Descriptive analysis used medians with interquartile range

Descriptive analysis used medians with interquartile range (IQR) values. Statistical comparisons between the 3 NK groups (<50, 50–99, ≥100) were based on the non-parametric Kruskal–Wallis test for continuous variables and were based on Pearson\'s chi-square test or Fisher\'s exact test when appropriate for comparison of proportions of patients in multiple groups. Given the large list of clinical and biological characteristics compared between the 3 NK groups, we report Benjamini–Hochberg adjusted P values (P), which maintains the false discovery rate at the nominal alpha 0.05 level (Benjamini–Hochberg, 1995). Overall survival (OS) was calculated from inclusion in DEFI study until the last visit or death from any cause. Follow-up ended in April 2014. Survival was estimated by the Kaplan–Meier product-limit method. For OS comparisons, the groups considered were severe, mild and no NK cell lymphopenia, and then a second analysis was performed in 4 groups, as a function of NK cell and/or CD4+ T cell deficiency, and the log-rank test was used. Statistical analyses were performed with STATA Statistical Software version 11.1 (Stata Corp., College Station, TX, USA).
Results
Discussion Large cohorts for studying the correlation between NK cell deficiency and clinical events in humans are extremely rare. The study by Imai and colleagues (Imai et al., 2000) reporting correlation between low NK cell cytotoxicity and the incidence of cancer in a cohort of 3625 patients is the only such study published to date. Our study is original in addressing the role of NK cells in a large cohort of immunodeficient patients through exploration of the phenotype of patients with quantitative NK cell deficiency. Different classifications and subgroups of CVID patients have been proposed, defined on the basis of the quantitative abnormalities of JNJ-26481585 and T cells (Malphettes et al., 2009; Wehr et al., 2008), but data regarding the NK cell compartment of these patients and the phenotype of CVID patients with NK cell deficiency are rare. Few studies have already evaluated number of NK cells in CVID patients and observed a decreased number of NK cells in these patients (Aspalter et al., 2000; Berrón-Ruiz et al., 2014; Kutukculer et al., 2015). Nevertheless, our study is unique in reporting clinical manifestations and phenotype of patients with NK cell deficiency, especially in a cohort as large as more than 450 CVID patients. In 2000 Aspalter and colleagues reported a decrease in absolute and relative numbers of NK cells (defined as CD3−CD16+ cells) in 55 CVID patients compared to healthy controls, but without study of clinical manifestations associated with this NK cell deficiency (Aspalter et al., 2000). In 2015 Kutukculer and colleagues explored innate immunity in a small cohort of 20 pediatric CVID patients, with decreased percentage and increased cytotoxicity of NK cells in patients with severe disease (depending on the presence of splenomegaly, granuloma and/or bronchiectasis), but without detail about each of these clinical manifestations and other complications of the disease (Kutukculer et al., 2015). In the study of Aspalter et al., a decrease of invariant NK T cells has also been found in CVID patients, with again difficulties to conclude without available clinical data in this study (Aspalter et al., 2000). Finally, type 1 innate lymphoid cells (ILCs) can share phenotypical markers, such as CD56, with bona fide NK cells (Hazenberg and Spits, 2014) and a type 1 ILCs defect could potentially contribute to the observed phenotype of patients with NK cell deficiency. Interestingly, although there is a large body of evidence concerning the role of NK cells in defense against viral pathogens (Lugli et al., 2014; Della Chiesa et al., 2014; Lee et al., 2007), the proportion of patients with viral infections did not differ between groups defined on the basis of their circulating NK cell counts in this large population of CVID patients. We also found no qualitative evidence for susceptibility to a particular type of virus in patients with severe NK cell lymphopenia. For example, Human papilloma virus (HPV) infections were equally frequent in all groups of patients, despite reports implicating NK cells in the control of infections with this pathogen (Kamili et al., 2014; Bere et al., 2014). Similarly, CMV infections did not appear to be more frequent in patients with severe NK cell deficiency than in patients with mild or no NK cell deficiency in our study. Difficult formal diagnosis of viral pathogens in clinical practice in these hypogammaglobulinemic patients, underestimation of benign viral episodes, or compensation by T cell compartment in this adult population could be as many explanations for this apparent discrepancy between earlier reports and our present study. In any event, these data highlight that the demonstration of a non-redundant role of NK cells in the control of viral infections has been obtained in the mouse, but not formally in humans.