Johns Hopkins’ Tragedy: Could Librarians Have Prevented a Death?
Posted On August 7, 2001
In a tragic situation that could have been averted, Ellen Roche, a healthy, 24-year-old volunteer in an asthma study at Johns Hopkins University, died in June because a chemical she inhaled led to the progressive failure of her lungs and kidneys. In the aftermath of this loss, it would appear that the researcher who conducted the experiment and the ethics panel that approved it allegedly overlooked numerous clues about the dangers of the chemical, hexamethonium, given to Roche to inhale.
Adding particular poignancy to the case is that evidence of the chemical's dangers could easily have been found in the published literature. The Baltimore Sun concluded that while the supervising physician, Dr. Alkis Togias, made "a good-faith effort" to research the drug's adverse effects, his search apparently focused on a limited number of resources, including PubMed, which is searchable only back to 1966. Previous articles published in the 1950s, however, with citations in subsequent publications, warned of lung damage associated with hexamethonium. The Sun's investigation included an article that asked the question, "Could Librarians' Help Have Prevented Hopkins Tragedy?"
The answer to that question is a resounding "Yes." Canvassing medical librarians and consulting the Medical Library Association's listserv (MEDLIB-LISTSERV.ACSU.BUFFALO.EDU) revealed that medical librarians around the country immediately searched various sources and by using just online resources discovered disturbing information on problems associated with this drug. For instance, a quick search of Toxline showed many citations identifying relevant articles. Consulting Micromedex's PoisonIndex Toxicologic Management on ganglionic blockers yielded relevant citations in the references sections. These citations are available online, but in other resources than those the Johns Hopkins physician researcher relied upon.
An internal investigation of the event disclosed that, in addition to print and scholarly online resources, such as PubMed, Dr. Togias searched the open Web. For his Web search strategy, the tools he chose were Google, Yahoo!, LookSmart, and GoTo.com. With the exception of Google, professional searchers would probably not have chosen the latter three for a thorough, open Web survey.
This kind of tragedy offers a terrible lesson in the importance of using professional searchers and medical librarians in critical searching situations.
The Sun reported that Togias had researched PubMed, which covers back to 1960. Actually the PubMed version of MEDLINE covers back to 1966, though requiring a double search of a divided database. Some versions of MEDLINE cover earlier years—e.g., Medscape's back to 1960. The Sun article quotes other doctors and medical librarians as charging Togias with "lazy" and "foolish" literature-searching practices. The article asserts that librarians could have performed manual searches of earlier medical indexes and tapped strong print collections to help avert this tragedy.
Before I continue, notice that a letter to the editor appeared in the July 16, 2001 issue of Newsweek that reads: "I read with interest your article ‘The New Patient Power.' There is a great deal of information available on the Internet, and so much information can be confusing to the patient. But there is also an entire profession dedicated to sorting it all out: medical librarians who have specialized training in how to find credible medical Web sites or other resources. Many hospitals have medical libraries on site that may be open to patients and their families as well as to the general public." (Written by Leann Isaac, Jameson Health System Library, New Castle, Pennsylvania.)
Following Roche's death, the Food and Drug Administration reported irregularities (http://www.fda.gov/ora/frequent/483s/JohnHopkins483.html) in the use of hexamethonium bromide by Alkis Togias, M.D. of the Johns Hopkins Asthma and Allergy Clinic. The study, "Mechanisms of Deep Inspiration-Induced Airway Relaxation," required three human subjects to inhale the hexamethonium in a study of how the lungs of healthy people protect against asthma attacks. Roche's death drew worldwide press coverage. Subsequently, the federal government suspended all research contracts or grants involving human subjects at Johns Hopkins—this was a severe blow to the institution that is the largest medical school recipient of medical research grants. A gradual reinstatement of grants seemingly began to occur within 72 hours, however.
Since the articles relevant to the case were published in the 1950s, were manual searches of print tools the only solution? Did online resources fail or only this online searcher? While researching this article, two online searchers—both experienced veterans in dealing with online medical literature and both information brokers without any print collections—conducted searches independently for the dangers of hexamethonium and found references in the online literature from an array of resources.
Before there were end-users, librarians were trained to do mediated searches for patrons. We knew we were not subject specialists and never would be, so we were taught the ins and outs of how the MEDLINE database was constructed so that we could go in and retrieve what our patrons needed by utilizing all the features the folks at the National Library of Medicine had so thoughtfully built into the system. As professional searchers experienced with multiple online search services, we also learned to access other prominent medical and pharmaceutical databases using powerful commercial search engines. As the Web arrived with its plethora of new information, we learned how to search that literature effectively and critically.
In approaching searches for clients, the professional searchers ask a series of questions. How can we most comprehensively research the question presented? How exhaustive should this research assignment be? What are the best print and nonprint sources to use? Are there agencies and organizations that we should look at when approaching the research strategy? As managers, should we devise a checklist of sources that all staff should use under certain circumstances?
Just looking at the MEDLINE database, one of the major features is the Medical Subject Headings (MeSH) vocabulary, which, as well as listing descriptors, also establishes a hierarchy of relationships between search terms. It also utilizes subheadings that enable the simple framing of quite complex queries. Professional searchers are also aware of the importance of knowing when terms entered the MeSH vocabulary and how to search for a concept before it became a MeSH term, as well as the value of free-text search strategies. Medical librarians are taught to always question any zero-citation result and to re-examine any search strategy producing negative results for logical as well as other errors.
A librarian trained in and utilizing such methods should have been able, by searching only PubMed, to retrieve enough citations to at least cast suspicion on the use of hexamethonium in the study at Johns Hopkins.
Consulting the MeSH would reveal that Hexamethonium became a MeSH descriptor only in 1995. Prior to that year and back to 1980, it was searchable as a supplementary chemical term. Supplementary chemical terms could not carry subheadings, so searching for adverse effects, poisoning, or toxicity as subheadings for this substance was not possible during this time period. We could have used free-text strategies or we could have tried broadening our search by "backing up the MeSH tree" when this kind of situation arose. In this case, the latter approach would have revealed the MeSH descriptor "Methonium Compounds" as a term in use at least back to 1966.
Using this term with the subheadings "adverse effects" and "toxicity" would have brought up four particularly significant citations among the 26 citations retrieved. [Available upon request from the author.] Of course, hindsight (always 20/20) tells us to look for respiratory problems. The remaining citations may have alerted the researcher to other potential problems with using the substance.
This list brings to mind another thing: the quickness with which patrons limit their searches to only English-language material. MEDLINE always includes English translations of the titles of foreign-language articles, and the German-language article we found did have an English abstract in the database. Even if this were not the case, when lives are in the balance, it might behoove researchers to get copies of foreign-language articles and have them translated.
It could be said that anyone who reads the Report of Internal Investigation relative to this tragedy (http://www.hopkinsmedicine.org/press/2001/JULY/report_of_internal_investigation.htm) would conclude that medical librarians have a responsibility to approach their IRBs to assist in the development of standards for the required scientific and safety reviews. The report includes a description of the literature-searching techniques used by the principal investigator, Dr. Togias, an end-user searcher:
Was the Scientific and Safety Review of the Study Protocol by the JHBMC IRB Appropriate? A medical librarian searching the Micromedex PoisonIndex Toxicologic Management section on ganglionic blockers found a number of citations—including several from the 1950s—covering the dangers of hexamethonium.
The principal investigator subsequently stated to the investigation committee that he had performed a standard PubMed search for potential hexamethonium toxicity and consulted standard, current edition textbooks of pharmacology and pulmonary medicine before submitting the application to the IRB. None of these sources mentioned hexamethonium-related pulmonary toxicity. Although the PI made a good-faith effort to uncover previous reports of hexamethonium-related toxicity, the vagaries of performing such a search are illustrated vividly by the following performed as part of this investigation: (1) None of the last four editions of Goodman and Gilman, or the current edition of Fishman's Pulmonary Medicine mentioned this toxicity, but an older edition of Fishman's text did; (2) During PubMed searches, "hexamethonium inhalation lung injury" gave 0 hits, "hexamethonium inhalation" gave 42 hits (but none referring to pulmonary toxicity), "hexamethonium lung" yielded 3 useful articles, "hexamethonium lung toxicity" gave 4 hits, but 0 useful articles, "hexamethonium lung hypersensitivity" gave 16 hits with 3 useful articles, and "hexamethonium lung fibrosis" gave 3 hits and 2 useful articles; (3) the Google search engine and Yahoo uncovered the pneumotox web site (see below), but LookSmart and GoTo.com did not; (4) the Micromedex data base had lung toxicity as the first adverse effect of hexamethonium.
Following another approach to the search problem, one professional searcher turned to the Chemical Sources database on STN, where she found several forms of hexamethonium, including hexamethonium bromide, listed along with the names and addresses, phone numbers, fax numbers, e-mail identifications, and URLs of the drugs' manufacturers. One could assume that pharmaceutical manufacturers might also collect and make available reports on the efficacy and dangers of using their products.
Looking at this case, it would appear that medical librarians may currently be better prepared than biomedical researchers to conduct medical literature searches and that the best searches would involve researchers and their librarians working closely together. Perhaps it's time to develop standards of what constitutes a reasonable search. Medical librarian listserv messages included suggestions that the Medical Library Association and/or the National Library of Medicine sponsor this effort.
[In producing this report, I'd like to thank Aggi Raeder, a veteran online searcher who's well experienced with medical and pharmaceutical databases, and Barbara Quint, Editor of Searcher and Contributing Editor, NewsBreaks. I also greatly appreciate Dialog's help in providing complimentary passwords to their array of biomedical and health sciences databases. Thanks to Sandy Scherer and Rose Pyle at Dialog, in particular.]