For doctors at the University of California, San Diego, it was seemingly a no-lose proposition: A 73-year-old patient’s bladder cancer was slowly progressing but he was generally stable and strong.

He seemed like the ideal candidate for an immunotherapy drug, atezolizumab, or Tecentriq, that had just been approved to treat bladder cancer patients.

Doctors started the patient on the drug in June. It was a spectacular failure: Within six weeks, he was removed from the drug, and he died two months later.

In a troubling phenomenon that researchers have observed in a number of cases recently, the treatment appeared not only to fail to thwart the man’s cancer, but to unleash its full fury. It seemed to make the tumor grow faster.

The patient’s case was one of a handful described last week in the journal Clinical Cancer Research. Of the 155 cases studied, eight patients who had been fairly stable before immunotherapy treatment declined rapidly, failing the therapy within two months.

Six saw their tumors enter a hyperactive phase, where the tumors grew by between 53 percent and 258 percent.

“There’s some phenomenon here that seems to be true, and I think we cannot just give this therapy randomly to the patient,” the author of the study, Dr. Shumei Kato, an oncologist at UC San Diego, said in an interview with STAT. “We need to select who’s going to be on it.”

Kato’s article, on which the noted cancer researcher Dr. Razelle Kurzrock is named as a principal investigator, strains to make the point that the findings are more suggestive than conclusive, and that “further investigation is urgently needed.”

But similar findings were published last year by cancer researchers at the Gustave Roussy Institute in France. These results were considered controversial by some, since they hadn’t been widely confirmed by other oncologists.

Some remain unconvinced, including Dr. Vinay Prasad, assistant professor in the Division of Hematology Oncology at Oregon Health and Science University.

“Tumor growth is not a precise measurement, and if you measure lots of people, some will have faster growth just because of the error in the test,” he said, adding that Kato’s research does not determine whether the growth happens “beyond the chance rate.”

“At the end of the day, while I find this interesting, I think the main point is if you use drugs where” randomized control trials “show benefit, you will be good,” Prasad said. But immunotherapies are often approved based on early results, without randomized control trials. Given that, “don’t be surprised if you harm patients. True for all drugs. Even, almost surely, immunotherapy,” Prasad said.

Dr. Antoni Ribas, who researches immunotherapies at the Ronald Reagan UCLA Medical Center, said these reports should not change patients’ and doctors’ approach to immunotherapy.

“The potential benefit of immunotherapy far outweighs the worry by miles at this time, so nobody should give up that option because we now have anecdotal evidence that they may have an adverse effect the tumors,” he said. “I can always be corrected by data, but this is something to be explored further.”

The observations are rippling through an oncology community that is trying to better understand the limitations of immunotherapy, a much-vaunted therapy that seeks to unleash the immune system against cancer.

In the latest Clinical Cancer Research findings, those who experienced the hyperprogression of tumors, as the phenomenon is known, shared specific genetic characteristics. In all six patients with so-called amplifications in the MDM2 gene family, and two of 10 patients with alterations in the EGFR gene, the anti-PD-1 or anti-PD-L1 immunotherapies quickly failed, and the patients’ cancers progressed rapidly.

Aside from atezolizumab, immunotherapies in this class include avelumab (Bavencio); pembrolizumab (Keytruda); and nivolumab (Opdivo). The other major class of immunotherapies are known as anti-CTLA-4 treatments, such as ipilimumab (Yervoy), which target a different mechanism to unleash immune cells to fight tumors.

Doctors who prescribe immunotherapies may be able to identify at-risk patients by submitting tumors for genetic testing, Kato and his coauthors suggested.

The findings published last year by the Gustave Roussy team also appeared in Clinical Cancer Research. In that study, of 131 patients, 12 patients, or 9 percent, showed hyperprogressive growth after taking anti-PD-1 or anti-PD-L1 immunotherapies.

The lead author of that study, Stephane Champiat, acknowledged that the research so far raises more questions than it answers. Prior to the latest publication from Kato, Champiat said he was unsure if the phenomenon is related to the immunotherapy drugs.

The new study, he said, “makes me more confident.”

Champiat suggested factors that could be associated with the effect. In his study’s patients, for instance, those who were older than 65 showed hyperprogressive growth at twice the rate of younger patients.

“Is it specific to older patients? I don’t think so. Do they have higher risk? Maybe,” Champiat said. “And I think it’s probably different from one tumor type to another tumor type.”

Oncologists studying this phenomenon said it could complicate treatment strategies, because some patients who receive immunotherapies can exhibit what’s known as “pseudo-progression,” in which tumor scans reveal apparent growth. In reality, however, the scans are instead showing areas where the cancer is being attacked by armies of immune cells.

Roughly 10 percent of melanoma patients on immunotherapies, for instance, experience this phenomenon.

Champiat said that he and his colleagues had mistakenly believed that patients whose cancer had entered a hyper-growth mode were merely experiencing pseudo-progression, so they continued the immunotherapy treatment.

“We were probably deleterious because we maintained that,” he said.

That’s also what Kato and his colleagues were hoping, as their patients’ cancer appeared to grow in their first weeks on immunotherapy.

The findings, if they continue to receive validation across the oncology community, could offer more reason for caution among cancer doctors who have witnessed exceptionally strong results from immunotherapies.

Jimmy Carter is perhaps the best-known immunotherapy success story.

But most patients do not respond to the immunotherapy treatments, for reasons that remain largely unknown.

In a study by Prasad and Dr. Nathan Gay, also of Oregon Health and Science University, nearly 70 percent of Americans die from forms of cancer for which there is no immunotherapy option, and for the rest who do qualify for immunotherapy, only 26 percent actually see their tumors shrink.

Read more: Few people actually benefit from ‘breakthrough’ cancer immunotherapy

And while immunotherapies typically include less intrusive side effects than chemotherapy, those side effects, when they happen, can be life-threatening.

Researchers have reported cases in which immunotherapies attacked vital organs, including the colon, liver, lungs, kidney, and pancreas, with some patients experiencing acute, rapid-onset diabetes after receiving the treatments.

But in those cases, the treatments were at least attacking the cancer. Such reports didn’t raise the specter of these treatments possibly working on the cancer’s behalf to shift it into overdrive.

“I’m a little bit nervous, to be honest with you, reporting this,” Kato said.

Dr. Jennifer S. Temel, clinical director of thoracic oncology at Massachusetts General Hospital, alluded to the emerging research on immunotherapy in a presentation to palliative care and hospice clinicians in February.

“I’m treating 30 patients with immunotherapy, and when it works, it is so awesome,” she said. But the new data showing the drugs might accelerate cancer, she said, has introduced a new set of worries for patients and doctors.

“How are we doing to deal with that?” she asked. “We just don’t know yet. But I am scared.”

“The challenge for oncologists,” Temel added in an email Monday, “is to balance our excitement and enthusiasm with this reality and be sure to communicate honestly and effectively with our patients.”

Read more: For cancer patients, newest treatments force the ultimate decision, with no room for error

Kato said the patient whose bladder cancer grew out of control after receiving atezolizumab was “very intelligent, very knowledgable.”

The man’s wife accompanied him to appointments, and together they decided to move ahead with the treatment. Within weeks he grew more fatigued and his sodium levels plummeted. When they scanned his tumors, they had grown by 258 percent.

“He was angry,” Kato said. “At me, at first, and to the team in general.”

“It was very, very hard,” he said. “We get very encouraged by good outcomes, and we struggle to try to make the same result, but unfortunately it’s not always the case.”

Republished with permission from STAT. This article originally appeared on April 3, 2017