Journal of the National Cancer Institute Advance Access originally published online on July 10, 2007
JNCI Journal of the National Cancer Institute 2007 99(14):1068-1069; doi:10.1093/jnci/djm073
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© Oxford University Press 2007.
NEWS |
NOT GETTING WORSE?
More Clinical Cancer Treatments Judged by Progression-Free Rather Than Overall Survival
Cancer research aims to help patients live longer, healthier lives. But waiting for patients to die to determine how well a drug works tries the patience of doctors and researchers. Measuring the effectiveness of a drug by how long it keeps cancer at bay may offer a faster way to find good interventions, but such methods have their own problems.
Known as progression-free survival (PFS) or the related time to progression (TTP), these methods measure the effectiveness of drugs faster than waiting for study participants to die from their cancers. The U.S. Food and Drug Administration has held meetings to discuss the advantages and pitfalls of PFS (sometimes used as an umbrella term for both concepts). Researchers like PFS because studies can be shorter and smaller, making them potentially less costly. While the FDA appreciates the shorter time it takes to rule therapeutics in or out, it warns that researchers must be careful when using PFS.
"The FDA is demonstrating a flexibility to move towards PFS. We encourage researchers to have a discussion with us about using it as an endpoint," says medical oncologist Richard Pazdur, M.D., director of the FDA's office of oncology drug products. "But it won't replace overall survival. Overall survival is the ‘gold standard.’"
|
The difference between PFS and TTP is mainly in how deaths are tallied. PFS measures the time that patients live without their cancer getting worse, from some arbitrary starting point, often the date of the intervention. Since survival is measured (as the name suggests), deaths such as those from drug toxicity or unrelated accidents are included in the final count. TTP measures the time from the arbitrary start date until the patient progresses, but deaths are not included of the TTP equation. In either case, overall survival often is measured along with the PFS and TTP endpoints.
"I prefer PFS because it is a better surrogate for overall survival," says hematologist Kyle Robert, M.D., of the Mayo Clinic in Rochester, Minn. "If you have an effective treatment, but half the patients died because of the toxicity, then you could have a nice, long TTP, but half would be dead," Robert says.
Some researchers prefer the shorter endpoint over the long wait. "PFS is cleaner than overall survival," says statistician Karla Ballman, Ph.D., also of the Mayo Clinic. "If patients progress, they can go off the protocol and get treated by a doctor. If you have a survival endpoint, you don't know how people got treated outside of the trial, and other factors might contribute to their survival or death."
However, perhaps the biggest stumbling block is measuring success, says biostatistician Katherine Panageas, Dr.P.H., of Memorial Sloan-Kettering Cancer Center in New York. "People are much more subjective in assessing progression as opposed to death."
Cancer at Bay
When using PFS, oncologists will measure the size of a tumor or the presence of a marker at predetermined intervals. The physicians still follow certain criteria to measure progression of the cancer or tumor, such as Response Evaluation Criteria in Solid Tumors (RECIST). When criteria for progression have been met, such as a tumor having grown by a certain percentage, PFS is reported as the time that has passed rather than how much the tumor has grown in that time.
The biggest problem with PFS is that any interval can be chosen for measuring progression, Panageas says (J Natl Cancer Inst 2007;99:428–32). Deciding how often to look for progression can be problematic because of the nature of the disease. For example, a tumor might grow steadily larger over the course of a year. One oncologist decides to measure the size of the tumor every 8 weeks and another every 6 months. "If you look every 6 months, then the earliest documented progression can be 6 months. If you are looking every 8 weeks, then the earliest you can get progression is 8 weeks," Panageas says.
Also, tumors that grow at similar rates could result in drastically different PFS values depending on when researchers assess the tumor's status. "The reason you picked it up at 12 weeks is because you looked at 12 weeks," Panageas says. "There is no push for standardization for this endpoint." There are ways to address different measurement intervals statistically, but they are complicated. She advocates presenting the data as a range between the earliest possible PFS time detected to the longest, which would allow oncologists a wider view of what potentially is happening to the disease. Researchers could also standardize intervals across studies and present the data as a binary, yes–no option of whether progression has occurred.
Using PFS, especially if the endpoints are not the same, could make comparing different trials more complicated than using overall survival as an endpoint, says oncologist Ramzi Dagher, M.D., also at the FDA. "You have to consider it as a factor, like you would the population of the study," he says.
The Measurement Problem
The lack of standardization leads to what Pazdur calls the measurement problem. "Using PFS requires meticulous planning on what needs to be done on taking measurements before the trial begins," Pazdur says. Considerations include how often the measurements will be taken and how the data will be reviewed. For example, will the researchers convene an independent review panel that has no expectations of what they are supposed to see on an x-ray film? And if so, how will discrepancies that arise between study authors and independent radiologists be resolved?
"It's not automatic that in a specific cancer, using PFS always means you'll need less resources," Dagher says. It's not clear how changing the endpoint translates to cost, but one thing Pazdur is sure about: "Is it an easier endpoint? No."
Especially important is measuring cancer growth between the different arms of the study in the same manner. Checking control subjects every 8 weeks and assessing those taking the drug every 7 could lead to inaccurate estimations. "If there are unequal time periods, any benefit might be due to timing of measurements," Pazdur says.
These issues don't necessarily arise because someone is trying to purposely bias the work. Reality interferes with clinical cancer trials as easily as it does with daily life: Patients might not make it to their scheduled appointments, or multiple scans may need to be performed for unexpected reasons.
And sometimes masking the identity of case patients and controls from researchers is hard. The toxicity of a drug could reveal who's on what compound, for example. Ballman says that anecdotally, it seems like physicians who believe the experimental treatment is beneficial tend to identify progression earlier in the control arm than in the experimental arm. That way the patient can go off the standard treatment and switch to the other drug. Ballman says statisticians are pushing for central unbiased reviews of progression. "That would fix a lot of problems. But how do you do that in real time?" Ballman asks.
Pazdur says that considering the magnitude of the increase in PFS is also important in determining whether the therapeutic is of direct clinical benefit. "A median improvement of 2 weeks may be a statistically significant finding, but you have to ask if it's of clinical significance." If a compound makes people ill, and they have to take it for a week to keep cancer at bay for 2 weeks, then that PFS would not be as important a result as one in which a week's nausea translated into 3 months of healthy living.
Works for Some, Not for All
Experts say that PFS is a better way to measure effectiveness in some cancers but not others. "Multiple myeloma is a nice example. Even though these patients progress and relapse, they may live for many years. To say we are going to rely on a survival endpoint means that the trial could go on for many, many years," Dagher says. "It doesn't mean we can't ask for survival data to be submitted later when it's collected." Pancreatic cancer, on the other hand, tends to progress rapidly. Then PFS wouldn't offer such clear advantages.
Another question is whether PFS can even serve as a good proxy for overall survival. "With ovarian cancer trials, there's a good correlation with those interventions that extend PFS and overall survival," says ovarian oncologist Robert Bast, M.D., of the University of Texas M. D. Anderson Cancer Center in Houston, of including both endpoints in studies. Panageas points out that the question hasn't been strictly evaluated in many cases.
The FDA, however, spent several years compiling data and expert opinions on endpoints for many different cancers (see http://www.fda.gov/cder/drug/cancer_endpoints) and released general guidelines in mid-May called Clinical Trial Endpoints for the Approval of Cancer Drugs and Biologics. How well PFS and overall survival line up depends on the kind of cancer, Ramzi says. In a 2006 study, Ballman looked at PFS at 6 months and overall survival at 12 months in brain cancer patients and found a strong association between the two endpoints, for example.
Pazdur worries that too many trials using PFS might shrink the total number of research subjects over time, which would reduce statistical significance for survival data. "If we focus on PFS, we will get smaller and smaller trials that can't demonstrate any survival effect," he says. "That is a controversial area. Is PFS a surrogate for survival?" Oncologists are usually interested in multiple endpoints and are collecting that data, Ramzi says, so ongoing studies will no doubt reveal that answer. "From a statistical standpoint, you have to pick one as the primary endpoint," he says. It looks like choosers can be picky.
Related Article in JNCI
CiteULike 
Connotea 
Del.icio.us What's this?
J Natl Cancer Inst 2007 99: 428-432.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||