Classics
Eysenbach, Tuische and
Diepgen’s Evaluation of Web Searching for Identifying Unpublished Studies for
Systematic Reviews: An Innovative Study Which is Still Relevant Today
A
Review of:
Eysenbach, G., Tuische, J.
& Diepgen, T.L. (2001). Evaluation of the usefulness
of Internet searches to identify unpublished clinical trials for systematic
reviews. Medical Informatics and the
Internet in Medicine, 26(3), 203-218. http://dx.doi.org/10.1080/14639230110075459
Reviewed
by:
Simon Briscoe
Information Specialist
National Institute for Health Research (NIHR) Collaboration for Leadership in
Applied Health Research and Care (CLAHRC) South West Peninsula
University of Exeter Medical School
Exeter, United Kingdom
Email: [email protected]
Received: 7 Mar. 2016 Accepted: 11 May 2016
2016 Briscoe. This is an Open Access article
distributed under the terms of the Creative
Commons-Attribution-Noncommercial-Share Alike License 4.0 International (http://creativecommons.org/licenses/by-nc-sa/4.0/), which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is
properly attributed, not used for commercial purposes, and, if transformed, the
resulting work is redistributed under the same or similar license to this one.
Abstract
Objective – To consider whether web searching is a useful
method for identifying unpublished studies for inclusion in systematic reviews.
Design – Retrospective web searches using the AltaVista
search engine were conducted to identify unpublished studies – specifically,
clinical trials – for systematic reviews which did not use a web search engine.
Setting – The Department of Clinical Social Medicine,
University of Heidelberg, Germany.
Subjects – n/a
Methods – Pilot testing of 11 web search engines was
carried out to determine which could handle complex search queries.
Pre-specified search requirements included the ability to handle Boolean and proximity
operators, and truncation searching. A total of seven Cochrane systematic
reviews were randomly selected from the Cochrane Library Issue 2, 1998, and
their bibliographic database search strategies were adapted for the web search
engine, AltaVista. Each adaptation combined search terms for the intervention,
problem, and study type in the systematic review. Hints to planned, ongoing, or
unpublished studies retrieved by the search engine, which were not cited in the
systematic reviews, were followed up by visiting websites and contacting
authors for further details when required. The authors of the systematic
reviews were then contacted and asked to comment on the potential relevance of
the identified studies.
Main Results – Hints to 14 unpublished and potentially
relevant studies, corresponding to 4 of the 7 randomly selected Cochrane
systematic reviews, were identified. Out of the 14 studies, 2 were considered
irrelevant to the corresponding systematic review by the systematic review
authors. The relevance of a further three studies could not be clearly
ascertained. This left nine studies which were considered relevant to a
systematic review. In addition to this main finding, the pilot study to
identify suitable search engines found that AltaVista was the only search
engine able to handle the complex searches required to search for unpublished
studies.
Conclusion –Web searches using a search engine have the
potential to identify studies for systematic reviews. Web search engines have
considerable limitations which impede the identification of studies.
Commentary
Background
Eysenbach, Tuische,
and Diepgen’s study is the first evidence-based evaluation of how searching the
Internet using a web search engine can contribute to the identification of
studies for systematic reviews, in particular, unpublished clinical trials. The
study deserves the status of classic due to its originality and continuing
significance; in particular, for proposing and evaluating a systematic approach
to web searching which to date is referenced in prominent guidelines for
conducting systematic reviews (Lefebvre,
Manheimer & Glanville, 2011).
Web searching is a
common activity for information professionals in almost all library and
information settings. Systematic reviews, however, are perhaps more familiar to
information professionals in health care research settings. Systematic reviews
answer research questions by identifying and appraising all the relevant
studies (using pre-specified eligibility and quality criteria) and synthesizing
the accumulated evidence (Higgins & Green,
2011). They are important in health care settings because there is too
much research literature for practitioners to appraise individually. In
addition, the methods and conclusions of systematic reviews are less biased
than narrative reviews or expert opinion (Higgins
& Green, 2011). It is important to identify unpublished studies, the
focus of Eysenbach et al., because they may contain findings which are more
up-to-date than published studies. There is also evidence suggesting that
studies with negative findings are less frequently published or take longer to
reach publication (Fanelli, 2010).
Information
professionals contribute to systematic reviews by identifying studies (Harris, 2005). Research has shown that their
contributions improve the quality of systematic reviews (Rethlefsen, Farrell, Osterhaus Trzasko, & Brigham, 2015). At
the time Eysenbach et al. was published in 2001, there had been several years
of research on the identification of studies for health care systematic reviews
using bibliographic databases. Early examples of this research include studies
by Dickersin et al. (1994) and Wilczynski
et al. (1993) – see also the historical
survey of methodological developments in this area by Lefebrve et al. (2013). There were also established
supplementary search methods for identifying studies, including checking
reference lists, hand searching, and searching company trials registries, all
of which were detailed in the systematic review guidance manual, the Cochrane Reviewers’ Handbook (now titled
the Cochrane Handbook for Systematic
Reviews of Interventions, hereafter, the Cochrane Handbook) (Clarke &
Oxman, 1999). Web searching did not have a prominent place amongst these
search methods. This is a view Eysenbach et al. verify with reference to the
lack of a web searching section in the otherwise comprehensive Cochrane Handbook .
Eysenbach et al.
addressed the lack of research and guidance on web searching for systematic
reviews, focusing on the use of web search engines to identify unpublished
studies. The authors tested the hypothesis that retrospectively conducted web
searches, which were adapted from the bibliographic database search strategies
of completed systematic reviews, would retrieve previously unidentified and
unpublished studies (specifically, clinical trials). They also set out to
address practical issues such as the suitability of various search engines for
the task.
Main Results
Following the
identification of 14 unpublished studies relating to 4 of the 7 included
systematic reviews in the study, Eysenbach et al. recommended that web
searching using a search engine with appropriate search features should be
conducted alongside other search methods. They also, however, noted that there
was no evidence the searches they conducted affected the outcome of a
systematic review. In particular they emphasized that none of the studies they
identified contained results that remained unpublished due to negative results.
(This would have contributed to the aforementioned evidence that studies with
negative results are hard to publish and less likely to be included in
systematic reviews (Fanelli, 2010).) The
authors concluded that web searching using a search engine should be conducted
as it has the potential to affect the
outcome of a systematic review.
This conclusion is
important for being the first evidence-based recommendation on web searching
for systematic reviews. The conclusion has been noted in subsequent editions of
the Cochrane Handbook, which
currently states that “[t]here is little empirical evidence as to the value of
using general internet search engines such as Google to identify potential
studies”, citing Eysenbach et al. as evidence (Lefebvre
et al., 2011). A forwards citation search on the citation index Web of
Science reveals a total of eighteen citations of Eysenbach et al. The Cochrane
Handbook citation is enough to ensure that health care information
professionals with systematic review experience are likely to have seen, or
learnt from mentors and on training courses, the main result and conclusion.
The web searching
section in the Cochrane Handbook also
advises that searchers might have more success identifying studies by targeting
known key websites, such as pharmaceutical companies, than using web search
engines. This is an important point considering the inaccessibility of a large
portion of the web, known as the invisible or deep web, to the automated
web-crawlers which index webpages for search engines (Devine & Egger-Sider, 2013).This is highlighted by Eysenbach
et al. To improve the efficacy of using search engines the authors recommended
that organizations involved in carrying out and funding trials should publish
details “on a robot [i.e. web crawler] accessible web page…. using the standard
format ‘randomized trial on (intervention) in (condition)’ … so that they can
be indexed by search engines and found by systematic reviewers” (p. 216).
Eysenbach et al.
advocated for the establishment of prospective and ongoing trials registries.
This would remove some of the difficulties of finding unpublished trials using
web search engines, though the authors anticipated that the web would play an
important part in “linking the evidence” between different registries (p. 215).
Recent developments in this area are detailed below in the discussion of
specialized web resources.
Pilot Study Results
In addition to the
enduring impact of the main finding of Eysenbach et al., the findings from the
pilot study remain relevant. In order to effectively adapt bibliographic
database search strategies for web search engines, the search engines require
similar search features. To this end, the search features of 11 web search
engines were assessed: AltaVista, Excite, FAST search, Google, HotBot,
InfoSeek, Lycos, Northern Light, WebCrawler, Medical World Search, and MedHunt.
Only AltaVista offered all the required search features, i.e., Boolean
operators, phrase, proximity, and truncation searching, and capitalization
recognition. Subsequently, AltaVista was the only search engine used in the
main study.
It remains the case
today that bibliographic databases have more advanced search features than web
search engines. There have been some improvements to the latter since Eysenbach
et al. was published. For example, Google did not offer Boolean searching when
Eysenbach et al. was published but it does at the time of writing, albeit with
limitations. However, the main developments in web search engines have been
moving away from complex searches where the user retains a degree of control,
towards simple searches where the user increasingly relinquishes control to
undisclosed algorithms which determine the relevancy and ranking of the
webpages retrieved (Granka, 2010; Pariser, 2011).
This is a challenge for information professionals with complex and detailed information
needs, in that search strategy development is limited, frequent changes to
algorithms compromise the reproducibility of searches, and bias is introduced
in cases where the search history of the user informs the webpages which are
retrieved (Briscoe, 2015).
The problem of
identifying relevant studies with a simple search interface has been
exacerbated by the growth of the web. When Eysenbach et al. carried out their
research in December 1998 there were approximately 2,400,000 websites, whereas
in March 2016 there were approximately 1,000,000,000 websites ("Total number," 2016). Subsequently,
the search string (study or trial or random*) near asthma* near (education* or
(self near management)), which retrieved 159 hits using AltaVista in December
1998 (p. 210), retrieved 389,000 hits using Google on 4 March 2016. AltaVista
was terminated in 2013 and is unavailable for testing ("Yahoo to shut," 2013). The same search on 4 March 2016
in Google Scholar, which limits results to scholarly literature, retrieved a
more focused 37,800 hits, although it is unclear whether the unpublished studies
which Eysenbach et al. searched for would be indexed in Google Scholar. The
high numbers retrieved indicate that the approach Eysenbach et al. used would
need to be adapted in order for the results to be manageable. Either the
searches would need to be made more focused, or the screening of hits would
need to be limited to a manageable number (Godin,
Stapleton, Kirkpatrick, Hanning, & Leatherdale, 2015).
The relatively simple
search capabilities of web search engines and the growth of the web highlight
the importance of assessing the tools and strategies used for web searching,
following the example of Eysenbach et al. In particular, in an age dominated by
Google, information professionals should be mindful to seek out and assess
other search engines.
The Development of
Specialized Web Resources
As a solution to the
limitations of using web search engines for systematic reviews, Eysenbach et
al. advocated the creation of “specialized search engines” containing “expert
knowledge on which [web]sites ongoing studies are published and [able to]
access dynamic databases [i.e. the deep web] and meta-trial registers” (p.
214). No such search engine exists to date, although the launch of the
web-based databases ClinicalTrials.gov and the ISRCTN registry (both in 2000)
have made it easier to identify unpublished studies, specifically, unpublished
clinical trials.
Google Scholar is a
specialized web search engine but it is unable to access the deep web as
advocated by Eyenbach et al. Nonetheless, Google Scholar is an advance in web
searching for the systematic review community, and in recent years there has
been research on how it can contribute to systematic reviews. In the health
care literature there has been research and debate about whether Google Scholar
can replace bibliographic databases as the main source of studies for
systematic reviews (Boeker, Vach, &
Motschall, 2013; Gehanno, Rollin, & Darmoni, 2013; Giustini & Boulos,
2013), general comparisons (not primarily
related to systematic review methods) of Google Scholar with the PubMed
database (Anders & Evans, 2010; Nourbakhsh,
Nugent, Wang, Cevik, & Nugent, 2012; Shultz, 2007), and in the
environmental science literature, its ability to identify grey literature (Haddaway, Collins, Coughlin, & Kirk, 2015).
There are varying views on how much Google Scholar can contribute to systematic
reviews, but in most studies the inadequacy of the Google Scholar search
interface for writing complex search strategies is a predominant theme,
reflecting the pilot study findings of Eysenbach et al.
Conclusion
Despite the
limitations of web search engines and the underwhelming result of Eysenbach et
al., information professionals who contribute to systematic reviews are likely
to continue to use them to identify literature. Although there are web-based
databases for health care literature, such as the ClinicalTrials.gov and ISRCTN
trials registries, web searches using search engines have the potential to
retrieve literature not indexed in these resources, or which exist in web
resources unknown to the searcher. More research is needed on the potential
role of web searching for different types of literature and different types of
systematic reviews. Evaluations of search engines launched since Eysenbach et
al. was published are also required. Eysenbach et al. will remain a benchmark
for future research in these areas, and deserves to be recognized as a classic
of the information science literature.
As an aid to future
research, Eysenbach et al. advocated that systematic review authors should
“carefully document their Internet search strategy in reports of systematic
reviews (rather than just mentioning that ‘Internet searches have been
performed’) so that factors influencing the effectiveness and necessity of
Internet searches can be identified” (p. 215). This is a recommendation which
research suggests requires more adherence (Briscoe,
2015).
References
Anders, M. E., & Evans,
D. P. (2010). Comparison of PubMed and Google Scholar literature searches. Respiratory Care, 55(5), 578-583.
Boeker, M., Vach, W., &
Motschall, E. (2013). Google Scholar as replacement for systematic literature
searches: good relative recall and precision are not enough. BMC Medical Research Methodology, 13, 131. http://dx.doi.org/10.1186/1471-2288-13-131
Briscoe, S. (2015). Web
searching for systematic reviews: a case study of reporting standards in the UK
Health Technology Assessment programme. BMC
Research Notes, 8, 153.
http://dx.doi.org/10.1186/s13104-015-1079-y
Clarke, M., & Oxman, A.
D. (Eds.) (1999). Cochrane Reviewers' Handbook 4.0 (updated July 1999) In: The Cochrane Library, Issue 4. Oxford:
Update Software.
Devine, J., &
Egger-Sider, F. (2013). Going Beyond
Google Again: Strategies for Using and Teaching the Invisible Web. London:
Facet Publishing.
Dickersin, K., Scherer, R., & Lefebvre, C.
(1994). Identifying
relevant studies for systematic reviews. British
Medical Journal, 309(6964), 1286-1291.
Fanelli, D. (2010). Do
pressures to publish increase scientists' bias? An empirical support from US
States Data. PLoS One, 5(4), e10271. http://dx.doi.org/10.1371/journal.pone.0010271
Gehanno, J. F., Rollin, L.,
& Darmoni, S. (2013). Is the coverage of Google Scholar enough to be used
alone for systematic reviews. BMC Medical
Informatics and Decision Making,13, 7. http://dx.doi/10.1186/1472-6947-13-7
Godin, K., Stapleton, J.,
Kirkpatrick, S. I., Hanning, R. M., & Leatherdale, S. T. (2015). Applying
systematic review search methods to the grey literature: a case study examining
guidelines for school-based breakfast programs in Canada. Systematic Reviews, 4, 138. http://dx.doi.org/10.1186/s13643-015-0125-0
Granka, L. A. (2010). The
Politics of Search: A Decade Retrospective. The
Information Society, 26(5), 364-374. http://dx.doi.org/10.1080/01972243.2010.511560
Haddaway, N. R., Collins, A.
M., Coughlin, D., & Kirk, S. (2015). The Role of Google Scholar in Evidence
Reviews and Its Applicability to Grey Literature Searching. PLoS One, 10(9), e0138237. http://dx.doi.org/10.1371/journal.pone.0138237
Harris, M. R. (2005). The
librarian's roles in the systematic review process: a case study. Journal of the Medical Library Association,
93(1), 81-87.
Higgins, J. P. T., &
Green, S. (Eds.) (2011). Cochrane
Handbook for Systematic Reviews of Interventions Version 5.1.0 (updated
March 2011). The Cochrane Collaboration. Available from www.cochrane-handbook.org
Lefebvre, C., Glanville, J.,
Wieland, L. S., Coles, B., & Weightman, A. L. (2013). Methodological
developments in searching for studies for systematic reviews: past, present and
future? Systematic Reviews, 2, 78. http://dx.doi.org/10.1186/2046-4053-2-78
Lefebvre, C., Manheimer, E., & Glanville, J.
(2011). Searching
for studies. In J. P. T. Higgins & S. Green (Eds.), Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0
(updated March 2011). The Cochrane Collaboration. Available from www.cochrane-handbook.org
Nourbakhsh, E., Nugent, R., Wang, H., Cevik, C.,
& Nugent, K. (2012). Medical
literature searches: a comparison of PubMed and Google Scholar. Health Information and Libraries Journal, 29(3), 214-222. http://dx.doi.org/10.1111/j.1471-1842.2012.00992.x
Pariser, E. (2011). The
troubling future of internet search. The
Futurist, 45(5), 6.
Rethlefsen, M. L., Farrell,
A. M., Osterhaus Trzasko, L. C., & Brigham, T. J. (2015). Librarian
co-authors correlated with higher quality reported search strategies in general
internal medicine systematic reviews. Journal
of Clinical Epidemiology, 68(6), 617-626. http://dx.doi.org/10.1016/j.jclinepi.2014.11.025
Shultz, M. (2007). Comparing
test searches in PubMed and Google Scholar. Journal
of the Medical Library Association, 95(4), 442-445. http://dx.doi.org/10.3163/1536-5050.95.4.442
Total number of websites.
(2016). Retrieved on 4 March 2016 from http://www.internetlivestats.com/total-number-of-websites/#trend
Wilczynski, N. L., Walker, C.
J., McKibbon, K. A., & Haynes, R. B. (1993). Assessment of methodologic
search filters in MEDLINE. Proceedings of
the Annual Symposium on Computer Applications in Medical Care, 601-605.
Yahoo to shut down pioneering
AltaVista search site. (2013). Retrieved on 4 March 2016 from http://www.bbc.co.uk/news/technology-23127361