Emulating a vaccine target trial under unmeasured confounding: single and double negative control methods
- 1Center for Communicable Disease Dynamics, Department of Epidemiology, Harvard T.H. Chan School of Public Health, Boston, MA, USA
Abstract
Estimating vaccine effectiveness (VE) for symptom-driven outcomes in observational settings remains challenging due to unmeasured differences in health-seeking behavior between vaccine groups. This challenge persists even with target trial emulation, which primarily addresses design-related biases. The test negative design, a popular alternative design for estimating VE, can reduce bias due to unmeasured confounding by health-seeking but often deviates from the design principles of trial emulation. Here, we emulate a target trial to estimate relative VE of COVID-19 bivalent booster in a cohort of Kaiser Permanente of Northern California members and provide evidence for residual unmeasured confounding through negative controls. We then propose two approaches for estimating VE that operate within the target trial framework but allow for unmeasured confounding. The first uses test-negative illness as a negative control outcome and corrects for bias under the assumption of equi-confounding on the multiplicative scale. The second relaxes equi-confounding but requires an additional (valid) negative control. The two approaches yield similar estimates of 1-year relative VE (11.1% vs. 15.9%) but are both qualitatively different from traditional methods (VE: -14.5%). These approaches illustrate the flexibility of the target trial framework which can retain sound design principles while accommodating alternative identification strategies.
Related
- Apparent Survival Benefit of Postoperative Radioactive Iodine in Lymph Node-Positive Papillary Thyroid Carcinoma: A Negative-Control Analysis for Unmeasured Confounding.
- Influenza vaccine protection and contact patterns among outpatients with influenza-like illness during the 2024-2025 season in China: a test-negative case-control study.
- Fibula–Tibia Distance and Revision for Instability After Primary Total Knee Arthroplasty: An Exploratory Case–Control Study Addressing Rotational Confounding
- Effectiveness of 20-valent pneumococcal conjugate vaccine against pneumococcal pneumonia hospitalizations among U.S. adults: Test-Negative Design
- ICONIC: An R Package for Integrating Instrumental Variable- and Negative-Control-Informed Causal Discovery and Diagnostics in Multiomic Studies
BibTeX
@misc{schaeffer2026emulating,
title = {{Emulating a vaccine target trial under unmeasured confounding: single and double negative control methods}},
author = {B. Schaeffer and T. Chervo and E. P. Elkin and W. P. Hanage and J. R. Nugent and I. N. Rubin and J. Shi and J. Skarbinski and C. B. Boyer},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.21.26366144},
url = {https://www.medrxiv.org/content/10.64898/2026.10.21.26366144v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Emulating a vaccine target trial under unmeasured confounding: single and double negative control methods
AU - B. Schaeffer
AU - T. Chervo
AU - E. P. Elkin
AU - W. P. Hanage
AU - J. R. Nugent
AU - I. N. Rubin
AU - J. Shi
AU - J. Skarbinski
AU - C. B. Boyer
PY - 2026
DA - 2026/10/21
PB - medRxiv
DO - 10.64898/2026.10.21.26366144
UR - https://www.medrxiv.org/content/10.64898/2026.10.21.26366144v1
AB - Estimating vaccine effectiveness (VE) for symptom-driven outcomes in observational settings remains challenging due to unmeasured differences in health-seeking behavior between vaccine groups. This challenge persists even with target trial emulation, which primarily addresses design-related biases. The test negative design, a popular alternative design for estimating VE, can reduce bias due to unmeasured confounding by health-seeking but often deviates from the design principles of trial emulation. Here, we emulate a target trial to estimate relative VE of COVID-19 bivalent booster in a cohort of Kaiser Permanente of Northern California members and provide evidence for residual unmeasured confounding through negative controls. We then propose two approaches for estimating VE that operate within the target trial framework but allow for unmeasured confounding. The first uses test-negative illness as a negative control outcome and corrects for bias under the assumption of equi-confounding on the multiplicative scale. The second relaxes equi-confounding but requires an additional (valid) negative control. The two approaches yield similar estimates of 1-year relative VE (11.1% vs. 15.9%) but are both qualitatively different from traditional methods (VE: -14.5%). These approaches illustrate the flexibility of the target trial framework which can retain sound design principles while accommodating alternative identification strategies.
ER -