Multiple hypothesis testing in allergy and hypersensitivity diseases investigation: a pedagogical perspective - 22/11/24

Doi : 10.1016/j.jahd.2024.100014

Hugo Guillermou ^a,⁎ , Christophe Abraham ^a, Isabella Annesi-Maesano ^a,^b, Nicolas Molinari ^a
^a Desbrest Institute of Epidemiology and Public Health, université de Montpellier, Iserm, Montpellier, France
^b Division of Respiratory Medicine, Allergology, and of Thoracic Oncology, centre hospitalier universitaire de Montpellier, Montpellier, France

^⁎Corresponding author.

connectez-vous ou créez un compte

Bienvenue sur EM-consulte, la référence des professionnels de santé.
Article gratuit.

Connectez-vous pour en bénéficier!

Graphical abstract

Le texte complet de cet article est disponible en PDF.

Highlights

•	We evaluated the performance of four common methods for controlling global alpha risk (Bonferroni, Sidak, Holm–Bonferroni, Benjamini–Hochberg) using simulated data under different scenarios.
•	These methods should be considered when analyzing real-life exposomic datasets examining associations between environmental exposures and health outcomes like allergy and hypersensitivity diseases or obesity.
•	Choice of control method depends on study objectives, tests conducted, and assumptions on dependence between comparisons.
•	The Holm–Bonferroni method offers a favorable balance between statistical power and stringent control of false positives akin to Bonferroni and Sidak, providing researchers confidence in any statistically significant findings.
•	Benjamini–Hochberg method is well-suited for exploratory analyses with positively correlated or independent comparisons, as was the case for the real-life dataset analysis.

Le texte complet de cet article est disponible en PDF.

Abstract

Background

Allergy and hypersensitivity diseases (AHD) are multifactorial diseases affecting different organs of the human body and with different level of severity. Therefore, multiple hypothesis testing is a common practice in AHD investigation. However, this increases the probability of rejecting a true null hypothesis, meaning there's no effect, the so-called risk of type 1 error or alpha risk. We present here how to control global alpha risk in the case of multiple comparisons or multitesting in AHD investigations to minimize the risk of drawing false positive conclusions.

Methods

Four methods for controlling global alpha risk, namely Bonferroni, Sidak, Holm–Bonferroni, and Benjamini–Hochberg, were applied to simulated and real data. Their performance was assessed through false negative, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV).

Results

The Bonferroni method was found to be the most conservative, while the Benjamini–Hochberg method had the most power. The Holm–Bonferroni method was a compromise between statistical power and control of false positives.

Conclusions

Controlling global alpha risk is crucial in multiple comparisons like they are needed in AHD investigation, and different methods are available to achieve it. Researchers should choose the method that best suits their study, considering the assumptions and objectives.

Le texte complet de cet article est disponible en PDF.

Keywords : Probability, Biostatistics, Methods, Multiple hypothesis testing, Alpha risk control, Multiple comparisons, Bonferroni, Sidak, Holm–Bonferroni, Benjamini–Hochberg