Gene-Guided Chemotherapy Research Questioned as Three NCI Trials Are Halted

July 27, 2010

Three ongoing cancer trials funded by the National Cancer Institute have been suspended after the validity of the technology being used was called into question by a large group of US scientists.

Developed at Duke University, the technology now under question uses gene signatures to predict responses to chemotherapy. Two of the trials involve patients with non-small cell lung cancer (NCT00545948 and NCT00509366), and the third is in patients with breast cancer (NCT00636441).

The trials were suspended on July 22 and 23.

The move was made after a group of 31 scientists called on the National Cancer Institute to suspend the trials because of concerns over the prediction models that were being used. The models were developed on the basis of research reported by Anil Potti, MD, and Joseph Nevins, PhD, from Duke University, Durham, North Carolina, but the validity of those models has been questioned by peer-reviewed reanalyses of their work, the scientists note.

In a letter dated July 19 and addressed to the new National Cancer Institute director, Harold Varmus, MD, the group of researchers called for the trials to be suspended until a "fully independent review is conducted of both the clinical trials and of the evidence and predictive models being used to make cancer treatment decisions."

At the same time, one of the Duke scientists involved in developing the technology has been suspended from his place of work. Dr. Potti was placed on administration leave while the university investigates allegations that he had falsely claimed to be Rhodes scholar, according to a report in the New York Times.

In addition, one of the published papers that reported this technology has now come under scrutiny. The Lancet Oncology has issued an "expression of concern" over a paper published in the journal in 2007, which described the validation of gene signatures to predict the response of breast cancer to neoadjuvant chemotherapy (Lancet Oncol. 2007;8:1071-1078).

That research was praised by an independent expert contacted by Medscape Medical News at the time, as it showed for the first time that gene signatures could predict responses to individual chemotherapy regimens.

However, since its publication in 2007, the methodology used to generate the response predictions has been questioned by statisticians from the M.D. Anderson Cancer Center in Houston, Texas, the journal notes.

The Lancet Oncology was contacted by senior author Richard Iggo, PhD, from the Swiss Institute for Experimental Cancer Research in Epalinges, Switzerland, and first author Herv Bonnefoi, MD, from the Institut Bergoni, University of Bordeaux, France. They "expressed grave concerns about the validity of their report in light of evolving events," and said they had repeatedly tried to contact their coauthors at Duke University (including Dr. Potti) without success.

The journal notes that the 15 European coauthors of the paper concur with the "expression of concern" notice that the journal has posted online and said that the 4 coauthors from Duke University have been contacted separately.

Controversy Surrounding Dr. Anil Potti and Duke University

The controversy surrounding Dr. Potti and his team's research at Duke University is outlined in exhaustive detail in a report published in the July 16 issue of The Cancer Letter. This publication found Dr. Potti's false claim of being Rhodes scholar in multiple grant applications submitted by him, and notes that the claim was also featured in a Duke newsletter in January 2007. However, this credential "disappeared" from Dr. Potti's biography later in 2007. The publication also found mentions of 2 other awards that it was unable to verify.

In addition to questions about Dr. Potti's credentials, The Cancer Letter notes that research coming out of his group has been "marred by corrections and even corrections of corrections," and points out that "errors in genomics research could have direct implications for patients."

Dr. Potti is considered to be a pioneer of personalized medicine because of his team's work on using gene signatures to predict responses to chemotherapy, and he has been featured in Duke University commercials aimed at the general public, the publication notes.

However, this work has been questioned by other scientists, it points out.

Two biostatisticians at the M.D. Anderson Cancer Center, Keith Baggerley, PhD, and Kevin Coombes, PhD, attempted to verify this work but found a series of errors, including mislabeling and mismatching of gene probe identifiers. They published their findings in November 2009 in the Annals of Applied Statistics (2009;3:1309-1334) and concluded: "Unfortunately, poor documentation can shift from an inconvenience to an active danger when it obscures not just methods but errors."

The biostaticians also suggested that the errors they found in the technology which was being used in ongoing clinical trials to allocate patients to treatment group may be putting patients at risk.

The Cancer Letter reports that as a result of that publication, Duke University temporarily suspended 3 clinical trials that were using gene signatures to assign patients to treatment these are the same 3 trials that were suspended again a just few days ago.

However, even though Duke suspended those trials in October 2009, they were restarted again in January 2010 after an internal investigation by Duke's Institutional Review Board confirmed the research and concluded that this approach was "viable and likely to succeed."

When contacted by The Cancer Letter and shown documents obtained under the Freedom of Information Act, the 2 statisticians from M.D. Anderson who had questioned the technology said they were not satisfied by the internal review. "Duke's statement implies that other members of the scientific community should be able to replicate the reported results with the data available," they told the publication. "Having tried, we can confidently state that this is not yet true."

The letter to the National Cancer Institute from the group of 31 scientists, which comprises many professors of statistics and biostatistics from prestigious US universities, including Johns Hopkins, Harvard, and Princeton, refers both to the Annals of Applied Statistics paper and The Cancer Letter reports.

"It is absolutely premature to use these prediction models to influence the therapeutic options open to cancer patients," the letter says, as independent experts have been unable to substantiate the researchers' claims using the researchers' own data. If the data and analysis can be validated, then it would be appropriate to reinitiate the trials, but until then, suspension of the ongoing trials is necessary, "given the potential of patients being assigned to improper treatment arms...[and] the associated potential risk posed to these patients."

Authors and Disclosures
Journalist
Zosia Chustecka

Zosia Chustecka is news editor for Medscape Hematology-Oncology and prior news editor of jointandbone.org, a Web site acquired by WebMD. A veteran medical journalist based in London, UK, she has won a prize from the British Medical Journalists Association and is a pharmacology graduate. She has written for a wide variety of publications aimed at the medical and related health professions. She can be contacted at ZChustecka@webmd.net.

Zosia Chustecka has disclosed no relevant financial relationships.

Gene-Expression Signatures in Lung Cancer: Not Ready Yet

It is the hope is that any patient with cancer would have their tumor biopsied and profiled. The profile would then be displayed as a unique genetic signature, which would in turn predict which therapy is most likely to work. However.....

Gene-Expression Signatures in Lung Cancer: Not Ready Yet

Roxanne Nelson - Medscape Medical News

March 17, 2010 — The identification of prognostic markers could assist in the clinical management of nonsmall-cell lung cancers (NSCLC). Although molecular profiling of tumors has led to the identification of gene-expression patterns, a new review has found "little evidence" that any of the signatures are ready for use in the clinical setting.

In addition, the researchers reported that they found "serious problems in the design and analysis of many of the studies" that were included in their review, published online March 16 in the Journal of the National Cancer Institute.

Even in its earliest stages, lung cancer has a very high recurrence rate and mortality, the authors note. Current clinical staging techniques have limitations in terms of predicting recurrence and guiding treatment, but the ability to identify new molecular targets using techniques such as microarray-based gene-expression profiling has the potential to improve patient care.

Inconclusive Results Thus Far

Studies have reported mixed results. As previously reported by Medscape Oncology, one recent review article found that gene-expression profiling failed to outperform standard histologic examinations. However, another study reported that a "5-gene signature" was closely associated with relapse-free and overall survival among patients with NSCLC.

More recently, at the 2010 Joint Conference on Molecular Origins of Lung Cancer, researchers reported that a mutated epidermal growth-factor receptor (EGFR) gene signature was a validated therapeutic target in NSCLC, and suggested that this gene signature might provide "predictive value and biological insights" into EGFR inhibitor responses in lung adenocarcinomas.

For the current review, Jyothi Subramanian, PhD, and Richard Simon, DSc, from the Biometric Research Branch at the National Cancer Institute in Bethesda, Maryland, conducted a literature search of studies published from 2002 to 2009 to critically evaluate studies that reported prognostic gene-expression signatures in NSCLC.

Little Evidence of Gene Signatures

The authors selected 16 studies as being most relevant, and closely assessed them for a number of criteria, including the appropriateness of the study design, the statistical validation of the prognostic signature on independent datasets, the presentation of results in an unbiased manner, and the demonstration of medical utility for the new signature beyond that obtained using existing treatment guidelines.

They noted that one of the "striking findings" is that none of the studies succeeded in showing that gene-expression signatures had better predictive power "over and above known risk factors." In fact, they note, the majority of the risk factors outlined by the National Comprehensive Cancer Network (NCCN) guideline were not even considered by most of the studies they reviewed.

For example, the extent of residual tumor after resection is the most important variable, after stage, when making decisions about adjuvant chemotherapy, according to the NCCN guideline. But only 7 of the studies stated that completeness of resection was a criterion for patient selection.

Drs. Subramanian and Simon point out that "the most important medical question that needs to be answered by a new prognostic signature in NSCLC is whether it can identify the subset of stage IA patients who might benefit from adjuvant chemotherapy." But only 2 studies in their survey included validation results for this subpopulation.

The majority of papers presented overall validation results for stage I patients, and some of the signatures were successful in identifying high-risk stage I patients. However, whether or not the signature was better at predicting overall survival than tumor size or other standard risk factors was not adequately addressed and was unclear from most of these studies, the authors report. Only 1 study, they note, reported a marginal improvement in the predictive accuracy for their gene-expression signature, compared with tumor size, for stage I patients

Another important medical need is the ability to identify the subset of stage IB and stage II patients who are at a low risk for disease recurrence without chemotherapy, the authors explain. But only one of the studies presented separated validation results for this subgroup of patients; a second study was the only one that reported the statistical significance of the prognostic signature for validation in stage II samples. The lack of predictiveness for stage II patients could be the result of the small number of such patients in the study samples, they note.

Most of the studies presented validation results on data that were not used for developing the predictive signatures.

"Most of the studies presented validation results on data that were not used for developing the predictive signatures," they write; in addition, "none of the 16 studies reviewed adequately addressed the question of the predictive power that could be attained by using easily measurable clinicopathological factors for stage I samples."

On the basis of their observations and analyses, the authors suggest a set of guidelines to aid the design, analysis, and evaluation of prognostic gene-expression studies, with a focus on NSCLC.

"Clinical validity of a prognostic signature implies demonstrating that the test result correlates with clinical outcome," they write, whereas "medical utility of a prognostic signature means that the test result is actionable, leading to patient benefit."

Therefore, the ultimate test of clinical validity for a prognostic signature is how well it performs in a prospective clinical trial. Several such trials are currently underway, including the CALGB 30506 trial that was recently initiated to clinically test the lung metagene prognostic signature in lung cancer, the authors point out.

"Regardless of clinical validation, unless a new prognostic signature provides additional risk stratification within the stage and risk-factor groupings on which current treatment guidelines are based, its broad acceptance in medical practice is unlikely," the authors conclude.

J Natl Cancer Inst. Published online March 16, 2010

New Paradigms of Cancer Treatment

Gene expression (signature) assays are panels of markers that can predict the likelihood of cancer recurrence in various populations. Functonal profiling assay is a test for drug activity against a tumor. Pharmacogenomic testing is a test to identify patients who are likely to have the most toxicity.

By testing the gene expression markers of a patient, oncologists can identify those patients unlikely to benefit from adjuvant chemotherapy from those that would. If the patient needs adjuvant chemotherapy, by testing the patient's tumor cells and testing the patient toxicity tolerance, the oncologist can select drugs that have a higher probability of being effective for an individual patient rather than selecting drugs based on the average responses of many patients in large clinical trials.

What a cancer patient would like ideally, is to know whether they would benefit from adjuvant chemotherapy. If so, which active drugs have the highest probability of working and are relatively non-toxic in a given patient.

Whether a patient would benefit from adjuvant therapy depends on two things: (1) whether the tumor is "destined" to come back in the first place and (2) whether the tumor is "sensitive" to drugs which might be used to keep it from coming back.

The gene expression (signature) marker assays actually could be calibrated to provide information both about the possibility of recurrence and also chemosensitivity. The problem is dissecting one from the other. Studies to date have just looked at whether people had a recurrence.

You can identify gene expression patterns (via assays) which correlate with this. But it can be hard and even impossible to tell what exactly you are measuring: is it intrinsic aggressiveness of the tumor? sensitivity to adriamycin? sensitivity to cyclophosphamide? sensitivity to taxol? sensitivity to tamoxifen? You find a gene expression panel which correlates with something, but picking apart the pieces is hard.

You can begin to do this if you combine gene expression studies (molecular profiling) with cell culture studies (functional tumor cell profiling). Use the functional profiling as the gold standard to define the difference between sensitivity and resistance. Then see which pattern correlates with which for individual tumors and individual drugs.

When the decision is made to treat a patient with chemotherapy, most patients are treated with a combination of drugs. The "functional profiling" method differs from existing DNA and RNA tests in that it assesses the activity of a drug upon combined effect of all cellular processes, using several metabolic and morphologic endpoints. Other tests, such as those which identify DNA or RNA sequences or gene expression signatures of individual proteins often examine only one component of a much larger, interactive process.

No gene-based test can discriminate differing levels of anti-tumor activity occurring among different therapy drugs. Nor can available gene-based tests identify situations in which it is advantageous to combine the new "targeted" drugs with other types of cancer drugs. So far, only cell-based functional profiling has demonstrated this critical ability.

Not only is this an important predictive test, it is also a unique tool that can help to identify newer and better drugs, evaluate promising drug combinations, and serve as a "gold standard" correlative model with which to develop new DNA, RNA, and protein-based tests that better predict for drug activity.

Genomic testing is not the answer, without cell "function" analysis. Functional tumor cell profiling has its own very sophisticted program to discover gene expression microarrays which predict for responsiveness to drug therapy. The way to identify informative gene expression patterns is to have a gold standard and that cell-based functional profiling assays are by far the most powerful, efficient, useful gold standard to have. It grasps the potential value of the assays today to individualize therapy.

And then you come to the 1,000 pound gorilla of a question: What effect will the different individual drugs have in combination in different, individual tumors? This is where cell-based functional profiling assays will always be able to provide uniquely valuable information. But it's not one versus the other. The best thing is to combine these different tests in ways which make the most sense. One month's worth of herceptin + avastin costs $8000. That's without any docetaxel and blood cell growth factors and anti-emetics. If nothing else, we can't afford too much trial and error treatment.

There are hundreds of different therapeutic drug regimens which any one or in combination can help cancer patients. The system is overloaded with drugs and underloaded with the wisdom and expertise for using them. We have produced an entire generation of investigators in clinical oncology who believe that the only valid form of clinical research is to perform "well-designed," prospective, randomized trials in which patients are randomized to receive one empiric drug combination versus another empiric drug combination.

The problem is not with using the prospective, randomized trial as a research instrument. The problem comes from applying this time and resource-consuming instrument to address hypotheses of trivial importance (do most cancers prefer Coke or Pepsi?). The failure of 30 years' worth of clinical trials research into "one-size-fits-all" therapy will eventually force a consideration of new approaches. All the more reason to "test the tumor" first - properly.