Juno Therapeutics, a pioneer in the sizzling field of treating cancer by revving up the immune system, on Thursday said it had halted development of its lead treatment after three patient deaths, dealing a blow to a promising but still unproven approach to oncology called CAR-T immunotherapy.

The treatment, dubbed JCAR015, is created by harvesting a patient’s own immune cells and rewiring them to home in on cancer in the blood. Three patients—all younger than 25—died after excess fluids accumulated in their brains, Juno said.

The first death came in May. Juno Therapeutics, which is based in Seattle, reported it to the Food and Drug Administration. Both the company and federal regulators agreed that it wasn’t clear what caused the death “and a change at that time was not warranted,” Juno CEO Hans Bishop said in a conference call.

But in the last week, two more patients died of the same issue, cerebral edemas.

The FDA then placed the study on clinical hold, meaning no new patients can be enrolled. The trial design called for 90 patients with advanced acute lymphoblastic leukemia to enroll; so far, just 20 have. Those 20 have already received their treatment, as  JCAR015, as a one-time infusion.

“Events like those in the last week are difficult and humbling for everyone involved,” Bishop said.

He acknowledged that the deaths “may lead you to question more broadly the safety of CAR-T cells.” However, he urged people not to rush to that conclusion, stating: “This is not the case.”

Juno executives blamed the deaths on their recent decision to add the chemotherapy fludarabine to the cocktail of drugs used to treat patients before administering JCAR015. The goal: To wipe out the patient’s existing immune cells, so that the genetically engineered immune cells can do their job with minimal interference.

In early safety trials, Juno used a single chemotherapy agent to do that job. But adding fludarabine in other immunotherapy trials improved outcomes with no additional risks, Bishop said, so the company decided to try it with JCAR015.

Chief Medical Officer Mark Gilbert said adult lymphocytic leukemia “may be a tougher disease to treat with a narrower therapeutic window.”

Juno has proposed to continue the trial without fludarabine and plans to submit supporting documents to the FDA this week, the company said—including updated consent forms for the patients being treated. The trial, designed to test JCAR015 in patients with advanced acute lymphoblastic leukemia, will remain on hold in the meantime. The delay could make it impossible for Juno to win approval to begin marketing the therapy next year, as the company once expected.

Juno still has a number of other CAR-T therapies in its pipeline, treating various forms of blood cancer. Bishop said Juno now expects “to be able to have the first approval in the US as early as 2018.” And the FDA plans on an expedited review process for the JCAR015 trial paperwork, he said. It normally takes about 30 days.

The news of patient deaths sent Juno’s share price down roughly 30 percent after hours on Thursday. Kite Pharma, a rival working with similar technology, fell about 14 percent. Both use modified T cells, called CAR-Ts, to treat cancer and have charted promising clinical results in blood malignancies.

Republished with permission from STAT. This article originally appeared on July 5, 2016.