Content uploaded by bhaynes@mcmaster.ca Haynes
Author content
All content in this area was uploaded by bhaynes@mcmaster.ca Haynes on Jul 04, 2016
Content may be subject to copyright.
Annals of Oncology 9: 377 383, 1998.
1 1998 Kluwer Academic Publishers. Primed
in
the Netherlands.
Review on evidence-based cancer medicine
Searching the medical literature for the best evidence to solve
clinical questions
D.
L. Hunt, R. B. Haynes & G. P. Browman
Health Information Research Unit. Department of Clinical Epidemiology and Biostatistics, Me Master University Faculty of Health Sciences,
Hamilton, Ontario, Canada
Key words: continuing education, evidence-based medicine, information retrieval, information technology, Internet, keeping
up-to-date
Introduction
Questions about the best care for our patients' problems
arise frequently in clinical practice. The pace of devel-
opment of new evidence from research is too quick
for standard textbooks to be of dependable help. Most
often, these questions relate to the best means of con-
firming a diagnosis or the optimal therapeutic approach
for a given condition. At other times, risk factors, screen-
ing, or prognosis may be the principal concerns. Or the
questions may relate to the co-morbid conditions of the
patient, such as the interactions of their medications for
heart disease or diabetes with the treatments that they
might receive for their cancer. When these questions
arise,
it is unlikely that they will be answered accurately
(or at all) unless we are able to find the answers quickly
and accurately. As the current best evidence on a given
topic changes at unpredictable times, even the most
experienced clinician cannot assume that she knows
the answer without looking. Fortunately, the advent of
better research, better information resources, and better
information technology makes it possible and worth-
while for all clinicians to respond to these challenges by
learning some basic literature search skills and acquir-
ing access to key evidence resources in the hospital and
clinic or at home. In this article, we will describe and
illustrate some of the skills and resources for answering
questions of relevance to the care of patients with cancer.
Clinical scenarios
Consider the following situations. You have just finished
seeing a 62-year-old man who was recently diagnosed
with locally advanced squamous cell carcinoma of the
oral cavity. The patient is a smoker and drinks alcohol
moderately; he has no other significant medical condi-
tions and the tumour appears to be amenable to surgery.
A visiting medical student is doing an elective with
you. She recalls from a recent lecture that chemotherapy
can be effective at improving survival for patients with
squamous cell head and neck cancers and asks you
about offering this patient a short course of chemo-
therapy before surgery. You are quite certain that neo-
adjuvant chemotherapy is not beneficial for patients
with locally advanced oral cancer but as you are about
to explain this, you realize that you do not know whether
this has been properly evaluated. You quickly turn the
tables on the student and suggest that she provide a
reference from the literature for your next meeting. To
avoid being upstaged (again!), you set off on your own
search.
Before you have had a chance to begin finding an
answer, a colleague who coordinates the breast cancer
screening program at your hospital approaches you with
another question. She is concerned about one of her
recent cases. After having a mammogram that was
interpreted as being suspicious for malignancy, the
woman went on to have a breast biopsy. This showed no
evidence of cancer. The patient was naturally elated with
the good news. On reviewing her mammogram with a
second radiologist, it was suggested that the biopsy was
not needed in the first place: the mammogram was
negative. Your colleague wonders whether any changes
could be made to the mammography screening program
to decrease the number of unnecessary breast biopsies,
without missing additional cases of breast cancer. She is
considering proposing a policy at the next medical staff
meeting that would require all mammograms to be read
by two radiologists before being reported, but is inter-
ested in your thoughts.
How would you go about trying to address these
questions? The first step is to carefully define the ques-
tions.
The second step is to retrieve the best current
evidence that pertains to the question, a task that
includes choosing an appropriate evidence resource;
developing and executing an effective search strategy;
critically appraising the results; and refining the search
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
378
strategy or moving on to a different database if neces-
sary. Finally, the findings must be applied in a way that
fits the clinical circumstances of the patient and respects
their wishes.
Defining the question
The first step for any evidence search is to formulate
a 'well-built clinical question'
[1].
This entails identifying
a question that is important to the patient's well-being,
is interesting to you, and that you are likely to encounter
on a regular basis in your practice. (For practical pur-
poses,
it is more efficient and usually better for your
patient if you seek consultants for questions that you
seldom address in your practice.)
To be answerable, the question must be specified
clearly so that it includes a specific patient group, the
diagnostic test or treatment or other clinical issue that
you are addressing, and the outcome that you are inter-
ested in. For example, asking whether chemotherapy
helps patients with breast cancer is a very broad ques-
tion that would be difficult to answer. Asking whether
post-menopausal women with Stage
1
breast cancer that
has been treated with lumpectomy and local radiation
therapy have prolonged survival if treated with tamox-
ifen is an example of a focussed question that is more
amenable to answering. For the two scenarios above, the
questions could be: For patients with locally advanced
and surgically resectable squamous cell cancer of the
head and neck, what is the effect of neoadjuvant chemo-
therapy on survival? For women undergoing screening
mammography, what are the effects of duplicate inde-
pendent interpretation on the diagnostic accuracy and
costs of screening?
Finding best evidence
MEDLINE and CANCER LIT
MEDLINE is one of the most readily available resour-
ces for locating important studies. This multipurpose
database of medical literature citations and abstracts
is produced by the US National Library of Medicine
(NLM). Over 7,000,000 clinical and pre-clinical studies
are indexed in MEDLINE. CANCERLIT is a special
subset of MEDLINE prepared by the National Cancer
Institute's International Cancer Information Center.
CANCERLIT includes almost all of the cancer-related
citations in MEDLINE, but also has abstracts from
meeting proceedings, book citations and theses.
Accessing CANCERLIT is generally quite easy.
Many hospital and academic libraries have CD-ROM
based systems such as OVID, Aries, CD-Plus, or Silver-
Platter. Online access using a modem or internet line
is available through a number of vendors, including
MEDLARS, DIALOG, and HealthGate. Some ven-
dors,
such as MEDLARS, Ovid, CDP Technologies,
Knight-Ridder Information, Inc., and HealthGate pro-
vide both online access and access over the Internet.
Ovid also provides access to a growing number of full
text journal articles. A more complete listing of infor-
mation about CANCERLIT distributors is available at
http://www.graylab.ac.uk/cancernet/400006.html.
The majority of these vendors also provide MED-
LINE access. Recently, however, MEDLINE access
became available free of charge, for anyone who has
access to the Internet, through PubMed (http://
www.ncbi.nlm.nih.gov/PubMed) and Internet Grateful
Med (http://igm.nlm.nih.gov). Both services provide
access to all MEDLINE citations as well as the Pre-
MEDLINE database, which includes citations and their
abstracts before they have been indexed by the US Na-
tional Library of Medicine. PubMed also provides links
to a small but growing number of fulltext articles via the
internet home pages of
journals.
Internet Grateful Med
also includes several additional databases such as AIDS-
LINE and SDILine (which permits storing your own
search strategies for periodic updates on specific topics).
MEDLINE and CANCERLIT both contain an enor-
mous number of citations, describing research from
'bench to bedside'. This comprehensive approach has its
price. Searching for high quality, clinically relevant stud-
ies requires thought and preparation and even then may
miss important studies (low sensitivity) while retrieving
many studies that are not relevant to the searcher's
purpose (low precision). PubMed provides direct access
to special search strategies that filter the literature
for studies that are most likely to be applicable to clin-
ical practice (http://www.ncbi.nlm.nih.gov/PubMed/
clinical.html). Nevertheless, a basic understanding of
how articles are indexed in MEDLINE is highly desir-
able for clinical users.
Luckily, some libraries offer training courses in MED-
LJNE searching, and the approach to using CANCER-
LIT is identical. Librarians have a wealth of experience
in using these services and are valuable consultants for
clinicians. Clinicians in the United Kingdom, United
States, and Canada can call the Health Care Informa-
tion Service (0171 412 7477), the National Library of
Medicine (1-800-272-4787), or the Canada Institute for
Scientific and Technical Information (1-800-668-1222),
respectively, to inquire about regional medical libraries
and programs that have been established to provide
MEDLINE training.
Returning to the therapeutic question at hand, you are
interested in finding studies that have evaluated the role
of neoadjuvant chemotherapy for patients with squamous
cell head and neck cancer. MEDLINE indexes citations
for content and methodology using a set of over 14,000
specific terms and over 18,000 synonyms and other terms.
This is known as the Medical Subject Heading (MeSH)
vocabulary. During the indexing process, an article is
assigned a number of MeSH terms. Topics that are the
major focus of the article are specially coded to indicate
this.
Some MEDLINE systems do this by referring to
these topics as major subject headings, while other sys-
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
379
terns place an asterisk (*) in front of the medical subject
heading.
It is important to note that when the appropriate
subject heading for an article is being chosen, the most
specific index terms available are used. For example, if
an article is about apples and oranges (and assuming
that the MeSH vocabulary actually included such edible
products), it would be indexed using these terms. Doing
a search using the MeSH term 'fruits' would not retrieve
the article because only articles that deal with fruits in
general would be indexed using the term 'fruits'. To find
the article on apples and oranges, the MeSH term 'apple'
or 'orange' would need to be used, or a special feature of
MEDLINE called 'exploding' could be used. Asking
MEDLINE to 'explode' a term indicates that you want
all articles that include that term to be retrieved, along
with all articles that have been indexed using more
specific topics. To continue with the food theme, a search
using the term 'explode fruits' would retrieve all citations
that discussed fruits in general, as well as articles on
apples and oranges (and pomegranates and so on).
Using Internet Grateful Med (IGM) for your search,
you could begin by using the MeSH term 'head and neck
neoplasms'. IGM automatically explodes this term to
include all head and neck neoplasms. This retrieves over
16,000 citations in the 1994 to September 1997 version of
MEDLINE. To whittle this down, you could do a second
search using the MeSH term 'chemotherapy, adjuvant'.
This second search retrieves over 3000 citations. How-
ever, you are only interested in articles that deal with
both chemotherapy and head and neck cancer. To limit
the list of citations to these articles, you will want to
combine them. MEDLINE allows citations from differ-
ent searches to be combined in various ways. For exam-
ple,
if only articles that appear in both sets are desired,
you can instruct MEDLINE to combine the sets using
the term 'AND'. On the other hand, if you want to pool
all of the citations that appear in either of the sets of
retrieved articles, you can combine the sets using the
term 'OR'. For the head and neck cancer search, you
should combine the search results using the term 'AND'.
This produces a much smaller list of articles, but there
are still over 400, too many to read through in a limited
amount of time!
As you are most interested in finding information
about how neoadjuvant chemotherapy affects survival
in patients with head and neck malignancies, narrowing
your search to include only articles that deal with mor-
tality would be appropriate. One approach could be to
use the MeSH term 'mortality'. Alternatively, a textword
search could be used. Textword searching involves ask-
ing MEDLINE or CANCERLIT to search all of the
titles and abstracts in its database for any occurrence of
a term. This approach is especially useful when search-
ing for information about a relatively new topic, such as
a new drug or procedure, before it has its own MeSH
term. Also, it is the only way to search for citations in
the PreMEDLINE database because these references
have not been assigned MeSH terms. An important
feature of textword searching is the ability to search for
all occurrences of
a
certain series of letters, regardless of
what letters come afterwards. For example, a textword
search using the term 'neoplas:' (the symbols'for ""are
used by many, but not all, CANCERLITand MEDLINE
systems to denote this concept) would retrieve citations
with a variety of terms including 'neoplasm','neoplasms'
and 'neoplastic'. Searching using the textwords 'mortality
or survival' and then combining the results with the
previous search using 'AND' produces a list of about 27
citations in the 1994-1997 MEDLINE database. You now
have a manageable list of citations to review.
As an alternative to the textword search, you could
limit the search to meta-analyses, that is, review studies
that summarize the evidence across relevant trials. IGM
provides a point-and-click feature for 'publication types',
including reports that use meta-analysis. Combining
this term with the first two search terms above retrieves
eight articles. Three of these reports include meta-anal-
yses of studies of adjuvant chemotherapy for head and
neck cancer. Retrieving the first relevant article that
your search has identified [2], you find that the paper is
a meta-analysis of previous studies that have evaluated
the role of chemotherapy in the management of squa-
mous cell carcinoma of the head and neck region. Un-
fortunately, the meta-analysis publication type is not
consistently applied in MEDLINE as yet, as is the case
for several related terms for review articles.
The temptation, naturally, will be to scan the abstract
and then jump to the conclusions. This approach, how-
ever, can contribute to drawing inappropriate conclu-
sions from studies. Rather, it is important to begin by at
least quickly assessing how a study was conducted, and
then to proceed to evaluate the findings if the study
methodology is acceptable. Several publications [3-9]
have presented systematic approaches to assessing ar-
ticles about many different types of studies, including
therapeutic interventions, diagnostic tests, prognosis
studies, and systematic reviews. A simplified approach
to critical appraisal appears in Table 1. In this case, the
article 'passes muster' as a systematic review. Its evi-
dence-based conclusion is that chemotherapy signifi-
cantly increases treatment toxicity and only improves
survival when used concurrently with local definitive
therapy, but not when used as induction treatment.
While in this case it is not crucial to limit the set yet
further, at times you will find it very helpful to combine
a content search, such as the one just completed, with a
methodological quality search, intended to limit the
number of studies retrieved to those that are most likely
to be methodologically sound. This can be accomplished
by adding methodological terms into search strategies
[10,
11]. Different terms are helpful for identifying studies
pertaining to questions of therapy, diagnosis, prognosis,
and etiology (see Table 2). PubMed has these methodo-
logical filters built in so that you do not need to keep a
copy of them in your lab coat. Most MEDLINE systems
will allow users to store such filters, to be evoked when
desired.
Using the most sensitive strategy in Table
1
for identi-
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
380
Table I. Guidelines for critical screening of journal articles on clinical research topics."
TherapyDiagnosisPrognosisCausationReviews
Random allocation of
patients to comparison
groups
Clearly identified com-
parison groups, one being
free of the disorder
Outcome measure of Objective or reproducible
known or probable clinical diagnostic standard,
importance applied to all participants
Inception cohort, early in
the course of the disorder
and initially free of the
outcome of interest
Objective or reproducible
assessment of clinically
important outcomes
Clearly identified com-
parison group for those
at risk of, or having, the
outcome of interest
Blinding of observers of
outcome to exposure;
blinding of observers of
exposure to outcome
Hollow-up of at least 80%Blinded assessment of test
and diagnostic standardFollow-up of at least 80% Follow-up of at least 80%
Comprehensive search for
relevant articles
Explicit criteria for rating
relevance and merit
Inclusion of all relevant
studies
" Based on references 3-9.
Table
2.
Search strategies for identifying studies relating to treatment,
diagnosis, prognosis, or etiology using MEDLINE.a
Treatment
Combination of terms with placebo.tw. ('tw' indicates textword)
best specificity: OR double.tw. AND blind:.tw.
Combination of terms with Randomized controlled trial.pt.
best sensitivity: OR random:.tw.
OR drug therapy (as a subheading
of the subject)
OR therapeutic use (as a subheading
of the subject)
Diagnosis
Combination of terms with
best specificity:
Combination of terms with
best sensitivity:
Prognosis
Combination of terms with
best specificity:
Combination of terms with
best sensitivity:
Etiology or cause
Combination of terms with
best specificity:
Combination of terms with
best sensitivity:
Explode 'sensitivity and specificity'
OR predictive.tw. AND value:.tw.
Explode 'sensitivity and specificity'
OR explode diagnosis
OR sensitivity.tw.
OR specificity.tw.
OR diagnostic use (as a subheading
of the subject)
Prognosis
OR survival analysis
Incidence
OR explode mortality
OR follow-up studies
OR prognos:.tw.
OR predict:.tw.
OR course:.tw.
OR mortality (as a subheading of the
subject)
Cohort studies
OR case-control studies
Explode cohort studies
OR explode risk
OR odds.tw. AND ratio:.tw.
OR relative.tw. AND risk.tw.
OR case.tw. AND control.tw.
" Based on references 10 and 11.
fying all therapy trials, more than 200,000 citations are
retrieved from 1994-September 1997 in MEDLINE.
Combining ('ANDing') these with the results of the
previous content search reduces the number of citations
to 19. Reviewing these, 16 have to do with esophageal
carcinoma, one is about nasopharyngeal cancer, and
two are about the treatment of head and neck cancer.
One of the two papers on head and neck cancer treat-
ment is a non-randomized study assessing the role of
chemotherapy and radiotherapy in organ preservation.
This leaves but one potentially relevant study, the same
one that was retrieved by the search using meta-analysis
as a publication type [2].
An alternative search strategy is to determine whether
someone has already done all the work for you, looking
for evidence-based practice guidelines based on a cur-
rent systematic literature review. MEDLINE has a pub-
lication type, 'practice guidelines', but a search for 'head
and neck neoplasms' did not yield any citations for
locally advanced head and neck cancer in this instance.
MEDLINE indexing for practice guidelines is improv-
ing but is not yet perfect, and some ingenuity and luck in
searching are required.
EMBASE
Excerpta Medica produces a comparable database to
MEDLINE that may be more accessible in Europe.
EMBASE and MEDLINE overlap in coverage but are
not coincident, so that some valuable information may
be in EMBASE that is not included in MEDLINE,
particularly for pharmaceutical trials. EMBASE is con-
siderably more expensive to use than MEDLINE (the
production of which is supported by US taxpayers and
subsidised by foreign end-users). Clinical trials are not
as consistently indexed in EMBASE.
The Cochrane Library
While the MEDLINE searches were successful in locat-
ing a relevant article, you may be wondering what other
electronic resources are available to facilitate locating
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
381
high quality studies. The Cochrane Library is one such
resource, produced by the Cochrane Collaboration, an
international organization that prepares, maintains, and
disseminates systematic reviews of randomized trials of
many health care interventions. The Cochrane Library,
available on CD-ROM from UpDate Software in Ox-
ford, is updated quarterly and contains four biblio-
graphic sections: the Cochrane Database of Systematic
Reviews (CDSR), the Database of Reviews of Effective-
ness (DARE), the Cochrane Controlled Trials Registry
(CCTR), and the Cochrane Review Methodology Data-
base (CRMD). CDSR consists of complete reports of
Cochrane Collaboration systematic reviews. This section
also lists the protocols for Cochrane systematic reviews
under development. DARE comprises systematic reviews
that have been published outside of the Collaboration.
Certain methodological quality criteria have to be met for
a systematic review to be included in DARE. Unfortu-
nately, some of the citations in DARE lack abstracts.
The third section of the Library, the CCTR, contains a
growing list of over 160,000 references to therapeutic
intervention trials. Once again, though, many of these do
not have abstracts available within the Library. CDSR is
also available via the Internet (http://www.medlib.com;
http://www.hcn.net.au/healthbase/cochrane/intro.htm).
The Cochrane Collaboration also assists the US National
Library of Medicine to improve the consistency of index-
ing randomised controlled trials and controlled clinical
trials that are not randomised.
Searching The Cochrane Library is easy. A simple
search mechanism permits words and terms to be en-
tered for full text searching across all databases. For
example, entering 'head and neck cancer' retrieves no
citations in CDSR, four in DARE, and 232 in CCTR.
The first of the citations in DARE is the El-Sayed meta-
analysis article [2]. The Cochrane Library also includes
an advanced search engine that allows both textwords
and MeSH terms to be used, and sets of citations can be
combined using the terms AND and OR. To answer the
question of the role of neoadjuvant chemotherapy in
patients with head and neck cancers, one can begin by
clicking on the 'MeSH' button on the advanced search
screen. Enter the term 'neoplasms'and then select 'Head
and neck neoplasms' from the listing by double-clicking
on it. Asking the program to 'Explode and search' using
this term retrieves 578 citations (using Issue 3 of the
1997 Cochrane Library). A search for the MeSH term
'chemotherapy, adjuvant' yields no citations, so a text-
word search using the term 'chemotherapy' is appro-
priate. This retrieves 4421 citations. Combining the two
sets using the AND function produces a smaller collec-
tion of 178 citations. Double-clicking on this line allows
you to see the citations. The Cochrane Library always
indicates how many 'hits' there were in each of the four
different sections in the top panel of the screen. In this
case,
it found no citations in the CDSR, four citations
in DARE, and 174 citations in the CCTR. To see the
citations in the DARE, double-click on the appropriate
line.
A further subclassification appears indicating that
one citation is in the 'Abstracts of quality assessed sys-
tematic reviews', one citation is in the 'Other assessed
reviews' section, while two citations appear under 'Other
reviews'. The first of these is the El-Sayed paper [2].
Moreover, this version includes an independently pre-
pared structured abstract that summarises the method-
ology and results of the review and indicates that the
meta-analysis has some weaknesses. Nevertheless, the
commentary accepts the authors' conclusions that che-
motherapy has significantly increased treatment toxicity
and only improved survival when used concurrently
with local definitive therapy, but not when used as induc-
tion treatment.
Best evidence
Another valuable resource for locating high quality
studies quickly is Best Evidence, available from the
American College of Physicians and BMJ Publishing
Group. This is the electronic version of two paper-based
abstract
journals:
ACP Journal Club and Evidence-Based
Medicine. Best Evidence only includes studies that are
methodologically-sound [12] and summarizes the ma-
jority of them using a structured abstract. In addition,
each abstract is followed by a commentary, written by
a clinical expert, that is designed to place the study
findings into clinical perspective. Updated annually,
Best Evidence has articles relating to general internal
medicine dating back to 1991. Since 1995, a broader
range of articles encompassing other fields of medical
care,
such as obstetrics and gynecology, family medi-
cine,
pediatrics, psychiatry and surgery, have been
added. Some articles specifically addressing oncology-
related issues are also included. Best Evidence, however,
will generally be more useful for answering those patient
care questions that are not oncology-specific.
Turning to the clinical question on mammography, a
search using Best Evidence is straightforward. Best
Evidence supports both textword searching (using the
'Search' option) and MeSH term searching (using the
index' function). After clicking on the 'Search' button
and entering the textword 'mammography', Best Evi-
dence provides a listing of 23 citations that include this
term. Reviewing the list reveals an article specifically
dealing with double reading of mammograms. Double-
clicking the reference title reveals the full structured
abstract and commentary. This indicates that consensus
double reading of mammograms resulted in fewer false
positives and false negatives than single reading and,
because of the reductions in errors, was ultimately less
costly. Non-consensus double reading, on the other
hand, led to detection of more tumours than single
reading (but not than consensus double reading), but
also had more false positives and higher costs.
The Cochrane Library also has numerous citations
concerning breast cancer screening. Searching using the
term 'mammography' retrieves 126 references, including
two that are directly pertinent to the question of improv-
ing the yield of screening [13, 14],
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
382
Textbooks
Finally, we come to textbooks. These can be very useful
for reviewing the pathophysiology of a condition, or the
mechanism of action of a medication. Texts may also
help to broaden the differential diagnosis in more com-
plex cases. Unfortunately, texts are less useful when it
comes to topics for which the evidence can change
rapidly, such as the optimal treatment for a given con-
dition. This is because the most recent editions of many
textbooks are often two to three years out-of-date, and
many important new studies may have been published in
the interim. Also, textbooks are seldom explicit about
the evidence or the quality of the evidence used for their
recommendations.
These limitations suggest that textbooks, by and
large, should not be relied upon for therapeutic decision
making. If you find yourself managing a condition that
you have not dealt with recently, however, you may
wish to refer to a textbook. For general medical con-
ditions, an electronic textbook that is updated regularly,
such as Scientific American Medicine, is an option. For
oncology-specific topics, a good choice would be PDQ.
This is the US National Cancer Institute's cancer infor-
mation database. It is available on-line and over the
Internet (e.g., http://wwwicic.nci.nih.gov/health.htm),
as well as on CD-ROM. Regularly updated, this database
includes information on cancer treatment and screen-
ing, new investigational drugs, and ongoing clinical
trials,
as well as a listing of physicians and organizations
involved in cancer care. Information for patients is also
available.
Finding information about a malignancy in PDQ is
primitive but efficient. A table lists the types of cancers
that are discussed, and by selecting the appropriate line,
a complete description is available. Recalling the first
clinical scenario, selecting 'oropharyngeal cancer' leads
to several documents about oral cavity tumours includ-
ing diagnostic work-up, cellular classification, staging,
and treatment options. Note, however, that the sys-
tematic review that we found using CANCERLIT and
The Cochrane Library is not mentioned in the refer-
ences.
Applying the findings
One of the harshest criticisms of 'evidence-based medi-
cine'
is that it places evidence from research above all in
clinical decision making. This criticism is unfair, as
advocates of evidence-based care have clearly pointed
out that clinical decision making includes careful con-
sideration of the clinical circumstances of the patient
(including, for example, co-morbidity, other treatments,
availability of various treatment options) and the pa-
tient's wishes, preferences and rights (15). Nevertheless,
in the scenarios that we began with, the evidence seems
clear that adjuvant chemotherapy should not be used for
induction for locally advanced head and neck squamous
cell carcinoma, and that consensus duplicate readings of
mammograms should be used. Unfortunately, despite
the win-win evidence for consensus duplicate mammo-
gram reading, it is unlikely to be widely implemented
unless payors and managers can be induced to shift
resources to permit it.
Conclusion
With an ever increasing number of trials evaluating
different aspects of medical care, being able to quickly
locate valid, up-to-date information is becoming more
and more important. Textbooks simply cannot keep up
with such advances. General purpose evidence databases
such as MEDLINE, CANCERLIT and EMBASE pro-
vide access to almost all relevant studies in medicine,
but their size makes searches for clinically pertinent and
sound studies clumsy at best. New resources like The
Cochrane Library and Best Evidence that only include
high quality clinical studies and summarize the findings
are beginning to improve this process for the fields and
types of studies that they cover. But being familiar with
how to use larger bibliographic databases, such as CAN-
CERLIT, is still essential for finding the current best
evidence about many clinical questions.
Acknowledgements
This work was supported by the Health Evidence Appli-
cation and Linkage Network (HEALNet, a Canadian
Network of Centres of Excellence Program). Dr. Hunt is
supported by a HEALNet fellowship. Dr. Haynes is
supported in part by a National Health Scientist award
from the National Health Research and Development
Program, Health Canada. Dr. Browman is supported by
the Cancer Care Ontario Program in Evidence-Based
Care.
References
1.
Richardson WS, Wilson MC. Nishikawa J. Hayward RSA. The
well-built clinical question: A key to evidence-based decisions
(Editorial). ACP J Club 1995; 123: A12-13.
2.
El-Sayed S. Nelson N. Adjuvant and adjunctive chemotherapy in
the management of squamous cell carcinoma of the head and
neck region. A meta-analysis of prospective and randomized
trials.
J Clin Oncol 1996; 14: 83S-47.
3 Guyatt GH, Rennie D. Users guides to the medical literature.
JAMA 1993; 270 (17): 2096-7
4.
Guyatt GH, Sackett DL, Cook DJ. Users' guides to the medical
literature. II. How to use an article about therapy or prevention:
A. Are the results of the study valid? Evidence-Based Medicine
Working Group. JAMA 1993; 270: 2598-601.
5.
Guyatt GH. Sackett DL. Cook DJ. Users' guides to the medical
literature. II. How to use an article about therapy or prevention:
B.
What are the results and will they help me in caring for my
patients? Evidence-Based Medicine Working Group. JAMA 1994;
271:59-63.
6. Jaeschke R, Guyalt G. Sackett DL Users' guides to the medical
literature. III. How to use an article about a diagnostic test.
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
383
10.
n.
A. Are the results of the study valid? Evidence-Based Medicine
Working Group. JAMA 1994;
271:
389-91.
Jaeschke R, Guyatt GH, Sackett DL. Users'guides to the medical
literature. III. How to use an article about a diagnostic test.
B.
What are the results and will they help me in caring for my
patients? Evidence-Based Medicine Working Group. JAMA 1994;
271:703-7.
Laupacis A, Wells G, Richardson WS, Tugwell P. Users'guides to
the medical literature: V. How to use an article about prognosis.
Evidence-Based Medicine Working Group. JAMA 1994; 272:
234-7.
Oxman AD, Cook DJ, Guyatt GH. Users' guides to the medical
literature: VI. How to use an overview. Evidence-Based Medicine
Working Group. JAMA 1994; 272:
1367-71.
Haynes RB, Wilczynski N, McKibbon KA et al. Developing
optimal search strategies for detecting clinically sound studies in
MEDLINE. J Am Med Informatics Assoc 1994; 1. 447-58.
Wilczynski NL, Walker CJ, McKibbon K.A, Haynes RB. Assess-
ment of methodological search filters in MEDLINE. Proc
ANNU Symp Comp Appl Med Care 1994; 17: 601-5.
12.
Haynes RB. The origins and aspirations of ACP Journal Club
(Editorial). ACP J Club 1991; 114: A18.
13.
Warren RM, Duffy SW, Bashir S. The value of the second view in
screening mammography. Br J Radiol 1996; 69: 105-8.
14.
Wald NJ. Murphy P. Major P et al. UK.CCCR multicentre
controlled trial of one and two view mammograohy in breast
cancer screening. BMJ 1995;
311:
1189-93.
15.
Sackett DL, Rosenberg WMC, Gray JAM, Haynes RB. Evidence-
based medicine: What it is and what it isn't. BMJ 1996: 312: 71 2.
Received 6 February 1998: accepted 18 February 1998.
Correspondence to:
Dr. R. B. Haynes
Room 3H7. McMaster University Medical Center
1200 Main St W
Hamilton
Ontario L8N 3Z5
Canada
E-mail: bhaynes(nsfhs.mcmaster.ca
by guest on July 4, 2016http://annonc.oxfordjournals.org/Downloaded from
