In addition, we performed receiver operating characteristic (ROC) analysis to assess the accuracy of EBV DNA load as a predictive marker of lymphoma [as estimated by the area under the curve (AUC)]. The optimal cut-off value of EBV DNA load for differentiating patients at risk of lymphoma from other patients was determined as the point of the ROC curve with the shortest distance to the 100/100% sensitivity/specificity angle (upper left corner) [i.e. lowest value for the term (1 – sensitivity)2 + (1 – specificity)2, assuming equal costs of false positive and false negative results]. The sensitivity, specificity and OR for developing lymphoma were then provided for the identified

cut-off point. All statistical analyses were performed using sas 9.2 (SAS Institute Inc., VEGFR inhibitor Cary, NC). EBV DNA was positive in PBMC samples from all lymphoma cases collected over the 3 years preceding the Obeticholic Acid diagnosis, while it was positive in 78 to 81% of samples from controls collected during the same period of time (Fig. 1a). Interestingly, eight of the 37 controls had undetectable EBV loads in PBMC1 while none of the 20 cases had undetectable EBV loads in PBMC1 (P = 0.04) (Fig. 1a). EBV load in PBMCs measured a median of 10 months before diagnosis was associated with an increased risk of B lymphoma [OR 2.48 (95% CI 1.16; 5.32) per increase in EBV

load of 1 log copies/106 PBMCs] (Table 2). Similar results were obtained when the OR was adjusted for CD4 cell count nadir instead of CD4 cell count at sample date (OR 2.33; 95% CI 1.12; 4.81). The OR associated with EBV load quantified in a sample collected earlier (median of 24 months before diagnosis) was of borderline significance, probably because of a smaller number of PBMC samples available for that period. When we restricted the analysis to the patients with a CD4 cell count > 300 cells/μL, the median EBV load was still lower in controls (median 2.69) than in cases (median 3.63), mainly because four out of 14 controls had undetectable EBV load vs. none of

seven cases. EBV DNA was more often detectable (> to EBV PCR threshold value or detectable but < to EBV PCR threshold value) in sera from cases than in sera from controls (with 24 to 25% positive detection in the last 3 years for cases vs. 8 to 10.5% for Unoprostone controls) (Fig. 1b); however, this difference was significant only for serum 2 samples (collected a median of 15.3 months before the diagnosis of lymphoma) (Table 2). EBV DNA was positive in PBMC samples from all tested cases during the 3 years preceding the diagnosis of cerebral lymphoma, but it was also positive in 87 to 94% of controls during the same period (Fig. 2a). EBV DNA was not more often detectable (> threshold or detectable < threshold) in sera from cases than in sera from controls (with 0 to 23.1% positive detection in the last 3 years for cases vs. 4.8 to 12% for controls) (Fig. 2b).