SAN FRANCISCO — Pancreatic cancer remains a highly fatal malignancy, and over the past three decades, improvement in survival has been modest.
One of the reasons for the lack of new therapeutic agents for this disease may have much to do with the clinical trial process, according to new findings presented here at the Gastrointestinal Cancers Symposium (GICS) 2017.
Many new agents show promise in phase 2 studies, but the promise all too often fizzles out during phase 3 trials. And that's if they even make it to the phase 3 level.
Despite a great number of phase 2 studies, only 15% have gone on to phase 3 for the treatment of locally advanced and metastatic pancreatic cancer, explained lead author Monica Tang, MD, from the Prince of Wales Hospital, Sydney, Australia.
A total of 148 phase 2 clinical trials for first-line systemic treatment in metastatic pancreatic cancer were conducted and published between 1978 and 2015, she reported at the meeting.
Of these trials, only 11% were for new investigational agents.
The results were a little disconcerting.
"These trials do not conform well with recommendations from the National Cancer Institute (NCI) for pilot trials in pancreatic cancer," said Dr Tang. "There was limited reporting of prognostic factors and substantial heterogeneity in the characteristics that were well reported."
Few of the trials of biological agents were enriched for biomarkers, and there was poor reporting of statistical methods, she explained.
"Achievement of statistical target effect size did not consistently determine whether or not a trial was considered successful, or whether it proceeded to phase 3 testing," Dr Tang said. "These findings may partly explain the limited success of clinical trials in advanced metastatic pancreatic cancer, and they add weight to the NCI recommendations."
Understand the Past, Improve the Future
"In order to improve clinical trials of the future, it's important to understand and learn from clinical trials of the past," Dr Tang commented.
For the initial part of their study, the authors conducted a systematic review of phase 3 trials in advanced metastatic pancreatic cancer and evaluated 32 trials with a total cohort of 13,675 patients.
These trials resulted in three clinically meaningful regimens, and an analysis revealed that a benchmark of 50% improvement in overall survival in phase 3 trials accurately predicted clinically meaningful phase 3 trials in 78% of cases.
The NCI Trials Planning Meeting on Pancreas Cancer Treatment has published comprehensive recommendations for pilot trial design. Some of these empirical recommendations include restricting trials to patients with good performance status, the use of predictive biomarkers for enrichment, and using survival as the preferred primary endpoint.
For transitioning to phase 3, the NCI recommends considering input from multiple studies and implementing phase 3 testing only after pilot data show robust signals.
The next step was to review and evaluate the quality of phase 2 trials and determine the correlation between outcomes of phase 2 and 3 trials for investigational agents.
A total of 148 studies with a cohort of 7505 patients were included in their final analysis.
In a breakdown of the trial design, the authors noted that 25 (16.9%) had multiple arms and that reporting of prognostic factors was limited. Of the 37 (25%) trials that tested biological agents, only 1 was enriched for biomarkers and 7 included biomarkers as exploratory or a translational endpoint.
There was also significant heterogeneity in reported baseline factors: The proportion of patients with locally advanced disease ranged from 0% to 80% among trials, and the proportion of patients with European Clinical Oncology Group status of 0 or 1 ranged from 14% to 100%.
The authors also analyzed the reporting of statistical methods and outcomes. The primary endpoint was defined in more than two thirds (68.9%) of trials, but the most common one was objective tumor response (41.2%), followed by progression-free survival (15.5%).
Only 10% used overall survival as a primary endpoint of the trial.
The sample size calculation and statistical assumption were reported in 63.5% of trials, and trials conducted after 2000 were more likely to publish their statistical methods (78.7%).
Of the 148 trials, noted Dr Tang, more than half (55.4%) were reported as successful by the investigators.
"But only 26% actually achieved their target effect size for a specified endpoint and even fewer, 15%, proceeded to phase 3 testing, which is arguably the most demonstrable evidence of success in phase 2 trials," she said.
The researchers also reported on 27 investigational agents that were tested in both phase 2 and 3 trials, and they found that phase 3 results tended to be less positive. There was a weak correlation for objective tumor response and progression-free survival, as well as a moderate correlation for overall survival.
"Overall survival had the strongest correlation between phase 2 and phase 3 trials," Dr Tang pointed out.
Make Juice Worth the Squeeze?
In a discussion of the paper, Vincent Picozzi, MD, director of the Pancreas Cancer Center of Excellence at Virginia Mason Cancer and Digestive Diseases Institute in Seattle, Washington, noted that this is a topic that "seems to generate more frustration than fruit, so I've titled my remarks as 'How Do We Make the Juice Worth the Squeeze?'"
He pointed out that when reviewing the general experience in drug development, and looking at all indications in oncology, the probability of a phase 2 trial leading to drug licensure is 16% or 1 in 6.
"But for pancreatic cancer, that rate is far lower, only about 3% or 1 in 30, or 5 times less than the oncologic average," he said. "Dr. Tang's abstract gives us some clues as to why that might be."
These "clues" may lie in the disappointing correlation between phase 2 and phase 3 trials, and also the limited use of biomarker strategies, Dr Picozzi explained. In addition, patient prognostic factors are heterogeneous, and there was suboptimal selection of primary endpoints.
Of note, the achievement of statistical target size did not correlate with the reported success or progression to phase 3 evaluation.
"Overall this is a fairly grim report, but I think things are worse than that," he emphasized.
For starters, even though biomarker strategies lead to more successful phase 3 outcomes, "a single biomarker effect is difficult to interpret given the complex pancreatic cancer biology," Dr Picozzi said.
Patient selection is also complex, as pancreatic cancer involves a very heterogenous group of people who are influenced by disease expression, psychosocial factors, comorbidities, and variations in patient and supportive care — all of which add to dimensions of heterogeneity beyond the simple biology itself.
"Overall survival correlated best with paired phase 2/3 results, but the statistical bar should be set much higher — such as a 50% overall survival improvement, or 80% to 100% progression-free survival improvement," Dr. Picozzi noted.
And finally, interpretation of results can be "overly optimistic, and often at variance with proof of concept." The decision to proceed to phase 3 could also be affected by other factors, he pointed out, such as financial concerns, priorities, or intellectual properties or even legal constraints.
"This abstract was selected not only because of its own merit but because it is reflective of a year in which there were a number of promising and highly touted trials in pancreatic cancer which unfortunately failed," he added.
No funding source was disclosed. Dr Tang reports expenses from Pfizer; the other authors have disclosed no relevant financial relationships. Dr Picozzi reports relationships with AbbVie, Aduro Health, Amgen, Celgene, Clovis Oncology, FibroGen, Halozyme, Immunomedics, Incyte, Johnson & Johnson, OncoMed, Taiho, and Theranostics Health.
Gastrointestinal Cancers Symposium (GICS) 2017. Abstract 227. Presented January 20, 2017.