What to do when the wrong end points are selected

Expert offers advice

It's not uncommon in research for the clinical trial to end without the expected positive outcomes, but with some positive results that were not an anticipated end point.

When this happens, what should sponsors and investigators do, especially when the findings suggest the study product is more effective than the study's end points show?

This question recently arose with a cancer vaccine study, says Philip M. Arlen, MD, director of Clinical Research Group at the National Cancer Institute's Laboratory of Tumor Immunology and Biology in Bethesda, MD.

Cancer vaccine trials use tumor shrinkage as study end points, Arlen says.

"We try to help people live longer, and the ultimate goal is survival," Arlen says. "The shrinkage of the tumor is a surrogate [end point]."

The theory is that if the treatment results in the tumor shrinking, then that will lead to the patient surviving longer with cancer, he explains.

"But what we're finding with cancer vaccines is that patients with the more advanced disease who are given the vaccine might not have actual tumor shrinkage," Arlen says.

Instead, investigators were surprised to find that patients were living longer after being given the vaccine, but this positive outcome was not picked up with the surrogate end point of tumor shrinkage, he adds.

In other words, the tumor response was different from the patient response.1

"Traditionally in cancer treatment we have been using drugs that are cytotoxic," Arlen says. "These chemotherapy drugs basically result in tumor shrinkage."

The drugs were very toxic to tumor cells, but also were toxic to other human cells, resulting in the side effects of gastrointestinal problems, including vomiting and diarrhea, Arlen says.

"But you don't see these side effects with the vaccine," he notes.

Arlen and other investigators looked at five different prostate cancer clinical studies that were conducted at different stages of the disease, including some phase II studies. The data suggest that survival may be changed with the vaccines.

"We're seeing a 20 to 30 percent prolongation of survival," Arlen says. "So if you expect an 18-month survival, then you'd have six months longer survival on average per cohort."

Investigators theorized that this phenomena was due to cancer vaccines initiating a dynamic immune process that was exploited in later therapies and that both radiation and some chemotherapy agents alter the tumor cells' phenotype, which makes them more susceptible to T-cell mediated killing.1

"When patients go from one therapy to another type of therapy, it's possible they will respond better to the second therapy because of the dynamic vaccine," Arlen explains. "And the patient may have a better response to the next therapy after the vaccine, and this could translate into survival."

The results showing the vaccine's positive impact on survival suggest that it would be worth the time and cost to enroll patients in a larger study that looks at survival end point, Arlen says.

Researchers are studying whether the cancer vaccines may be dynamic, so that even when you stop it there may be activity for years, Arlen says.

The vaccines don't act like drugs, but may be training the immune system to attack cancer, he notes.

"Maybe what we should be looking at is there isn't one primary end point," Arlen suggests. "Maybe there should be two end points built into the study."

One end point would show progression, and the other would show overall survival, he adds.

"Obviously, the progression end point is the one you'll reach much sooner," Arlen says. "And if you see that it is a surrogate to seeing overall survival, then you possibly could get approval for commercial use based on the surrogate."

But even when the progression end point doesn't show positive outcomes, the survival outcome, which is the one that is more important, may be reached, he says.

A chief problem with designing clinical trial studies this way has to do with the statistics, Arlen says.

"The current dogma is that there should be one end point to build the statistics and number of patients on," he says. "The second issue is when do you pick the second end point?"

For instance, an end point of survival requires a much longer time period to obtain, Arlen says.

"Because of the time frame, it requires a larger number of patients to be put on the study," he says. "The larger the group, the longer it takes to accrue patients and complete the study."

Often investigators and sponsors want answers quickly, so they select end points that may result in answers sooner, Arlen says.

The cancer vaccine trials' experience with having examined the wrong end points is not that unusual, and investigators should keep the phenomenon in mind when starting clinical trials, Arlen notes.

"Often times with [traditional] drugs, you will give it once, and then it's out of the system, so the benefit is only during the time you've given it," Arlen says.

"But with the vaccine and new therapies, you may see benefits continue well past the actual time the person is receiving the therapy," he adds.

Molecular-targeting therapies change the biology of the tumor. These therapies may have benefits from combining with another therapy or giving a more conventional or traditional therapy afterward. When this happens, investigators might see continued benefits from the first therapy, Arlen explains.

Investigators who study cancer might consider using survival time as an end point, but so should investigators studying infectious diseases such as hepatitis C and HIV, Arlen says.

"People are living longer with these diseases, although we may not be curing them," he says.

Survival as an end point should definitely be considered.

"The statistics could be arranged to incorporate the end points and still keep things relatively simple," Arlen says. "It may be complex with larger numbers of patients, but I think that's something that could be worked out."

Reference:

  1. Schlom J, et al. Cancer vaccines: moving beyond current paradigms. Clin Can Res. 2007;13(13):3776-3782.