Although initially viewed as one of the most promising cancer therapies, CAR-T has had a number of setbacks, the most prominent being patient deaths. Data released released by Novartis (NVS) regarding it own CAR-T therapy, CTL019, at the European Haematology Association meeting this week noted an emphasis on safety.

In March 2017, Juno Therapeutics (JUNO) announced it was shuttering its lead CAR-T oncology program, JCAR015, because of patient deaths. Down but not out, Juno came back with positive interim data from its other CAR-T program, JCAR017, at the ASCO meeting held in Chicago in June.

On May 31, Novartis released excellent data for CTL019 in patients with relapsed/refractory chronic lymphocytic leukemia (CLL) who had been on Imbruvica (ibrutinib) for at least six months, but were not in complete remission.

CAR-T stands for chimeric antigen receptor T-cells. CARs are proteins that let T-cells recognize a specific antigen on cancer cells. They are drawn from patients and engineered in the laboratory to identify the patient’s specific cancer, then infused back into the patient, where they then attack the cancer.

The problem with the therapy is cytokine release syndrome, which is a common and life-threatening adverse effect of CAR-T therapy. Also called an infusion reaction, it is an overreaction by the immune system that creates severe inflammation and often looks like severe allergic reactions.

Juno had 18 deaths in trials not related to disease progression. Fifteen of them were related in some way to CAR therapy. EP Vantage noted, "The toxicity of CAR-T therapies has been problematic ever since the first signs of truly robust activity were seen—it is known that cytokine release syndrome and neurotoxicities correlate with efficacy and tumor burden. However, nurses have become very skilled at dealing with these types of toxicities, and several important centers have developed detailed plans to minimize their effects. Nonetheless, the specific finding of cerebral edema posed very serious questions, and ultimately did for JCAR015, which until recently was Juno’s lead."

Similarly, Kite Pharma (KITE) had issues with neurotoxicity and cerebral edema. It reported nine deaths not due to disease progression, with four of them related to CAR. EP Vantage does say, "And of course it must be remembered that subjects on CAR-T studies are very ill, so are already at high risk of death. While this does not reduce the need for clinical rigor, several sources have commented to EP Vantage off the record that Juno might—scientifically if not reputationally—have been rash to discontinue JCAR015."

It’s also important to note that there is a fair amount of granularity involved in cause of death in these cases—cytokine release syndrome versus neurotoxicity versus brain edema versus dying from the cancers or other causes. Nonetheless, the EP Vantage analysis is reasonably thorough.

Novartis, on its part, did not report any new unexpected deaths, and no cerebral edema.

Seeking Alpha writes, "Meanwhile, it is one of the many mysteries of CAR-T therapy that there is still no agreement on what precise mechanism might be responsible for causing cerebral edema, though it is largely accepted that, being a cytokine-mediated event, it is something that arises from cytokine release syndrome."

Some researchers think it involves activated CAR-T cells crossing the blood-brain barrier, but others disagree. The U.S. Food and Drug Administration (FDA) is creating its own database of CAR-T toxicities. Seeking Alpha writes, "An even deeper mystery is what aspect of CAR-T therapy might give rise to this kind of stimulation. Some have suggested the role of the co-stimulatory domain that each construct uses, others the manufacturing process, others still the indication studied, but given that all the CAR constructs differ in multiple ways, and that the numbers are still small, there is no way to be sure."

Seeking Alpha also suggests that as the numbers and details emerge, investors should pay attention.

Also, CAR-T, which when successful, is dramatically successful, is also a very laborious and expensive process, largely because it’s uniquely developed for each individual patient. Juno’s trial of JCAR017 in B cell non-Hodgkin lymphoma had an overall response rate (ORR) of 86 percent and a complete response rate of 59 percent, or 26 of 44 patients. That’s pretty dazzling. But at this time, the success rates are going to compete with the risks and expense.