© 2007 Nature Publishing Group http://www.nature.com/naturebiotechnology
CORRESPONDENCE databases for identifying human candidate genes based on expression variation, I would need to pay the BLAST database for finding homologs, pay the ENSEMBL database for its mouse gene build and chromosome mapping and compensate the Gene Ontology consortium as well as Mouse Genome Informatics for providing functional annotation on candidate genes. Budget justification in grant applications would be filled with URLs (universal resource locators) and database names. Money issues aside, practical problems would arise. Modern biological databases are not isolated entities. They integrate information, annotations and predictions as cross references from other databases. Imagine if you wanted to click a link to an external database or download data. In Busa’s world, you would have to waste valuable research time handling lengthy agreements, not to mention the mental frustration of dealing with tech-transfer guys. More importantly, your results wouldn’t be easily replicated by other researchers, unless they would be willing to pay what you have paid. You might also be unwilling to disseminate or even present your primary research data because of concerns about legal liability. (I have a bit of unpleasant experience in this domain, so I say this with some insight.) Academia and the public sector build databases to increase accessibility and solicit recognition from others, whereas the private sector builds database to attract more buyers—there is an intrinsic difference here. Unlike databases specializing in a single disease or a small group of genes, protein interaction databases appeal to a broad set of users among the scientific community, and by necessity must be incorporated into, and cross-linked with, other databases. The importance of BIND is evident in the fact that over 77 scientific journals had signed on to publish BIND accession numbers in their papers in just two years. In addition, a freely downloadable raw data set (no longer available) provides the basis for algorithmic development for computational prediction of protein interactions. Researchers may not mind paying for the luxury of specialized databases, but data registries that cater to a broad set of users should be broadly and freely accessible to the research community. Although the initial development of databases, such as BIND, requires caution and close oversight of budgets, an equally important aim should be to ensure that data repositories of particular utility to the
972
research community remain sustainable and publicly accessible. Databases, such as BIND, should not be left to the private sector. Ensuring public accessibility to data essential for research progress is the responsibility of the central planner, not Adam Smith’s invisible hand in the marketplace.
COMPETING INTERESTS STATEMENT The author declares no competing financial interests.
Kai Wang University of Pennsylvania, Department of Genetics, 415 Curie Boulevard, Philadelphia, Pennsylvania 19104, USA. e-mail:
[email protected] 1. Busa, W.B. Nat. Biotechnol. 24, 1065 (2006).
The Genasense ‘learning curve’ To the editor: I have been involved in the field of antisense therapeutics for 20 years and find the views presented in the May issue by Carol Potera in her article “Antisense—down but not out”1 to be both misleading and significantly inaccurate, especially concerning Genta’s (Berkeley Heights, NJ, USA) phosphorothioate antisense compound Genasense (oblimersen sodium). It is at best disingenuous to quote a direct competitor that “Genasense’s [failure] is not a failure of the drug, but [of] the regulatory strategy.” Moreover, the contrast in tone evident in the description of Genta’s efforts (entirely dismissive) versus those of Isis (Carlsbad, CA, USA) is strikingly inappropriate, especially considering that no cancer-related Isis product that has completed phase 3 clinical trials has ever been subjected to the US Food and Drug Administration’s withering scrutiny. Indeed, no other antisense compound than Genasense has undergone such extensive testing or yielded more clear-cut results, which explains why the drug currently is under active review in the United States and Europe by regulatory agencies in two different indications (chronic lymphocytic leukemia [CLL] and melanoma). Thus, I find it surprising that Potera failed to check her facts against the published data2. Potera states that in CLL “…the FDA flunked Genasense because the primary endpoint of overall survival missed the required statistical cutoff.” Here are the facts: the study’s primary endpoint was complete plus nodular partial response (not survival), and it was achieved with statistical significance (P = 0.025). The FDA Oncology Drug Advisory Committee,
which contained no CLL experts, ‘flunked’ Genasense due to their unaccountably preferred secondary endpoint of time to progression (TTP), which was not significantly different. No significant safety issues were identified. The efficacy basis for initial regulatory approval of every drug for every form of leukemia has always been response rate. Complete responses that last for extended duration, as observed in the Genasense trial, have always been considered as the most reliable evidence of activity. Conversely, no leukemia drug has ever been initially approved on the basis of TTP, overall survival or symptomatic improvement. Potera’s comments about survival are especially misguided because no randomized trial at any stage of CLL has ever shown a significant increase in overall survival. Indeed, it is unclear what regulatory credentials or evidence allowed Potera to opine on the appropriateness of endpoint selection or clinical trial design. I cannot recall any other oncology product that met its primary endpoint in a randomized controlled trial and was denied approval, particularly for seriously ill patients in an orphan-sized population with high unmet medical need. Perhaps these are some of the reasons why Genta has chosen to appeal the initial FDA decision. Furthermore, Potera compounds the confusion by stating that “….patients with elevated levels of lactate dehydrogenase (LDH), a marker previously established to predict a poor prognosis in CLL patients, also respond poorly to Genasense, so LDH has become a marker for optimizing patient selection.” This statement is simply false. LDH as a marker was never part of the CLL trial; it was, however, a stratification
VOLUME 25 NUMBER 9 SEPTEMBER 2007 NATURE BIOTECHNOLOGY
© 2007 Nature Publishing Group http://www.nature.com/naturebiotechnology
CORRESPONDENCE criterion in the GM301 trial3 of Genasense plus dacarbazine versus dacarbazine alone in advanced melanoma because it is the best known prognostic marker in this disease. Unfortunately, the substantial therapeutic accomplishments of Genasense in a randomized, global, 771-patient trial in advanced melanoma (and specifically among 508 patients with a normal baseline LDH level, which showed statistically significant benefits in survival (P = 0.018), progression-free survival (P = 0.0007) and response rate (P = 0.009)) were ignored by Potera in favor of open-label, preliminary, phase 2 human data, and an additional single data set that involved antisense therapeutics in sick penguins. The most accurate assessment of the current situation in oligonucleotide therapeutics in Potera’s article came from John Rossi, who stated that “There’s a place for RNAi and antisense.” I hope that Potera takes careful note of this opinion, especially with respect to human cancer, and most importantly, that she takes greater care in her reporting of facts and their assessment. Readers should note that although I am not a Genta stockholder and have no contractual relationships with the company, I was formerly a member of their scientific advisory board (but have not been since 2005). COMPETING INTERESTS STATEMENT The author declares competing financial interests: details accompany the full-text HTML version of the paper at http://www.nature.com/ naturebiotechnology/.
C. A. Stein Medicine Urology and Molecular Pharmacology, Albert Einstein College of Medicine, Montefiore Medical Center, 111 East 210th Street, Bronx, NY 10467, USA. e-mail:
[email protected] 1. Potera, C. Nat Biotechnol. 25, 497–499 (2007). 2. O’Brien S. et al. J. Clin. Oncol. 25, 1114–1120 (2007). 3. Bedikian, A.Y. et al. J. Clin Oncol. 24, 4738–4745 (2006).
Carol Potera responds: If it appears that a positive tone was taken with some companies and not with others, it was unintentional. Not only the management at Isis, but also several of the other companies and academic antisense experts with whom I spoke expressed the common view that Genta has been treated very badly by the FDA. This was said with concern, not avarice. My comment “failure of the regulatory strategy” represents this common view. Perhaps it could have been
strengthened with an added thought, such as “and does not reflect inferior science.” My comment “missed the required statistical cutoff ” was intended to reflect exactly what Stein mentions. Although there are other markers, such as partial nodular response, that show that Genasense is effective in certain patient populations, the FDA ignored them and only focused on overall survival. As for LDH, in cutting a paragraph about Genasense in melanoma trials to shorten the article, the sentence about LDH was
left in and combined with the preceding paragraph about CLL. This was a mistake I did not catch. As Stein says, LDH is a marker used in melanoma trials, not CLL. As for the paper by O’Brien et al.1 that Stein mentions, that study was published in the March 20, 2007 issue of the Journal of Clinical Oncology, several months after I had submitted my original draft. I was not aware that this current study was available. 1. O’Brien S. et al. J. Clin. Oncol. 25, 1114–1120 (2007).
Ethical framework for previously collected biobank samples To the editor: An increasing trend in biobank research is to pool collections of biological samples in international scientific studies, thereby amplifying their potential scientific value. The pooling of samples, however, poses several challenges to national/international legislation and ethical guidelines relevant to biobank samples and also introduces new issues for patient/donor interests—as highlighted by correspondence in your May issue last year1, which indicated that for human biobanks “legal comparisons between regulations in different countries are laborious and defy generalizations.” Indeed, routines that have been used for information and consent vary greatly between existing biobanks. A majority of samples stored in clinical biobanks have been collected without expressed consent for research from sample donors. When consent has been obtained, it has often taken different forms. Here, we present an ethical framework for research on previously collected biobank samples constrained by some of the above issues. On this basis, we also provide recommendations for adapting existing consent procedures on such samples. The need to resolve the potential conflict between research interests, the safety, personal integrity (including privacy) and autonomy of research subjects, and the preservation of public trust in biomedical research is central to any discussion on ethics in research involving human subjects, human biological samples or personal information (as one of us (M.G.H.) has previously noted2,3). Biobank research aiming to improve knowledge, prevention and therapy of disease is clearly in the interest
NATURE BIOTECHNOLOGY VOLUME 25 NUMBER 9 SEPTEMBER 2007
of researchers, sample donors and the population as a whole. Having a biological sample stored in a biobank involves no direct physical risk to the donor once the sample has been obtained. Information may be derived from the samples or be traceable to them, as in combined register and biobank research. It is the inappropriate distribution of this information (e.g., to insurance companies or employers) that can potentially harm the donor. Thus, an ethical platform should minimize risks to the interests of individual research subjects while ensuring optimal scientific value of the research performed. This balance must not be reduced to a conflict between research and patient/donor interests, as sample donors also have interests in research (for reference, see paper by M.G.H.4). To balance the interests at stake, we propose that each research project using established biobanks should be preceded by careful assessment by both the researchers themselves and an ethical review board (ERB) to ascertain predictable risks and burdens in relation to foreseeable benefits to the subject and others. This assessment would include a consideration of good practice in storing, coding and using samples, as well as appropriate procedures for obtaining consent and counseling. Routines for coding and storage of biobank samples, with restricted access to personal information, must be in place to promote the safety and personal integrity of sample donors. Storage conditions should optimally preserve the usefulness of the sample while protecting against unauthorized access. Sample identities must be coded,
973
