Bayesian methods as a complementary tool: balancing innovation and rigor in critical care research

Abstract

The perspective by Patel and Green, “Death by P-value: The Overreliance on P-values in Critical Care Research” [1], offers a timely critique of rigid statistical thresholds in critical care trials. By advocating for hybrid approaches that integrate Bayesian methods with traditional frequentist analysis, the authors highlight the potential of probabilistic reasoning to uncover clinically meaningful effects obscured by borderline p-values. While their argument is thought-provoking, several considerations warrant further discussion to ensure a balanced application of Bayesian methods in this field.

The authors rightly emphasize that Bayesian analysis should complement—not replace—frequentist frameworks. Their examples (e.g., hydrocortisone in traumatic brain injury, β-blockers in septic shock) demonstrate how posterior distributions can contextualize findings when p-values are near 0.05. However, while compelling, such re-analyses must not be conflated with definitive evidence. For instance, the reported 87% posterior probability of hydrocortisone reducing ventilator-associated pneumonia (VAP) risk by ≥ 10% remains hypothesis-generating. Bayesian results should be interpreted as one component of a broader evidentiary hierarchy, alongside biological plausibility, trial design, and external validation.

A key concern in Bayesian analysis is the influence of prior distributions. While the authors employed neutral priors (e.g., mean effect = 0, standard deviation = 10%), even these choices introduce assumptions. Using a standard deviation of 10% in the β-blocker mortality analysis presumes that true effects beyond ± 20% are implausible—a debatable premise in sepsis research. To enhance objectivity, future studies should:

• Pre-specify prior distributions in trial protocols, informed by systematic reviews or expert consensus

• Conduct sensitivity analyses using skeptical priors (e.g., centered on harm) or enthusiastic priors (e.g., larger expected benefits)

• Adhere to guidelines such as the ISBA bulletin [2] on Bayesian Hypothesis Testing with transparent reporting of prior justification and Bayes factors

The critique of p-values should not overshadow their utility in controlling Type I error rates. More specifically, the continuous versus interrupted chest compressions trial reported a posterior probability of 75% for survival benefit with interrupted compressions. Yet, the frequentist 95% confidence interval (− 1.5 to 0.1%) and corresponding credible interval remind us that the effect could plausibly be null or harmful. Rather than abandoning p-values, a hybrid approach could:

• Use Bayesian posterior probabilities to prioritize interventions for further study

• Reserve frequentist analyses for confirmatory endpoints in pre-registered trials

• Report both Bayesian and frequentist results in interim and final analyses

Critical care research often faces small sample sizes due to patient heterogeneity and practical constraints. While Bayesian methods can extract insights from limited data, they are not immune to overfitting. In the tracheotomy timing study [1], the 7% absolute reduction in VAP (P = 0.07) corresponds to a wide 95% confidence interval (hazard ratio [HR] 0.42–1.04). A posterior probability of > 75% benefit must be weighed against the frequentist evidence suggesting the true effect spans from a 58% reduction to a 4% increase. Here, Bayesian analysis serves best as a bridge to targeted trials—particularly through adaptive designs identifying subgroups where the signal is strongest.

To harness the strengths of both paradigms, we propose:

• Co-primary endpoints: Pre-specify both Bayesian posterior probabilities and frequentist Type I error thresholds in trial designs

• Replication standards: Validate Bayesian analyses in independent cohorts before clinical implementation

• Education initiatives: Train clinicians to interpret both posterior probabilities and power analyses within clinical context

Patel and Green’s perspective rightly challenges the dogma of p < 0.05. However, the solution lies not in discarding p-values but in enriching our analytical toolkit. By combining Bayesian flexibility with frequentist rigor, critical care research can better navigate the tension between statistical precision and clinical urgency. Let us embrace hybrid methods—but with the same scrutiny we demand of traditional approaches.

No datasets were generated or analysed during the current study.

1. Patel S, Green A. Death by p-value: the overreliance on p-values in critical care research. Crit Care. 2025;29(1):73.

   PubMed PubMed Central Google Scholar

2. ISBA Bulletin (2011). The Official Bulletin of the International Society for Bayesian Analysis. 2011. Retrieved from https://bayesian.org/wp-content/uploads/2016/09/1103.pdf

Download references

None

None.

Author notes

1. Sen Lu and Kai Liu contributed equally to this article and are co-first authors.

Authors and Affiliations

1. Department of Critical Care Medicine, Sichuan Academy of Medical Sciences and Sichuan Provincial People’s Hospital, University of Electronic Science and Technology of China, Chengdu, China

   Sen Lu & Jing-chao Luo

2. Department of Critical Care Medicine, Zhongshan Hospital, Fudan University, Shanghai, China

   Kai Liu

3. School of Data Science, Chinese University of Hong Kong, Shenzhen, China

   Xin-yun Chen

Authors

1. Sen LuView author publications

   You can also search for this author inPubMed Google Scholar

2. Kai LiuView author publications

   You can also search for this author inPubMed Google Scholar

3. Xin-yun ChenView author publications

   You can also search for this author inPubMed Google Scholar

4. Jing-chao LuoView author publications

   You can also search for this author inPubMed Google Scholar

Contributions

SL, KL, JL and XC wrote and revised the manuscript.

Corresponding authors

Correspondence to Xin-yun Chen or Jing-chao Luo.

Ethical approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

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

Lu, S., Liu, K., Chen, Xy. et al. Bayesian methods as a complementary tool: balancing innovation and rigor in critical care research. Crit Care 29, 135 (2025). https://doi.org/10.1186/s13054-025-05380-0

Download citation

• Received: 18 March 2025

• Accepted: 19 March 2025

• Published: 25 March 2025

• DOI: https://doi.org/10.1186/s13054-025-05380-0

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