On his article on PubMed Central, the NIH's new electronic archive of biomedical research, Harvey Blume paints critics like me as quixotically trying to hold back the Internet (he refers to the controversy as "a tale of new technology versus old, of innovation and inertia"). In fact, my arguments against PubMed Central have nothing to do with the medium and everything to do with the message. I do not think it's a good idea to disseminate bad medical research--on paper, on the Internet, or by any other means--and PubMed Central, as presently envisioned, would make that not only likely, but inevitable.
Determining the validity of medical research is no easy matter. To evaluate the papers we receive, The New England Journal of Medicine employs six full-time physician editors, six part-time associate editors in various medical specialties, three statistical consultants, and a consultant in molecular biology, in addition to some 5,000 peer reviewers. We reject over 90 percent of the papers submitted to us, primarily because the research is of poor quality. The design or methodology of the study may be inadequate to address the hypothesis, the analysis of the data may be inappropriate, the conclusions may not be supported by the data or the data may support alternative conclusions, and so forth. The possible flaws, many of them fatal, are virtually endless. Of those papers we publish, nearly all require some revision to address scientific concerns.
PubMed Central would do two things. First, it would post all biomedical research reports after publication in journals. I have no objection to that, although we put our papers on our own Web site the day of publication. (We post them immediately on acceptance if they have practical implications that might make doctors change their practice.) Our Web site has about 800,000 visitors each month. It is true that we charge readers for electronic access to the full text of all our articles. PubMed Central would charge authors instead.
But PubMed Central would also permit authors to post their work without undergoing peer review and revision. That is where the trouble lies. I have seen too much bad research to think it is a good idea for the NIH to establish a magnet for it. Almost certainly, this part of PubMed Central would be an uninterpretable mix of junk science and promotional material disguised as research. The fact that it would be labeled as unreviewed would not help much, and the notion of peer review after publication is backwards.
If the idea is so bad, why did Harold Varmus and his colleagues come up with it? I believe the reason is that they come from the world of basic science, and the idea reflects the biases of basic scientists. To them PubMed Central is a way for investigators to exchange information with one another as rapidly as possible. If the information is preliminary or simply wrong, other investigators will correct it. As Blume points out, the Los Alamos National Laboratory electronic archive for physicists was their model.
But clinical science, such as cardiology, is different from basic biomedical science, such as molecular genetics, and it is different from physics. The audience for clinical research consists not just of other investigators, but of practicing physicians, most of whom are not trained in research methods. They simply want to know the best way to care for their patients, and they rely on journals to screen the evidence for them. The media and the general public are also far more interested in clinical research than in basic research or physics because of the health implications. They, too, cannot be expected to sort out good research from bad. The risk is that physicians, reporters, and the public will make medical decisions on the basis of invalid research in the Junk Section of PubMed Central.
Blume and others like him are so giddy with the technology that they seem to believe it can somehow make dubious, unreviewed information accurate. They also assume that critics of PubMed Central are equally concerned with the technology. In fact, the technology is irrelevant. Just as the printing press did not make all books good, so the Internet cannot make all research reliable.
Marcia Angell, M.D., is editor in chief of The New England Journal of Medicine.
Mr. Angell writes that I and others like me are "giddy with the technology" of the Internet, whereas for her, the "technology is irrelevant." It's not that the technology is unimportant--she compares the significance of the Internet to that of the printing press--but that she will not be distracted by it from her commitment to the publication of reliable medical research. To this, I can only say that if Dr. Angell is correct in comparing the Internet to the printing press, and I think she is, then it is all the more peculiar for her to act as if the delivery systems for medical research can and should remain basically as they were before. Others I mention in my piece--including Dr. Varmus, Dr. Bell, and Dr. Ruskin--have taken the initiative in making the Internet a medium for the sharing and review of medical information. Dr. Angell and The New England Journal of Medicine seem instead to have chosen a lofty attitude of nonparticipation.
Otherwise, why does the NEJM continue to refuse to share its contents with PubMed Central? The original plan for PubMed Central (then known as E-biomed) was modified precisely to satisfy concerns expressed by Dr. Angell and others that they could not submit material from their journals prior to publication without jeopardizing their journals' existence, in accordance with which PubMed Central now requests such content at any time the journals choose to submit it, including post-publication. Still, as I reported, the NEJM has "no immediate plans" to participate in PubMed Central. This attitude is unnecessarily adversarial. Does Dr. Angell believe that if the NEJM refuses to bestow recognition on PubMed Central, it will sooner or later wither away?
As for Dr. Angell's claim that the unreviewed section of PubMed Central will become a depository for "junk science and promotional material disguised as research," it may simply be too soon to tell how much, if any, truth there is to this charge. Dr. Angell's argument would be more convincing if she cited any evidence to back it up, which she does not.
Otherwise, Dr. Angell's rejoinder repeats, nearly without modification, the same points I quote her as making in my piece. Yes, the NEJM is a superb example of a peer reviewed research journal. None of the advocates of PubMed Central I talked to thought otherwise. It is not necessary to devalue the NEJM to stress the advantages of new media, and it is unfortunate that here again Dr. Angell sets up a dichotomy.
Her argument that Dr. Varmus, by basing PubMed Central on the design of the Los Alamos National Laboratory's physics archive, was mistakenly assuming that doctors (like physicists) are research scientists when they are in fact primarily clinicians, is also well represented in my piece and, in part, incontrovertible. But has it occurred to Dr. Angell that the Internet will make physicians more research savvy, just as it has already helped make patients more sophisticated about their medical options? Of course, there has been misleading material distributed online as well, but the positive effects of making medical material widely available to those in need of medical services surely outweigh the negative. Or does Dr. Angell disagree? These days, in any case, doctors are being routinely trained in medical school to evaluate research techniques, and the Internet is only part of a process that may render the divide between physician and researcher less absolute than Dr. Angell assumes it is.
It is not giddy to believe PubMed Central offers opportunities, as well as presenting dangers, and that the NEJM could do better than to take an adversarial stance toward it. ¤
Harvey Blume is a writer in Cambridge, Massachusetts.