To speed completion of cancer clinical trials, drug companies and regulators now often rely on surrogate measures of effectiveness, such as tumor shrinkage or delay of tumor growth, rather than waiting for direct evidence of improved survival or quality of life. But a study by researchers at OHSU in Portland, Oregon, raises new questions about the shortcut measure. The researchers found that using less-reliable surrogate endpoints only shortens clinical trial completion by about 11 months.
“That is not a long time to wait for more definite data,” said first author Emerson Chen, M.D., a hematology and medical oncology fellow in the OHSU Knight Cancer Institute, OHSU School of Medicine. To put the number in context, Chen and co-authors said 11 months is 12 percent of the average development time of current cancer drugs, which is 7.3 years.
“Whether this tradeoff of faster but less reliable results is worthwhile for patients deserves further scrutiny,” Chen said. The paper was published in JAMA Internal Medicine on April 1 with senior author Vinay Prasad, M.D., M.P.H., an associate professor of medicine (hematology and medical oncology) in the OHSU School of Medicine, and medical student co-authors Sunil Joshi and Audrey Tran.
“We were trying to document what is the tradeoff between speed and uncertainty with surrogate endpoints,” said Prasad. “We found that surrogates are associated with a lot of uncertainty and only a little bit of speed.”
A majority of new cancer drugs in recent years have gained approval by the U.S. Food and Drug Administration based on a surrogate endpoint clinical trial. This shortcut strategy makes sense if the surrogate reliably predicts improvements in survival or quality of life. It’s also reasonable in rare disease situations with no existing treatment options.
But surrogate endpoints do not reliably predict improvements in survival or quality of life. And unacceptable safety issues may be more likely to go undiscovered in clinical trials using surrogate endpoints. Such trials tend to enroll fewer patients and monitor them for a shorter time than trials designed to measure overall survival.
The drug bevacizumab (Avastin) illustrates the problem. It gained an accelerated approval for metastatic breast cancer in 2008 based on dramatic gains in progression-free survival, a widely used surrogate endpoint that measures how long before a cancer resumes its progression (or until death, should a trial participant not survive). In subsequent randomized trials, bevacizumab did not help breast cancer patients live longer even as it caused a range of severe side effects, some of them life-threatening. The FDA withdrew its approval for the breast cancer indication in 2011.
In the new study, Chen and co-authors analyzed 107 cancer drugs approved for 188 indications from 2006 to 2017. They found that 38 percent of these cancer drugs were approved based on the surrogate endpoint of response rate, which is the percentage of patients whose cancer shrinks. Another 34 percent were approved based on progression-free survival. The remaining 28 percent were approved based on more direct measures of benefit: overall survival or patient-reported quality of life.
The authors asked how much time is saved by the use of progression-free survival or response rate in clinical trials. They used statistical models to answer the question, taking into account how quickly different trials recruited participants, how late in the disease course subjects entered the study, and other factors affecting the speed of trial completion.
If progression-free survival were used instead of overall survival, Chen and colleagues estimate that study duration would be shorter by an average of 11 months. If response rate were used, they estimate study duration would be shorter by 19 months
If only clinical endpoints like overall survival were used, they estimate that 38 percent of trials would be 19 months slower, 34 percent would be 11 months slower and 28 percent would be unchanged. Combining these calculations, the median time to develop a modern cancer drug using only definite endpoints would be associated with an increased duration of 11 months and extend average drug development time from 7.3 years to 8.2 years.
This research was funded by the Oregon Clinical and Translational Research Institute (grant 1 UL1 RR024140).