Oncology: January 2011 Archives
Today's Duke Chronicle reports the retraction of yet another article by geneticist Anil Potti. (For important story background, go here, here, and here.) On Friday, editors of The Lancet Oncology and the article's European coauthors justified their retraction of "Validation of gene signatures that predict the response of breast cancer to neoadjuvant chemotherapy," expressing "concern over the validity of the results."
The chemotherapy sensitivity predictions reported in the Article were based on an approach described by Anil Potti and colleagues in Nature Medicine. Re-examination of the validation datasets used for the Nature Medicine study has uncovered errors in the labeling of the clinical response in some of the datasets. Reanalysis of the predictive accuracy with correctly labeled data has shown that in two instances the reported signatures do not predict the response of the validation samples to chemotherapy.
In other words, the article was retracted because the results were founded on the flawed (ie, mislabeled) data from another retracted article by Potti et al (in Nature Medicine). An investigation into the validity of the Lancet Oncology article was begun last October, when editor David Collingridge relayed an "expression of concern" from 15 European investigators who were coauthors with Potti and 3 other Duke researchers. After a damning report in July from biostatisticians at M. D. Anderson about data-labeling errors in another article by Potti et al, the Lancet Oncology coauthors repeatedly attempted to contact their Duke colleagues, but they were ignored, wrote Collingridge.
This is the third article by Potti to be retracted since his work was called into serious question last year. A 2007 article in the Journal of Oncology was retracted in December, and the Nature Medicine article was pulled this month. In October, Collingridge also revealed that "a large group of scientists" wrote to NCI director Harold Varmus on July 19th, expressing their concerns about the validity of a) Potti's cancer-treatment prediction models and b) 3 clinical trials that were based on these prediction models.
On a related note, Friday's The Cancer Letter (subscription required), by way of The Great Beyond Blog, reveals that FDA auditors have visited Duke to determine if the university obtained proper approval for 3 cancer trials, which were based on the demonstrably flawed work of geneticist Anil Potti. At issue, from the agency's standpoint, is whether Duke obtained an Investigational Device Exemption (IDE), which was necessary because genomic chips were used to identify genetic biomarkers on which treatments was based.* There was evidently some confusion at Duke about whether an IDE was needed.
* The trials used Potti's (mislabeled) data to determine what anticancer drugs or combination of drugs cancer patients would receive on the basis of genetic biomarkers.
We made the decision because the drug has a marginal effect on tumor growth in breast cancer and in light of Avastin's severe side effects, the risks outweigh the limited benefit.
Pazdur cited "issues" and "problems" with a pivotal study, E2100, which showed that Avastin (with paclitaxel) delays the growth of advanced breast cancer for 5 months. Not surprisingly, drugmaker Genentech, a subsidiary of Roche, rejects the FDA's view of the controversial study and its results. But follow-up studies failed to confirm the favorable results of E2100, reports the WSJ. Some insurance companies, acting independently of the FDA, have already denied coverage for the use of Avastin in metastatic breast cancer.