Not Specific Enough
Breaking data down by individual solid tumor type may not even be enough to truly demonstrate validity of surrogate endpoints. It may be necessary to break down data to disease subtype.
As Dr Gyawali and colleagues pointed out, there was a strong correlation between treatment effects in disease-free survival and that in overall survival for HER2-positive early stage breast cancer.
“There are multiple subtypes of breast cancer, and when we lump them all together and try to show benefit for a single therapy, sometimes you can’t make that distinction,” said Douglas Yee, MD, of the Masonic Cancer Center at the University of Minnesota in Minneapolis.
As an example, the I-SPY 2 trial used pathologic complete response as a primary endpoint to identify investigational drugs that would improve outcomes in women with stage II/III breast cancer.6 The study showed that among all women, regardless of treatment arm, pathologic complete response was highly associated with 3-year event-free survival. The rate of pathologic complete response varied by breast cancer subtype, with the highest rates seen in women with HER2-positive disease.
This trial also excludes hormone receptor-positive tumors with genomic low-risk scores as determined by the 70-gene assay (MammaPrint), Dr Yee pointed out. Patients with “clinical high-risk, genomic low-risk” tumors do not benefit from adjuvant chemotherapy when compared with endocrine therapy alone. Thus, their inclusion in previous neoadjuvant trials further weakens the association between pathologic complete response and long-term benefit.
A study by Aditya Bardia, MD, of Massachusetts General Hospital, Boston, and colleagues examining outcomes in stage II/III breast cancer also demonstrated that pathologic complete response was strongly associated with improved event-free survival and overall survival, supporting “the FDA’s decision to use pCR rate as a surrogate marker of efficacy from neoadjuvant treatment.”7 This was particularly true for patients with triple-negative breast cancer and HER2-positive breast cancer, according to the study. The analysis included studies with patient-level data measuring the association between pathologic complete response and overall or event-free survival.
That demonstrates the difference between looking at trial-level data and patient-level data, according to Dr Yee.
“If you are a patient on a neoadjuvant trial and your tumor disappeared, no matter what kind of tumor it is, that leads to improvement in event-free and overall survival,” Dr Yee said.
Dr Gyawali agreed. “That is where many of us get confused between prognostic marker for a patient and surrogate endpoint for trials,” he said. “For example, pathologic complete response can be a prognostic marker, meaning that a patient with pathologic complete response may live longer than a patient who did not achieve pathologic complete response — but whether a drug that improves pathologic complete response also improves overall survival is a different question altogether. An endpoint can be considered a surrogate for trials only if treatment effects on the endpoint correlated with treatment effects on overall survival.”
Application of surrogate endpoints to data in metastatic breast cancer can also present challenges, according to Dr Bardia. With the Accelerated Approval Program, drug companies are required to conduct confirmatory studies to confirm clinical benefit.
“Crossover has a major impact on overall survival. In other words, if the trial permits [that] patients could cross over to the other arm at time of disease progression, or if a drug gets approved based on progression-free survival in the metastatic setting, patients in the control arm have access to that investigational agent, which could impact overall survival,” Dr Bardia explained.
That is why many therapies considered blockbuster drugs have not shown improvement in overall survival, he said. Poly-(ADP)-ribose polymerase (PARP) inhibitors are an example, he noted.
“PARP inhibitors have a major impact on patients with metastatic breast cancer who have germline BRCA mutations, but we have not seen an overall survival benefit,” Dr Bardia said.
In contrast, Dr Gyawali suggests that if these drugs did not improve overall survival, perhaps they were not “blockbuster” drugs at all.
“I acknowledge the problem of confounding effect on overall survival due to crossover, and therefore, we propose in our paper that surrogacy probably differs by lines of treatment and is not ‘mechanism-agnostic’ as the current FDA Table proposes,” Dr Gyawali said. “Also, if the overall survival effect of a drug is masked due to crossover, it would mean that the drugs could have been given in later lines of therapy and produce the same effect.”
The second thing to consider is whether the study was powered to look at overall survival in the first place.
“Many of the studies in the metastatic setting are powered for progression-free survival and not overall survival, which can impact the ability to detect a significant overall survival benefit,” Dr Bardia said.