Hello Tom,
Before all else, of course I do not feel bad that you are directing
some questions at another researcher!!! :) Firstly, I am honoured
that you have directed some of your questions specifically at me, and,
secondly, I know that pafalafa is an excellent researcher. While we
researchers obviously do compete for open questions, a significant
number of us have also succeeded in building a warm and supportive
community, and we certainly do not begrudge or envy our colleagues.
The main thing anyway is that Google Answers is a service, and it is
important to us that our clients use it in the way that best suits
them and obtain the information they need.
I will start with a general discussion, and then move to the specific
points.
The two classification systems you mention were really designed mainly
to assist in the organization of libraries and their catalogues, ie to
provide a way of indicating where documents are located. I doubt
whether either goes down to the level of detail to fully describe
information on adverse events. The UDC does have an advantage in this
respect. Being a numerical decimal-based system, it can be extended
ad infinitum to ever lower levels of detail, since all that is needed
is to go down a further level after the decimal point. It is also
good at expressing more than one facet, because numbers can be joined
together in various ways. However, as a retrieval system it is not
user-friendly precisely because it uses numbers which are far more
difficult to work with than words. So yes, in a customised form, it
probably could work as an in-house classification and/or indexing
system, but I cannot really see it being used in any extensive way for
external access databases, except maybe linked to a thesaurus or
glossary.
If your interest is retrieving information from outside sources, as
opposed to organizing and storing it for your own purposes or those of
other users of your services, then all the various
thesauri/glossaries/dictionaries/vocabularies you mention will only
serve two basic purposes:
1. As a general aid in helping you define ways in which to carry out a
free text search of any resource. A free text search is one in which
you do not restrict your search terms to any particular field, so the
words might occur in any one or more of the title, abstract (or full
text if the whole article is stored) or indexing terms. Looking at
the list of preferred and non-preferred terms will enable you to
determine various synonyms, which all need to be used to be sure of an
exhaustive search strategy if you are unable to use the power of a
specific indexing system (because you cannot access the indexing
system used in a particular resource, because you are trying to search
simultaneously across several resources, because the resource does not
have a specific indexing system: the prime example of this being the
Internet). For example, you might need to determine the generic name
of a drug, its chemical name, the codes used for it while it was being
trialed, and the various trade names under which it is marketed around
the world.
The other general way in which you can use a hierarchical thesaurus
such as MeSH is to find the relationship between concepts. For
example, you are searching for X, but cannot find much information.
Looking at the thesaurus, you see that X is a subset of Z. Searching
for Z or other subsets of Z might give you related information, which
could possibly be useful. For example, if you cannot find much about
a specific drug, searching for adverse effects associated with other
drugs of the same class might give you some clues of what to expect
with your drug of interest.
2. The second use of thesauri etc is as a way of making your search
very precise. However, this applies ONLY if the resource you are
searching uses that particular thesaurus to index its records. In
this case, once you know how a particular concept is indexed in that
resource, you do not need to worry about finding all possible
synonyms, because you know that index term will appear in the indexing
field of (almost) all records dealing with that concept. I say
almost, because, as I mentioned before, there is a certain degree of
inconsistency between indexers which does sometimes result in relevant
records not being retrieved because an indexing term which should have
been applied was not. For this reason, even when I do use thesaurus
terms as a basis for a search (eg MeSH terms in a Pubmed search), I
tend to start by using them within a free text search, to mop up any
articles which the indexers might have missed. However, sometimes
this can result in too many hits, which will include irrelevant
material. For example, my search term might appear within a sideline
comment. To illustrate this, imagine an article whose abstract starts
like this, “Thalidomide achieved notoriety when its use in pregnancy
resulted in multiple birth defects. However, it is now finding new use
in a number of severe conditions, such as leprosy… “ In this case,
the article will not be about the birth defects caused by thalidomide.
However, if you had been free text searching on “birth defects”, you
would have retrieved it. One way to narrow down the search is to
limit some or all of your search terms to the indexing terms field.
I’m going to use MeSH as an example, since I will merge this into an
answer to your requests for tips on using MeSH. Before I do I just
want to make a couple more general points.
Information scientists talk about the concepts of recall and
precision. Recall is the percentage of relevant documents retrieved
out of the complete population of documents being searched. So, if
you have a collection of 1000 documents, of which 500 are relevant to
your search, and you retrieve 200 of these, the recall rate of your
search will be 40%. Precision is the percentage of relevant documents
among the documents you have retrieved in your search. So, if your
search retrieved 400 documents in total, including the 200 relevant
ones, the precision rate will be 50%. Obviously the ideal is 100%
recall and 100% precision, in this example getting out all 500
relevant documents and no others. However, studies on information
retrieval have shown that even in the best-designed and best-indexed
databases, with the best possible search strategies, you maximise one
at the expense of the other. Therefore, the more precise and narrow
you make your search, the more risk you run of not retrieving relevant
material. On the other hand, the more you attempt to make your search
as exhaustive as possible, the more you will have to wade through lots
of irrelevant hits. Searches are never perfect. They are always a
compromise somewhere along the line between the two ideals of maximum
recall and maximum precision.
Another point I want to make is that when you do use a bibliographical
database such as Medline on Pubmed or Embase, you are not going to get
the most current information. The time for an article to be published
after being submitted to a journal can be as little as a couple of
months and as much as 2 years. To this you also have to add the time
it takes for an issue of a journal to be indexed into the database
after it has appeared, which can be anything from a few days to
several months, depending on whether it is seen as a high or low
priority publication. Also, these databases will not include any
routine reporting of adverse events, but only reports of case
histories, clinical trials, clinical studies and other observations
that have been deemed sufficiently noteworthy to be published as a
scientific paper. So these databases will really be more useful as a
backup rather than as the first port of call, if your main interest is
to obtain timely and exhaustive information. Something like the AERS
data, which has already been mentioned in previous questions, might be
more useful, although even that is only updated quarterly.
Finally, you also need to remember that big thesauri are only updated
maybe once a year, maybe even less frequently, and that maintainers of
thesauri need time to decide which new terms should be included. That
means that some of the latest scientific concepts and newly discovered
drugs and illnesses might take a while to filter into the approved
indexing terms. That is why it is always useful to use free text as
well as controlled vocabulary searching, adjusting your strategy as
you go, on the basis of the results obtained.
To return to narrowing down the search through the use of the indexing
terms:
On the NLM web site, you can go into the MeSH Browser at
http://www.ncbi.nlm.nih.gov/entrez/meshbrowser.cgi and put “birth
defects” into the browser search box. Here are the results:
http://www.ncbi.nlm.nih.gov/entrez/meshbrowser.cgi?retrievestring=&mbdetail=n&term=birth+defects
Notice that MeSH tells you it does not include the term “birth
defects” in its lists, but it tells you that it is associated with the
MesSH term “abnormalities” and includes a hierarchical display of
where that term occurs in the thesaurus. You will notice that as well
as a list of abnormalities which are given specific names, eg Cri du
Chat Syndrome, and a list of abnormalities classified by the organ or
system affected, there is also a term which covers drug-induced
abnormalities. If you are actually wanting to find articles about
drug-induced birth defects in your search, this is the term to go for.
Click on the link for that term, and you will be taken to a page with
a definiton of exactly what topics are indexed by this term:
“Congenital abnormalities caused by medicinal substances or drugs of
abuse given to or taken by the mother, or to which she is
inadvertently exposed during the manufacture of such substances. The
concept excludes abnormalities resulting from exposure to
non-medicinal chemicals in the environment.”
http://www.ncbi.nlm.nih.gov/entrez/meshbrowser.cgi?term=Abnormalities,+Drug-Induced&retrievestring=&mbdetail=n
If you click the ADD button on this page, the term will be
automatically inserted into your Medline search query. However, it is
often useful to find out if there are any specific ways in which this
thesaurus term can be used.
Click on “detailed display” and you are taken to
http://www.ncbi.nlm.nih.gov/entrez/meshbrowser.cgi?term=Abnormalities,+Drug-Induced&retrievestring=&mbdetail=y
On this page you can see when the term was introduced into MeSH
(useful to know if you want to be sure that older information will
also be indexed with it).
You can also find a whole list of subheadings, which can be used to
narrow down your search even more: blood, classification,
complications, diagnosis, drug therapy, economics, embryology,
enzymology, epidemiology, ethnology, etiology, genetics, history,
immunology, metabolism, mortality, nursing, pathology,
physiopathology, prevention and control, psychology, radiography,
radiotherapy, rehabilitation, surgery, therapy, ultrasonography,
urine, veterinary
You can select one or more of these as you want.
Different MeSH terms will offer a choice of different subheadings. I
already showed you previously how a drug name that is a MeSH term will
enable other subheadings, such as “adverse effects” and “toxicity”.
You will have noticed that the definition of the term we are looking
at excludes congenital abnormalities resulting from environmental
exposure. At the bottom of the detailed description page you will
find a “see also” reference to the MeSH term “teratogens”. That means
using this term will retrieve related information. If you go to the
description http://www.ncbi.nlm.nih.gov/entrez/meshbrowser.cgi?term=Teratogens&retrievestring=&mbdetail=y
you will see that this term enters the MeSH hierarchy under
“Environmental Pollutants, Noxae, and Pesticides” ie it will cover
those other types of birth defects not covered by the term
“Abnormalities, drug-induced.”. Note, however, that it also appears
in the hierarchy under “Chemical Actions” and will therefore be used
to index articles which discuss the mechanisms through which
congenital defects induced by all kinds of chemicals are induced.
You will also see that the detailed description page has an option to
select not to “explode” your search, by including all the terms under
the term you have selected. The reason for having an “explode”
facility, which is the default on a Pubmed search using MeSH terms,
is because the usual policy in database indexing is to assign the most
specific terms possible at the lowest level of the hierarchy.
“Abnormalities, drug-induced” does not have any terms below it, so let
us go one level up to “Abnormalities” to see what this means.
http://www.ncbi.nlm.nih.gov/entrez/meshbrowser.cgi?term=Abnormalities&retrievestring=&mbdetail=y
You will note that there are a number of lower level headings to this
term. Suppose an indexer processes a review paper which deals with
congenital abnormalities of all kinds. This might well be assigned
the term “Abnormalities”, especially if a lot of other indexing terms
are needed to describe the rest of the content. However, a paper
which deals specifically with drug-induced abnormalities will be
assigned ““Abnormalities, drug-induced”. If you specify the MeSH term
“Abnormalities”, you will get all documents which have been assigned
that term, and all documents which have been assigned any of the terms
below it in the hierarchy. The assumption is that you are interested
in retrieving all and everything about all types of abnormalities.
Why might you choose not to explode? Really, only if you find you
have retrieved more information that you can cope with. If you have
no other way in which to restrict your search, this might be the only
way to bring it down to a manageable size.
There is one further way to restrict a search from the detailed
description page. That is to limit the MeSH term to Major Topic
headings only. This means you will only retrieve articles in which
the term is linked to the main topic of the paper, and will exclude
more passing treatment.
You can also restrict your search from the main Pubmed search page:
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed If you click
on “Limits” you will be taken to a series of drop-down menus, through
which you can limit your search to human or animal studies, by date of
publication, age-group, type of paper (article, review, case study
etc) and various other criteria, including individual fields.
There is a tutorial about searching PubMed:
http://www.nlm.nih.gov/bsd/pubmed_tutorial/m1001.html
There is also a help page with information on how to use more
sophisticated tricks such as truncation, Boolean logic, field tags etc
http://www.ncbi.nlm.nih.gov/entrez/query/static/help/pmhelp.html#SearchFieldQualification
I will be happy to discuss any of these in answer to a request for
clarification if you require any further explanations.
If all this is too much to cope with, there is another trick you can
use, which works on any database. Go through articles which you
already have, and find one which contains the sort of information that
you are seeking. Go back into Pubmed and search for that article, by
putting in author names and/or key words from the title. When you
display the results, you will see there is a drop-down menu marked
“Display”. Select “MEDLINE” Here is an article I chose at random
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&list_uids=12444541&dopt=Abstract
It does not work if I try to give you a URL for the Medline display,
so you will have to perform that selection yourself from the Abstract
display I have cited. You can now see all the MeSH terms that have
been used to index this article. This will give you ideas for which
ones to use in your searches. Doing this exercise showed me a term
absolutely relevant to your interests, of which I was not aware:
“Product Surveillance, Postmarketing” All database will give you the
option of viewing the full record with assigned indexing terms. This
is also a good way to use a database when no thesaurus is available to
you.
Elsevier does have a thesaurus that is used for its EMBASE database It
is called EMTREE. MeSH terms have been included into EMTREE. Like
MeSH, EMTREE is a general thesaurus for all medical and biomedical
topics that are covered in the database.
This is available in hard copy:
EMTREE 2003
The Life Science Thesaurus
Paperback ISBN: 0-444-51303-5 3340 pages (in 3 volumes)
“a sophisticated, hierarchically structured controlled vocabulary
arranged in a cascading "tree" and used for subject indexing in
EMBASE, EMBASE.com, and customer-specific solutions. The printed
EMTREE includes more than 45,000 drug and medical terms, 10,000
numeric codes and over 190,000 synonyms. It incorporates MeSH terms
used by the National Library of Medicine, as well as almost 20,000 CAS
registry numbers assigned by the Chemical Abstracts Service. In 2003 a
significant number of general biological terms have been added.”
http://www.elsevier.nl/inca/publications/store/6/7/2/8/3/6/index.htt
The thesaurus is also included in the various subsets of the EMBASE
database that can be obtained on CD or via the Internet by
subscription, with quarterly updates, through Ovid/Silver Platter.
For example, there is a Drugs and Pharmacology subset:
http://www.ovid.com/site/catalog/DataBase/66.jsp?top=2&mid=3&bottom=7&subsection=10
“contains only information pertinent to the specialty of drugs and
pharmacology. The database contains more than 1,560,000 abstracts and
citations, providing comprehensive coverage of drug and pharmacology
literature. Topics included are the effects and uses of drugs and
potential drugs, clinical and experimental aspects of pharmacokinetics
and pharmacodynamics, side effects, and adverse effects.
Manufacturers' and drug trade names are included where available. The
database includes EMTREE”
Unfortunately, price information is not given for either of these
products.
COSTART (Coding Symbols for Thesaurus of Adverse Reaction Terms) seems
to have been supplanted by MedDRA. Here is some information from
Coding Plus, a company involved in providing MedDRA training and
coding solutions:
“MedDRA will be the language of communication among companies,
regulators, and CROs… The MedDRA dictionary itself poses significant
challenges. It is new, complex, and roughly 10 times larger than
COSTART or WHO-ART (WHO Adverse Drug Reaction Terminology)…
http://www.codingplus.com/focus.htm
Some of the most salient points from the FAQ on the site:
“MedDRA was not designed as an adverse reaction thesaurus (ART), as
compared to COSTART or WHO-ART. MedDRA contains additional clinical
data elements such as procedures, investigations, and social history.
I recommend the creation of a MedDRA adverse reaction subset
containing only those terms that are likely to be reported as adverse
events. For example, an abnormal lab test result would qualify but the
name of the test would not…. The 12 COSTART Body Systems have been
replaced by 26 System Organ Classes (SOCs) in MedDRA…. The most
practical method to avoid the pitfalls of MedDRA is by learning the
dictionary. It is unreasonable to learn 45,000+ terms, so I recommend
identifying those SOCs of particular interest and thoroughly
understanding them”
“When should a company convert to MedDRA? … Mergers create the unique
opportunity to switch to MedDRA as a common dictionary. New clinical
projects and postmarketing databases are reasonable candidates for
high-priority MedDRA implementation. Companies which plan electronic
transmission of safety reports to regulatory agencies will have to use
MedDRA…. Many potential users are awaiting a specific regulatory
mandate, even though the FDA has been using MedDRA since 1997 in the
AERS system, and other regulators receive complimentary copies of
MedDRA. The strategic implications of MedDRA implementation should
stimulate companies to begin planning, with the same degree of urgency
and organizational support as for Y2K. Some large organizations will
require a MedDRA project plan of as much as 2 years, while smaller
companies without marketed products may get up to speed within 6-9
months. Recent presentations at industry meetings indicate that MedDRA
is imminent, and 2 years is a long estimate for MedDRA
implementation.” http://www.codingplus.com/faq.htm
Maintenance and Support Services Organization (MSSO) is the sole
licensee and distributor of MedDRA subscriptions.
http://www.meddramsso.com./
“Major global regulatory authorities (within the US, Europe and Japan)
are adopting MedDRA and moving toward requiring its use. The FDA, for
example, has already implemented MedDRA within its Adverse Event
Reporting System (AERS). European authorities are beginning to use
MedDRA as a key part of their electronic database systems.”
MedDRA subscriptions vary with the annual revenue of the company, the
range being $3,605- $86,994
http://www.meddramsso.com./newwebaug2001/meddramsso/subscriptions/rates2003.htm
“Pharmaceutical manufacturers in Europe, Japan, and the United States
will soon have to comply with new regulatory policies. In the next two
years, all expedited (serious) adverse drug reaction reports will be
submitted electronically to major global regulatory agencies (FDA,
EMEA, MHW), using MedDRA terminology. With over 60,000 terms, MedDRA
is more voluminous and complex than previous baseline adverse event
terminologies, which could substantially alter standard operating
procedures (SOPs) throughout the pharmaceutical industry, including
pharmacovigilance procedures, labeling, systems, and coding.”
http://www.drugsafety.com/DrugSafety/meddra/default.asp
You might be interested in the NLM’s Unified Medical Language System
(UMLS), which includes terminology from COSTART, MedDRA, and WHOART
“The purpose of the UMLS is to aid the development of systems that
help health professionals and researchers retrieve and integrate
electronic biomedical information from a variety of sources and to
make it easy for users to link disparate information systems,
including computer-based patient records, bibliographic databases,
factual databases, and expert systems…. NLM encourages broad use of
the UMLS products by distributing quarterly editions free-of-charge
under a license agreement.”
http://www.nlm.nih.gov/pubs/factsheets/umls.html
The UMLS consists of three resources:
1. UMLS Metathesaurus “contains information about biomedical concepts
and terms from many controlled vocabularies and classifications used
in patient records, administrative health data, bibliographic and
full-text databases and expert systems. It preserves the names,
meanings, hierarchical contexts, attributes, and inter-term
relationships present in its source vocabularies; adds certain basic
information to each concept; and establishes new relationships between
terms from different source vocabularies… The January 2003AA edition
includes 875,255 concepts and 2.14 million concept names in over 100
biomedical source vocabularies, some in multiple languages.
Compared to last year's January release, there are 98,315 more
concepts with 226,729 more names in the concepts file, MRCON, from 100
vocabulary sources, counted as the number of source families by
language. New vocabularies in the 2003AA Metathesaurus are: Portuguese
and Spanish versions of MedDRA, Version 5.1; the NCBI Taxonomy from
the National Center for Biotechnology Information; and the developing
RxNorm Clinical Drug vocabulary from the National Library of
Medicine.”
“NLM does not charge for the Metathesaurus (or other UMLS products)
and it is available to both U.S. and international users… [who]… must
sign a License Agreement… Some uses of some Metathesaurus source
vocabularies require separate agreements, which may involve fees, with
the individual vocabulary producers. The UMLS Metathesaurus is
available to licensees via ftp, Web interface, and applications
program interface (API) from the UMLS Knowledge Source Server. It is
also available on CD-ROM by explicit request.”
http://www.nlm.nih.gov/pubs/factsheets/umlsmeta.html
The sources are listed at
http://www.nlm.nih.gov/research/umls/METAB2.HTML
2. Specialist Lexicon “Lexical information includes syntactic
category, inflectional variation (e.g., singular and plural for nouns,
the conjugations of verbs, the positive, comparative, and superlative
for adjectives and adverbs), and allowable complementation patterns
(i.e., the objects and other arguments that verbs, nouns, and
adjectives can take). The lexicon recognizes eleven syntactic
categories, or parts of speech: verbs, nouns, adjectives, adverbs,
auxiliaries, modals, pronouns, prepositions, conjunctions,
complementizers, and determiners”
http://www.nlm.nih.gov/pubs/factsheets/umlslex.html
3. Semantic Network “provides a consistent categorization of all
concepts represented in the UMLS Metathesaurus. The 54 links between
the semantic types provide the structure for the Network and represent
important relationships in the biomedical domain. All information
about specific concepts is found in the Metathesaurus; the Network
provides information about the basic semantic types that are assigned
to these concepts, and it defines the relationships that may hold
between the semantic types. The Semantic Network serves as an
authority for the semantic types that are assigned to concepts in the
Metathesaurus. It defines these types, both with textual descriptions
and by means of the information inherent in its hierarchies.”
http://www.nlm.nih.gov/pubs/factsheets/umlssemn.html
These might also be of interest:
Australian Adverse Drug Reactions Bulletin available online at
http://www.health.gov.au/tga/adr/aadrb.htm
WHO Drug Alerts http://www.who.int/medicines/organization/qsm/activities/drugsafety/orgqsmalerts.shtml
WHO Drug Information http://www.who.int/druginformation/
Adverse Drug Reaction Bulletin Bimonthly Annual subscription is $73
for individuals, $115 for institutions
http://www.lww.com/product/0,0,0044-6394,00.html
Canadian Adverse Drug Reaction Newsletters. Quarterly, online at
http://www.hc-sc.gc.ca/hpb-dgps/therapeut/zfiles/english/publicat/adrindex_e.html
To receive the Newsletter and Advisories free by email, join Health
Canada's Health_Prod_Info mailing list. Go to
www.hc-sc.gc.ca/hpb-dgps/therapeut/htmleng/adr.html and click on
"subscribe".
Adverse Reactions Titles (Section 38 EMBASE)
“Each year, approximately 400,000 articles appearing in over 4,000
biomedical and 200 chemical serials are scanned at Elsevier by medical
specialists and pharmacologists. Each year, an average of 100,000
articles dealing with drugs and other biologically active substances
are included in EMBASE, the Excerpta Medica database, of which an
average of about 8,000 fall within the scope of Adverse Reactions
Titles.”
It is expensive!!! Year 2003, Volume 38, 12 issues, ISSN: 0001-8848,
Institutional price: USD 13,084 for all countries except Europe and
Japan, JPY 1,550,700 for Japan, EUR 11,696 for European countries
http://www.elsevier.nl/inca/publications/store/5/0/6/0/0/7/
Drug Safety
“Expert evaluations in pharmacovigilance, pharmacoepidemiology and
benefit-risk assessment..
Articles focus on the practical aspects of ensuring the safest
possible use of drugs, including analysis and interpretation of
adverse effects, post-marketing surveillance and
pharmacoepidemiology.”
http://www.adis.com/page.asp?ObjectID=46
EUDRA Vigilance “the new European data-processing network and database
management system for the exchange, processing and evaluation of
Individual Case Safety Reports related to medicinal products
authorised in the Community. The launch of EudraVigilance is a
significant milestone with major implications on the conduct of
pharmacovigilance and the implementation of the electronic exchange of
pharmacovigilance data” http://www.eudravigilance.org/
I hope this information is the sort you were seeking, but please ask
for further clarification if required.
Search strategy: Information on search strategies and how to use MeSH
and Pubmed is from my professional knowledge. I conducted individual
searches on names of products, such as EMTREE and MedDRA, in order to
obtain supplementary information, as well as using the search
expressions: adverse drug bulletin and adverse drug newsletter to find
details of printed and online alerts. |
,
8 Comments
)