Estimating additive interaction in 2-stage individual participant data meta-analysis

Maartje Basten; Lonneke A. Van Tuijl; Kuan Yu Pan; Adriaan W. Hoogendoorn; Femke Lamers; Adelita V. Ranchor; Joost Dekker; Philipp Frank; Henrike Galenkamp; Mirjam J. Knol; Nolwenn Noisel; Yves Payette; Erik R. Sund; Aeilko H. Zwinderman; Lützen Portengen; Mirjam I. Geerlings

doi:10.1093/aje/kwae325

Estimating additive interaction in 2-stage individual participant data meta-analysis

Maartje Basten^*
, Lonneke A. Van Tuijl
, Kuan Yu Pan
, Adriaan W. Hoogendoorn
, Femke Lamers
, Adelita V. Ranchor
, Joost Dekker
, Philipp Frank
, Henrike Galenkamp
, Mirjam J. Knol
, Nolwenn Noisel
, Yves Payette
, Erik R. Sund
, Aeilko H. Zwinderman
, Lützen Portengen
, Mirjam I. Geerlings

^*Corresponding author for this work

Vrije Universiteit Amsterdam
Amsterdam Public Health
University Medical Center Utrecht
University of Groningen
Karolinska Institutet
University College London
University of Amsterdam
National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands.
CHU Sainte-Justine
Nord University
Norwegian University of Science and Technology
Nord-Trøndelag Hospital Trust
Amsterdam Neuroscience
Amsterdam University Medical Center

Research output: Contribution to journal › Article › Academic › peer-review

Abstract

Individual participant data (IPD) meta-analysis provides important opportunities to study interaction and effect modification for which individual studies often lack power. While previous meta-analyses have commonly focused on multiplicative interaction, additive interaction holds greater relevance for public health and may in certain contexts better reflect biological interaction. Methodological literature on interaction in IPD meta-analysis does not cover additive interaction for models including binary or time-to-event outcomes. We aimed to describe how the Relative Excess Risk due to Interaction (RERI) and other measures of additive interaction or effect modification can be validly estimated within 2-stage IPD meta-analysis. First, we explain why direct pooling of study-level RERI estimates may lead to invalid results. Next, we propose a 3-step procedure to estimate additive interaction: (1) estimate effects of both exposures and their product term on the outcome within each individual study; (2) pool study-specific estimates using multivariate meta-analysis; (3) estimate an overall RERI and 95% confidence interval based on the pooled effect estimates. We illustrate this procedure by investigating interaction between depression and smoking and risk of smoking-related cancers using data from the PSYchosocial factors and Cancer (PSY-CA) consortium. We discuss implications of this procedure, including the application in meta-analysis based on published data.

Original language	English
Pages (from-to)	1661-1667
Number of pages	7
Journal	American Journal of Epidemiology
Volume	194
Issue number	6
DOIs	https://doi.org/10.1093/aje/kwae325
Publication status	Published - 1 Jun 2025

Bibliographical note

Publisher Copyright:
© 2024 The Author(s).

Funding

This work was supported by the KWF Dutch Cancer Society (VU2017-8288). ELSA is funded by the National Institute on Aging (R01AG017644) and the National Institute for Health and Care Research (198/1074-02). The infrastructure of the NESDA study (www.nesda.nl) is funded through the Geestkracht program of the Netherlands Organization for Health Research and Development (ZonMw; Grant No. 10-000-1002) and financial contributions by participating universities and mental health care organizations (VU University Medical Center, GGZ inGeest, Leiden University Medical Center, Leiden University, GGZ Rivierduinen, University Medical Center Groningen, University of Groningen, Lentis, GGZ Friesland, GGZ Drenthe, Dimence, Rob Giel Onderzoekscentrum).

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

SDG 3 Good Health and Well-being

Keywords

additive interaction
effect-modification
individual participant data meta-analysis
relative excess risk due to interaction (RERI)

Access to Document

10.1093/aje/kwae325Licence: CC BY

kwae325Final published version, 656 KBLicence: CC BY

Cite this

Basten, M., Van Tuijl, L. A., Pan, K. Y., Hoogendoorn, A. W., Lamers, F., Ranchor, A. V., Dekker, J., Frank, P., Galenkamp, H., Knol, M. J., Noisel, N., Payette, Y., Sund, E. R., Zwinderman, A. H., Portengen, L., & Geerlings, M. I. (2025). Estimating additive interaction in 2-stage individual participant data meta-analysis. American Journal of Epidemiology, 194(6), 1661-1667. https://doi.org/10.1093/aje/kwae325

@article{d555d07bbd0842ba840dc1629e33b531,

title = "Estimating additive interaction in 2-stage individual participant data meta-analysis",

abstract = "Individual participant data (IPD) meta-analysis provides important opportunities to study interaction and effect modification for which individual studies often lack power. While previous meta-analyses have commonly focused on multiplicative interaction, additive interaction holds greater relevance for public health and may in certain contexts better reflect biological interaction. Methodological literature on interaction in IPD meta-analysis does not cover additive interaction for models including binary or time-to-event outcomes. We aimed to describe how the Relative Excess Risk due to Interaction (RERI) and other measures of additive interaction or effect modification can be validly estimated within 2-stage IPD meta-analysis. First, we explain why direct pooling of study-level RERI estimates may lead to invalid results. Next, we propose a 3-step procedure to estimate additive interaction: (1) estimate effects of both exposures and their product term on the outcome within each individual study; (2) pool study-specific estimates using multivariate meta-analysis; (3) estimate an overall RERI and 95\% confidence interval based on the pooled effect estimates. We illustrate this procedure by investigating interaction between depression and smoking and risk of smoking-related cancers using data from the PSYchosocial factors and Cancer (PSY-CA) consortium. We discuss implications of this procedure, including the application in meta-analysis based on published data.",

keywords = "additive interaction, effect-modification, individual participant data meta-analysis, relative excess risk due to interaction (RERI)",

author = "Maartje Basten and \{Van Tuijl\}, \{Lonneke A.\} and Pan, \{Kuan Yu\} and Hoogendoorn, \{Adriaan W.\} and Femke Lamers and Ranchor, \{Adelita V.\} and Joost Dekker and Philipp Frank and Henrike Galenkamp and Knol, \{Mirjam J.\} and Nolwenn Noisel and Yves Payette and Sund, \{Erik R.\} and Zwinderman, \{Aeilko H.\} and L{\"u}tzen Portengen and Geerlings, \{Mirjam I.\}",

note = "Publisher Copyright: {\textcopyright} 2024 The Author(s).",

year = "2025",

month = jun,

day = "1",

doi = "10.1093/aje/kwae325",

language = "English",

volume = "194",

pages = "1661--1667",

journal = "American Journal of Epidemiology",

issn = "0002-9262",

publisher = "Oxford University Press",

number = "6",

}

Basten, M, Van Tuijl, LA, Pan, KY, Hoogendoorn, AW, Lamers, F, Ranchor, AV, Dekker, J, Frank, P, Galenkamp, H, Knol, MJ, Noisel, N, Payette, Y, Sund, ER, Zwinderman, AH, Portengen, L & Geerlings, MI 2025, 'Estimating additive interaction in 2-stage individual participant data meta-analysis', American Journal of Epidemiology, vol. 194, no. 6, pp. 1661-1667. https://doi.org/10.1093/aje/kwae325

TY - JOUR

T1 - Estimating additive interaction in 2-stage individual participant data meta-analysis

AU - Basten, Maartje

AU - Van Tuijl, Lonneke A.

AU - Pan, Kuan Yu

AU - Hoogendoorn, Adriaan W.

AU - Lamers, Femke

AU - Ranchor, Adelita V.

AU - Dekker, Joost

AU - Frank, Philipp

AU - Galenkamp, Henrike

AU - Knol, Mirjam J.

AU - Noisel, Nolwenn

AU - Payette, Yves

AU - Sund, Erik R.

AU - Zwinderman, Aeilko H.

AU - Portengen, Lützen

AU - Geerlings, Mirjam I.

PY - 2025/6/1

Y1 - 2025/6/1

N2 - Individual participant data (IPD) meta-analysis provides important opportunities to study interaction and effect modification for which individual studies often lack power. While previous meta-analyses have commonly focused on multiplicative interaction, additive interaction holds greater relevance for public health and may in certain contexts better reflect biological interaction. Methodological literature on interaction in IPD meta-analysis does not cover additive interaction for models including binary or time-to-event outcomes. We aimed to describe how the Relative Excess Risk due to Interaction (RERI) and other measures of additive interaction or effect modification can be validly estimated within 2-stage IPD meta-analysis. First, we explain why direct pooling of study-level RERI estimates may lead to invalid results. Next, we propose a 3-step procedure to estimate additive interaction: (1) estimate effects of both exposures and their product term on the outcome within each individual study; (2) pool study-specific estimates using multivariate meta-analysis; (3) estimate an overall RERI and 95% confidence interval based on the pooled effect estimates. We illustrate this procedure by investigating interaction between depression and smoking and risk of smoking-related cancers using data from the PSYchosocial factors and Cancer (PSY-CA) consortium. We discuss implications of this procedure, including the application in meta-analysis based on published data.

AB - Individual participant data (IPD) meta-analysis provides important opportunities to study interaction and effect modification for which individual studies often lack power. While previous meta-analyses have commonly focused on multiplicative interaction, additive interaction holds greater relevance for public health and may in certain contexts better reflect biological interaction. Methodological literature on interaction in IPD meta-analysis does not cover additive interaction for models including binary or time-to-event outcomes. We aimed to describe how the Relative Excess Risk due to Interaction (RERI) and other measures of additive interaction or effect modification can be validly estimated within 2-stage IPD meta-analysis. First, we explain why direct pooling of study-level RERI estimates may lead to invalid results. Next, we propose a 3-step procedure to estimate additive interaction: (1) estimate effects of both exposures and their product term on the outcome within each individual study; (2) pool study-specific estimates using multivariate meta-analysis; (3) estimate an overall RERI and 95% confidence interval based on the pooled effect estimates. We illustrate this procedure by investigating interaction between depression and smoking and risk of smoking-related cancers using data from the PSYchosocial factors and Cancer (PSY-CA) consortium. We discuss implications of this procedure, including the application in meta-analysis based on published data.

KW - additive interaction

KW - effect-modification

KW - individual participant data meta-analysis

KW - relative excess risk due to interaction (RERI)

UR - https://www.scopus.com/pages/publications/105008341278

U2 - 10.1093/aje/kwae325

DO - 10.1093/aje/kwae325

M3 - Article

C2 - 39218424

AN - SCOPUS:105008341278

SN - 0002-9262

VL - 194

SP - 1661

EP - 1667

JO - American Journal of Epidemiology

JF - American Journal of Epidemiology

IS - 6

ER -

Estimating additive interaction in 2-stage individual participant data meta-analysis

Abstract

Bibliographical note

Funding

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this