An Association Between ICP-Derived Data and Outcome in TBI Patients: The Role of Sample Size

Brenno Cabella, Joseph Donnelly, Danilo Cardim, Xiuyun Liu, Manuel Cabeleira, Peter Smielewski, Christina Haubrich, Peter J A Hutchinson, Dong Ju Kim, Marek Czosnyka

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

Background: Many demographic and physiological variables have been associated with TBI outcomes. However, with small sample sizes, making spurious inferences is possible. This paper explores the effect of sample sizes on statistical relationships between patient variables (both physiological and demographic) and outcome. Methods: Data from head-injured patients with monitored arterial blood pressure, intracranial pressure (ICP) and outcome assessed at 6 months were included in this retrospective analysis. A univariate logistic regression analysis was performed to obtain the odds ratio for unfavorable outcome. Three different dichotomizations between favorable and unfavorable outcomes were considered. A bootstrap method was implemented to estimate the minimum sample sizes needed to obtain reliable association between physiological and demographic variables with outcome. Results: In a univariate analysis with dichotomized outcome, samples sizes should be generally larger than 100 for reproducible results. Pressure reactivity index, ICP, and ICP slow waves offered the strongest relationship with outcome. Relatively small sample sizes may overestimate effect sizes or even produce conflicting results. Conclusion: Low power tests, generally achieved with small sample sizes, may produce misleading conclusions, especially when they are based only on p values and the dichotomized criteria of rejecting/not-rejecting the null hypothesis. We recommend reporting confidence intervals and effect sizes in a more complete and contextualized data analysis.

Original languageEnglish
Pages (from-to)1-5
Number of pages5
JournalNeurocritical Care
DOIs
Publication statusAccepted/In press - 2016 Nov 7

Fingerprint

Intracranial Pressure
Sample Size
Demography
Arterial Pressure
Logistic Models
Odds Ratio
Head
Regression Analysis
Confidence Intervals
Pressure

Keywords

  • Autoregulation
  • Intracranial pressure
  • Outcome prediction
  • Statistical inference
  • Traumatic brain injury

ASJC Scopus subject areas

  • Critical Care and Intensive Care Medicine
  • Clinical Neurology

Cite this

Cabella, B., Donnelly, J., Cardim, D., Liu, X., Cabeleira, M., Smielewski, P., ... Czosnyka, M. (Accepted/In press). An Association Between ICP-Derived Data and Outcome in TBI Patients: The Role of Sample Size. Neurocritical Care, 1-5. https://doi.org/10.1007/s12028-016-0319-x

An Association Between ICP-Derived Data and Outcome in TBI Patients : The Role of Sample Size. / Cabella, Brenno; Donnelly, Joseph; Cardim, Danilo; Liu, Xiuyun; Cabeleira, Manuel; Smielewski, Peter; Haubrich, Christina; Hutchinson, Peter J A; Kim, Dong Ju; Czosnyka, Marek.

In: Neurocritical Care, 07.11.2016, p. 1-5.

Research output: Contribution to journalArticle

Cabella, B, Donnelly, J, Cardim, D, Liu, X, Cabeleira, M, Smielewski, P, Haubrich, C, Hutchinson, PJA, Kim, DJ & Czosnyka, M 2016, 'An Association Between ICP-Derived Data and Outcome in TBI Patients: The Role of Sample Size', Neurocritical Care, pp. 1-5. https://doi.org/10.1007/s12028-016-0319-x
Cabella, Brenno ; Donnelly, Joseph ; Cardim, Danilo ; Liu, Xiuyun ; Cabeleira, Manuel ; Smielewski, Peter ; Haubrich, Christina ; Hutchinson, Peter J A ; Kim, Dong Ju ; Czosnyka, Marek. / An Association Between ICP-Derived Data and Outcome in TBI Patients : The Role of Sample Size. In: Neurocritical Care. 2016 ; pp. 1-5.
@article{e06523bdf040458895a76b7811a95dd8,
title = "An Association Between ICP-Derived Data and Outcome in TBI Patients: The Role of Sample Size",
abstract = "Background: Many demographic and physiological variables have been associated with TBI outcomes. However, with small sample sizes, making spurious inferences is possible. This paper explores the effect of sample sizes on statistical relationships between patient variables (both physiological and demographic) and outcome. Methods: Data from head-injured patients with monitored arterial blood pressure, intracranial pressure (ICP) and outcome assessed at 6 months were included in this retrospective analysis. A univariate logistic regression analysis was performed to obtain the odds ratio for unfavorable outcome. Three different dichotomizations between favorable and unfavorable outcomes were considered. A bootstrap method was implemented to estimate the minimum sample sizes needed to obtain reliable association between physiological and demographic variables with outcome. Results: In a univariate analysis with dichotomized outcome, samples sizes should be generally larger than 100 for reproducible results. Pressure reactivity index, ICP, and ICP slow waves offered the strongest relationship with outcome. Relatively small sample sizes may overestimate effect sizes or even produce conflicting results. Conclusion: Low power tests, generally achieved with small sample sizes, may produce misleading conclusions, especially when they are based only on p values and the dichotomized criteria of rejecting/not-rejecting the null hypothesis. We recommend reporting confidence intervals and effect sizes in a more complete and contextualized data analysis.",
keywords = "Autoregulation, Intracranial pressure, Outcome prediction, Statistical inference, Traumatic brain injury",
author = "Brenno Cabella and Joseph Donnelly and Danilo Cardim and Xiuyun Liu and Manuel Cabeleira and Peter Smielewski and Christina Haubrich and Hutchinson, {Peter J A} and Kim, {Dong Ju} and Marek Czosnyka",
year = "2016",
month = "11",
day = "7",
doi = "10.1007/s12028-016-0319-x",
language = "English",
pages = "1--5",
journal = "Neurocritical Care",
issn = "1541-6933",
publisher = "Humana Press",

}

TY - JOUR

T1 - An Association Between ICP-Derived Data and Outcome in TBI Patients

T2 - The Role of Sample Size

AU - Cabella, Brenno

AU - Donnelly, Joseph

AU - Cardim, Danilo

AU - Liu, Xiuyun

AU - Cabeleira, Manuel

AU - Smielewski, Peter

AU - Haubrich, Christina

AU - Hutchinson, Peter J A

AU - Kim, Dong Ju

AU - Czosnyka, Marek

PY - 2016/11/7

Y1 - 2016/11/7

N2 - Background: Many demographic and physiological variables have been associated with TBI outcomes. However, with small sample sizes, making spurious inferences is possible. This paper explores the effect of sample sizes on statistical relationships between patient variables (both physiological and demographic) and outcome. Methods: Data from head-injured patients with monitored arterial blood pressure, intracranial pressure (ICP) and outcome assessed at 6 months were included in this retrospective analysis. A univariate logistic regression analysis was performed to obtain the odds ratio for unfavorable outcome. Three different dichotomizations between favorable and unfavorable outcomes were considered. A bootstrap method was implemented to estimate the minimum sample sizes needed to obtain reliable association between physiological and demographic variables with outcome. Results: In a univariate analysis with dichotomized outcome, samples sizes should be generally larger than 100 for reproducible results. Pressure reactivity index, ICP, and ICP slow waves offered the strongest relationship with outcome. Relatively small sample sizes may overestimate effect sizes or even produce conflicting results. Conclusion: Low power tests, generally achieved with small sample sizes, may produce misleading conclusions, especially when they are based only on p values and the dichotomized criteria of rejecting/not-rejecting the null hypothesis. We recommend reporting confidence intervals and effect sizes in a more complete and contextualized data analysis.

AB - Background: Many demographic and physiological variables have been associated with TBI outcomes. However, with small sample sizes, making spurious inferences is possible. This paper explores the effect of sample sizes on statistical relationships between patient variables (both physiological and demographic) and outcome. Methods: Data from head-injured patients with monitored arterial blood pressure, intracranial pressure (ICP) and outcome assessed at 6 months were included in this retrospective analysis. A univariate logistic regression analysis was performed to obtain the odds ratio for unfavorable outcome. Three different dichotomizations between favorable and unfavorable outcomes were considered. A bootstrap method was implemented to estimate the minimum sample sizes needed to obtain reliable association between physiological and demographic variables with outcome. Results: In a univariate analysis with dichotomized outcome, samples sizes should be generally larger than 100 for reproducible results. Pressure reactivity index, ICP, and ICP slow waves offered the strongest relationship with outcome. Relatively small sample sizes may overestimate effect sizes or even produce conflicting results. Conclusion: Low power tests, generally achieved with small sample sizes, may produce misleading conclusions, especially when they are based only on p values and the dichotomized criteria of rejecting/not-rejecting the null hypothesis. We recommend reporting confidence intervals and effect sizes in a more complete and contextualized data analysis.

KW - Autoregulation

KW - Intracranial pressure

KW - Outcome prediction

KW - Statistical inference

KW - Traumatic brain injury

UR - http://www.scopus.com/inward/record.url?scp=84994430576&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84994430576&partnerID=8YFLogxK

U2 - 10.1007/s12028-016-0319-x

DO - 10.1007/s12028-016-0319-x

M3 - Article

C2 - 27822739

AN - SCOPUS:84994430576

SP - 1

EP - 5

JO - Neurocritical Care

JF - Neurocritical Care

SN - 1541-6933

ER -