna Chatterjee and Anna Goodman and Christopher Green and Alexander Hicks and Vincenzo Libri and Sunil Sharma and Alastair McGregor and Angela Minassian and Patrick Moore and Orod Osanlou and Adrian Palfreeman and Stephan Saich and Dinesh Saralaya and Ray Sheridan and Matthew Stokes and Emma Thomson and Chris Twelves},
journal = {Trials},
year = {2026},
month = oct,
doi = {10.1186/s13063-026-10058-1},
url = {https://link.springer.com/article/10.1186/s13063-026-10058-1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Identifying adverse event clusters and treatment effects in randomised trials using latent class analysis: an application in the COV-BOOST trial
AU - Lea Maria Khoueiry
AU - Eleanor Van Vogt
AU - Leila Janani
AU - Victoria Cornelius
AU - Saul N. Faust
AU - Alasdair P. S. Munro
AU - Xinxue Liu
AU - Suzie Cro
AU - the COV-BOOST study group
AU - Katrina Cathie
AU - David Baxter
AU - Marcin Bula
AU - Krishna Chatterjee
AU - Anna Goodman
AU - Christopher Green
AU - Alexander Hicks
AU - Vincenzo Libri
AU - Sunil Sharma
AU - Alastair McGregor
AU - Angela Minassian
AU - Patrick Moore
AU - Orod Osanlou
AU - Adrian Palfreeman
AU - Stephan Saich
AU - Dinesh Saralaya
AU - Ray Sheridan
AU - Matthew Stokes
AU - Emma Thomson
AU - Chris Twelves
PY - 2026
DA - 2026/10/28
JO - Trials
DO - 10.1186/s13063-026-10058-1
UR - https://link.springer.com/article/10.1186/s13063-026-10058-1
AB - Background: Randomised controlled trials (RCTs) are essential for evaluating treatment efficacy and safety, particularly in vaccine trials, where adverse events are systematically collected and assessed. The high volume of reported adverse events often presents challenges in discerning meaningful associations between treatments and adverse events. Traditional analytical approaches rely on multiple pairwise treatment comparisons of individual adverse events, which can be difficult to interpret. This study explores the use of latent class analysis (LCA) to identify adverse event profiles and estimates the average treatment effects (ATE) for these in an RCT setting. Methods: Data from the COV-BOOST trial (ISRCTN registration 73765130), which assessed the safety and efficacy of seven COVID-19 booster vaccines, was used. This trial monitored a predefined set of 17 adverse events, graded by severity, over a 7-day period following immunisation. LCA was applied to classify participants into adverse event subgroups based on shared event profiles. After identifying these classes, the ATE was estimated for each identified adverse event profile to quantify the causal impact of different vaccine boosters. This was achieved by regressing treatment on the latent adverse event class membership outcome. Results: Fatigue, pain, and headache were commonly reported adverse events across all vaccine arms. LCA identified three distinct adverse event profiles among participants: (class 1) minimal AEs, (class 2) mild AEs, and (class 3) severe AEs reports. Four vaccines, ChAd after BNT/BNT primer, Ad26, CVn, and MOD after both primer vaccines, were associated with higher reactogenicity, with most recipients falling into class 2 or 3. Conclusion: This study demonstrates the utility of LCA in identifying adverse event profiles within an RCT and uniquely integrates the latent adverse event class membership of participants as an outcome for ATE estimation. By examining the varying effects of booster vaccines in the COV-BOOST trial, this method advances the understanding of these vaccines’ safety. The successful application of this innovative approach in this study could pave the way for its adoption by researchers in future clinical trials across various medical domains. More broadly, integrating LCA with ATE estimation provides a novel framework for enhancing vaccine pharmacovigilance and improving safety assessments for medical interventions.
SN - 1745-6215
ER -