2018 Section 5 - Rhinology and Allergic Disorders

TOMASSEN ET AL

J ALLERGY CLIN IMMUNOL MAY 2016

A

0.2

Cluster 1 2 3 4 5 6 7 8 9 10

-1.0 -0.5 0.0 0.5 1.0

0.1

0.0

IL-17

IL-1β

Dimension 2

MPO IL-8

−0.1

Dimension 2 (15.47%)

IL-6

−0.2

−0.2

−0.1

0.0

0.1

0.2

0.3

albumin ECP

Dimension 1

IgE

IL-5

Analogously, clinical parameters (which were not used in the cluster analysis) were analyzed within clusters and compared with control values ( Fig 3 ). The proportion of CRSwNP was significantly different between clusters, with clusters 1 to 3 having a significantly lower proportion than most clusters and clusters 8 to 10 having a significantly higher proportion than most clusters. Asthma prevalence was significantly different from that in control subjects and differed significantly between clusters. Age, sex, and prevalence of allergy and reported aspirin sensitivity were not significantly different between clusters (see Table E3 in this article’s Online Repository at www.jacionline.org ). Means and ratios of cytokine and phenotype data were calculated and tabulated as a heat map to characterize the clusters ( Fig 3 ). The most striking differentiation was formed by the IL-5/ECP/IgE/albumin axis, resulting in 4 clusters with undetectable or low concentrations of these biomarkers and 6 clusters with high concentrations. The endotype differentiation was also reflected in the clinical phenotypes, with 3 clusters (clusters 1-3) showing a predominant CRSsNP appearance, 4 clusters (clusters 4-7) showing a mixed phenotype, and 3 clusters (clusters 8-10) showing a predominant CRSwNP appearance. Clusters 1 to 3 showed no increase in asthma prevalence, whereas clusters 4 to 10 showed that asthma prevalence was increased compared with that in control subjects and clusters 1 to 3, with further increases in clusters 9 and 10. The 4 clusters with predominant noneosinophilic appearance could further be differentiated as follows: cluster 1 did not show FIG 1. Principal component analysis and clustering of variables. Vectors of each parameter are plotted against the first 4 rotated principal components. A, First 2 principal components. B, Third and fourth principal components. The fifth rotated component is not depicted. It consisted of IFN- g and SE-IgE. C, Dendrogram of clustered variables. The horizontal dashed line indicates the tree cut used in this analysis based on the Rand index, resulting in 6 variable clusters. = FIG 2. Cluster plot of cases. Clustering of cases illustrated in a clusplot in which patients are plotted in a 2-dimensional space after multidimensional scaling. Here, dissimilarities between patients, taking all variables into account, are represented by distances between the patients in 2 dimensions.

-1.0

-0.5

0.0

0.5

1.0

Dimension 1 (32.6%)

B

-0.5 0.0 0.5 1.0

IL-22

TNF-α

IL-17

SE-IgE

TGF-β1

Dim ension 4 (9.3%)

-1.0

-1.0

-0.5

0.0

0.5

1.0

Dimension 3 (10.0%)

Height C

0.0 0.5 1.0 1.5

IgE

IL-5

IL-6

IL-8

IL-1

ECP

MPO

IL-17

IL-22

IFN-γ

TNF-α

SE-IgE

TGF-β1

albumin

64