IVIG for Infection Prevention After CAR-T-Cell Therapy

Will use similar analytic approaches as in the primary outcome analysis to evaluate secondary outcomes (any serious infection, any infection) and all outcomes in pre-specified populations (mITT, per-protocol, ITT). Will explore outcomes restricted to early (days 0-28) and later (days 29-168) post-CARTx time periods. Plot cumulative incidence curves of time-to-first serious bacterial infection by study arm. Incidence rate ratios, hazard ratios (HR), and their 95% CIs will be reported. Will use one-sided testing to determine whether the rate of infections is significantly lower in the immunoglobulin replacement therapy (IGRT) versus placebo arms.

Will graphically display mean total IgG levels, IgG subclasses, and S. pneumoniae IgG titers over time, stratified by study arm. Will use generalized estimating equations (GEE) models to compare means by randomization arm. Will test the association of each IgG metric as time-varying covariates with risk of subsequent infection (serious bacterial, any serious, any) at any time post-randomization and separately during early (days 0-28) and later (days 29-168) periods post-CARTx using Cox proportional hazard models. Model estimates will be presented as HRs with 95% CIs.

Will describe HRU metrics by study arm (overall and by relationship to study drug infusion or infection). Will compute rates (per 1000 person-days) for each metric (e.g., days alive and out of the hospital, days of antibiotics, outpatient visits, emergency room visits, hospitalizations, adverse events) and compare between study arms using rate ratios with p-values and 95% CIs calculated by Poisson or negative binomial regression.

Will plot cumulative incidence curves, stratified by study arm, of any and grade >= 3 CRS, ICANS, or both, and compare hazard ratios using multivariable Cox proportional hazards models to adjust for stratification variables and potential confounders. To account for multiple comparisons, we will consider the Benjamini-Hochberg procedure for controlling the false discovery rate.

Will display and compare between study arms the distribution of Cmax based on flow cytometry and quantitative polymerase chain reaction (qPCR) at weekly intervals between days 0 and 28 after CAR-T cell infusion using the Wilcoxon rank-sum test.

Will display and compare between study arms the area under the curve (AUC) 0-28 CAR-T cell values (log10 cells/uL) based on flow cytometry and quantitative polymerase chain reaction (qPCR) at weekly intervals between days 0 and 28 after CAR-T cell infusion using the Wilcoxon rank-sum test. The AUC 0-28 will be computed among log10-transformed CAR-T cell counts using a trapezoidal rule computational algorithm. Will use linear regression to test for differences in AUC 0-28 by study arm after adjusting for stratification variables (CAR-T cell product, site), disease burden (pre-lymphodepletion % bone marrow blasts; >= 5% versus < 5%), and other potential confounders.

Will display the distribution of CAR-T cell values (log10 copies/ug deoxyribonucleic acid) per study arm based on monthly qPCR results and compare results using similar approaches as described for expansion. Timing of the loss of CAR-T cell persistence will be defined as the date of the first of two consecutive CAR-T cell values < 10 copies/ug. Will compute the cumulative incidence of loss of persistence and evaluate the association with study arm by day 168 using a Cox proportional hazards model including similar adjustment factors as previously described.

Compare CAR T-cell phenotype and function at day 14 after infusion between study arms. Cell subsets will be depicted using Uniform Manifold Approximation and Projection (UMAP), a dimension reduction technique to visualize and understand large, high dimensional datasets. Heatmaps will be created in R using the heatmap package to visualize differences in the proportion of patients with activated versus exhausted CAR-T cells and other select immune cell clusters, grouped for every marker and supervised by treatment arm. Will compare differences in absolute immune cell subset counts, the proportion of cell subsets with markers of activation versus exhaustion, and intracellular signaling (phospho-protein responses, stimulated and unstimulated) between study arms. Multivariable linear regression will be used to assess the relationship between treatment group and cell subset counts, as well as counts of activated versus exhausted cells, with adjustment for relevant confounders.

Compare all immune cell subset phenotypes and functional makers at the end of the study between study arms. Cell subsets will be depicted using UMAP, a dimension reduction technique to visualize and understand large, high dimensional datasets. Heatmaps will be created in R using the heatmap package to visualize differences in the proportion of patients with activated versus exhausted CAR-T cells and other select immune cell clusters, grouped for every marker and supervised by treatment arm. Will compare differences in absolute immune cell subset counts, the proportion of cell subsets with markers of activation versus exhaustion, and intracellular signaling (phospho-protein responses, stimulated and unstimulated) between study arms. Multivariable linear regression will be used to assess the relationship between treatment group and cell subset counts, as well as counts of activated versus exhausted cells, with adjustment for relevant confounders.

Compare IgA and IgM at end of study between study arms.

Compare HRQOL between study arms. Assessment of quality of life using the Patient Short Form-36 HRQOL electronic questionnaire.