Cochrane Hand Search Manual
Part I: Overview of Searching Activities
Contents
The aims of the Cochrane Collaboration are
"preparing, maintaining and disseminating systematic
reviews of the effects of health care". Before one can
prepare a systematic review, one has to find the relevant
research reports; in the case of the Cochrane Collaboration,
this usually means reports of randomized clinical trials [1].
This can be surprisingly difficult [2,3].
Cochrane reviewers rely on several means of searching for
relevant reports, including both electronic and manual
methods. For complete identification of published reports,
there appears to be no alternative to a page-by-page search
of the literature. This Manual is designed to aid this type
of manual searching.
The Manual chapters are written primarily with members of
Cochrane Centres, Collaborative Review Groups, Fields and
Networks in mind, but we hope that the information presented
will be helpful to all involved with searching for trials. Chapter
2 describes the development of the International
Register of RCTs of Health Care, and how Cochrane searching
activities as a whole will be coordinated to achieve its
completion. Chapter
3 presents a step-by-step checklist of searching
activities, while Chapter
4 addresses concerns regarding quality control. Part II
of this Manual contains an ever-growing set of examples,
step-by-step instructions, and model documents prepared for
and used by selected Cochrane searching teams who are
carrying out the tasks discussed here. Part III contains a
complete training course in hand searching that can be
customized to suit the particular needs of any Cochrane
searching group.
Note:Parts II and III of the Hand Search Manual
are not available in electronic form. Contact the New
England Cochrane Center at Providence for a complete printed
copy of Part II and Part III of the Hand Search Manual.
Hundreds of journals have been or are being hand searched
by members of the Cochrane Collaboration, some journals by
as few as one and others by as many as 32 individual
searchers. As of this writing, an estimated 1,000 searchers
are contributing to this effort. As more trials are
identified, it will be possible to prepare, maintain and
disseminate increasing numbers of systematic reviews.
A necessary starting point for performing systematic
reviews of RCTs is identification of the maximum possible
number of potentially relevant trials. Data collection for a
systematic review involves setting eligibility criteria for
all RCTs to be included in the review, identification of the
relevant RCTs, and extraction of relevant data from the
individual trials. Complete identification of RCTs for a
systematic review, even when fully published reports are
available, continues to be remarkably difficult [2,3].
In addition, the limitations of existing systems for
electronically and manually searching bibliographic
databases make the process time consuming and inefficient [4,5].
Therefore, it is essential to develop a register or
registers of all existing RCTs to reduce the searching
burden on individual reviewers, to help ensure that RCTs are
identified as completely as possible, and to facilitate the
production of increasing numbers of systematic reviews.
The Cochrane Collaboration is working with the National
Library of Medicine (NLM) to create an International
Register of RCTs of Health Care, which will be mounted and
maintained as a "virtual" register within NLM's
MEDLARS system of bibliographic databases. The project is
being coordinated by the New England Cochrane Center at
Providence.
Cochrane collaborators are systematically searching the
medical literature to identify trials that meet the Cochrane
inclusion criteria for the International Register (see Appendix
1). Trial reports will be sent to staff at the New
England Cochrane Center at Providence, who will in turn
communicate this information to the NLM, where citation
records for identified trials will be assigned specified
Medical Subject Headings (MeSH) tags within MEDLINE. Those
interested in identifying trials for a systematic review
will be able to search MEDLINE for citations with these MeSH
tags and have some confidence that the search will identify
all of the relevant trials. A diagram of the
register-building process, as it is currently envisioned, is
provided in Figure 1. Activities
shown on the left side of the figure are currently in
progress; those shown on the right are proposed for the
future.
The eligibility criteria established by the Cochrane
Collaboration for reports to be included in the
International Register of RCTs of Health Care are provided
in Appendix 1. To ensure that the
greatest proportion of published randomized trials will be
identified, the agreed eligibility criteria are designed to
be sensitive rather than specific. That is, all articles
that definitely or possibly describe trials that
prospectively assigned participants to one of two (or more)
alternative forms of health care using either random
allocation or some quasi-random method of allocation (such
as alternation or medical record number) are eligible for
inclusion.
The term "trial" is used in its broadest sense:
it includes any prospective study that compared two
or more interventions concurrently and was performed in
humans. Even when a study was performed in healthy people
and/or may not concern health care directly, it should be
included in the International Register because it may
contain information that is relevant to the evaluation of a
health care intervention.
Trials eligible for inclusion in the International
Register of RCTs of Health Care are classified according to
the reader's degree of certainty that random allocation was
used to form the comparison groups in the trial. If the
author(s) state explicitly (usually by some variant of the
term "random" to describe the allocation procedure
used) that the groups compared in the trial were established
by random allocation, then the trial is classified as an
"RCT" (randomized controlled trial). If an
eligible trial has not been explicitly described by the
author as randomized, then there is less certainty that it
is, in fact, an RCT. This uncertainty is reflected in a
different classification: "CCT" (controlled
clinical trial). The classification "CCT" is also
applied to quasi-randomized studies where the method of
allocation is known but is not considered strictly random.
The classification is based solely on what the author has
written, not on the reader's interpretation.
The Cochrane classification terms "RCT" and
"CCT" are intended to conform to the NLM's
definitions of the publication type MeSH terms
RANDOMIZED-CONTROLLED-TRIAL and CONTROLLED-CLINICAL-TRIAL (see
Figures 2 and 3).
The use of two classification terms balances the NLM's need
for precision in the application of index terminology with
the Collaboration's need for comprehensiveness in the
selection of reports to be included in the International
Register.
Since the classification terms are based on the author's
description of the method of allocation used, they are not
meant to reflect the actual quality of the allocation
procedure. A trial report which states that the study
participants were randomized is classified as "RCT",
despite the fact that quasi-random methods may actually have
been used. Conversely, although double-blind trials are
nearly always randomized, many trial reports fail to mention
this explicitly and such reports are classified as "CCT".
It is the responsibility of those performing systematic
reviews to decide which trials to include and exclude in any
given review.
National Library of Medicine Definition
PT RANDOMIZED-CONTROLLED-TRIAL
| MH |
Randomized Controlled Trial |
| DC |
2 |
| IDXSH |
NULL LIST |
| MH_TH |
NLM (1991) |
| MHLEX |
NON |
| GM |
publication type: for reports of
randomized controlled trials |
| AN |
publication type only; to designate a
type of clinical trial in which two or more groups
are chosen at random, one receiving the service, the
other not; for randomized controlled trials as a
subject or of value as research, index under
RANDOMIZED CONTROLLED TRIALS (MH); do not confuse
with Publication type CONTROLLED CLINICAL TRIAL; do
not interpret trial design: use term of author; if
in doubt, read MeSH definitions; coord IM or NIM any
other epidemiological or statistical method of
design present; Manual 17.36+, 26.26.3 |
| MS |
A clinical trial that involves at
least one test treatment and one control treatment,
concurrent enrollment and follow-up of the test- and
control-treated groups, and in which the treatments
to be administered are selected by a random process,
such as the use of a random numbers table. Treatment
allocations using coin flips, odd-even numbers,
patient social security numbers, days of the week,
medical record numbers, or other such pseudo- or
quasi-random processes, are not truly randomized and
a trial employing any of these techniques for
patient assignment is designated simply a CONTROLLED
CLINICAL TRIAL. |
| OL |
search policy: Online Manual; iuse:
main heading AND RANDOMIZED CONTROLLED TRIAL (PT) |
National Library of Medicine Definition
PT CONTROLLED-CLINICAL-TRIAL
| MH |
Controlled Clinical Trial |
| DC |
2 |
| IDXSH |
NULL LIST |
| MH_TH |
NLM (1995) |
| MHLEX |
NON |
| GM |
publication type: for reports of
controlled clinical trials |
| AN |
publication type only; to designate a
type of clinical trial of drugs, devices, procedures
for diagnosis, therapeutic or preventive effect
under a tightly designed protocol; for controlled
clinical trials as a subject, index under CONTROLLED
CLINICAL TRIALS (MH); do not confuse with
Publication Type RANDOMIZED CONTROLLED TRIAL; do not
interpret trial design: use term of author; if in
doubt, read MeSH definitions; Manual 17.11+ |
| MS |
A clinical trial involving one or more
test treatments, at least one control treatment,
specified outcome measures for evaluating the
studied intervention, and a bias-free method of
assigning patients to the test treatment. The
treatment may be drugs, devices, or procedures
studied for diagnostic, therapeutic, or prophylactic
effectiveness. Control measures include placebos,
active medicine, no-treatment, dosage forms and
regimens, historical comparisons, etc. When
randomization using mathematical techniques, such as
the use of a random numbers table, is employed to
assign patients to test or control treatments, the
trial is characterized as a RANDOMIZED CONTROLLED
TRIAL. However, trials employing treatment
allocation methods such as coin flips, odd-even
numbers, patient social security numbers, days of
the week, medical record numbers, or other such
pseudo- or quasi-random processes are simply
designated as controlled clinical trials. |
| OL |
search policy: Online Manual; iuse:
main heading AND CONTROLLED CLINICAL TRIAL (PT) |
Prospective identification and tagging of RCTs and CCTs
will continue to be part of regular MEDLINE indexing
operations. RCTs and CCTs will also be identified
prospectively by Cochrane collaborators and journal editors
who have agreed to search individual journals prospectively.
Both electronic (of MEDLINE and other bibliographic
databases) and manual searching of individual journals and
other literature, is being done by the Collaboration with
the assistance of volunteers, editors and publishers of
biomedical journals around the world. Each of the
Collaborative Review Groups, Fields and Networks are
responsible for searching the specialist literature in their
particular area of interest, while the Cochrane Centres are
responsible for searching the general medical literature
published in their country or region. The overall
coordination of all Cochrane search efforts is being handled
by the New England Cochrane Center at Providence.
During 1994, Carol Lefebvre and staff of the UK Cochrane
Center electronically searched MEDLINE's publication years
1985 through 1993, scanned the search output, and identified
20,000 "new" RCTs for the Register. Concurrently,
Cochrane Collaborators were engaged in hand searching: by
January, 1995, the search of nearly 500 journals was either
underway or completed.
Cochrane hand searching groups will prepare and document
electronic files of all citations to identified RCTs and
CCTs, and mail them on floppy disks to the New England
Cochrane Center at Providence (see Chapter
3). The New England Cochrane Center at Providence will
merge the file with others received and prepare a
comprehensive electronic re-tagging request file for the NLM.
As part of its annual MEDLINE maintenance procedures, the
NLM will tag all citations in the MEDLINE database (1966
forward) with the publication types "RANDOMIZED-CONTROLLED-TRIAL"
and "CONTROLLED-CLINICAL-TRIAL" as appropriate.
Individual Collaborative Review Groups, Fields and
Networks can search MEDLINE using subject matter terms plus
the two publication type terms, "RANDOMIZED-CONTROLLED-TRIAL"
and "CONTROLLED-CLINICAL-TRIAL", to build their
own specialized registers.
We plan that Cochrane-identified RCTs and CCTs not in
MEDLINE will be added manually by the NLM, either to MEDLINE
or to an ancillary database. The development of an ancillary
database to MEDLINE to hold newly created bibliographic
records for reports of RCTs and CCTs not currently in
MEDLINE is in the planning stage. Ideally, this ancillary
database will be able to adapt or make use of information
already present in other bibliographic databases, such as
PsychLit, whenever possible or permissible by copyright.
The intent of building a "virtual register" of
published reports of RCTs and CCTs in all subject areas
within MEDLINE is to enable the Collaborative Review Groups,
Networks and Fields to create specialist registers of trials
in their interest areas in as simple and efficient a manner
as possible. This is, however, a long term goal that will
take a number of years to attain. In the meantime, all
searching activities throughout the Collaboration must be
seen as serving a dual purpose: to contribute to the
comprehensive International Register and to contribute to
one or more individual specialist registers. This is why it
is essential that all journals and other publications be
searched comprehensively for all eligible reports,
and not merely for reports relevant to a particular topic or
specialist register.
1. Identify a search coordinator.
2. Determine which journal(s) to search.
3. Register the intent to search a journal with the New
England Cochrane Center at Providence, using the Journal
Hand Search Registration Form (Appendix
2).
4.Train and test searchers, using training materials
custom developed by your searching group or adapted from
materials given in Parts II and III of this Manual.
5. Search journal(s):
a. Examine each issue, page by page, to identify ALL
reports of RCTs and CCTs, whether they appear in articles,
abstracts, news columns, editorials, letters or any other
text, and regardless of the relevance of the subject matter
to the searcher.
b. Photocopy the first page of each report identified, as
well as any subsequent pages requested by a group's search
coordinator.
c. Annotate each photocopy, on either the front or back,
with complete citation information, wherever necessary.
d. Classify each report as RCT or CCT (or other as
relevant to the needs of the search group) and annotate each
photocopy with the classification codes. Highlighting (e.g.,
with a yellow marker) or underlining the key terms which
justify the classification may be required by a group's
search coordinator.
6. Whenever such citation records exist, download the
citation records corresponding with the search results
directly from MEDLINE.
7. Transfer (or enter, when there is none to transfer)
citations into Pro-Cite and add classification codes to any
specified, unused field.
8. Forward citations on 3 1/2" IBM-compatible floppy
disk to the New England Cochrane Center at Providence.
Include annotated photocopies of any reports without a
corresponding MEDLINE citation record.
Each of these steps is described in more detail in the
following sections.
3.2.1 Identify a search coordinator
Each of the Cochrane Centres has the responsibility of
searching the general medical literature of their country or
region, and most have now designated one or more staff
members to coordinate this effort. The CRGs, Fields and
Networks are responsible for establishing the mechanisms for
searching the specialist literature in their areas of
interest, and should coordinate these activities.
It is the responsibility of the editorial team of any
Collaborative Review Group (CRG) to coordinate searching
activities of specialist health care journals and to ensure
that all RCTs and CCTs within these journals are
identified as completely as possible and registered for
their own benefit and that of others in the Collaboration.
This essential task can and should be pursued concurrently
with efforts to establish the CRG.
Designating a search coordinator for the group should be
among the first actions taken by those planning to develop a
specialized register. Coordination of searching activities
is essential to avoid unnecessary duplication of effort, to
maximize the completeness with which trials are identified,
and to minimize the resources used to do this. The overall
coordination of searching activities is the responsibility
of the New England Cochrane Center at Providence. The New
England Cochrane Center at Providence is able to provide
guidance for organizing the work of assembling a register
based on the experience of other groups, and from the
results of methodological research relating to the
identification and registration of trials. The aim of the
New England Cochrane Center at Providence is not to assume
responsibility for searching activities, but to assist those
with interest in specific subject areas to do this work as
effectively and efficiently as possible.
A growing number of journal editors have agreed to search
their own journals, in cooperation with the Collaboration.
In addition, many individuals have volunteered to search
journals, independent of any Cochrane group, usually because
Review Groups or Fields have not yet been established in
their areas of interest. Most often, these searches are
coordinated by a Cochrane Center, but some are coordinated
by Review Groups, Networks or Fields. While the actual
methods of search coordination will vary according to the
resources and preferences of individual search teams, every
individual who hand searches a journal on behalf of the
Cochrane Collaboration should rely on and receive training,
guidance and support from a search coordinator.
Where individual members of searching teams are working
in close proximity to one another, communication is less of
a problem. However, the membership of many groups span the
continents. Preparing and sending a newsletter, whether by
electronic or regular mail, can be an excellent way to keep
members involved, encouraged and informed as progress is
made. The New England Cochrane Center at Providence would
appreciate receiving copies of newsletters or other
informational mailings sent out by your group.
3.2.2 Determine which journal(s) to search
Another essential preliminary activity that can and
should be pursued concurrently with efforts to establish the
CRG, Field or Network is the preparation of a listing and
prioritization of all of the journals the group hopes to
search. Directories of journals exist both as electronic
databases and in book form, from which subject listings of
health care journals can be derived.
A Master List of Journals Being Searched is maintained by
the New England Cochrane Center at Providence as part of an
overall progress tracking database of activities towards
building the International Register of RCTs of Health Care.
The Master List is updated regularly as new journal searches
are registered. Those planning or coordinating journal
searches for Centres, CRGs, Fields or Networks, especially
in interest areas likely to overlap with others, should
consult the Master List to avoid duplication of effort.
Copies of the most recent Master List may be downloaded
from this site or explored
on-line.
3.2.3 Register the journal search
Once a decision has been made to search a particular
journal, whether for all years from 1948 (or first issue) to
the present, from this time forward only, or for some other
specified interval of years, a Journal Hand Search
Registration Form (see Appendix 2)
must be forwarded to the New England Cochrane Center at
Providence. This form, which is also available as an
Appendix to Section V of the Cochrane Collaboration
Handbook, should be completed either by the person
agreeing to search the journal or the person responsible for
coordinating hand searching activities for that journal.
3.2.4 Train and test searchers
Whether or not searchers have prior experience with
identifying published reports of RCTs, all who plan to hand
search literature on behalf of the Cochrane Collaboration
should complete a training course tailored to the
requirements and needs of the Cochrane Collaboration. The
purpose of training is to ensure that all searchers apply
standard eligibility criteria (discussed in Chapter 2) and
classify reports in the same manner.
The Training Notes for Hand Searchers (Part III of this
Manual) is a training course designed for students,
volunteers, and others previously unfamiliar with randomized
trials and literature searching. Alexia Antczak-Bouckoms of
the Cochrane Oral Health Group has adapted this course to
produce a hand search manual appropriate for training
collaborators familiar with randomized trials (see Chapter 3
of Part II). Either of these courses can be further modified
to suit the requirements of a particular searching group.
It is strongly recommended that all searchers be required
to complete an appropriate training program and complete a
self-test, prior to embarking on their task. To prepare for
the self-test, search coordinators should develop a
"gold standard" search of one or several
publication years of a typical journal relevant to their
subject area. They will then be able to compare the results
of each searcher's search with the "gold standard".
3.3.1 Scope of the search
Full-text searching of journals and other literature
involves the page-by-page examination of the entire text
(i.e., articles, abstracts, letters, editorials, news items,
etc.) of the publication. All RCTs and CCTs contained
in the publication are to be identified, classified and
prepared for forwarding to the New England Cochrane Center
at Providence, irrespective of the relevance of the subject
matter to the individual or group performing the search.
Groups performing journal searches are free to develop
the administrative structure and operating procedures most
comfortable and convenient for them. Some examples of search
procedures used by individual Cochrane groups are provided
in Part II.
3.3.2 Make and annotate photocopies of eligible
reports
When a CRG, Field or Network has sufficient resources to
do so, it can be helpful to reviewers if photocopies of
entire articles are made as the articles are identified as
RCTs or CCTs, and several groups have elected to do this. At
a minimum, each searcher should photocopy the first page of
each eligible report identified, annotating each photocopy,
on either the front or back, with the full reference and the
report classification (i.e., RCT or CCT). Some groups have
chosen to widen the scope their search to include review
articles, meta-analyses or other sources of references to
RCTs and CCTs, and have developed additional classification
codes to accommodate these reports. Only the RCTs and CCTs
are to be sent to the New England Cochrane Center at
Providence for inclusion in the International Register.
It will be helpful to those who prepare the search
results for transfer to the New England Cochrane Center at
Providence if searchers underline or highlight (e.g., with a
permanent, yellow marking pen) the text that lead him or her
to classify the report as an RCT or CCT. If this is done,
then searchers should also photocopy additional relevant
pages when the key text occurs after the first page.
Once the page-by-page search of some number of journal
issues has been accomplished, the set of photocopies
resulting from the search must be checked for completeness
and for proper annotation and classification. Some methods
for doing this are discussed in Chapter
4. After it has been verified that the search results
are complete and accurate, they must be prepared for
forwarding to the New England Cochrane Center at Providence
for inclusion in the International Register, as well as for
inclusion in one or more specialist register, when
appropriate.
The appropriate procedure for forwarding reports of RCTs
and CCTs (see Chapter
2 for definitions), to the New England Cochrane Center
at Providence depends upon whether a bibliographic record
currently exists in MEDLINE for each of the identified
reports. If bibliographic citations to reports of trials
that have been identified through hand searching are present
in MEDLINE, the citations must be downloaded directly from
MEDLINE in the manner described in Section 3.4.1.
Information on accessing MEDLINE and on procedures for using
the downloading software can be obtained from your local
medical librarian.
If a journal is not currently indexed in MEDLINE (or was
not at the time a particular issue was published), there
will be no MEDLINE records to update for reports of RCTs and
CCTs found in that journal. In addition, some reports that
were published in MEDLINE-indexed journals do not have
corresponding MEDLINE records. There are several reasons why
this can occur. It may be that the report in question was
published during a year (e.g., prior to 1966) when the
journal was not indexed. It may be that the report is of a
type not generally indexed separately by NLM (e.g.,
abstracts from conference proceedings). Very rarely, a
report or issue may have been omitted by mistake. Regardless
of the reason, whenever there is no MEDLINE citation record
for an identified RCT or CCT report, a bibliographic record
for the report will have to be created and indexed manually
by NLM staff for future inclusion in a proposed ancillary
database. These citations must be documented in the manner
described in section 3.4.3 and sent to the New England
Cochrane Center at Providence.
3.4.1 Download MEDLINE citations
The entire MEDLINE record (all fields) for each
report is required for processing by the New England
Cochrane Center at Providence. The specific contents of the
entire MEDLINE record will vary according to the method used
to access MEDLINE. Please be aware that some versions of
search/download software used to access MEDLINE will
automatically default to an incomplete record unless
explicitly instructed otherwise.
Most versions of search/download software used to access
MEDLINE offer a choice of two or more record formats when
downloading. Choose the "MEDLARS" (sometimes
called "Reprint") option, which is the
format designed for transfer of the records into Pro-Cite or
other bibliographic database software. Each line of the
downloaded records will then be preceded by a two-letter
field label, needed to facilitate transfer of the record
into Pro-Cite.
The following MEDLINE record fields are essential:
- Unique Identifier This is the eight-digit
number which uniquely identifies the record to the
MEDLINE database management software. Depending upon the
search/download software and MEDLINE access method used,
this could also be called the "Accession Number"
(AN) or the "Call Number". No
re-tagging request can be sent to NLM without this
number.
- Publication Year This may be captured as a
separate field or in combination with the journal name
and citation information in a field called "Source".
- Journal Name (at the time the report was
published)
- Citation Information This should contain the
volume number, issue number, and both initial and
concluding page numbers.
- Name(s) of Author(s) Some RCT reports are
indexed as "anonymous".
- Title of Article Some letters and editorials do
not have titles.
- Abstract The text of the abstract is often
truncated to 250 words.
- MeSH Index Terms These will allow the records
to be copied into searchable Pro-Cite database(s) for
interim use by Collaborators and provide data for
research in coding and classification.
- Publication Types Records cannot be processed
for re-tagging requests without information about the PT
tags that have already been applied. Some MEDLINE
formats store this information in a field called "Notes".
The downloaded citation records may be grouped into files
in whatever manner is convenient. Records need not be
separated according to journal, publication year, or type of
report, unless such an arrangement would be useful to the
search group.
3.4.2 Prepare Pro-Cite database(s)
There are a number of different software applications
available for managing bibliographic references, and each
has different strengths and weaknesses. Pro-Cite provides
the greatest flexibility for importing references from any
tagged external source and consequently the Cochrane
Collaboration has decided to standardize on Pro-Cite for
managing the references being incorporated into the
International Register of Published RCTs of Health Care.
Therefore, the New England Cochrane Center at Providence
requests that, if at all possible, those working with the
Collaboration use this software to transfer data to the New
England Cochrane Center at Providence for inclusion in the
Register. Groups choosing to use other software should
format their transfer data so that it may be easily exported
into Pro-Cite.
Transfer of citations downloaded from MEDLINE in MEDLARS
format into a Pro-Cite database is easily accomplished using
the "Biblio-Links" software designed for this
purpose. The appropriate linking application, "SPLINK"
for use with SilverPlatter or "NLMLINK" for use
with CD-PLUS Ovid or Grateful Med, is available for purchase
along with Pro-Cite. Detailed instructions for using NLMLINK
are provided in Part II. The steps needed to use SPLINK are
virtually identical.
Once the citations have been transferred into one or more
Pro-Cite databases, the classification code (e.g., RCT or
CCT) may be entered into any empty field, by editing the
individual records. At the New England Cochrane Center at
Providence, we use the "Loc" field for this
purpose, but any empty field will do. Some searching groups
have chosen to design their own Pro-Cite formats (called
"Workforms") to accommodate their own
classification systems. Instructions for creating and using
the customized Pro-Cite Workform used by the Cochrane Field
Coordination in Emergency Medicine and Intensive/Critical
Care Medicine are included in Part II. The Pro-Cite
database(s), once the classification codes have been entered,
can form the basis for your group's specialized register.
We recognize that some searching groups have already
begun to construct their specialized registers using
application software other than Pro-Cite, and that
conversion to Pro-Cite at this time would present a hardship.
When this is the case, the citations downloaded from MEDLINE
may be forwarded to the New England Cochrane Center at
Providence in ASCII files (in MEDLARS format so that we may
transfer the citations into Pro-Cite for you). If your group
requires this accommodation, please notify the New England
Cochrane Center at Providence in advance, so that the
additional effort can be scheduled. In addition, we request
that the RCTs and CCTs be forwarded to us in separate files,
and that a hard copy bibliography indicating the report
classifications be included with the transfer package.
Regardless of the format used, the citations must be
forwarded to the New England Cochrane Center at Providence
on 3 1/2", DOS-formatted floppy disks. At present, the
New England Cochrane Center at Providence lacks the
facilities to process citations sent in any other format (e.g.,
MacIntosh). Each report must be classified as either an RCT
or a CCT in an accompanying hard copy bibliography (if the
data files are in ASCII format) or in any unused, specified
data field of the Pro-Cite database.
3.4.3 Document reports not in MEDLINE
As stated previously, whenever existing MEDLINE citation
records for identified RCT or CCT reports are unavailable,
bibliographic records for each report must be created and
indexed manually by NLM staff for future inclusion in a
proposed ancillary database. To accommodate their needs and
to make interim usage possible within the Collaboration,
each report must be documented in the following manner:
- Photocopy the title and/or Table of Contents page,
showing the authors, journal title and volume and issue
numbers, for the journal issue containing the report.
- Photocopy the title page of the report itself,
annotated to show the complete bibliographic citation (authors,
title, journal name, volume, pages, and year) and the
classification of the report as either an RCT or a CCT.
- Provide the bibliographic reference for each report in
a WordPerfect or other word processing software
document. Label the disk as to software used and file
contents.
The annotated photocopies, along with the bibliography on
3 1/2" DOS-formatted floppy disk, should be forwarded
to the New England Cochrane Center at Providence. In
addition to the items listed above, electronic copies of
citations downloaded from other bibliographic databases (e.g.,
PsychLit) should also be included, whenever available, in
the transfer package sent to the New England Cochrane Center
at Providence. These may be submitted as either a set of
Pro-Cite files, an ASCII-formatted file or a word processing
software document (e.g., WordPerfect) in addition to or
instead of the file of bibliographic information described
above.
3.4.4 Forward the citation package to the New England
Cochrane Center at Providence
At present, the preferred method for transfer of
citations to the Baltimore Cochrane Center is by mail or
courier service, on 3 1/2" DOS-formatted floppy disk(s).
Alternative arrangements can be made (e.g., transfer via
e-mail of very small files) by contacting, in advance, a
member of the New England Cochrane Center at Providence
staff. Accommodation of other forms of electronic transfer
(e.g., by ftp) is planned for the future.
New England Cochrane Center at Providence staff
include:
Kay Dickersin, PhD
Co-Director
Tel: 1 401-863-9950
Fax: 1 401-863-9944
Jeanette Downing-Park
Coordinator
Tel: 1 401-863-9950
Fax: 1 401-863-9944
Eric Manheimer
Methodologist
Tel: 1 401-863-9950
Fax: 1 401-863-9944
e-mail: cochrane@brown.edu
Citation packages should be sent to our new address:
New England Cochrane Center at Providence
Department of Community Health
Brown University
Box G-S2
Providence, RI 02912
tel: +1 401-863-9950
fax: +1 401-863-9944
e-mail: cochrane@brown.edu
Web: http://www.cochrane.org/neccp.htm
At first, it may appear that hand searching a journal is
more tedious than difficult. However, this effort is as
subject to human error as any other, perhaps even more so
because of the nature of the task. In our experience, the
progress lost to mislaid or illegible pieces of paper,
typographical errors, missed or missing citations, or simple
failure to follow directions, can be considerable. This
chapter lists some of the problems that have been
encountered by Cochrane searching groups in the past, and
gives some suggestions of ways to minimize or avoid them.
There are several ways that hand searchers can needlessly
duplicate another's effort. Fortunately, these situations
can be easily avoided.
If a searcher or search coordinator does not consult the
Master List of Journals Being Searched before beginning a
search, then effort can be wasted on searching a journal
that has already been searched by someone else. If a
searcher or search coordinator does not submit a Journal
Hand Search Registration Form (see Appendix
2) for each journal they plan to search, then the Master
List will be incomplete, and someone else could waste effort
searching a journal already searched by you. The remedy is
simple: always consult the Master List and submit a
Registration Form before beginning a journal search.
Searchers should not conduct an incomplete search, e.g.,
selecting only full length "original" articles or
reports relevant to a single topic area). When this has been
done, it has been necessary for someone else to repeat the
entire search of the journal. Examine every page
of journals searched and retrieve every eligible
report.
Search coordinators may assign several searchers to a
single journal, but it may be difficult to keep track of who
is searching which publication years. Duplicate search
assignments can be made unintentionally. Keep a record of
search assignments in a spreadsheet to simplify the
necessary record-keeping. Some examples are provided in
Part II.
You may find that a journal issue that you have agreed to
search is unavailable in your local library. Sometimes
journal volumes can be bound with sections removed, pages
missing, and even whole issues omitted. When this occurs, a
note should be made in a spreadsheet or on a form designed
for this purpose. Sometimes the missing materials can be
obtained through an interlibrary loan or copying service. Please
report all unavoidably incomplete searches to the New
England Cochrane Center at Providence.
Searching activities can generate quite a lot of paper,
and pieces of paper can get lost or mislaid. We recommend
that searchers photocopy the Table of Contents of each issue
searched and "check off" each article examined
while searching. The marked Tables of Contents help
guard against lost photocopies and can also serve as
evidence that an issue was searched if no eligible reports
were found.
4.5 Incomplete searches - missed reports
Once a journal issue has been hand searched, we hope that
it will never need to be searched again. If an eligible
report is missed by a hand search, it may be lost forever.
Therefore, regardless of whether searchers are experienced
collaborators, journal editors, students or volunteers, each
searcher should be required to undergo a training exercise.
Establishing a "gold standard" search of at least
several issues of a journal relevant to the search team
should be one of the first responsibilities of each search
coordinator. Every searcher should conduct a test search
to be compared against the group's "gold standard"
search. The results of the comparison should be
discussed with the searcher to resolve any misunderstandings
before he or she begins an actual journal search.
As several groups have discovered to their disappointment,
sometimes initial searcher training is not enough to ensure
that all eligible reports are being retrieved. We
recommend that search coordinators give a certain proportion
of duplicate search assignments for ongoing comparison
between searchers. Twenty percent duplication is
reasonable, unless a problem is uncovered. This can be done
by giving search assignments of five-year intervals that
overlap between searchers by one year, or by selecting a 20%
sample of completed issues for duplicate searching by
another searcher. Whatever method is chosen, this should be
done in a timely manner and on a continuing basis, so that
feedback (both positive and negative) on the result of each
comparison can be given to the searchers and
misunderstandings can be resolved promptly.
Even when all eligible reports have been identified and
retrieved by a searcher, errors can be made in classifying
them as either "RCT" or "CCT". These
errors can cause serious difficulties later on when the
citations are forwarded to the New England Cochrane Center
at Providence for inclusion in the International Register. Search
coordinators should check the photocopies to ensure that
each has been annotated with the correct classification and
complete reference information, before preparing the
citations for transfer to NECC at Providence.
Materials needed to document search results and permit
their inclusion in the International Register were discussed
in Chapter 3.6. Citations to reports of RCTs and CCTs that
do not already have MEDLINE records must be entered manually
into a Pro-Cite database or electronic word processing file.
Please proofread all bibliographic lists of search
results for typographical errors before sending them to NECC
at Providence.
(1)Chalmers I, Dickersin K, Chalmers TC.
Getting to grips with Archie Cochrane's agenda. BMJ
1992;305:7868.
(2)Dickersin K, Scherer R, Lefebvre C. Identification of
relevant studies for systematic reviews. BMJ
1994;309:1286-91.
(3)Adams CE, Power A, Frederick K, Lefebvre C. An
investigation of the adequacy of MEDLINE searches for
randomized controlled trials (RCTs) of the effects of mental
health care. Psychol Med 1994;24:741-8.
(4)Schuyler PL, Dickersin K, Scherer R, Wright ND.
Identification of randomized clinical trials using MEDLINE.
(unpublished material).
(5)Lefebvre C, Kelleher TA, Tihanov G. Identification of
randomized controlled trials using MEDLINE: the situation in
1993. Presented at "An Evidence-Based Health Care
System: The Case for Clinical Trial Registries".
National Institutes of Health, Office of Medical
Applications of Research, Bethesda, MD, Dec. 6, 1993.
Criteria for
registering studies with the Cochrane Collaboration's
International Register of Published RCTs of Health Care
The highest possible proportion of all reports of RCTs of
health care should be included in the International
Register. Thus, those searching the literature to identify
trials should give reports the benefit of any doubts.
Publications which simply mention the possibility of
undertaking an RCT should not be included, however.
Reviewers will decide whether to include a particular
report in a review. The aim of the Register is to provide
reviewers with all possible trials they may wish to include
in a review, not to decide whether a report is worthy or
relevant for inclusion.
Eligibility Criteria:
Relevant reports are reports published in any year, of
studies comparing at least two forms of health care (medical
treatment, medical education, diagnostic tests or techniques,
a preventive intervention, etc.) where the study is on
either living humans or parts of their body or human parts
that will be replaced in living humans (e.g., donor kidneys).
Studies on cadavers, extracted teeth, cell lines, etc. are
not relevant.
Studies are eligible for inclusion in the Register if
allocation to the intervention was random or
intended-to-be-random (e.g., alternation), or if a
concurrent control group was used in the trial and it is
possible that a random or intended-to-be-random method was
used to allocate participants to the study groups.
Judgements as to the quality of the methods used or whether
the authors actually did what they claimed should not be
used to decide eligibility for inclusion in the Register.
A trial should thus be included in the Register if, on
the basis of the best available information, it is judged
that:
- the individuals (or other units) followed in the trial
were definitely or possibly assigned prospectively
to one of two (or more) alternative forms of health care
using random allocation or some quasi-random
method of allocation (such as alternation, date of
birth, or case record number).
In addition:
- If one or more outcomes were assessed using "double
blinding" or "double masking" such that
neither the participant/patient nor the assessor was
aware of the intervention received, but randomization is
not mentioned explicitly in the text, a trial should be
included.
- Crossover trials, in which patients have been assigned
to the first intervention using random or quasi-random
allocation, should be included.
- Reports of trials dealing only with animals should not
be included.
- Units of randomization may be individuals, groups (such
as communities or hospitals), organs (such as eyes) or
other parts of the body (such as teeth).
- A report of a randomized trial should be included even
when no results are presented or when results are
limited to the analyses of baseline variables.
- Articles describing an intended or ongoing trial or
commenting on a trial (such as in an editorial)should be
brought to the attention of the New England Cochrane
Center at Providence, but are not eligible for inclusion
in the International Register of Published RCTs.
Journal
Hand Search Registration Form - from New England
Cochrane Center at Providence.
|
