Background: iAEs such as cytokine release syndrome (CRS) and neurotoxicity (NT) are common with CAR T cell therapies such as ide-cel; CRS and NT observed with ide-cel are generally low grade and manageable with tocilizumab (toci) and corticosteroids. Identifying iAE risk with sensitivity and specificity is challenging, suggesting multifactorial etiology and limitations evaluating dynamic processes at specific timepoints. Macrophage activation syndrome (MAS; per CTCAE) has a specific molecular profile representing a homogeneous subgroup within CRS, tending to be more severe/refractory to front-line management. Using longitudinal data on post-infusion soluble factors we aimed to develop a candidate post-infusion inflammatory risk score based on MAS molecular signature and determine associations with pt baseline characteristics. Methods: Statistical learning tools were applied to the longitudinal profile of 34 factors measured over 28 d after ide-cel infusion in 291 pts from KarMMa (NCT03361748) and KarMMa-2 (NCT03601078) to extract features describing the profiles. Unsupervised machine learning-based clustering was applied to these features and enrichment for MAS cases within each cluster was assessed. A metric describing membership in a cluster of interest was used as a candidate post-infusion inflammatory risk score for further correlative analyses. Results: Pts with MAS had cytokine profiles consistent with high levels of pro-inflammatory signalling across multiple inflammatory cell types. Cytokine profiles of MAS were qualitatively similar to those of CRS/NT; however, unsupervised clustering showed a distinct group of pts with MAS. A metric of similarity to the MAS-enriched cluster was defined as a candidate post-infusion inflammatory risk score, posited to identify pts more likely to have more severe/refractory iAEs. This inflammatory risk score positively correlated with CRS and NT grades and frequency of toci usage (Kruskal-Wallis test; all P< 0.001). A prototype predictive model based on baseline pt and clinical characteristics strongly correlated with post-infusion inflammatory risk score (Pearson; ρ = 0.88). Low blood counts, higher tumor burden, and elevated levels of inflammatory markers on the day of ide-cel infusion were associated with a higher inflammatory risk score. Conclusions: A molecular profile of severe/refractory iAEs was preliminarily identified in ide-cel treated pts with MM. This profile enabled derivation of a candidate post-infusion inflammatory risk score and a prototype predictive model of this risk score based on pt baseline characteristics was developed. These results may help identify pts at risk for severe/refractory iAEs, who may be potential candidates for closer monitoring or novel interventions, and warrant further study. Clinical trial information: NCT03361748, NCT03601078.