Re-evaluating the association between alcohol use and in-hospital survival: the role of collider bias

Recent studies have reported paradoxical associations between alcohol use and trauma outcomes, sparking debate in public health, trauma surgery, and emergency medicine [1,2,3]. In a recent paper published in Critical Care, Kazuma Sasaki et al. [4] explored the relationship between pre-injury alcohol consumption and traumatic brain injury (TBI) outcomes of hospitalized patients in Japan, reporting a potential direct protective effect of alcohol in conferring survival advantage. While the authors acknowledge the observational nature of their findings, we draw attention to an important methodological issue—namely, collider bias—which may have distorted the reported associations in terms of causal inference. In this Matters Arising, we demonstrate how collider bias can arise in such analyses and discuss the need for caution in the causal interpretation of non-randomized observational studies of this nature.

Collider bias arises when two exposure variables both influence a third collider variable W, and the analysis conditions on W. This causal relationship can be illustrated using a directed acyclic graph (DAG), where W is located at the collision point of the two arrowheads from two independent variables Z1 and Z2 (Fig. 1A). In this structure, adjusting for W, or selecting or stratifying the sample population based on W, can introduce a collider bias [5]. Such paths with colliders can be a source of selection bias, introducing a methodological flaw to observational research and misleading conclusions when unidentified in study design.

Fig. 1

A DAG illustrating W as a collider variable in the path Z1 → W ← Z2. Before conditioning on W, the relationship between X and Y may be confounded by the path X-Z1-W-Y. However, conditioning on W opens an otherwise closed non-causal pathway X-Z1-W-Z2-Y, creating an apparent association between X and Y [5]. B In the analytic model of this study by Sasaki et al., conditioning on hospitalization with TBI may introduce a non-causal association between alcohol use and in-hospital survival via covariates Z1 and Z2

Full size image

In this study by Sasaki et al., a simplified DAG can be constructed with alcohol use as X, hospitalization with TBI as W, and in-hospital survival as Y (Fig. 1B). Patient demographics may be a factor (Z1) that influences both alcohol use (X) and the likelihood of being hospitalized with TBI (W). Injury severity (Z2) would affect both the likelihood of hospitalization (W) and survival (Y). By making hospitalization with TBI (W) a necessary condition for inclusion in the study, W acts as a collider and can inadvertently open non-causal paths between X and Y. As a result, alcohol use may appear spuriously associated with improved in-hospital survival, even if no true protective effect or causal relationship exists. For example, this could occur if alcohol use (X) is associated with both reduced trauma severity and impaired consciousness, as noted by the authors [4]. This bias arises not from failure to adjust, but rather from conditioning on a variable that sits at the intersection of multiple causes.

While the authors use regression and propensity score methods to account for confounding, these methods may not address and potentially even magnify the structural issue of conditioning on a collider. Notably, performing mediation analysis on a population already conditioned on the collider variable (W), as done in this paper, does not address the collider bias introduced initially. We do not dispute the possibility that alcohol interacts with pathophysiological processes in trauma, but the present study design does not allow for drawing causal conclusions. More robust causal inference techniques, such as target trial emulation or structural causal models, are necessary to address these complexities. Moreover, suggestions of direct protective effects of alcohol must be interpreted with caution, given the well-documented harms of alcohol and the potential for public misinterpretation [6, 7].

In conclusion, this study by Sasaki et al. contributes important data to the ongoing exploration of alcohol and trauma outcomes. However, its causal interpretations are limited by collider bias introduced through selection on hospitalization with TBI. Recognizing and accounting for such structural biases is crucial for valid inference in observational research.

No datasets were generated or analysed during the current study.

TBI:

Traumatic brain injury

DAG:

Directed acyclic graph

1. Lau G, Gabbe BJ, Mitra B, Dietze PM, Reeder S, Cameron P, et al. Association between acute pre-injury alcohol use and 12-month health outcomes for survivors of major trauma: A registry-based study. Injury. 2024;55:111782. https://doi.org/10.1016/j.injury.2024.111782.

   Article PubMed Google Scholar

2. Ng C, Fleury M, Hakmi H, Bronson B, Vosswinkel JA, Huang EC, et al. The impact of alcohol use and withdrawal on trauma outcomes: A case control study. Am J Surg. 2021;222:438–45. https://doi.org/10.1016/j.amjsurg.2020.12.026.

   Article PubMed Google Scholar

3. Covarrubias J, Grigorian A, Nahmias J, Chin TL, Schubl S, Joe V, et al. Vices-paradox in trauma: Positive alcohol and drug screens associated with decreased mortality. Drug Alcohol Depend. 2021;226:108866. https://doi.org/10.1016/j.drugalcdep.2021.108866.

   Article PubMed Google Scholar

4. Sasaki K, Tagami T, Obinata H, Tanaka C, Otake K, Yoshino Y, et al. Influence of alcohol on in-hospital survival rate among patients with traumatic brain injury: A nationwide cohort study. Crit Care. 2025;29:133. https://doi.org/10.1186/s13054-025-05364-0.

   Article PubMed PubMed Central Google Scholar

5. Rothman KJ, Greenland S, Lash TL. Modern epidemiology. 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2008.

   Google Scholar

6. Anderson BO, Berdzuli N, Ilbawi A, Kestel D, Kluge HP, Krech R, et al. Health and cancer risks associated with low levels of alcohol consumption. Lancet Public Health. 2023;8:e6-7. https://doi.org/10.1016/S2468-2667(22)00317-6.

   Article PubMed PubMed Central Google Scholar

7. Cecchini M, Filippini T, Whelton PK, Iamandii I, Di Federico S, Boriani G, et al. Alcohol intake and risk of hypertension: A systematic review and dose-response meta-analysis of nonexperimental cohort studies. Hypertension. 2024;81:1701–15. https://doi.org/10.1161/HYPERTENSIONAHA.124.22703.

   Article CAS PubMed Google Scholar

Download references

Not applicable.

Not applicable.

Authors and Affiliations

1. Department of Emergency and Trauma Medicine, Kameda Medical Center, 929 Higashi-cho, Kamogawa, Chiba, 296-8602, Japan

   Tomoko Yasuda, Taisuke Shibata & Atsushi Shiraishi

Authors

1. Tomoko YasudaView author publications

   You can also search for this author inPubMed Google Scholar

2. Taisuke ShibataView author publications

   You can also search for this author inPubMed Google Scholar

3. Atsushi ShiraishiView author publications

   You can also search for this author inPubMed Google Scholar

Contributions

T.Y. wrote the main manuscript text and prepared the figure. All authors have made substancial contributions to the conception and revision of the manuscript.

Corresponding author

Correspondence to Atsushi Shiraishi.

Conflict of interest

The authors declare no competing interests.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Ethical approval

Not applicable.

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

Cite this article

Yasuda, T., Shibata, T. & Shiraishi, A. Re-evaluating the association between alcohol use and in-hospital survival: the role of collider bias. Crit Care 29, 203 (2025). https://doi.org/10.1186/s13054-025-05442-3

Download citation

• Received: 27 April 2025

• Accepted: 04 May 2025

• Published: 20 May 2025

• DOI: https://doi.org/10.1186/s13054-025-05442-3

Share this article

Anyone you share the following link with will be able to read this content:

Get shareable link

Sorry, a shareable link is not currently available for this article.

Copy to clipboard

Provided by the Springer Nature SharedIt content-sharing initiative