. 2015 Feb 6;58(4):781–790. doi: 10.1007/s00125-015-3509-8

Table 3.

Association analysis of the peripheral CD4⁺ T-cell compartment with T1D

Phenotype (gating strategy)	n	Healthy donors		Type 1 diabetes patients		% change (95% CI)^a	p value
Phenotype (gating strategy)	n	n	Mean frequency (95% CI)	n	Mean frequency (95% CI)	% change (95% CI)^a	p value
Cytokine production^b
IL-21⁺ memory T cells (% of CD4⁺) (CD4⁺ CD45RA⁻ IL21⁻)	114	53	4.88% (4.33, 5.50)	61	5.92% (5.44, 6.44)	21.9% (5.8, 40.2)	3.9 × 10⁻³
IFN-γ⁺ memory T cells (% of CD4⁺) (CD4⁺ CD45RA⁻ IFN-γ⁺)	116	54	20.12% (17.52, 23.11)	62	16.93% (14.67, 19.52)	−12.1% (−24.0, 1.7)	0.07
IL-17⁺ CCR6⁺ memory T cells (% of CD4⁺) (CD4⁺ CD45RA⁻ CCR6⁺ IL17⁺)	66	29	3.61% (2.68, 4.86)	37	4.26% (3.25, 5.58)	28.7% (−0.7, 65.9)	0.04
CD4⁺ T follicular helper cell (Tfh) frequency^c
Tfh (% of CD4⁺ CD45RA⁻ CXCR5⁺) (PD-1⁺ CCR6⁻)	62	32	26.02% (23.86, 28.15)	30	29.89% (27.53, 32.25)	14.9% (2.9, 26.9)	0.016

The p values for the cytokine production phenotypes were calculated by linear regression, including batch as a covariate, comparing the mean frequency of the assessed cytokine production phenotypes in type 1 diabetes patients and healthy donors matched as closely as possible for age, sex and time of sample preparation

^aPercent change and 95% CI of the respective T cell phenotype in type 1 diabetes patients relative to healthy donors

^bStatistical tests for the cytokine production phenotypes were performed on log-transformed data because some phenotypes showed a strong right skew. Mean frequencies represent the geometric means obtained from the transformation (Y = exp[Y]) of the log-transformed data (natural log)

^cThe p value for the Tfh cell immunophenotyping was calculated using an unpaired two-tailed Student’s t test