Deep immune cell phenotyping predicts influenza vaccine immunogenicity in HCT recipients

Annual influenza vaccination is recommended after allogeneic hematopoietic cell transplantation (HCT), but antibody responses vary widely. We asked whether pre-vaccination immune-cell composition predicts the antibody response beyond baseline titers and clinical factors. We studied the first vaccination season of 260 participants in 2 multicenter phase 2 trials of high- versus standard-dose influenza vaccine, 1 pediatric and 1 adult, and held out for validation 74 second-season episodes from participants who re-enrolled the following year. Thirty-marker mass cytometry measured blood cell composition, and models predicted the mean antibody log fold-rise across 3 antigens in recipients left out of nested cross-validation. Adding immune-cell composition to baseline antibody titers reduced mean squared prediction error by 15.3% (95% CI, 7.0-23.2%; P < .001), and adding it to titers plus clinical factors by 12.5% (4.6-20.3%; P = .001). A cellular composition score identified recipients unlikely to respond: in its lowest fifth, 15% had a fourfold antibody rise against 51% of the others. At a given baseline titer, each standard deviation of the score about doubled the odds of a fourfold rise (odds ratio, 2.2; P = .007), in children and adults alike; low scores reflected fewer B cells and more monocytes. In the held-out second season, composition reduced prediction error by 39%, and prediction was similar with 15 markers. Pre-vaccination immune-cell composition therefore predicts influenza vaccine immunogenicity after HCT, and simpler assays merit prospective validation. The trials are registered at www.clinicaltrials.gov as # NCT02860039 (pediatric) and # NCT03179761 (adult).