- arts and humanities
- medicine and health sciences
- natural sciences
- social sciences
- systematic review
Some of the examples below are tied to using specific databases, while others show more of the search process and how to think about your topic and choose different databases or approaches. Experiment with the examples to become familiar with the various strategies for efficient searching.
Arts and humanities
Research question: Heroes’ and villains’ use of alcohol and tobacco in Disney animated films 1937-2000
With research projects in the humanities, it is not a question of using only one approach or of searching only one database. In this case it is useful to divide the research question into components and search literature for each component seperately.
- Substance use in popular media
- Animated films
- Archetypes in popular culture
Not every aspect of relevance to the project is always evident in the research question. In this case, there could be an interest in how children’s exposure to substance use through popular culture affects their health choices later in life, which would also entail finding research literature on this topic.
The literature used to inform a project like this would be found through various sources.
- Cross-disciplinary reference databases like Scopus
- Library catalogues like Oria
- Archival sites like Entertainment Industry Magazine Archive
Research question: Narrative effects of thematizing truth in German autobiographical fiction in the 20th century
When you search literature for a PhD in the humanities, you will always need more than one database and more than one search strategy. In this case, it would be useful to search for literature regarding the different components of the research question separately.
- Narrative techniques
- Truth as literary motif
- Autobiographical fiction
- Literary critique of 20th century German literature
Though it is not explicit in the research question, you would also need to examine how the specific works of autobiographical fiction in your study were received when published and if there has been any previous research done on these works.
A project like this would require the use of many different types of literature. You would therefore consult several sources and apply different search techniques. It would be particularly important here to use reference tracing and possibly citation searches.
- Library catalogues like Oria
- Reference databases like Arts & Humanities in Web of Science
Medicine and health sciences
Research question: The effectiveness of light therapy interventions to treat winter depression.
Database: PubMed (Medline). Relevant in biomedicine, life sciences and social sciences.
Date of search: May 2013
Divide your search into several steps and combine results afterwards. When searching literature in medicine and health, make use of defined subject headings in MeSH.
Step 1: Searching for “winter depression”
The MeSH term for this is seasonal affective disorder. In Scandinavia this condition is commonly named winter depression, which is mentioned in a few abstracts. The terms are combined with OR, to retrieve articles where either is mentioned. The list below shows search terms and numbers of results in parentheses. Searches are marked with the hash symbol ( # ) in PubMed.
Search 1: seasonal affective disorder[MeSH Terms] 1039
Search 2: seasonal affective disorder[Title/Abstract] 976
Search 3: winter depression[Title/Abstract] 240
Search 4: #1 OR #2 OR #3 1414
Step 2: Searching for “light therapy interventions”
The correct MeSH term for light therapy intervention is phototherapy and this is combined with light therapy in the title and abstract as a synonym. Additional searches for the terms in titles or abstracts here add a few relevant articles. They are all combined with OR, to retrieve articles where either term is mentioned.
Search 5: phototherapy[MeSH Terms] 25680
Search 6: phototherapy[Title/Abstract] 4834
Search 7: light therapy[Title/Abstract] 1050
Search 8: #5 OR #6 OR #7 27496
Step 3: Combining search results
Finally, both sets of terms are combined with AND. This will retrieve a combination of the words above.
Search 9: #4 AND #8 656
Step 4: Limiting search results
Now you have 656 references to articles concerning your search terms. If you want to decrease that number further, you can limit by study design, for example to reviews, which will result in 131 studies
Search 10: (# 9) Filters: Review 131
An additional limit by publishing year will decrease the number even further. There are 42 reviews published in the last ten years.
Search 11: (# 9) Filters: Review; published in the last 10 years 42
For an initial check of relevance of the retrieved documents, browse titles, year of publication, contributing authors and journals involved. Keep revising your search strategy and also consider setting up an alert. As no database is exhaustive, consider including searches in other databases. Pubmed is often used for finding clinical information. If you plan a systematic review, we advise you to search MedLine (Ovid).
Research question: Symbiotic relationships between Lepiotaceaean fungi and leaf-cutter ants.
Database: BIOSIS Previews, which is a part of Web of Knowledge. Relevant for the life sciences.
Date of search: 10. January 2014
Divide your search into several steps, search separately for relevant keywords according to research topic, combine and refine afterwards.
Step 1: Searching for leaf cutter ants and Lepiotaceaea fungi
Make sure to include various spellings and relevant synonyms of a search term or search concept. In this case, Latin names as well as common names are combined in a Boolean OR search. Add a wildcard (here *) to the stem of a word to include all forms. In BIOSIS, searching by topic means searching by title, abstract or keywords. The list below shows search expressions and number of results in parentheses.
Search 1: Topic=(Atta OR Acromyrmex OR (leaf cutt* ant*) OR (leafcutt* ant*)) (1529)
Search 2: Topic=(fungus OR fungi OR lepiotaceae*) (488401)
Step 2: Combining search results
Go to your search history and combine the previous two searches with the Boolean operator AND. Searches are marked with the hash symbol (#) in BIOSIS.
Search 3: #1 AND #2 (385)
Step 3: Refine results to type of work
The left panel in the result list of the search interface lets you refine your search. Reviews summarize previous research and may be a good starting point. Choose Literature Types and tick off Literature Review if you would like to start with the eight papers retrieved in our example.
Search 4: Refined by: Literature Types=(LITERATURE REVIEW) (8)
For our topic Symbiotic relationships between Lepiotaceae fungi and leaf-cutter ants, a total of 385 documents were retrieved (Step 2) using BIOSIS on 10 January 2014.
While it is tempting to include a third search concept, symbiosis, this would exclude the documents dealing with symbiosis but without mentioning it in the title, abstract or keywords. If a paper is about both fungi and ants, it is also likely to relate to symbiosis.
For an initial check of relevance of the retrieved documents, try browsing titles, year of publication, contributing authors and journals involved. As no database is exhaustive, consider including searches in other databases. Keep revising your search strategy and also consider setting up an alert.
Research question: The impact of having divorced parents on self-esteem in adolescents.
Database: PsycINFO (Ovid). Relevant for psychology, health sciences, social sciences and education.
Date of search: 24. April 2019
Divide your search into several steps. Identify the main concepts in your research question and search separately for relevant keywords for each concept. In this example self-esteem, adolescents, and divorced parents are the main concepts. Individual searches for main concepts and synonymes/ alternative spellings of search terms within a main concept are combined. You can then consider using various possibilities to refine your search result.
Step 1: Searching for “self-esteem”
Make sure to include various spellings and relevant synonyms of a search term or search concept. You might want to consider whether the concepts “self-respect” and/or “self-confidence” ought to be included. Many databases treat the spelling alternatives “self esteem” and “self-esteem” identically. It is recommended that you check if both alternatives need to be included. It is also easy to test whether a search term consisting of more than one word should be delimited with quotation marks.
Check the thesaurus for possible subject headings that match your main search terms well enough to be included as search criteria. In PsycINFO, a subject heading search will be marked with a slash (/) in the search history. The Search History in the database presents the individual searches representing the main concept, “self-esteem”, and you combine them using the Boolean operator OR. The list below shows spelling variants of “self-esteem” and the numbers of results in parentheses.
Search 1: self-esteem.ti,ab. (41932)
Search 2: selfesteem.ti,ab. (83)
Search 3: self esteem/ (24471)
Search 4: Search 1 OR Search 2 OR Search 3 (46039)
Step 2: Searching for adolescents
Truncating after the ‘n’ in adolescen* will include adolescent, adolescents and adolescence. Truncating after the ‘n’ in teen* will include teens, teen-ager/-s and teenager/-s.
You might also want to consider including the search terms youth*, “young people” and student*.
Search 5: adolescen*.ti,ab. (220286)
Search 6: teen*.ti,ab. (20825)
Search 7: adolescent development/ (45687)
Search 8: Search 5 OR Search 6 or Search 7 (236189)
Step 3: Searching for divorced parents
Truncating after the ‘e’ in divorce* will include divorced (which will also identify “divorced parents” and “parents who are divorced”). Including terms other than divorce might be a good idea if it is not an important point that there has been a marriage before the break-up of the parents’ relationship. PsycINFO allows for truncation within a compound search term, such as “parent* breakup” which will include parental breakup, parents breakup and parent breakup. The searches representing the concept of “divorce” are combined with OR (Search 14).
Search 9: divorce*.ti,ab. (16833)
Search 10: “broken home*”.ti,ab. (528)
Search 11: “parent* breakup”.ti,ab. (10)
Search 12: “parent* break-up”.ti,ab. (15)
Search 13: divorce/ (8244)
Search 14: Search 9 OR Search 10 OR Search 11 OR Search 12 OR Search 13 (18194)
Step 4: Combining search results
Go to your search history and combine previous searches with the Boolean operator AND.
Search 15: Search 4 AND Search 8 AND Search 14 (106)
Step 5: Refining results to type of work
You can refine your search by choosing Additional Limits. If, for example, you want to identify the reviews among your search results, choose Methodology and mark the following options: 0800 Literature Review, 0830 Systematic Review, 1200 Meta Analysis and 1300 Metasynthesis. Limiting our research question by these types of applied methodology returns 3 documents.
Search 16: limit 15 to (0800 literature review or 0830 systematic review or 1200 meta analysis or 1300 metasynthesis) (3)
For our research question “The impact of having divorced parents on self-esteem in adolescents”, a total of 106 documents were retrieved (Step 4) using PsycINFO on 18th March 2019. Keep revising your search strategy and consider setting up an alert. As no database is exhaustive, consider including searches in other databases.
When working with systematic reviews, the search is an important part of the method and needs to be conducted according to specific criteria. The search will provide the data set for the investigation undertaken in the review. Documentation of the search will be published with the review.
Prepare your search
Preparing the literature retrieval process is crucial, as this will be the foundation for your further work.
- Identify the main elements of your research question
Defining which elements from your research question to include in the search will help you choose search terms effectively. Too few elements will lead to an unspecific search and many irrelevant hits. Too many elements can make it difficult to find anything at all, or lead to a biased selection of studies.
- Make a choice of relevant and accessible databases to search
Your library may subscribe to relevant databases that you are not aware of. Choose as many relevant ones as necessary. Even though there may be overlap in the search results, you will probably gain additional references. Consult the manuals of the chosen databases. Here you will find information about such advanced search options as truncation symbols, Boolean operators and syntax in each database.
- Unpublished material
In addition to databases of published research there are several databases of registered clinical trials. Here you can discover unpublished results from completed trials, or you can decide to wait for the results to be published.
- Grey literature
‘Grey literature’ is a common name for information or publications not appearing in regular channels for scholarly communication. These publications are typically from research institutes or government bodies, and are often not registered (or indexed) in reference databases. The publications can be located on the institutional websites, in library catalogues or in designated databases for grey literature.
- Strategies for tracing references
To be as exhaustive as possible, making sure all relevant literature is included, the reference lists of relevant studies should also be traced. Read more about Tracing references in Searching.
- Find search terms
In order to capture all previous research, use an exhaustive set of search terms. Include synonyms and use thesauruses and the subject headings relevant to the selected databases for your overall search.
In searching, we distinguish between text words and subject headings, which are in different searchable fields of a reference. Text words (also called free text) are an author’s own written words that appear in the title and abstract. Subject headings describe the content of an article and are added by the database providers; these are often standardized.
Strategies for finding search terms
- Off the top of your head. If you know the field, you know several relevant terms.
- Scan journal articles. A good exercise is to analyse the text of articles on the topic to identify common terms used in the field. Check for both words in the text and author’s keywords.
- In addition, do some test searches. See if any unexpected but relevant terms appear.
- Search dictionaries and thesauruses. The same concepts may be described differently in different thesauruses; they might have different subheadings or include several related concepts.
|Second-hand smoking||Tobacco Smoke Pollution||Passive smoking|
|Vaporizers||Nebulizers and Vaporizers||Vaporization|
This example shows subject headings in Medline (which uses MeSH) and Embase (which uses Emtree), using different terms to describe the same concepts.
- Look up relevant references in databases and see what subject headings they are indexed with. Check the same reference in databases with different thesauruses.
- Take advantage of others’ published search strategies. It has become more common to publish search strategies in review articles. Search for reviews on your topic or parts of it, and see if there is anything you can use. Systematic reviews from the Cochrane Collaboration usually present their full search strategies in an appendix to their published review.
- Text analysis programmes. There are various text analysers available online which can help you discover even more search terms. Reference managers such as Endnote also generate a term list based on library references, which can be useful.
- Be careful not to overlap search terms.
|women AND pregnancy||No need to search for women as only
women get pregnant.
|pregnancy AND preeclampsia||No need to search for pregnancy as
preeclampsia only occurs in pregnancy.
Language, publication year, demography and methodology filters
If your research question allows for limitations, these may be applied in the literature search as well. Most databases allow limitations by year, language, demography, research design, etc. Consider using these for increased precision.
Use a gold standard
If you have a set of key articles on your topic, use them to test your search. First, check whether they are indexed in the database, then check whether they come up in your search. If they are not included in the search, try to find out why and revise your strategy. Missed search terms? Too narrow limitations?
Test and revise your search
Part of the process of searching is to revise your search several times. By including or excluding terms and testing your search, you will end up with what best suits your research question.
Peer review of search strategy
When critically appraising a systematic review it is important to determine whether all relevant key studies have been included. This can be assessed by evaluating the literature search, that is, the sources and search terms used. Similarly, you can have a colleague or information specialist peer-review your search strategy during the review process to help improve the quality of the review. Others might identify missed terminology, syntax errors and other mistakes which are easily overlooked.
The PRESS checklist (Peer Review of Electronic Search Strategies) has been developed to identify elements of accuracy of literature searches.
Translate a search between databases
The precise way to build up a search strategy will depend on the researcher and on the database. Here we present an example of how the process of developing a search strategy can look and how the literature searches will look in different databases.
In the following, our research question is: Does exposure to smoke from e-cigarettes increase the risk of obstructive lung disease?
In this example, we will search PubMed, Embase and Web of Science. In a real-life systematic review on the topic, we would search more databases.
The main elements of this question are
- electronic cigarettes
- obstructive lung diseases
- second-hand smoking
To search as thoroughly as possible we need all the terms describing these elements. To systematize this work, all the words are organized in the table below.
- The most obvious search terms here are words from our research question, and they will be the starting point for the search. These words will be used as text words both in the singular and plural, truncated with an asterisk * in the table.
- Other natural language words and abbreviations should also be included as text words.
- We then search the MeSH database for relevant MeSH-terms and Embase for Emtree terms. These subject headings are also used as text words, as subject headings and text words are placed in different fields of a reference.
- By searching and surfing the databases, we can identify additional search terms. In this example, when searching for electronic cigarettes we stumbled upon an article on electronic nicotine delivery systems, which was a new term for us. As it is the ‘electronic nicotine’ that is the core of our question and not the ‘delivery system’, we chose to use only ‘electronic nicotine’.
- ‘Vaporizers’ is also a common term used for the smoking equipment. However, as vaporizers are also used in other contexts (e.g. inhaling medications), it is important to relate the term to smoking or nicotine to prevent retrieval of irrelevant references. Therefore, smoking/nicotine must be combined within brackets/parentheses with AND, or their respective search lines can be combined with AND.
- ‘COPD’, ’airway disease’, ‘lung disease’ and ‘pulmonary disease’ are synonyms to ‘obstructive lung disease’ and are therefore included as alternative text words.
MeSH-terms Emtree-terms Text words Electronic cigarettes -Electronic Cigarettes -electronic cigarette -electronic cigaret*
Vaporizers and tobacco smoking -Nebulizers and Vaporizers
-Tobacco Use Cessation Products
– smoking cessation
– smoking cessation program
Passive smoking -Inhalation Exposure
-Tobacco Smoke Pollution
– passive smoking
– passive smoking
– tobacco smoke pollution
– second hand smoking
Lung disease – Asthma
-Pulmonary Disease, Chronic Obstructive
– Lung Diseases, Obstructive
– obstructive airway disease
– chronic obstructive lung disease
– obstructive pulmonary disease*
– obstructive airway disease*
– obstructive lung disease*
- By combining text words and subject terms on the same concept with OR you will get either of the variations and probably retrieve all literature concerning that concept.
- Then combine all search terms across concepts with AND.
- Our search terms are then applied in the chosen databases. The database syntax is considered and we adapt the search strategy for each specific database.
- As electronic cigarettes are quite a new phenomenon, we have not applied any publication limitations. The topic will limit itself by date.
- To retrieve all literature on the topic we have not applied any language filters.
The following searches are in PubMed, Embase Ovid and Web of Science can look like this
Date of searches: 6th October 2015.
Example PubMed search strategy
|#1||electronic cigarettes[MeSH Terms]||351|
|#2||electronic cigaret*[Title/Abstract] OR e-cigaret*[Title/Abstract] OR electronic nicotine[Title/Abstract] OR e-nicotine[Title/Abstract]||2299|
|#3||#1 OR #2||2332|
|#4||Nebulizers and Vaporizers[MeSH Terms]||8887|
|#5||vapor*[Title/Abstract] OR vapour*[Title/Abstract] OR vaper*[Title/Abstract] OR vaping[Title/Abstract] OR vaporizer*[Title/Abstract]||41836|
|#6||#4 OR #5||50309|
|#7||Tobacco products[MeSH Terms] OR Smoking[MeSH Terms] OR Tobacco Use Cessation Products[MeSH Terms] OR Tobacco use[MeSH Terms] OR Smoking cessation[MeSH Terms] OR Nicotine[MeSH Terms]||149671|
|#8||smoking[Title/Abstract] OR nicotine[Title/Abstract] OR tobacco[Title/Abstract]||2327822|
|#9||#7 OR #8||238856|
|#10||#6 AND #9||946|
|#11||#3 OR #10||3059|
|#12||Inhalation Exposure[MeSH Term] OR Tobacco Smoke Pollution[MeSH Term]||17475|
|#13||passive[Title/Abstract] OR exposure[Title/Abstract] OR pollution[Title/Abstract] OR second-hand[Title/Abstract]||735510|
|#14||#12 OR #13||741069|
|#15||Asthma[MeSH Terms] OR Bronchitis[MeSH Terms] OR Lung Diseases, Obstructive[MeSH Terms] OR Pulmonary Disease, Chronic Obstructive[MeSH Terms]||178505|
|#16||Asthma[Title/Abstract] OR Bronchitis[Title/Abstract] OR COPD[Title/Abstract] OR obstructive pulmonary disease*[Title/Abstract] OR obstructive airway disease*[Title/Abstract] OR obstructive lung disease*[Title/Abstract]||174515|
|#17||#15 OR #16||230092|
|#18||#11 AND #14 AND #17||77|
Example Embase Ovid search strategy
(Electronic cigaret* OR E-cigarette* OR electronic nicotine OR e-nicotine).tw.
1 OR 2
(vapor* OR vapour* OR vaper* OR vaping OR vaporizer*).tw.
4 OR 5
Smoking/ OR smoking cessation/ OR smoking cessation program/ OR nicotine/
(Tobacco OR Nicotine OR Smoke OR Smoking).tw.
|9||7 OR 8||407264|
|10||6 AND 9||1264|
|11||3 OR 10||2517|
passive smoking/ OR exposure/
(passive OR pollution OR second-hand OR exposure).tw.
|14||12 OR 13||964912|
obstructive airway disease/ OR asthma/ OR bronchitis/ OR chronic obstructive lung disease/
(asthma OR COPD OR obstructive pulmonary disease* OR obstructive airway disease* OR obstructive lung disease*).tw.
|17||15 OR 16||342694|
|18||11 AND 14 AND 17||123|
Note: The / indicates a subject heading from Emtree and .tw. indicates text words.
Example Web of Science search strategy
|1||TOPIC: (e-cigarette* OR Electronic Cigarette* electronic nicotine OR e-nicotine)||973|
|TOPIC: ((vapour* OR vaper* OR vaping OR vaporizer*) AND (tobacco product* OR smoking OR Tobacco Use Cessation Product*s OR tobacco use OR smoking cessation OR nicotine))||357|
|3||TOPIC: (Tobacco Smoke Pollution OR second hand OR exposure OR passive)||995 886|
|4||TOPIC: (Asthma OR bronchitis OR COPD OR (obstructive pulmonary disease*) OR (obstructive airway disease*) OR (obstructive lung disease*))||199 043|
|5||1 OR 2||1258|
|6||3 AND 4 AND 5||35|
Note: Web of Science does not have a subject heading index, therefore all search terms are text words.
Processing the search result
There are 228 retrieved references in these searches. As this is a relatively small number, we chose not to limit our search to observational studies that investigate risk and causation. Had the number of hits been high, a relevant step would be to add a methodological filter to narrow our search to observational studies (via the Clinical Queries option available in PubMed and Embase). What constitutes a high number of hits is a subjective judgement: how many references do you want to review vs how specific should the search be?
While searching multiple databases, importing the result to a reference management program can help you create an efficient workflow. The reference manager can remove duplicate references, help you sort relevant references and organize your further work.
- The three databases have a total of 235 references (respectively 123, 77 and 35).
- After automatic deduplication in Endnote there are 166 references left.
- After manual deduplication there are 158 references left.
- The final result is 158 references.
- A high number of references will be excluded by employing the inclusion and exclusion criteria when screening the title and abstract, and by reading the full text of articles/publications.
- The final list of studies are those included in your review.
Document your search
Documentation of the search makes your research reproducible and indicates that your methodology is sound. You would want to describe the search strategy that led to the included studies.
- names of databases
- which vendor/host of databases
(did you use Medline through PubMed or Ovid? Embase on Ovid or Embase.com?)
- last date of the search or alert service
- the search terms used. Include both subject headings and text words
- how the search terms were combined
- the limitations where used: publication dates, study designs, languages, etc.
- whether reference lists were reviewed
- number of retrieved records
- number of internal duplicates/number of duplicates removed
- number of articles after removal of duplicates
Check the journals’ instructions for authors for any specific requirements on how to report the search strategy. It can also be a good idea to look at published reviews in your preferred journal to see what is customary. For instance, some journals encourage the full search strategy in an appendix.
Documenting the search strategies can be done by:
- Describing the search strategy in plain text in the methodology section of your manuscript.
This is where you inform your readers how you made your choices and proceeded.
- Documenting precisely how your literature search was performed, by saving or copying the search strategy from each database. This will fit in an appendix to your article and can be presented in a similar way to the examples in ‘Translate a search between databases’.
- A flow chart can illustrate the workflow from the initial search results to the final studies included. The PRISMA flow chart is a well-known chart for documenting the search process. The PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines are a set of minimum criteria for reporting systematic reviews and meta-analyses. It includes a chapter on reporting the literature search.
Example on how documentation can be done in a section of an article
Studies included in this review were located from searching PubMed, Embase Ovid and Web of Science (last search 6th October 2015). The literature search included MeSH-terms and text words in the following combination: electronic cigarettes OR tobacco vaporizers AND tobacco exposure AND obstructive lung diseases. The full search strategy in PubMed included the following MeSH-terms and text words: (electronic cigarettes[MeSH Terms] OR electronic cigaret* OR e-cigaret* OR electronic nicotine) OR ((Nebulizers and Vaporizers[MeSH Terms] OR vapor* OR vapour* OR vaper* OR vaping OR vaporizer*) AND (Tobacco products[MeSH Terms] OR Smoking[MeSH Terms] OR Tobacco Use Cessation Products[MeSH Terms] OR Tobacco use[MeSH Terms] OR Smoking cessation[MeSH Terms] OR Nicotine[MeSH Terms] OR smoking OR nicotine ))) AND (Inhalation Exposure[MeSH Term] OR Tobacco Smoke Pollution[MeSH Term] OR passive OR exposure OR pollution OR second hand) AND (Asthma[MeSH Terms] OR Bronchitis[MeSH Terms] OR Pulmonary Disease, Chronic Obstructive[MeSH Terms] OR Asthma OR Bronchitis OR Chronic Obstructive Pulmonary Disease* OR COPD OR Obstructive pulmonary disease* OR Chronic Obstructive airway Disease* OR Obstructive airway disease* ).
The search was adapted to Embase with the following Emtree terms: electronic cigarette, vaporization, smoking, exposure, obstructive airway disease, asthma, bronchitis and chronic obstructive lung disease, and to Web of Science by searching by topic. There were no restrictions on language or publication dates. The reference lists of the included articles were screened for additional references.
For further input and advice on systematic searching, consult your library, help pages in databases and literature on the topic. Some libraries offer courses in systematic searching, or they will do such searches on request.
Haraldstad, A. M., & Christophersen, E. (2015). Literature searches and reference management. In P. Laake, H. B. Benestad, & B. R. Olsen (Eds.), Research in medical and biological sciences from planning and preparation to grant application and publication (pp. 125–166). Amsterdam: Elsevier.
Higgins, J. P. T., Green, S., & Cochrane, C. (2008). Cochrane handbook for systematic reviews of interventions. Chichester: Wiley-Blackwell.