IDENTIFYING RELEVANT LITERATURETHE REVIEW SHOULD STATE HOW PRIMARY...

2. IDENTIFYING RELEVANT LITERATURE

The review should state how primary accuracy studies were identified. This is done in

several steps. These steps should be documented and their conduct should be

transparent. Typically, once the question has been formulated, the next step is to

construct a strategy for electronic database searching. Search strategy should

explicitly state how widely the internet has been cast in an attempt to identify

primary studies. These may include, in addition to searching electronic databases,

searching the grey literature, searching the reference lists of primary studies and

review articles, and contacting the experts (and manufacturers of the test) for

unpublished studies. There should be no language restriction. Restriction in the

search, either of databases or of languages, has potential to bias accuracy reviews,

(4).

General guidelines on methods of electronic searching are available (5, 6, 7).

Essentially, it consists of formulation of an appropriate combination of search terms,

pilot searches to refine the search term combination, selection of relevant databases

(e.g. Medline, Embase, Pascal, Biosis, and BioBase) and citation retrieval from the

refined searches for selection of potentially relevant citations. This is done by

scrutinising the title and abstract of citations retrieved from the electronic searching

using selection criteria derived from the review question Table 1.

Narrative question

Among pregnant women, what is the accuracy of cervico-vaginal fetal fibronectin test

in predicting preterm birth?

Structured question and selection criteria

Population

Pregnant women at low or high risk of preterm birth

(The people at risk of having the condition of interest)

Test

Antenatal cervico-vaginal fetal fibronectin

(The test which purports to predict the presence or absence of the

condition)

Spontaneous birth with known gestation either at term or preterm

Gold

standard

(The condition of interest whose existence is confirmed or refuted

beyond reasonable doubt independently of the test being evaluated)

Full papers of all potentially relevant citations are examined to make final inclusion

and exclusion decisions based on the explicit selection criteria. The process of

literature identification can be a long and drawn out one. An example flow chart

representing this process is shown in Figure 2. A flow chart for identification of the

literature,

Once potentially relevant papers have been obtained, information is then extracted

on methodological quality and accuracy data.