Clinical Trials, Predictive Factors, and Endpoints: Current and Future Challenges

Oct 29, 2012


Hope S. Rugo, MD
University of California, San Francisco, Helen Diller Family Comprehensive Cancer Center

In the article that follows, Dr. Anthony Provenzano discusses some of the more challenging problems facing clinical research in this decade. An overarching theme is the inexact process of drug evaluation and approval, which is closely associated with a lack of predictive factors that would help truly individualize cancer care. Indeed, recent and ongoing studies have highlighted the importance of treatment in the neoadjuvant setting as a pathway to drug approval, a concept now embraced by regulatory agencies. Using a backbone of standard chemotherapy for breast cancer, the ISPY2 multicenter trial is testing the addition of a series of novel agents to paclitaxel with serial biopsies and assessment of pathologic response at the time of surgery. This approach offers the potential to determine predictive biomarkers in real time, without starting with predetermined single (and therefore more limited) markers. ISPY2 is the largest of many trials utilizing this general approach and has taken the novel approach of partnering with pharmaceutical companies through the Foundation of the National Institute of Health.

Along with new and hopefully more efficient mechanisms of evaluating novel therapeutics, it is critical to involve the patient as a partner in decision making and provide the tools to navigate increasingly complex therapeutic options. Each treatment decision is associated with toxicity and cost; goals of therapy need to be clearly defined as each new therapy adds complexity to therapy. Indeed, we need to hold drug development to a high standard, including an assessment of efficacy that is meaningful to those suffering from the disease under study, and a careful and prioritized exploration for predictive biomarkers.

Looking forward, it does indeed appear that oncology requires more of a Renaissance approach—the simple country doctor just won’t do.

Dr. Rugo is a Professor of Medicine and Director of Breast Oncology and Clinical Trials Education at the University of California, San Francisco, Helen Diller Family Comprehensive Cancer Center. She serves on the ASCO Connection Editorial Board and is an Associate Editor for Cancer.Net, ASCO’s patient information website.

The Cross Roads of Clinical Trials, The Promise of Personalized Medicine, and Surrogate Endpoints: “To Boldly Go Where No Man Has Gone Before...”

Anthony F. Provenzano, MD
Lawrence Medical Associates, P.C., Lawrence Hospital Center

We are at the crossroads where the clinical trial process, the promise of personalized medicine, and surrogate endpoints (perceived clinical benefit) have finally intersected. All three must be reconciled so they can be translated into true clinical benefit. Furthermore, these algorithms must coexist within an economic environment that will allow for the continual and efficient allocation of research funding. These goals can be further realized through the strategic collaboration of scientific cross-disciplines, such as biomedical engineering, computer science, physics, and other scientific disciplines that have contributed recently to improvements in diagnosis and treatment of cancer.1

Clinicaltrial process
Only one in every 10 new molecular therapeutic agents that enters clinical development receives U.S. FDA approval.2 Phase I/II/III models have been used to test drugs commonly in the metastatic setting before they are used in operable disease. In the era of personalized medicine, we may need criteria other than Response Evaluation Criteria in Solid Tumors (RECIST) to assess true drug benefit in both the primary (neoadjuvant and adjuvant) and metastatic setting.

In the preoperative setting, the use of phase 0 (window of opportunity)trials can help to facilitate the drug development process and help target select patients in subsequent clinical trials.3 We are still left with the problem of how best to validate markers and assess response criteria in an integrated fashion. This is where the aforementioned collaborative approach could expedite the process, perhaps with the aid of complex computer models that could predict how these different parameters should be weighted.

The promise of personalized medicine
Marker identification and response do not guarantee true clinical benefit, even though RECIST criteria may be losing ground to other surrogates such as metabolic imaging. Nonetheless, the holy grail of drug assessment will be the discovery of a valid subset of markers at each stage of tumor “evolution” whose phase 0response to a cocktail of drugs will be not only predictive but prognostic. We are not even remotely close to this realization (except, of course, in Star Trek reruns with Mr. Spock and Dr. McCoy in command).

Surrogate endpoints
A surrogate endpoint can be either an intermediate clinical endpoint or a biomarker.4 Two conditions must be fulfilled to qualify as an intermediate endpoint: strong association between the surrogate and true endpoint from a patient perspective, and hazard ratios of the effect of treatment on the surrogate and true endpoint must be comparable.

As oncologists, we must be able to translate for our patients the meaning of true clinical benefit as a balance between patient- and tumor-centered outcomes.5 Overall survival (OS) and quality of life(QOL) are endpoints suitable for patients with either early or advanced disease. Progression-free survival (PFS) and time to progression (TTP) are endpoints used for patients with advanced disease. PFS measures the time from a patient’s random assignment to a treatment until either the disease progresses or the patient dies from his or her cancer. OS is the time from random assignment to death from any cause. PFS measures the effect of only one drug until disease progression. OS is diluted by subsequent treatments. PFS captures tumor shrinkage or stabilization.

Surrogate endpoints bring ambiguity. Statisticians, for instance, censor deaths when calculating TTP but count them as events for PFS. In addition, some investigators use these two surrogates interchangeably when they are clearly different.

Progression-free survival, examined
We measure PFS by new lesions, increased size of the target measurable lesions by RECIST, clear increase in nontarget disease, and increase in signs or symptoms of disease without quantitative change in measurements. The advantages of PFS as an endpoint are shorter duration of studies, the effect of only the study drug is measured, subsequent treatment has less effect on the outcome, and it is advantageous for phase II studies. The disadvantages are more measurement errors, the frequency of assessments may introduce bias, and it is more difficult to measure or reproduce measurements.5 PFS is defined as time to first progression or death from any cause if disease progression did not occur, and cannot evaluate treatment strategy beyond firstprogression.4

Few studies have found associations between PFS and QOL.6-8 Panitumumab in metastatic colorectal cancer and sunitinib in metastatic renal cell carcinoma are two of them. Initial approval and later revocation by the FDA of bevacizumab in the treatment of metastatic breast cancer were based upon smaller increase in PFS and a higher incidence of adverse events without improvement in OS or QOL.9,10 Is the increased PFS with maintenance bevacizumab in ovarian cancer worth it, since patients are asymptomatic and being followed with serial CTs and CA125s?11 What if the duration of therapy is so long to achieve that small PFS benefit but is associated with a longer period of side effects?12 When is watchful waiting better than maintenance therapy?13 Patients are fearful of watchful waiting but, on the other hand, do we really spend enough time learning from patient-reported outcomes?14

We know that PFS is a surrogate for OS in metastatic colorectal cancer but not for many other cancers.15 Insurers know this as well. Will economic issues drive not only the approval process but also the discontinuation process for expensive drugs?

Future considerations
PFS is subjective, and other biologic surrogates not confounded by how progression is defined and measured are needed.16 During stable disease, for example, how does the tumor microenvironment change and what effect do stem cells have on this complex entity we call “stable disease”? These interactions with “measurable” disease are not assessed by RECIST.

We should probably coin a new acronym: TWOTRT (time without treatment-related toxicity). Intraperitoneal (IP) chemotherapy for ovarian cancer does not lead to significant increase in PFS or OS, and yet neurotoxicity can last for more than 12 months beyond the end oftreatment.17 When patients are asked about acceptable minimal survival benefits, 46% reported greater than 12 months, 17% reported 10 to 12months, and 10% reported 1 to 2 months.18

How can we improve the correlation of true clinical benefit with our perceived notions of benefit based upon surrogacy?

  1. We need ways to better predict projected life span before a diagnosis of cancer is made.
  2. We need better ways to predict noncancer-related mortality after a diagnosis of cancer is made.
  3. We need better ways to standardize QOL assessments before and after diagnosis.
  4. New surrogate endpoints are required in the rapidly changing world of genomic (personalized) medicine—even though we are far from realizing that Spockian achievement.
  5. Embedded in surrogates such as PFS and OS are inherent biologic variables that defy a way to achieve accurate measurements.
  6. A new collaboration with basic and applied science across disciplines is necessary to expedite the discovery process and make it more efficient and cost effective. We must be able to answer our patients’ simple questions, such as, “How long do I have to live?”, “Will this treatment help me?”, and “How will this medicine make me feel? Is it worth it?” We must answer them without having the patient resort to a medical dictionary or cling to farfetched hopes that seem more connected to science fiction than reality.

The fictitious Dr. McCoy once said (and I paraphrase):“I’m a doctor—not a moon shuttle conductor, bricklayer, psychiatrist, mechanic, engineer, scientist, physicist, magician, miracle worker, flesh peddler, or veterinarian.” Contrary to this sentiment, I believe we sometimes must wear other hats in order to make progress.

Dr. Provenzano is the Chief of Medical Oncology and Director of Oncology at Lawrence Medical Associates, P. C., and Lawrence Hospital Center in New York, and a Clinical Assistant Professor at New York Medical College. He currently serves on the Editorial Board ofCancer.Net, ASCO’s patient information website.

Technical assistance provided by Henry H. Cheung.

1. Provenzano AF. Simone’s OncOpinion on Clinical Trials(12/25/08). Oncology Times.2009:31:3-4.
2. Kola I, Landis J. Nat Rev Drug Disc.2004;3:711-5. PMID: 15286737.
3. Kalinsky K, Hershman DL. J Clin Oncol.2012;30:2573-5. PMID:22565006.
4. Chibaudel B, Bonnetain F, Shi Q, et al. J Clin Oncol.2011;29:4199-204. PMID: 21969501.
5. Fallowfield LJ, Fleissig A. Nat Rev Clin Oncol.2011;9:41-7. PMID:22009075.
6. Siena S, Peters M, Van Cutsem E, et al. Br J Cancer.2007;97:1469-74. PMID: 18040272.
7. Odom D, Barber B, Bennett L, et al. Int J Colorectal Dis.2011;26:173-81. PMID: 21190026.
8. Motzer RJ, Hutson TE, Tomczak P, et al. N Engl J Med.2007;356:115-24. PMID: 17215529.
9. Miller K, Wang M, Gralow J, et al. N Engl J Med.2007;357:2666-76.PMID: 18160686.
10. Jones A, Ellis P. BMJ.2011;343:d4946. PMID: 21816748.
11. Hensley ML. J Clin Oncol.2011;29:1230-2. PMID: 21383287.
12. Ozols RF. J Clin Oncol.2003;21:2451-3. PMID: 12829660.
13. Ciuleanu T, Brodowicz T, Zielinski C, et al. Lancet.2009;374:1432-40. PMID: 19767093.
14. Basch E. Ann Oncol.2009;20:1905-6. PMID: 19934250.
15. Minasian LM, O’Mara AM, Reeve BB,et al. JClin Oncol. 2007;25:5128-32.PMID: 17991932.
16. Mayfield E. Life Raft AccessedAugust14, 2012.
17. Wenzel LB, Huang HQ, Armstrong DK,et al.J Clin Oncol. 2007;25:437-43.PMID: 17264340.
18. Sheik-Yousouf A, Gandhi S, Dukhovny S, et al. Eur J Cancer Suppl.2010;8:77(abstr 63).


Christopher G. Azzoli, MD

Dec, 03 2012 8:18 AM

Beam me up, Tony!  I am with you.  I agree completely that if we are going to harness new technology in a timely fashion, we need to reinvent the way we practice oncology, break with traditional clinical trial designs, and bend efficacy endpoints.  Solid tumor oncology was traditionally a surgeon's disease, defined by histology, stage, and the possibility of surgical resection.  The rise of new drugs, based on molecular classification of disease, is giving the medical oncologist an increasingly important role in the diagnosis and treatment of solid tumors.  It is more important than ever for practicing oncologists to communicate, and collaborate, when working in this frontier.  We must anticipate that molecular tests of blood and tissue will one day surpass pathologic stage in terms of determining how patients are managed.  ASCO is the perfect forum to help develop these bold new pathways to test and validate new technologies together.

Robert G. Maki, MD, PhD, FACP

Dec, 04 2012 12:53 PM

Thank you for these meaningful insights, Tony. We are in need of improvements across the board, from understanding the dynamics of cancer growth in the models systems we study (the modeling of angiogenesis in vivo as shown by Dr Jain at the ASCO 2012 plenary is a perfect example of how we need to do this more broadly) to the evaluation of tumor shrinking and growth in our patients. For example, with PFS endpoints we cannot be sure that what we do helps people live longer, but this may be by design (allowing crossover of patients from one treatment to another, e.g.) or be due to true lack of benefit.

With the additional difficulties of understanding the dynamics of tumor growth after treatment and what ends up being the not-so-simple assessment of someone's health on a treatment, we have much to learn.

It should be an important goal for every clinical trialist in oncology to incorporate biomarkers (even things as simple as serum albumin levels) as well as novel endpoints (e.g. the "Choi" criteria that provide an alternative to RECIST), so that with each study we are able to say that much more about the most important metrics we study, patient survival and quality of life.

Back to Top