Systematic Review

While Cari doesn’t set any date limits on her searches, she documents the oldest article found in each search, or the date range given by the database, where available. As noted, once all the records are loaded into EndNote, Cari de-duplicates them, ensuring that only one record of each article is used in the screening process.

She notes that this is difficult, as in a few places author names are presented differently, especially those with accented names. (In some cases the same article was found in four or more databases, while other articles were only located in a single database).

Cari downloads key information about each of her selected articles into a spreadsheet, so she can easily print it and share it, and so she  and Lee have separate versions of the document.

Now that Cari and Lee are ready to start screening their articles, she adds columns to the spreadsheet for each criterion, so they always know why an article screened out, and at what stage. Here’s what the table looked like at first:

Cari and Lee put an ‘E’ in a column indicating why an article was excluded at title- screening stage. The last column indicates whether or not an article would proceed to the abstract screening phase (Y) or not (N).  This allows the spreadsheet to be sorted or filtered quickly. (Lee set up the spreadsheet to automatically change ‘Y’ to ‘N’ if an article was excluded for some reason, using a ‘nested if’ function.

Initially, Cari planned to do title and abstract screening at the same time, but she quickly found it was easy to get distracted, so she decided to start with the titles, and then move on to the abstracts for those articles passing the first check. This meant that an article should only be excluded if there was a clear reason to do so; if there wasn’t a clear reason to exclude an article based on its title, it would proceed to the next stage.

Here are a few of the articles found by Cari and Lee. If you would exclude the article, drag it to the relevant column indicating the reason for exclusion. If you would include the article for full-text screening, drag it to the ‘Include’ column. Remember, Cari and Lee only excluded articles that were clearly not relevant.

(Cocoa extracts include polyphenols, flavonoids, flavanols, catechins and epicatechins). The journals below all use peer review for their research articles.

Articles that were noted to be reviews, but which were on closely related topics, were used for forward searching. Cari and Lee skimmed through these review articles for mentions of PICO keywords, checking out any related articles cited. Cari and Lee were relieved to see that they had already collected many of the sources mentioned, but the forward search still added a further 27 articles to the pool to be screened. These additional papers were often published in less popular, or more specialised, journals, so were not found in the main databases. Working with their subject librarian, they were able to track down the full texts of these articles so they could be screened along with the others. 

 

 

Extract from Rimbach, Melchin, Moehring and Wagner (2009), who critically reviewed the role of cocoa polyphenols in vascular health. The review included a discussion of studies of cocoa polyphenols in humans (p. 4299), which included a number of references to primary studies. Cari and Lee used this to help them find relevant  additional articles in the review’s reference list. 

Cari and Lee then repeated the screening process for articles that had passed the title screening. The abstracts now needed to be read before a decision could be made. Abstracts were compared against the same inclusion and exclusion criteria used for the title screening. In some cases a bit more detail was available that indicated an article should be excluded, e.g. the abstract revealled the study had been an animal trial, used extracts rather than whole-cocoa products, or was based on mathematical modelling rather than human trials.  If an article clearly didn’t meet the inclusion criteria, it was excluded, otherwise it proceeded to the final full-text screening stage. 
A further 83 articles were excluded through abstract screening. > resulting in …. articles 

Once the located article titles and abstracts had been screened, it was necessary to look in detail at those articles remaining. The initial screening steps had greatly reduced the number of articles requiring this detailed reading phase. 

During the full-text screening stage, Cari and Lee read through each article in detail, looking to see exactly what was recorded and compared, and gathering more information about the sampling and interventions used. A few articles were rejected at this stage, when more detail revealed, e.g., that the intervention had been part of a wider dietary intervention, or that the necessary cholesterol data were not measured.