On July 12, the US Food and Drug Administration (FDA) reinstated Juno Therapeutics' ROCKET phase 2 clinical trial of CAR-T cell therapy (JCAR015) for acute lymphocytic leukemia (ALL), just days after it had put it on clinical hold. Three patients out of between seven and ten in one arm of the trial died of cerebral edema within days of receiving the therapy. The arm at issue was testing a combination of preconditioning drugs commonly used in cell-based therapies to prepare the host for the incoming cells: the DNA-alkylating agent cyclophosphamide (the most commonly used one) and the purine analog Fludara (fludarabine). Juno had evidence (Blood 126, 3773, 2015) from previous trials that treating patients with such preconditioning drugs can increase the efficacy of CAR-T cell therapies, and no additional safety risks arose during a phase 1 trial in which ten patients were treated with JCAR014 and preconditioning agents at the same doses as in the ROCKET trial. But JCAR014 differs from JCAR015 in the co-stimulatory molecule used: JCAR014 contains the 41BB moiety, whereas the JCAR015 cell therapy has CD28. In a conference call with investors on July 7, Juno executives proposed that it was the combination of Fludara and CD28 that led to rapid amplification of the signal and neurotoxicity, which is a common side effect of CAR-T cell therapy, along with cytokine storms. They proposed to resume the trial using only cyclophosphamide, arguing that their program is an important treatment option for refractory ALL; in trials including more than 40 patients, they saw a 35–40% durable response (no deaths). The argument apparently held sway with the FDA, which lifted the hold only five days after imposing it; this is possibly the shortest clinical trial hold on record, according to BioMedTracker. Whether this change will reduce the efficacy of the therapy and what effect this series of events will have on the otherwise red hot field of immunotherapy remains to be seen. Needless to say, Juno's stock underwent a roller coaster ride as these events unfolded, plunging 31% when the hold was announced and bouncing back to not quite its original level when the hold was lifted.